SOCIAL AND FISCAL FRAUD IN BELGIUM A PILOT STUDY ON DECLARED AND UNDECLARED INCOME AND WORK: SUBLEC

SOCIAL AND FISCAL FRAUD IN BELGIUM A PILOT STUDY ON DECLARED AND UNDECLARED INCOME AND WORK: SUBLEC

Studies over sociale en fiscale fraude Dit onderzoeksrapport kwam tot stand in opdracht van de POD Wetenschapsbeleid ten behoeve van de FOD Sociale Zekerheid in het kader van het programma ‘Actie ter ondersteuning van de strategische prioriteiten van de federale overheid’. Dit programma werd in het leven geroepen om snel en efficiënt te kunnen inspelen op de behoeften van de federale overheidsdepartementen inzake gerichte onderzoeksacties van bepaalde duur (6 maanden tot 1 jaar) en/of verkennend onderzoek met betrekking tot strategische gebieden. Het betreft een “horizontale” actie: ze staat open voor de financiering van onderzoeksprojecten binnen de verschillende beleidsthema’s die in het kader van de regeringsbeslissingen naar voren worden geschoven.

Etudes sur la fraude sociale et fiscale La présente recherche a été commanditée et financée par le SPP Politique scientifique en appui à la politique du SPF Sécurité sociale dans le cadre du programme ‘Action en soutien aux priorités stratégiques de l’autorité fédérale’. Ce programme est conçu pour répondre rapidement et efficacement aux besoins des départements de l’Autorité fédérale en matière d’actions de recherche ciblées d’une durée déterminée (6 mois à 1 an) et/ou d’actions d’investigation concernant des domaines stratégiques. Il s’agit d’une action “horizontale”, elle est ouverte aux projets de recherche au sein des différents thèmes de politique mis en avant dans le cadre des décisions gouvernementales.

“The more people know about fraud, the more they discuss it, and the better society can fight it.” (OLAF, Deterring Fraud by Informing the Public, 2005, 2006).

Social and fiscal fraud in Belgium A pilot study on declared and undeclared income and work: SUBLEC

Jozef Pacolet, Sergio Perelman, Frederic De Wispelaere, Jérôme Schoenmaeckers, Laurent Nisen, Ermano Fegatilli, Estelle Krzeslo, Marianne De Troyer and Sigrid Merckx

Dit onderzoeksrapport kwam tot stand in opdracht van de POD Wetenschapsbeleid ten behoeve van de FOD Sociale Zekerheid in het kader van het programma AGORA – AG/00/137: SUBLEC – De organisatie van een microsurvey met het oog op een beschrijvende en verklarende analyse van de problematiek van de sociale en fiscale fraude. Ce rapport d’étude a été rédigé à la demande du SPP Politique scientifique pour le SPF Sécurité sociale dans le cadre du programme AGORA –AG/00/137: SUBLEC – Organisation d’une micro-enquête en vue d’effectuer une analyse descriptive et explicative de la problématique de la fraude sociale et fiscale.

Acco

Leuven / Den Haag

Eerste druk: 2012 Gepubliceerd door Uitgeverij Acco, Blijde Inkomststraat 22, 3000 Leuven (België) E-mail: [email protected] – Website: www.uitgeverijacco.be Voor Nederland: Acco Nederland, Westvlietweg 67 F, 2495 AA Den Haag E-mail: [email protected] – Website: www.uitgeverijacco.nl Omslagontwerp: www.frisco-ontwerpbureau.be © 2012 by Acco (Academische Coöperatieve Vennootschap cvba), Leuven (België) Niets uit deze uitgave mag worden verveelvoudigd en/of openbaar gemaakt door middel van druk, fotokopie, microfilm of op welke andere wijze ook zonder voorafgaande schriftelijke toestemming van de uitgever. No part of this book may be reproduced in any form, by mimeograph, film or any other means without permission in writing from the publisher. D/2012/0543/154

NUR 756

ISBN 978-90-334-8923-5

Contents

Ten geleide

9

Avant-propos

11

Samenvatting

13

Résumé

27

Foreword

41

Abbreviations

43

Introduction

45

Chapter 1. Scope of the survey and applied methodology

47

1. 2. 3.

4. 5.

6.

The definition of the underground economy The choice for a face-to-face survey Development of the questionnaire 3.1 Scope 3.2 Strategy 3.3 Communication of the questions Sample selection Roll-out 5.1 Pretest 5.2 Briefing of the interviewers 5.3 Method of contacting the sample 5.4 Face-to-face interviews 5.5 Debriefing Data weighting method

47 50 52 52 52 53 54 55 55 57 57 58 58 58

6

X

Contents

Chapter 2. Characteristics of respondents and non-respondents

61

1. 2. 3. 4.

61 61 66 67

The different characteristics First analysis Logistic Regression The impact of a monetary incentive

Chapter 3. Socio-demographic and socio-economic information

69

1. 2.

69 71

Socio-demographic information Socio-economic information

Chapter 4. Demand for undeclared work 1. 2.

3. 4.

General Most important categories of expenditures 2.1 Construction and housing 2.2 Car repairs 2.3 Domestic work and service vouchers Probit analysis Conclusion

Chapter 5. Supply of undeclared work 1.

2.

3. 4. 5.

General 1.1 Frequency 1.2 Intensity and volume Specific socio-economic categories 2.1 Employees 2.2 Social benefit recipients Coherence between the supply of and the demand for undeclared work Probit analysis Conclusion

Chapter 6. Other forms of fiscal fraud 1. 2.

3. 4.

General Frequency and volume of several other forms of fiscal fraud 2.1 Frequency 2.2 Volume Probit analysis Conclusion

77 77 82 82 85 86 87 90

91 91 91 91 95 95 99 100 101 102

105 105 107 107 108 109 110

Contents

Chapter 7. Opinion questions 1.

2. 3. 4. 5.

Opinion on undeclared work 1.1 In the SUBLEC-sample 1.2 Comparing the SUBLEC sample with the experimental group of students 1.3 Comparing SUBLEC with the Eurobarometer 1.4 Opinions integrated in the fraud triangle Social benefit fraud Fiscal fraud General Conclusion

W

7

113 113 113 115 115 119 122 123 125 126

Chapter 8. Lessons learnt from the survey

127

1. 2. 3. 4.

127 128 129 129

Influence of the data request on the results Methodology of interviewing The use of a face-to-face interview Sensitivity of the questions

Chapter 9. Summary and conclusions

133

1. 2. 3.

134 134 136

Strengths and weaknesses of the present pilot study Some summary of first tentative results Relevance for policy making and how to continue

Annex 1. Programme kick-off seminar SUBLEC – 25 February 2008

139

Annex 2. Steering committee

141

References

143

Ten geleide

Onder het motto “De aanhouder wint”, bevestigt de FOD Sociale Zekerheid met dit derde deel in onze reeks “Studies over sociale en fiscale fraude” zijn vaste wil om het debat over fenomenen zoals “zwartwerk” en “ondergrondse economie” levendig te houden. De FOD wil daarbij innovatief te werk gaan Voor dit boekdeel werd gekozen voor een directe onderzoeksmethodologie. Een enquête werd het instrument voor een rechtstreekse bevraging van de Belgische bevolking. Een gevarieerde doorsnede van de Belgische bevolking (actieven, niet-actieven, werknemers, zelfstandigen en uitkeringstrekkers) werd bevraagd naar hun gedragingen en hun houding ten opzichte van sociale en fiscale fraude. Deze enquête, die zowel kwalitatieve als kwantitatieve vragen bevatte, is uniek in zowel haar onderwerp als haar reikwijdte. Door de bevolking rechtstreeks te ondervragen over een dergelijk gevoelig onderwerp en door te erkennen dat sociale en fiscale fraude tegelijk behandeld kunnen worden, draagt deze studie bij tot het in kaart brengen van deze moeilijk meetbare fenomenen. Door een inzicht te verschaffen in de mentaliteit van de bevolking en door bij te dragen aan geloofwaardig cijfermateriaal, hoopt de enquête de beleidsmakers te ondersteunen in hun strijd tegen fraude. De FOD wil ook het debat opentrekken. Fraude valt moeilijk af te bakenen binnen een bepaald domein. Het is moeilijk om sociale en fiscale fraude af te lijnen en de parallellen tussen beide aspecten valt niet te ontkennen. Daarom ook wordt bij deze studie geopteerd om te peilen naar het “aangeven en niet-aangeven van inkomen en arbeid”. Hiermee hopen wij de samenwerking tussen de verschillende actoren nog sterker te maken en het beleid hiertoe effectief te ondersteunen. Daarnaast kunnen wij ook niet naast de Europese dimensie kijken. “Undeclared work” en “tax evasion” zijn ook binnen de Europa 2020 strategie een belangrijke factor, ze bedreigen immers de herlancering van de economie en de bevordering van werkgelegenheid. Europa is vragende partij om informatie over de omvang van de fiscale en sociale fraude vanuit de Lidstaten te ontvangen. Met dit initiatief dragen we hieraan bij. Tot slot kunnen we niet naast de maatschappelijke relevantie van het debat kijken. De ongunstige economische en financiële situatie maakt dat fraudulent gedrag meer in de kijker geplaatst wordt. Studies zoals “SUBLEC” geven een gezicht aan fraude en fraudeurs, wat dan weer de bespreekbaarheid verhoogt en op termijn een meer gerichte aanpak moet bewerkstelligen. Vertrouwen in de overheid is hierbij cruciaal.

10

X

Ten geleide

Dit alles zou niet kunnen zonder de medewerking van vele betrokkenen. In de eerste plaats bedanken we de POD Wetenschapsbeleid, die deze studie financieel ondersteund heeft, maar ook de talrijke andere leden van de stuurgroep die mee richting hebben gegeven aan deze studie. Hierbij is het verhaal echter niet ten einde. De FOD Sociale Zekerheid zal zijn rol van voortrekker in dit debat blijven opnemen en zal verschillende nieuwe initiatieven blijven lanceren en ondersteunen. Didier Verbeke Adviseur-generaal Coördinator domein “Undeclared Work”

Avant-propos

Reprenant à son compte la devise “La persévérance vient à bout de tous les obstacles”, le SPF Sécurité sociale confirme, avec ce troisième numéro de la série “Etudes sur la fraude sociale et fiscale”, sa ferme volonté de continuer à entretenir le débat relatif aux phénomènes que sont le “travail au noir” et l’“économie souterraine”. Dans ce cadre, le SPF entend adopter une approche novatrice. Pour le présent numéro, c’est une méthodologie d’investigation directe qui a été choisie. Une enquête est devenue l’instrument d’une consultation directe de la population belge. Un échantillon diversifié de la population belge (actifs, non actifs, travailleurs salariés, travailleurs indépendants et bénéficiaires de prestations) a été interrogé à propos de ses comportements et attitudes par rapport à la fraude sociale et fiscale. Cette enquête, qui comportait des questions à la fois qualitatives et quantitatives, est unique de par son sujet et sa portée. En interrogeant directement la population à propos d’un sujet à ce point sensible et en admettant la possibilité de traiter conjointement la fraude sociale et fiscale, la présente étude contribue à dresser l’inventaire de ces phénomènes difficilement mesurables. Grâce à l’apport de connaissances relatives à la mentalité de la population et de chiffres crédibles, l’enquête espère soutenir les décideurs politiques dans leur lutte contre la fraude. Le SPF entend également élargir le débat. Il n’est pas aisé de délimiter la fraude dans le cadre d’un domaine bien déterminé. Circonscrire la fraude sociale et fiscale est une entreprise complexe et les parallèles entre les deux fraudes sont indéniables. C’est la raison pour laquelle il a été décidé pour cette étude d’investiguer sur les raisons qui poussent à “déclarer ou ne pas déclarer un revenu ou un travail”. Nous espérons de la sorte renforcer encore la coopération entre les divers acteurs et soutenir efficacement la politique à mener en la matière. Il est clair, par ailleurs, que la dimension européenne ne peut être omise. Car les facteurs “Undeclared work” et “Tax evasion” sont également importants dans le cadre de la stratégie Europe 2020, étant donné qu’ils mettent en péril la relance de l’économie et la promotion de l’emploi. L’Europe demande à recevoir des Etats membres des informations concernant l’ampleur de la fraude fiscale et sociale. La présente initiative entend y contribuer. Enfin, il n’est pas possible de passer à côté de la pertinence sociale de ce débat. La situation économique et financière défavorable a pour effet de focaliser davantage l’attention sur les comportements frauduleux. Des études telles que “SUBLEC” permettent de donner un visage à la fraude et aux fraudeurs, ce qui, partant, accroît la possibilité d’en débattre et

12

X

Avant-propos

doit à terme permettre une approche plus ciblée. A cet égard, la confiance en l’autorité est fondamentale. Cette étude n’aurait pas pu être menée à bien sans la collaboration de bon nombre d’intervenants. Nous tenons à remercier en premier lieu le SPP Politique scientifique, qui l’a soutenue financièrement, mais aussi les très nombreux autres membres du groupe de pilotage qui a co-orienté le cours de cette étude. L’histoire n’en est cependant pas terminée pour autant. Le SPF Sécurité sociale continuera à assumer son rôle de chef de file dans ce débat et à lancer et soutenir plusieurs nouvelles initiatives. Didier Verbeke Conseiller général Coordinateur du domaine “Undeclared Work”

Samenvatting Sociale en fiscale fraude in België. Een pilootstudie omtrent aangegeven en niet-aangegeven inkomen en arbeid

Inleiding Het meten van de ondergrondse economie lijkt een ‘mission impossible’. In de literatuur worden er verschillende onderzoeksmethoden naar voren geschoven, afgezien van het feit dat er ook nog uiteenlopende definities bestaan van ondergrondse economie en fraude. Zo kan men de omvang van de ondergrondse economie ramen op basis van macro-economische modellen, of op basis van een bevraging bij de bevolking, of van de bedrijven, of op basis van administratieve gegevens of via inschatting in de nationale rekeningen. Zij leveren uiteenlopende en vaak controversiële resultaten op. Ook de Europese Commissie heeft de voorbije jaren de verschillende mogelijkheden verkend, zonder tot een bevredigend eindresultaat te komen. Vermoedelijk zal enkel een combinatie van diverse methoden ons een stap vooruit helpen. Want om de strijd tegen de sociale en fiscale fraude te kunnen voeren, moet men wel weten wat de omvang van het probleem is, de verschijningsvormen, de oorzaken en motieven. In twee vroegere studies gepubliceerd in deze reeks werden deze ramingmethoden toegelicht en wordt er gepleit voor een ‘observatorium’ over de ondergrondse economie dat dient als observatiepost en draaischijf van informatie over de fraude en de strijd er tegen.1 In dit rapport publiceren wij de resultaten van een nieuwe enquête over het fraudegedrag van de Belgen, hun opinies daarover en hun motieven. Het onderzoek werd geïnitieerd en gefinancierd door de FOD Sociale Zekerheid en de POD Wetenschapsbeleid BELSPO.2 Het

1. 2.

Pacolet J., Perelman S., Pestieau P., Baeyens K. & De Wispelaere F. (2009); Pacolet J. & De Wispelaere (2009). Wij danken deze instanties voor de geboden onderzoeksmogelijkheden in het kader van het AGORA-project AG/00/137: SUBLEC – De organisatie van een microsurvey met het oog op een beschrijvende en verklarende analyse van de problematiek van de sociale en fiscale fraude. Bijzondere dank gaat hierbij uit naar Aziz Naji van BELSPO en Didier Verbeke en Koen Vleminckx van de FOD Sociale Zekerheid voor hun geloof in dit soort onderzoek en het vertrouwen dat zij ons schonken. Ook Chris Brijs van de Kruispuntbank van de Sociale Zekerheid was een onmisbare steun en toeverlaat voor de realisatie van dit project. Tevens willen wij de leden van de stuurgroep danken voor hun inhoudelijke suggesties en steun aan dit project. Voor allle mogelijke vergissingen en misinterpretaties in deze studie zijn uiteraard enkel de auteurs verantwoordelijk.

14

X

Samenvatting

werd een samenwerkingsverband tussen de KU Leuven (HIVA), de ULg (CREPP) en de ULB (METICES).3 Zowel de omvang van de bevraging, als de methode van de bevraging en de inhoud van de vraagstelling werd grondig voorbereid. Maar ‘tussen droom en daad staan ook wetten en praktische bezwaren’, zodat de oorspronkelijk beoogde ruime uitrol van de bevraging niet is kunnen gerealiseerd worden. De bevraging is beperkt gebleven tot een steekproef van een 246 personen. Daarom noemen wij het een pilootstudie. Dit is een beperking. Maar over de inhoud en de bruikbaarheid van het onderzoeksinstrument hebben wij geen twijfel. Daarom hebben wij deze bevraging verwerkt met al de beperkingen van dien, en maken wij het nodige voorbehoud bij bepaalde resultaten, maar durven wij toch ook bepaalde conclusies te formuleren ten behoeve van de strijd tegen de fraude. ‘Full reports of pilot studies are rare in the research literature’.4 Misschien zijn wij nu met voorliggend rapport bij die uitzonderingen, en is dit misschien te verklaren doordat wij oorspronkelijk gestart waren met een grootschalige versie. Maar het doel van elke rapportering is te leren uit de opgedane ervaring. Dit leermoment willen wij hier vasthouden.

Opzet In de zomer van 2010 werd de pilootstudie uitgerold omtrent de bevraging van het ‘aangegeven en niet-aangegeven inkomen en arbeid’. Het was het voorlopig eindpunt van een lange voorbereiding. Vooreerst werd de wenselijkheid en doenbaarheid nagegaan van een bevraging bij de bevolking om de omvang, structuur en determinanten van niet-aangegeven werk en inkomen te meten. Een ruime definitie van fraude is aangewezen: zowel fiscale fraude als sociale fraude werden bevraagd, en bij de sociale fraude kwam zowel bijdragefraude als uitkeringsfraude ter sprake. De voorbereidende werkzaamheden omvatten de analyse van het internationale gebruik van de enquête als instrument om de omvang van de fraude te meten, welke enquêtemethode het best werd gebruikt, hoe de steekproef diende getrokken te worden, welk soort vragen dienden gesteld te worden, welke toestemmingen er nodig waren om de steekproef te trekken met respect voor de privacy, en wie uiteindelijk de gebruiker zou kunnen worden van onze resultaten. Voor ons was dat in eerste instantie de diverse overheden en sociale parastatalen geconfronteerd met de strijd tegen de fraude. Dit rapport stelt het voorlopig eindpunt voor van dit proces. Om talrijke redenen was de uiteindelijke ‘roll out’ van de bevraging verschillend van wat het oorspronkelijk plan was. In plaats van een bevolkingsenquête, representatief voor de totale bevolking maar ook voor diverse relevante subcategorieën, werd de schaal gereduceerd tot de omvang van een fatsoenlijke pilootstudie die zou toelaten de methodologie van dergelijke bevraging verder op punt te zetten. Het doel was om uiteindelijk een brede definitie van de vraag en het aanbod van niet-aangegeven activiteiten, inkomen, sociale fraude, fiscale fraude, uitkeringsfraude,

3. 4.

HIVA, Onderzoeksinstituut voor Arbeid en Samenleving; CREPP, Center of Research in Public Economics and Population Economics; METICES, Migrations, Espaces, Travail, Institutions, Citoyenneté, Epistémologie, Santé. Van Teijlingen & Hundley, 2001.

Samenvatting

W

15

en de karakteristieken en mogelijke determinanten te identificeren, dit binnen de Belgische bevolking tussen 18 en 75 jaar, en naar socio-professionele groepen onderscheiden. De studie werd gelanceerd onder het acroniem SUBLEC ‘Survey on the black economy’, vragenlijst omtrent de ondergrondse economie. Dit was echter niet ingeperkt tot zwartwerk of ontduiking van sociale zekerheidsbijdragen, maar sloeg ook op uitkeringsfraude, en alle mogelijk vormen van fiscale fraude. Door de beperkte omvang is het een pilootstudie geworden, en is de bespreking maar partieel, tentatief en soms riskeert het zelfs om speculatief te worden. Het is in alle geval niet definitief. Het is eerder een validatie van de haalbaarheid van dit soort bevraging, de vragenlijst zelf en de maatschappelijke relevantie. Sommige van de inzichten zullen wij confronteren met een vroegere gelijkaardige studie op Europees, vlak, met name de Eurobarometer No 284 van 2007.5 Voor andere elementen, met intrigerende conclusies, zou verdere externe validatie verantwoord zijn, maar dit kan beter gebeuren wanneer de vragenlijst op grotere schaal zal worden uitgevoerd. De survey over niet-aangegeven inkomen en arbeid wenst de beleidsmakers te informeren over de omvang van deze fenomenen, en om aanbevelingen te maken hoe zij de strijd tegen de fraude kunnen verbeteren. Gezien de omvang van de enquête die vandaag voorligt zijn onze aanbevelingen eerder in de aard van hoe verder kan gegaan worden, zonder de definitieve antwoorden te geven (voor zover onderzoek ooit definitieve antwoorden kan geven) over de omvang, structuur en hoe de strijd te voeren. De aanbevelingen moeten als tentatief beschouwd worden. Soms zijn zij contra-intuïtief, soms contesteren zij bestaande evidentie en opinies, en zijn zij waardevol als hypothese voor verdere verificatie en onderzoeken. Op diverse plaatsen kunnen wij echter de relevantie voor het debat over zwartwerk en niet-aangegeven inkomen wel aantonen. Door de lengte van de vragenlijst en de oriëntatie naar bepaalde doelgroepen en deelfenomenen, welke zijn ingebed in de ene moedervragenlijst, zijn veel deelonderzoeken verborgen. Wanneer de vragenlijst op voldoende schaal zal worden uitgerold zal zij daar zijn bijkomend voordeel tonen. Wij bespreken hierna de sterkten en zwakten van het ontwerp en de roll out van onze vragenlijst, en de voornaamste resultaten die voorlopig zijn, maar illustratief voor de waarde voor het beleidsdebat. De slotconclusie is om verder te gaan langs de weg die wij ingeslagen waren met dit onderzoek. In voorliggend geval is de opdracht voor deze survey afkomstig van de FOD Sociale Zekerheid en de POD Wetenschapsbeleid. Zij waren overtuigd om deze methode ook in België toe te passen. Als onderzoeker moet men vaak de overheid overtuigen om bepaalde enquêtes te organiseren. Het is soms opboksen tegen vooroordelen dat het bevragen van inkomen, of vermogen, of in dit geval zwart inkomen en werk, niet doenbaar is. Dan moeten wij ons zelf en de collega’s en medewerkers overtuigen dat dit doenbaar is. En daarna moeten wij de enquêteurs overtuigen om de vragen te stellen, en te blijven aandringen bij de respondenten om deel te nemen aan de enquête. En dan moeten zij die respondenten overtuigen om op alle vragen te antwoorden. ‘Boost interviewers’ confidence about their ability to ‘sell’ the survey’ was de terechte aanbeveling die wij lazen in de vermogensenquête van de ECB (Europese

5.

European Commission (2007), Special Eurobarometer 284. Undeclared Work in the European Union, Brussels, 90 p.

16

X

Samenvatting

Centrale Bank).6 Eens al die stappen gelukt zijn, kan het resultaat nog aanzienlijk verbeterd worden via datacleaning of ‘editing’ en het ‘imputeren’ van ontbrekende informatie.7 Wij hopen met dit rapport aangetoond te hebben dat deze inspanningen ook hier zouden lonen.

Sterkten en zwakten van de huidige pilootstudie Op basis van een literatuuroverzicht concludeerden wij dat een ‘face to face’ mondelinge bevraging van de respondent het beste resultaat zou opleveren. Een gestructureerde steekproef was getrokken uit de populatie van de diverse parastatalen van de sociale zekerheid, gespreid over het gehele grondgebied. De vragenlijst werd door de drie equipes gezamenlijk opgesteld en in twee landstalen. Door de onderzoekers en nadien door een enquêteur werden proefenquêtes afgenomen waarna de vragenlijst werd gefinaliseerd. Er werd beslist om met een ‘open vizier’ de vragenlijst af te nemen. Deze werd dan ook aangekondigd als een ‘grondige bevraging van aangegeven en niet-aangegeven inkomen en arbeid’. De vragenlijst was deels volledig nieuw gedefinieerd, maar sommige vragen werden overgenomen of waren vergelijkbaar met bestaande enquêtes, onder meer de Eurobarometer omtrent zwartwerk van 2007. Tabel 1. Populatie en steekproef voor de bevolkingsenquête over aangegeven en niet-aangegeven inkomen en arbeid, België. Belgische populatie (2007-2008)

% aandeel totaal

Totale brutosteekproef

% aandeel totaal

Werknemers private sector Werknemers publieke sector RSZ-PPO Werknemers publieke sector RSZ Zelfstandigen in hoofdberoep Zelfstandigen in bijberoep Zelfstandigen na pensioenleeftijd Helpers Werklozen Tijdelijke werklozen Invaliden Primaire arbeidsongeschikten Gepensioneerde werknemers of zelfstandigen tot 74 jaar Pensioen ambtenaren tot 74 jaar Personen met een leefloon Personen met een handicap Huisvrouw/man (tot 64 jaar)

2 656 308 347 876 735 576 584 836 192 473 61 979 84 658 658 589 134 736 246 159 109 786 959 394

34,9% 4,6% 9,7% 7,7% 2,5% 0,8% 1,1% 8,7% 1,8% 3,2% 1,4% 12,6%

1 712 176 307 507 163 53 77 470 96 160 44 880

32,9% 3,4% 5,9% 9,7% 3,1% 1,0% 1,5% 9,0% 1,8% 3,1% 0,8% 16,9%

152 025 103 258 135 552 441 120

2,0% 1,4% 1,8% 5,8%

139 44 52 322

2,7% 0,8% 1,0% 6,2%

Totaal (populatie 18-75)

7 604 325

100,0%

5 202

100,0%

Bron:

6. 7.

Jaarverslagen parastatalen, FOD Economie en eigen definitie steekproef.

ECB, Household Finance and Consumption Network (2008), Reducing non-response bias. ECB, Household Finance and Consumption Network (2008), Imputation and data-editing.

Samenvatting

W

17

De steekproeftrekking verliep via de Kruispuntbank van de Sociale Zekerheid. Zij zou een afspiegeling vormen van de totale Belgische bevolking tussen 18 en 75 jaar, volgens sociale zekerheidssituatie. Actieven, niet actieven, werknemers, zelfstandigen, uitkeringstrekkers, zij zouden zowel globaal, als naar deelcategorieën bestudeerd kunnen worden, en voor elk van deze deelcategorieën waren ook aparte modules met vragen opgesteld. De totale populatie en de bruto-steekproef zijn in tabel 1 gegeven. Oorspronkelijk was de steekproef nog groter voorzien, maar omwille van budgettaire redenen was deze steekproef al gereduceerd. Oorspronkelijk ambieerden wij een netto-steekproef van ongeveer 4500 respondenten, de normale omvang van een bevolkingsenquête. Normaal zou rechtstreeks en herhaaldelijk contact worden opgenomen door de enquêteurs met de respondenten. Omwille van de privacy redenen liet het Sectoraal comité van de Sociale Zekerheid en de Gezondheid binnen de Commissie voor de bescherming van de persoonlijke levenssfeer dit niet toe. De geselecteerde adressen moesten via de Kruispuntbank van de Sociale Zekerheid aangeschreven worden met de vraag of zij bereid waren om mee te werken aan de enquête. Enkel wie daarop positief antwoordde kon aangesproken worden door de enquêteurs. Dit leidde tot een aanzienlijk lager dan verwachtte respons, en ook kan er een aanzienlijke vertekening zijn van de respons in functie van het onderzoeksvoorwerp, met name fraude. Om de respons te verbeteren werd bij de herinneringsbrief met de uitnodiging tot medewerking aan de enquête een cadeaubon beloofd van 10 €. De respons was hierna meer dan het dubbele, maar het is niet verder onderzocht welke groep extra werd aangetrokken. Uiteindelijk konden wij werken met een netto-responsgroep van 246 personen, de omvang voor een goede pilootstudie. Informatie werd via de KSZ bekomen omtrent de oorspronkelijke steekproef zodat enig zicht kon verkregen worden op het profiel van de non-respons. De netto-respons is in tabel 2 weergegeven. Dit kan gecorrigeerd worden via wegingcoëfficiënten (wat ook gebeurd is in de verwerking). De vertekening in het soort van respondenten die zelf beslisten of zij wilden meewerken of niet, is niet te corrigeren. In de deelname blijkt reeds een zekere antwoordselectiviteit. Zelfstandigen, werklozen en huisvrouwen/mannen, hebben minder deelgenomen; gepensioneerden, mensen werkzaam in de horeca, in de overheidsadministratie, het onderwijs en de gezondheidszorg relatief meer. Wat is bijgevolg de waarde van het aldus verzamelde materiaal? ‘The proof of the pudding is in the eating’. Wij hebben hier een eerste analyse gemaakt van onze beperkte steekproef alsof het een grootschalige enquête betreft. Want ook een pilootstudie verdient een grondige verwerking.8 Hiermee konden wij mogelijke inconsequenties en tegenstrijdigheden ontdekken in de antwoorden maar ook in de vraagstelling. Maar vooral wilden wij de relevantie van het bevraagde materiaal naar voren brengen. Tegelijk moesten wij onze rapportering inperken omdat bepaalde detailanalyses, waarvoor de vragenlijst toch ontworpen was, onvoldoende observaties bevatte.

8.

Van Teijlingen & Hundley, 2001.

18

X

Samenvatting

Tabel 2. Netto-steekproef en responsgraad.

Totaal Regio – Vlaams Gewest – Waals Gewest – Brussels Hoofdstedelijk Gewest Geslacht – Man – Vrouw Socio-economisch statuut – Werknemer – Zelfstandige – Gepensioneerde – Werkloze – Ander statuut Leeftijdsgroep – 15-24 jaar – 25-34 jaar – 35-44 jaar – 45-54 jaar – 55-64 jaar – 65-74 jaar Bron:

Netto-steekproef

Responsgraad

246

4,76%

114 108 24

4,43% 5,63% 3,53%

123 123

4,36% 5,23%

106 29 74 13 24

4,83% 3,63% 7,26% 2,77% 3,49%

6 45 36 45 50 64

1,96% 4,83% 3,60% 4,31% 5,11% 7,04%

Eigen berekeningen op basis van SUBLEC-enquête en gegevens KSZ over de bruto-steekproef.

Eerste tentatieve resultaten omtrent vraag en aanbod van zwartwerk 38,8% van de Belgische bevolking kocht in de loop van de laatste 12 maanden een goed of een dienst in het zwart. Dit percentage van de vraag naar zwartwerk is veel hoger dan wat in de Eurobarometer voor België was gerapporteerd, en lag ook boven de Europese cijfers. Maar niet alleen het percentage van personen dat goederen en diensten in het zwart aankoopt is belangrijk, maar ook de grootte van deze aankopen. Een gemiddeld bedrag van de grootste uitgave voor goederen of diensten in het zwart aangeschaft is 1 553 €. In de Eurobarometer was dit voor België slechts 1 050 € en voor de EU27 was dit 1 028 €. Ook het aanbod van zwartwerk was in onze enquête aanzienlijk hoger dan in de Eurobarometer. Niet minder dan 14,1% van de respondenten antwoordde dat zij arbeid in het zwart hadden gepresteerd in de loop van de laatste 12 maanden, tegen maar 6% van de Belgen in de Eurobarometer en 5% voor de bevolking in de EU 27. Het gemiddeld ontvangen inkomen op basis van deze zwartarbeid in de laatst 12 maanden was 1 332 €, iets hoger dan het resultaten in de Eurobarometer (België 1 000 €; EU27 1 119 €). Het percentage van de bevolking dat soms goederen of diensten koopt of aanbiedt maal het gemiddelde bedrag levert een gemiddeld volume aan fraude in de economie per persoon

Samenvatting

W

19

op, en kan verder uitgedrukt wordt ten opzichte van het BBP. De bovenstaande cijfers leveren ons 1,9% van het BBP op als gemiddeld bedrag in het zwart besteedt aan goederen en diensten, en 0,6% van het BBP aangeboden werk in het zwart. Normaal zouden beide geraamde volumes gelijk moeten zijn. Blijkbaar rapporteert men meer gekochte goederen en diensten in het zwart, als dat men toegeeft dat men zelf in het zwart gewerkt heeft. Blijkbaar is een bevraging van het gebruik van goederen en diensten in het zwart minder gevoelig dan vragen over het aanbod. Deze laatste zullen zeker een onderschatting opleveren. Maar waarschijnlijk zijn beide cijfers een onderschatting. Opmerkelijk is dat de in onze enquête gerapporteerde vraag en aanbod naar goederen en diensten in het zwart finaal hoger liggen dan de cijfers van de Eurobarometer. Dit is des te merkwaardig omdat onze steekproef vertekend is (zelfs na weging) naar de eerlijke respondent. Het topje van de ijsberg? Het hanteren van de oorspronkelijke methode, om de respondent rechtstreeks, en met aandrang te benaderen zou ons waarschijnlijk een groter deel van de ijsberg boven water gebracht hebben.

Tabel 3. Omvang van het zwartwerk. SUBLEC

Eurobarometer: België

Eurobarometer: EU27

Vraag naar zwartwerk: diensten/goederen Algemeen – Diensten – Goederen Gemiddeld bedrag (€) % BBP

38,8% 35,2% 14,1% 1 553 1,9%

15% 8% 1 050 0,6%

11% 9% 6% 1 028 0,5%

Aanbod van zwartwerk: diensten/goederen Algemeen Gemiddeld bedrag (€) % BBP

14,1% 1 332 0,6%

6% 1 000 0,2%

5% 1 119 0,2%

Bron:

Eigen berekeningen op basis van data SUBLEC; EC (2007), Special Eurobarometer 284.

Eerste resultaten over andere vormen van fraude De SUBLEC-vragenlijst bevatte verder een aantal bijkomende modules die andere vormen van fraude in beeld wensen te brengen, met name fraude in de fiscale aangifte, inkomen dat ‘onder de tafel’ wordt betaald, andere fiscale fraude in roerend en onroerend inkomen, en bij erfenis- en registratierechten. Het was immers de bedoeling om een exhaustief beeld te geven van alle vormen van sociale en fiscale fraude. Ook uitkeringsfraude en de combinatie van uitkeringsfraude met zwartwerk werd bijgevolg gedetailleerd bevraagd, met afzonderlijke modules naar gelang de diverse categorieën uitkeringstrekkers. In tabel 4 zijn sommige antwoorden op deze vragen samengevat. Telkens wordt de frequentie van het voorkomen van deze fraudevorm weergegeven, als % van de totale populatie,

20

X

Samenvatting

en indien van toepassing ook het volume, in % van het onderliggende bedrag. Wij kunnen hieruit afleiden, dat naast het reeds besproken voorkomen van gebruik en aanbod van zwartwerk, nog 2% van de loontrekkende respondenten zegt dat zij onder de tafel werden betaald, 5,6% van de uitkeringstrekkers zeggen dat hun uitkering niet volledig strookt met waarop zij zouden mogen recht hebben, en 4,3% van hen combineert een uitkering met zwartwerk. Hiermee situeert het zwartwerk bij de uitkeringstrekkers zich lager dan bij de totale groep respondenten. Wat de fiscale aangiften betreft zegt 24% dat de aangifte niet helemaal in orde is, en dat dit voor de totale groep kan slaan op ongeveer 2,3% van het aan te geven inkomen. Voor roerend inkomen antwoord 3,5% van de respondenten dat de aangifte niet correct is, wat slaat op 3% van hun kapitaalinkomen, en voor onroerend inkomen is dit 0,3% (de mogelijkheden tot fraude zijn beperkter), wat goed is voor 1,8% van hun inkomen. Ontduiking bij erfenisrechten komt dan weer frequenter voor, maar de tijdsperiode is ruimer (5 jaar), en voor een groter aandeel (gemiddeld 1/3 van hun aangifte). De registratierechten worden bij ongeveer 2% van de respondenten gefraudeerd, en voor 5,1% van de waarde.

Tabel 4. Samenvatting van de Belgische ondergrondse economie. Frequentie (% van de bevolking) Vraag naar zwartwerk Aanbod van zwartwerk Loon onder tafel Sociale uitkeringsfraude Uitkering gecumuleerd met zwartwerk Belastingaangifte niet volledig correct ingevuld Roerende inkomsten Onroerende inkomsten Erfenisrechten Registratierechten * **

38,8% 14,1% 2,0% 5,6% 4,3% 24,1% 3,5% 0,3% 5,5% 1,9%

Volume (% of totaal bedrag)

Kent iemand 79,2% 78,5% 52,1%

2,3%* 3,0%** 1,8%** 33,2%** 5,1%**

33,6% 27,3% 41,3% 40,3%

Als % van het inkomen van alle respondenten samen. Als % van het inkomen van de respondenten die antwoordden dat zij dit inkomen niet correct hadden aangegeven.

Bron:

Eigen berekeningen op basis van data SUBLEC.

Het hemd is nader dan de rok, dus men zal minder snel zijn eigen fraudegedrag opbiechten dan dat men dit van de andere wenst te signaleren. Op de vraag of men iemand kent die in het zwart werkt, of goederen en diensten koopt waarin zwartwerk steekt, of alle andere vormen van fraudegedrag, zijn de cijfers altijd aanzienlijk hoger. Die percentages worden ook gegeven in tabel 4. Maar die kans dat men iemand kent die fraudeert, kan inderdaad hoger zijn. Een alternatieve vraag is de inschatting van hoeveel mensen dat men denkt dat fraudeert. Deze vragen werden gesteld bij de opinievragen. Deze vragen kunnen nuttig zijn om na te gaan wat de impact is van dergelijke opinies over het gedrag van de andere op het eigen gedrag. Indien de respondent een professional zou zijn, en gevraagd wordt naar zijn

Samenvatting

W

21

opinie als expert van zijn eigen beroep of sector, zou het ook een schatting kunnen opleveren van het werkelijke fraudegedrag.9

Een tentatieve verklaring van de omvang van de fraude Op basis van een meer uitgebreide steekproef zou een differentiëring naar socio-professionele groep, inkomen, regio, geslacht, kunnen gemaakt worden van al deze variabelen. Rapportering van deze puntschattingen is hier niet verantwoord. Op basis van de beperkte steekproef is het wel mogelijk om een eerste verkenning te maken van welke factoren het fraudegedrag bepalen. Wij vergelijken in onderstaande tabel via een probit analyse een aantal determinanten voor de kans dat men zwartwerk vraagt, zwartwerk aanbiedt en fiscale fraude pleegt (hier gedefinieerd als niet volledig correct invullen van de belastingaangifte). Geslacht, regio, socio-economische groep, inkomen, het gedrag van de anderen en de eigen moraliteit zijn weerhouden als mogelijke verklarende factor.

Tabel 5. Probit analyse. Parameter Intercept Geslacht Regio Socio-economische groep Kent iemand (vraag) Kent iemand (aanbod) Kent iemand (fiscale fraude) Inkomen (1) Moraliteit (2)

Variabele

Man Franstalig Zelfstandige Uitkeringsgerechtigde Inactief Ja

Vraag naar zwartwerk

Aanbod van zwartwerk

Fiscale fraude

-0,3613 -0,0331 -0,0896 0,8488** -0,4383** -0,4079 0,8468***

-2,1466*** 0,6068** 0,424* 0,1701 -0,7391*** 0,3395

0,5952* 0,0162 -0,2762 0,1999 0,3426* 0,0457

1,1406** -0,0152 Moeilijk Totaal akkoord Redelijk akkoord Niet akkoord

-0,3393* -0,5279* 0,261 -0,1914

0,1508 -0,652* -0,5856* -0,23

0,0667 0,3207 -0,0158 -0,1203

Opmerking: *, **, en *** geven een significantieniveau van respectievelijk 10%, 5% en 1% weer. 1 Komt men rond op het einde van de maand? 2 De belastingen zijn te hoog? Bron:

Eigen berekeningen op basis van data SUBLEC.

Iemand anders kennen die ook goederen of diensten koopt in het zwart, of in het zwart werkt, lijkt een significante invloed te hebben op het eigen gedrag. Dit wordt bevestigd

9.

Het HIVA heeft dergelijke methode toegepast in de bouwsector. Pacolet J. & Baeyens K. (2007).

22

X

Samenvatting

door experimenteel onderzoek omtrent belastingfraude en uitkeringsfraude.10 Het kan een belangrijk aangrijpingspunt zijn voor het beleid, met name het doorbreken van deze spiraal van ‘de andere doet het ook’ of ‘iedereen doet het’, dus ‘ik ook’. Geslacht heeft geen invloed op de consumptie van goederen of diensten in het zwart, maar wel op het aanbod van zwartwerk. Mannen zijn wel significant meer aanbieder van zwartwerk. De afwezigheid van de invloed van het geslacht op de vraag kan te verklaren zijn door het feit dat veel aankopen in feite gezinsaankopen zijn, zodat mannen en vrouwen samen die beslissing nemen/ondergaan om bepaalde activiteiten in het zwart te laten gebeuren. Een uitkering trekken heeft een negatieve invloed op zowel de vraag als het aanbod van zwartwerk. Dit is het omgekeerde van wat doorgaans verwacht wordt, of van het beeld dat opgehangen wordt van zwartwerk als een overlevingsstrategie voor uitkeringstrekkers, om hun inkomen aan te vullen of goedkoper te consumeren. Zelfstandigen blijken wel meer zwartwerk te vragen, maar er is geen significant effect te merken op het aanbod, noch op de fiscale fraude. Beide soorten effecten gaan tegen gangbare opinies in, en maken dat men dit soort van enquêtes verder nodig heeft om met een grotere graad van zekerheid conclusies te trekken. Ook inkomen en opinies over fraude(als proxy voor de belastingmoraal van de betrokkene) leverden op basis van deze steekproef geen of beperkt significante resultaten op. Zo zou de vraag naar zwartwerk kleiner zijn naarmate men moeilijk rond komt met zijn inkomen, wat opnieuw indruist tegen de fraude als overlevingsstrategie.

Determinanten van fraude en strijd tegen fraude Regelmatig wordt een driehoek van krachten gehanteerd die de omvang van de fraude verklaren, en die tegelijk ook een aangrijpingspunt kunnen zijn van de strijd tegen fraude: het zijn de belastingmoraliteit (van een persoon, in een land); de belastingdruk (of het voordeel van het ontduiken van deze druk) en de controle (pakkans en boete). In de figuren 1 en 2 vatten wij het relatief belang samen dat de respondenten hechten aan deze drie factoren, zowel in de verklaring van het aanbod van zwartwerk, als in hun appreciatie welke factor het sterkst kan aangegrepen worden in de fraudebestrijding. Bij de oorzaken wordt bijna uitsluitend gewezen naar de belastingdruk, dus het voordeel om te frauderen. Bij de appreciatie van het beleid wordt toch ook aandacht gegeven aan de controlefactor. Het beleid zelf kan hierin besluiten dat niet alle heil kan verwacht worden van alleen de belastingdruk te verlagen, maar dat de bevolking ook verwacht dat er passende controle is. In beide grafieken valt op dat men weinig belang hecht aan de belastingmoraliteit, maar als wij de impact zien van het demonstratie-effect, is inwerken op deze belastingmoraliteit misschien ook een weg waarlangs men de fraude kan terugdringen. Dit strookt dan wel met een andere observatie uit dit onderzoek, met name dat de respondent de ‘belastingmoraliteit’ wel aangeeft als reden (16%) om niet te frauderen.

10. Lefebvre, Pestieau, Riedl, Villeval, 2011.

Samenvatting

W

23

Moraliteit, gedrag

10,9%

11,4% Belasting, regulering, ‘red tape’ Bron:

77,7% België

Controle, pakkans, handhaving

Eigen berekeningen op basis van data SUBLEC.

Figuur 1. Wat zijn volgens u de redenen om niet-aangegeven arbeid te verrichten? (n = 246)

Moraliteit, gedrag

6,8%

35,1% Belasting, regulering, ‘red tape’ Bron:

58,1% Belgium

Controle, pakkans, handhaving

Eigen berekeningen op basis van data SUBLEC.

Figuur 2. Welke maatregelen zijn volgens u het meest efficiënt in de strijd tegen niet-aangegeven arbeid? (n = 246)

Enkele saillante observaties als hypothese voor verder onderzoek Er is heel wat detail voorhanden in de bevraging waar omwille van de schaal eerder kan geconcludeerd worden dat het interessante onderzoekspistes zijn dan dat het definitieve conclusies zijn. Een intrigerende observatie is bijvoorbeeld dat men in Vlaanderen relatief meer aangeeft dat het zwartwerk van bedrijven afkomstig is, terwijl de Franssprekende respondenten relatief meer verwijzen naar het informele circuit van vrienden, collega’s, kennissen, familie. Dit kan in een grootschalige enquête een interessante morfologie van het zwartwerk opleveren. Nog een andere vraag is relevant voor de overheid die de strijd wenst aan te gaan tegen de fraude en die soms als tegenargument krijgt dat heel wat activiteiten zouden verdwijnen moest men het zwartwerk beteugelen: 2/3 van de respondenten zouden het goed of de dienst gekocht hebben op de reguliere markt als het daar alleen zou aangeboden zijn, en een kwart van de respondenten zou terugvallen op doe-het-zelf activiteiten.

24

X

Samenvatting

De specifieke vragen over de dienstencheque die wij opgenomen hadden in de vragenlijst tonen dan weeral een veel hoger percentage van huishoudelijk werk dat voorheen in het zwart gebeurde, dan tot nu toe aangenomen. Op de vraag wat de respondent zou aanzetten om de goederen of diensten in het formele circuit te kopen antwoordt de helft dat zij overtuigd zou worden door de garanties tegen fouten en gebreken. Het kan een aanknopingspunt vormen voor een informatiecampagne tegen zwartwerk. De antwoorden van de uitkeringstrekkers over de adequaatheid van de uitkering reveleren een confronterend beeld over een verzorgingsstaat onder druk: 58% van de genieters van een vervangingsinkomen vinden het te laag of veel te laag, en 35% vindt het juist genoeg in vergelijking met hun vroeger inkomen. Maar dit heeft, zie boven, deze respondenten blijkbaar niet aangezet tot verhoudingsgewijs meer in het zwart te werken of te consumeren. Zo vinden wij ook een opmerkelijk positief verband tussen het aanbieden van zwartwerk en de vraag naar zwartwerk. Maar niet omgekeerd. Misschien is dit weeral logisch: wie toegeeft in het zwart te werken zal ook toegeven om zaken aan te kopen in het zwart, maar niet omgekeerd. Combinatie van beide vragen zou nuttig kunnen zijn voor imputatie van bepaalde gegevens, of een correctie van de omvang van het zwartwerk. Hetzelfde zou kunnen gebeuren aan de hand van opinievragen: een redelijk hoog percentage van onze respondenten hadden een hoge inschatting van de fraude bij de rest van de bevolking. Aangezien de opinie over het fraudegedrag van de andere ook een invloed schijnt te hebben op het eigen fraudegedrag, zou het antwoord over het eigen fraudegedrag met deze informatie kunnen gecorrigeerd worden. Methodologisch werk aan de winkel, indien wij voldoende grondstof zouden hebben onder de vorm van een grootschalige enquête.

Beleidsrelevantie en hoe nu verder Hierboven geven wij enkele van de vele tentatieve bedenkingen die wij konden maken bij de interpretatie van de eerste resultaten. Verschillen naar socio-professionele groep, naar regio ook, wij willen zij niet uitvergroten. De pilootstudie over de fraudefenomenen in België moet gelezen worden als voorlopig, soms contra-intuïtief, soms uitdagend voor bestaande evidentie of opinies, en waardevol vooral als hypothese voor verder onderzoek. Op diverse plaatsen kon de relevantie voor het beleid worden aangestipt, onder meer welke risicogroepen er zijn of juist veel minder dan doorgaans aangenomen, wat de impact is van demonstratiegedrag, welke verwachtingen er zijn ten overstaan van het beleid en welke impact het kan hebben. In het rapport hebben wij bij sommige van onze commentaren gewaarschuwd dat zij riskeren niet alleen tentatief maar zelfs speculatief te zijn. Dit moet als dusdanig meegenomen worden in het debat. Voorbeelden daarvan zijn welke groepen meer of minder frauderisico vertonen, wat de fraude bepaalt (belastingdruk, gebrek aan controle, gedrag van anderen, eigen belastingmoraliteit) en wat de beste manier is om zij te bestrijden (belastingdruk, standvastigheid in het controle- en sanctiebeleid, bewustmakingscampagnes). Het is wel met dit soort onderzoeksmateriaal dat hierover een uitspraak kan gedaan worden. Door een ruime en exhaustieve definitie te hanteren van de sociale en fiscale fraude (alleen belastingontwijking hebben wij niet ter sprake gebracht – maar die kan misschien zelfs afge-

Samenvatting

W

25

leid worden uit de registers die de administraties zelf hebben) is dit onderzoek relevant voor diverse overheden en administraties en doelgroepen. Dat is ook het nut van een zogenaamde bevolkingsenquête. Zij moet representatief zijn voor deze diverse aspecten, en laat dan verdere analyse in detail toe. Dit was een redelijk exhaustieve bevraging, redelijk intensief (maar wij kennen zelfs meer intensieve bevolkingsenquêtes) maar onvoldoende extensief. De schaal was te klein. De voorgestelde methode is tijdsintensief en kostelijk. Het gebruik kan dan langdurig zijn, zoals dat ook het geval is met andere bevolkingsenquêtes als de huishoudbudgetenquête, de SILC, de arbeidskrachtentelling.11 Het kan ons cijfers bezorgen omtrent de ondergrond van onze economie en het verborgene van het economisch gedrag, zodanig dat zij op een regelmatige maar zeker niet jaarlijkse basis dient verricht te worden. De enquête organiseren met een zeker tijdsinterval lijkt ons dan ook aangewezen, maar eens moet het de eerste keer zijn. Op basis van deze pilootstudie lijkt ons nu het goede moment aangebroken om eindelijk die eerste volledige uitrol, ‘full blown’, te realiseren en het echte veldwerk te starten.

Referenties ECB, Household Finance and Consumption Network (2008a), Reducing non-response bias. ECB, Household Finance and Consumption Network (2008b), Imputation and data-editing. European Commission (2007), Special Eurobarometer 284. Undeclared Work in the European Union, Brussels, 90 p. Lefebvre M., Pestieau P., Riedl A. & Villeval M.C. (2011), ‘Tax Evasion, Welfare Fraud, and “The Broken Windows” effect: An Experiment in Belgium, France and the Netherlands’, IZA Discussion Paper No. 5609, 49 p. Pacolet J. & Baeyens K. (2007), Deloyale concurrentie in de bouwsector. Een terreinverkenning van mechanismen van sociale fraude, hun omvang en hun gevolgen voor de sector, HIVA-KU Leuven, Leuven, 149 p. Pacolet J. & De Wispelaere (2009a), Naar een observatorium ondergrondse economie. Een haalbaarheidsstudie, Acco, Leuven, 166 p. Pacolet J. & De Wispelaere (2009b), ‘The underground economy: designing an appropriate survey methodology to reveal sensitive behaviour (social and fiscal fraud)’, HIVA-KU Leuven, mimeo. Pacolet J. & Merckx S. (2008), SUBLEC: designing a survey methodology for fiscal en social fraud in Belgium: recent international comparative evidence and conclusions for Belgium, HIVA-KU Leuven, Leuven, mimeo. Pacolet J., Perelman S., Pestieau P., Baeyens K. & De Wispelaere F. (2009), Zwartwerk in België. Een indicator van omvang en evolutie – Travail au noir en Belgique. Un indicateur concernant l’étendue et l’évolution, Acco, Leuven, 195 p. Van Teijlingen, E. R. & Hundley V. (2001), ‘The importance of pilot studies’, Social Research Update, University of Surrey, winter, issue 35, p. 1-4.

11. Zo is ook nu pas, anno 2011, rond een ander belangrijk maatschappelijk domein, de vermogens, een bevolkingsenquête lopende in België en diverse andere landen, onder leiding van de ECB. Zie referenties hierboven.

Résumé Fraude sociale et fiscale en Belgique. Une étude-pilote relative au travail et aux revenus déclarés et non déclarés

Introduction Mesurer l’ampleur de l’économie souterraine semble être une ‘mission impossible’. Une des difficultés est que l’économie souterraine et la fraude ne font pas l’objet d’une définition univoque. Néanmoins, la littérature propose différentes méthodes d’enquête. On peut ainsi appréhender l’ampleur de l’économie souterraine sur base des modèles macro-économiques, ou sur base d’un sondage de la population, ou des entreprises, ou sur base de données administratives ou par évaluation des comptes nationaux. Ces méthodes produisent souvent des résultats discutables. La Commission européenne a également exploré différentes possibilités ces dernières années, sans obtenir de résultat satisfaisant. Il est probable que seule une association de différentes méthodes pourra faire avancer le problème. Car pour pouvoir lutter contre la fraude sociale et fiscale, il faut connaître l’ampleur du problème, savoir quelles en sont les manifestations, les causes et les motivations. Deux études précédentes publiées dans cette série expliquent ces méthodes d’estimation et plaident pour un ‘observatoire’ de l’économie souterraine, servant de poste d’observation et de plaque tournante des informations sur la fraude et la lutte contre celle-ci.1 Dans ce rapport, nous publions les résultats d’une nouvelle enquête sur les activités frauduleuses des Belges, sur leurs opinions et leurs motivations. L’étude a été commanditée et financée par le SPF Sécurité sociale et le SPP Politique scientifique BELSPO.2 Elle est le fruit

1. 2.

Pacolet J., Perelman S., Pestieau P., Baeyens K. & De Wispelaere F. (2009) ; Pacolet J. & De Wispelaere (2009), Nous remercions ces instances pour les possibilités d’enquête offertes dans le cadre du projet AGORA AG/00/137 : SUBLEC – Organisation d’une micro-enquête en vue d’effectuer une analyse descriptive et explicative de la problématique de la fraude sociale et fiscale. Nous remercions tout particulièrement Aziz Naji de BELSPO et Didier Verbeke et Koen Vleminckx du SPF Sécurité sociale pour leur foi dans ce type d’enquête et la confiance qu’ils nous ont témoignée. Chris Brijs, de la Banque Carrefour de la Sécurité Sociale, a été un soutien indispensable pour la réalisation de ce projet. Nous tenons également à remercier les membres du Comité d’accompagnement pour leurs suggestions et leur soutien tout au long de ce projet. Seuls les auteurs sont responsables des éventuelles erreurs et mauvaises interprétations dans le cadre de cette étude.

28

X

Résumé

d’une collaboration entre la KU Leuven (HIVA), l’ULg (CREPP) et l’ULB (METICES).3 L’ampleur, la méthode et le contenu de l’enquête ont été soigneusement préparés. Mais entre le rêve et la réalité, il y a des lois et des obstacles pratiques, de sorte que le déploiement à grande échelle de l’enquête tel qu’il était prévu n’a pas pu être réalisé. L’enquête est restée limitée à un échantillon de 246 personnes. C’est pourquoi nous la qualifions d’étude pilote. C’est une limitation, mais le contenu et l’utilité de l’instrument de recherche ne font aucun doute. C’est pourquoi nous avons traité cette enquête, avec toutes les limitations qui en découlent, et formulons les réserves nécessaires pour certains résultats. Cela ne nous a toutefois pas empêchés de formuler certaines conclusions dans le cadre de la lutte contre la fraude. ‘Full reports of pilot studies are rare in the research literature’.4 Il se peut que ce rapport fasse partie des exceptions, ce qui peut s’expliquer peut-être par le fait que nous avons commencé au départ avec une version à grande échelle. Mais le but de tout rapport est de tirer les leçons de l’expérience acquise. Et nous voulons retenir ici ces leçons.

Déploiement de l’étude A l’été 2010, l’étude pilote concernant l’enquête ‘travail et revenus déclarés et non déclarés’ a été déployée. Elle représente l’aboutissement provisoire d’une longue préparation. Nous avons tout d’abord déterminé l’opportunité et la faisabilité d’une enquête parmi la population en vue de mesurer l’ampleur, la structure et les déterminants du travail et des revenus non déclarés. Une définition large de la fraude était requise : les questions ont porté sur la fraude fiscale et sur la fraude sociale, et pour cette dernière, il a aussi bien été question des contributions que des allocations. Les travaux préparatoires concernaient l’analyse de l’utilisation internationale de l’enquête comme instrument pour déterminer l’ampleur de la fraude, quelle méthode d’enquête était la plus adéquate, comment l’échantillonnage devait être constitué, quels types de questions devaient être posées, quelles autorisations étaient nécessaires pour constituer l’échantillonnage en respectant la vie privée et qui serait l’utilisateur final de nos résultats. Il s’agissait en première instance pour nous de diverses instances et d’organismes parastataux sociaux confrontés à la lutte contre la fraude. Ce rapport met un terme provisoire à ce processus. Pour d’innombrables raisons, le ‘roll out’ final de l’enquête ne s’est pas déroulé comme planifié initialement. Au lieu de réaliser une enquête de population, représentative de la totalité de la population, mais également de différentes sous-catégories pertinentes, nous avons dû réduire l’échelle et limiter la portée à une étude pilote valable devant permettre de mettre davantage au point la méthodologie d’une telle enquête. L’objectif final étant d’identifier une définition large de la demande et de l’offre des activités non déclarées, des revenus, de la fraude sociale, de la fraude fiscale, de la fraude aux allocations, et des caractéristiques et déterminants éventuels, et ce parmi la population belge entre 18 et 75 ans, avec une distinction entre les groupes socio-professionnels.

3. 4.

HIVA, Onderzoeksinstituut voor Arbeid en Samenleving ; CREPP, Center of Research in Public Economics and Population Economics ; METICES, Migrations, Espaces, Travail, Institutions, Citoyenneté, Epistémologie, Santé. Van Teijlingen & Hundley, 2001.

Résumé

W

29

L’étude a été lancée sous l’acronyme SUBLEC ‘Survey on the black economy’, questionnaire relatif à l’économie souterraine. Cette étude ne se limitait pas au travail au noir et à la fraude aux contributions sociales, mais couvrait également la fraude aux allocations et à toutes les formes possibles de fraude fiscale. En raison de sa portée limitée, l’enquête est devenue une étude pilote qui ne peut qu’être partielle, expérimentale et même devenir spéculative. Elle n’est en tout cas pas définitive. Il s’agit plutôt une validation de la faisabilité de ce type d’enquête, du questionnaire proprement dit et de sa pertinence sociale. Nous confronterons certains résultats à une étude préalable similaire réalisée au niveau européen, l’Eurobaromètre n° 284 de 2007.5 Pour d’autres éléments aux conclusions étonnantes, une validation externe serait judicieuse mais mieux vaut l’effectuer lorsque l’enquête sera réalisée à plus grande échelle. L’enquête sur le travail et les revenus non déclarés a pour objectif d’informer les responsables politiques de l’ampleur de ces phénomènes et de formuler des recommandations d’amélioration de la lutte contre la fraude. Étant donné l’ampleur de l’enquête qui nous occupe aujourd’hui, nos recommandations vont plutôt dans le sens de savoir comment nous pouvons agir à l’avenir, sans donner de réponse définitive (pour autant qu’une enquête puisse donner des réponses définitives) quant à l’ampleur, la structure et la façon de lutter. Les recommandations doivent être considérées comme des tentatives. Parfois elles sont contre-intuitives, parfois elles contestent des évidences et des opinions existantes et peuvent servir d’hypothèses pour des vérifications et des études ultérieures. A divers endroits, nous pouvons en démontrer la pertinence pour le débat sur le travail au noir et les revenus non déclarés. À cause de la longueur du questionnaire et de son orientation sur certains groupes cibles et phénomènes partiels, intégrés dans le questionnaire de base, bon nombre de parties de l’enquête sont masquées. Lorsque l’enquête sera déployée à plus grande échelle, elle présentera son avantage supplémentaire à ce niveau. Nous discutons ensuite des forces et des faiblesses du projet et du roll out de notre questionnaire ainsi que des principaux résultats, qui sont provisoires mais illustratifs de la valeur qu’ils revêtent pour le débat politique. La conclusion finale est de poursuivre sur cette voie que nous avons empruntée avec cette enquête. Dans le cas qui nous occupe, cette enquête a été commanditée par le SPF Sécurité sociale et le SPP Politique scientifique. Ces deux services étaient convaincus qu’il fallait également appliquer cette méthode en Belgique. Les chercheurs doivent souvent convaincre les autorités d’organiser certaines enquêtes. Il faut parfois lutter contre le préjugé selon lequel il n’est pas possible d’interroger les gens sur leurs revenus ou leur patrimoine, ou dans ce cas-ci, sur les revenus et le travail au noir. Nous devons alors nous convaincre, ainsi nos collègues et collaborateurs, que c’est faisable. Ensuite, nous devons convaincre les enquêteurs de poser des questions et d’insister auprès des répondants pour qu’ils participent à l’enquête. Et ils doivent alors convaincre les répondants de répondre à toutes les questions. ‘Boost interviewers’ confidence about their ability to ‘sell’ the survey’ était la recommandation tout à fait pertinente que nous avons lue dans l’enquête sur le patrimoine de la BCE (Banque Centrale Européenne).6 Une fois toutes ces étapes franchies avec succès, le résultat peut

5. 6.

European Commission (2007), Special Eurobarometer 284. Undeclared Work in the European Union, Brussels, 90 p. ECB, Household Finance and Consumption Network (2008), Reducing non-response bias.

30

X

Résumé

encore être considérablement amélioré par datacleaning ou ‘editing’ et ‘l’imputation’ des informations manquantes.7 Nous espérons avec ce rapport avoir démontré que tous ces efforts porteraient leurs fruits ici aussi.

Forces et faiblesses de l’étude pilote actuelle Sur la base d’un aperçu de la littérature, nous avons conclu qu’une interview orale ‘face to face’ du répondant produirait le meilleur résultat. Un échantillon structuré a été constitué parmi la population des divers parastataux de la sécurité sociale, répartie sur tout le territoire. Le questionnaire a été élaboré simultanément par trois équipes et dans les deux langues nationales. Les chercheurs, et puis un enquêteur, ont effectué des enquêtes d’essai, et le questionnaire a ensuite été finalisé. Il a été décidé que les réponses au questionnaire se feraient ‘à visage découvert’. Celui-ci a été annoncé comme étant ‘une enquête approfondie sur le travail et les revenus déclarés et non déclarés’. Le questionnaire a été partiellement redéfini, mais certaines questions ont été reprises ou étaient comparables à des enquêtes existantes, notamment l’Eurobaromètre concernant le travail au noir de 2007.

Tableau 1.

Population et échantillon pour l’enquête de population sur le travail et les revenus déclarés et non déclarés en Belgique. Population belge (2007-2008)

% du total

Échantillon brut total

% du total

Salariés secteur privé Salariés secteur public ONSS-APL Salariés secteur public ONSS Indépendants à titre principal Indépendants à titre complémentaire Indépendants après l’âge de la pension Aidants Chômeurs Chômeurs temporaires Invalides En incapacité de travail primaire Salariés pensionnés ou indépendants jusqu’à 74ans Fonctionnaires pensionnés jusqu’à 74 ans Personnes ayant un revenu d’intégration Personnes ayant un handicap Femme/homme au foyer (jusqu’à 64 ans)

2 656 308 347 876 735 576 584 836 192 473 61 979 84 658 658 589 134 736 246 159 109 786 959 394

34,9% 4,6% 9,7% 7,7% 2,5% 0,8% 1,1% 8,7% 1,8% 3,2% 1,4% 12,6%

1 712 176 307 507 163 53 77 470 96 160 44 880

32,9% 3,4% 5,9% 9,7% 3,1% 1,0% 1,5% 9,0% 1,8% 3,1% 0,8% 16,9%

152 025 103 258 135 552 441 120

2,0% 1,4% 1,8% 5,8%

139 44 52 322

2,7% 0,8% 1,0% 6,2%

Total (population 18-75)

7 604 325

100,0%

5 202

100,0%

Source : Rapports annuels parastataux, SPF Economie et définition de l’échantillon.

7.

ECB, Household Finance and Consumption Network (2008), Imputation and data-editing.

Résumé

W

31

L’échantillon a été constitué par la Banque Carrefour de la Sécurité Sociale. Il devait être le reflet de la population belge totale entre 18 et 75 ans, selon la situation en matière de sécurité sociale. Actifs, non actifs, salariés, indépendants, allocataires devaient être étudiés globalement, mais aussi en catégories partielles. Et pour chacune de ces catégories, des modules spécifiques de questions avaient aussi été élaborés. La population totale et l’échantillon brut sont présentés au tableau 1. L’échantillon prévu à l’origine devait être plus large, mais pour des raisons budgétaires, il a été réduit. Au départ, nous voulions un échantillon net d’environ 4500 répondants, ce qui est normal dans le cas d’une enquête menée auprès de la population. Normalement, les enquêteurs devaient prendre directement et régulièrement contact avec les répondants. Mais pour des raisons de protection de la vie privée, le Comité sectoriel de la Sécurité sociale et de la santé au sein de la Commission pour la protection de la vie privée ne l’a pas permis. Les personnes sélectionnées devaient être contactées par écrit via la Banque Carrefour de la Sécurité Sociale afin de demander leur accord de participation à l’enquête. Seules les personnes ayant donné une réponse positive pouvaient être contactées par les enquêteurs. Cette restriction a produit un taux de réponse bien plus faible qu’escompté, et il peut y avoir aussi une distorsion considérable de la réponse à cause du sujet de l’enquête, à savoir la fraude. Pour augmenter le taux de réponse, la lettre de rappel avec l’invitation à participer à l’enquête promettait un chèque cadeau de 10 €. Le taux de réponse a ensuite plus que doublé, mais nous n’avons pas étudié la nature du groupe ainsi recruté. En fin de compte, nous avons pu travailler avec un groupe de réponse net de 246 personnes, ce qui correspond à une bonne étude pilote. Les informations ont été obtenues via la BCSS concernant l’échantillon original, de sorte que nous avons pu déterminer quelque peu le profil des non répondants. La réponse nette est également présentée en tableau 2. Une correction peut être apportée via les coefficients de pondération (ce qui a été fait lors du traitement). La distorsion au niveau du type de répondants ayant décidé eux-mêmes de collaborer ou non ne peut pas être corrigé. La participation présente déjà une certaine sélectivité de réponse. Les indépendants, les chômeurs et les femmes/hommes au foyer ont moins participé ; les pensionnés, les personnes actives dans l’horeca, dans l’administration publique, l’enseignement et les soins de santé relativement plus. Quelle est par conséquent la valeur du matériel collecté ? ‘The proof of the pudding is in the eating’. Nous avons effectué une première analyse de notre échantillon limité comme s’il s’agissait d’une enquête à grande échelle. Car une étude pilote mérite également un traitement approfondi.8 Nous avons ainsi pu détecter les éventuelles incohérences et contradictions dans les réponses mais aussi dans les questions. Mais surtout, nous voulions mettre en avant la pertinence du matériel étudié. Nous avons dû en même temps restreindre notre rapport parce que certaines analyses détaillées, pour lesquelles le questionnaire était conçu, ne comportaient pas suffisamment d’observations.

8.

Van Teijlingen & Hundley, 2001.

32

X

Résumé

Tableau 2.

Echantillon net et taux de réponse.

Total Région – Région flamande – Région wallonne – Région Bruxelles Capitale Sexe – Homme – Femme Statut socio-économique – Salarié – Indépendant – Pensionné – Chômeur – Autre statut Groupe d’âge – 15-24 ans – 25-34 ans – 35-44 ans – 45-54 ans – 55-64 ans – 65-74 ans

Echantillon net

Taux de réponse

246

4,76%

114 108 24

4,43% 5,63% 3,53%

123 123

4,36% 5,23%

106 29 74 13 24

4,83% 3,63% 7,26% 2,77% 3,49%

6 45 36 45 50 64

1,96% 4,83% 3,60% 4,31% 5,11% 7,04%

Source : Calculs réalisés sur la base de l’enquête SUBLEC et des données de la BCSS concernant l’échantillon brut.

Premiers résultats expérimentaux concernant la demande et l’offre de travail au noir 38,8% de la population belge ont acheté un bien ou un service au noir ces 12 derniers mois. Ce pourcentage de demande de travail au noir est bien plus élevé que celui que donne l’Eurobaromètre pour la Belgique, et est également supérieur aux chiffres européens. Le pourcentage de personnes achetant des services et des biens au noir est important, mais l’ampleur de ces achats l’est également. La somme moyenne de la plus grande dépense pour des services ou biens acquis en noir est de 1 553 €. L’Eurobaromètre n’indiquait que 1 050 € pour la Belgique, et 1 028 € seulement pour l’EU27. Dans notre enquête, l’offre de travail au noir était également nettement supérieure à ce qu’indique l’Eurobaromètre. Pas moins de 14,1% des répondants ont répondu avoir travaillé au noir durant les 12 derniers mois, contre seulement 6% des Belges dans l’Eurobaromètre et 5% de la population dans l’EU 27. Les revenus moyens perçus dans le cadre de ce travail au noir ces 12 derniers mois s’élevaient à 1 332 €, soit un peu plus que les résultats de l’Eurobaromètre (Belgique 1 000 € ; EU27 1 119 €). Le pourcentage de la population qui achète ou offre parfois des biens ou des services, multiplié par le montant moyen, donne un volume moyen de la fraude par personne dans

Résumé

W

33

l’économie et peut être exprimé par rapport au PIB. Les chiffres ci-dessus donnent 1,9% du PIB comme montant moyen consacré aux services et biens en noir, et 0,6% du PIB de travail proposé au noir. Normalement, ces deux chiffres devraient être identiques. Les Belges avouent manifestement plus facilement avoir acheté au noir des biens et des services que d’avoir eux-mêmes travaillé au noir. Des questions concernant l’utilisation de biens et de services en noir sont moins sensibles que les questions sur l’offre. Ces dernières vont certainement produire une sous-évaluation. Mais les deux chiffres sont probablement des sous-estimations. On remarque que dans notre enquête, la demande et l’offre de biens et de services sont supérieures aux chiffres de l’Eurobaromètre. C’est d’autant plus remarquable que notre échantillon est biaisé (même après pondération) dans le sens du répondant honnête. La partie immergée de l’iceberg? L’application de la méthode d’origine, pour approcher le répondant directement et avec insistance, aurait sans doute permis de faire émerger une partie plus importante de l’iceberg.

Tableau 3.

Ampleur du travail au noir. SUBLEC

Eurobaromètre : Belgique

Eurobaromètre : EU27

15% 8% 1 050 0,6%

11% 9% 6% 1 028 0,5%

6% 1 000 0,2%

5% 1 119 0,2%

Demande travail au noir : services/produits Général 38,8% – Services 35,2% – Produits 14,1% Montant moyen (€) 1 553 % PIB 1,9% Offre travail au noir : services/produits Général 14,1% Montant moyen (€) 1 332 % PIB 0,6%

Source : Calculs propres sur base des données SUBLEC ; CE (2007), Special Eurobarometer 284.

Premiers résultats sur d’autres formes de fraude Le questionnaire SUBLEC contenait en outre une série de modules complémentaires qui avaient pour but de présenter d’autres formes de fraude, à savoir la fraude dans la déclaration fiscale, les revenus qui sont payés ‘sous la table’, d’autres fraudes fiscales dans les revenus mobiliers et immobiliers ainsi que dans les droits de succession et d’enregistrement. L’objectif était en effet d’avoir une vue exhaustive de toutes les formes de fraude fiscale et sociale. La fraude aux allocations seule et combinée au travail au noir fait par conséquent aussi l’objet de questions détaillées, avec des modules spécifiques en fonction des différentes catégories d’allocataires. Dans le tableau 4, sont résumées certaines réponses à ces questions. La fréquence de survenance de cette forme de fraude est également indiquée en % de la population totale et

34

X

Résumé

le volume est aussi repris, le cas échéant, en % du montant sous-jacent. Nous pouvons en déduire qu’outre l’utilisation et l’offre de travail au noir dont il a déjà été question, 2% des répondants salariés déclarent avoir été payés sous la table, 5,6% des allocataires disent que leur allocation ne correspond pas totalement à ce dont ils devraient avoir droit, et 4,3% des allocataires combinent une allocation au travail au noir. Le travail au noir chez les allocataires se situe donc à un niveau plus bas que pour le groupe total de répondants. En ce qui concerne les déclarations fiscales, 24% disent que la déclaration n’est pas totalement correcte et que pour le groupe total, cela peut porter sur près de 2,3% des revenus à déclarer. Pour les revenus mobiliers, 3,5% des répondants disent que la déclaration n’est pas correcte ; cela concerne 3% de leurs revenus du capital. Pour les revenus immobiliers, cela correspond à 0,3% (les possibilités de fraude sont plus limitées), ce qui équivaut à 1,8% de leurs revenus. La fraude aux droits de succession est donc à nouveau plus fréquente mais la période est plus large (5 ans) et le pourcentage plus grand (en moyenne 1/3 de leur déclaration). Les droits d’enregistrement font l’objet d’une fraude chez environ 2% des répondants et pour 5,1% de la valeur.

Tableau 4.

Résumé de l’économie souterraine belge. Fréquence (% de la population)

Demande de travail au noir Offre de travail au noir Salaire sous la table Fraude aux allocations sociales Allocation cumulée au travail au noir Déclaration fiscale pas correctement remplie Revenus mobiliers Revenus immobiliers Droits de succession Droits d’enregistrement

38,8% 14,1% 2,0% 5,6% 4,3% 24,1% 3,5% 0,3% 5,5% 1,9%

Volume (% du montant total)

Connaît quelqu’un 79,2% 78,5% 52,1%

2,3%* 3,0%** 1,8%** 33,2%** 5,1%**

33,6% 27,3% 41,3% 40,3%

* Comme % du revenu de l’ensemble des répondants. ** Comme % du revenu des répondants qui ont répondu qu’ils n’avaient pas déclaré correctement ce revenu. Source : Calculs réalisés sur la base des données SUBLEC.

‘Charité bien ordonnée commence par soi-même’, on va donc plus rapidement signaler un comportement frauduleux chez autrui que confesser ses propres fraudes. À la question de savoir si l’on connait quelqu’un qui travaille au noir, achète des biens ou des services qui impliquent du travail au noir ou commet toutes les autres formes de fraude, les chiffres sont toujours nettement plus élevés. Ces pourcentages sont aussi indiqués dans le tableau 4. Mais la probabilité que l’on connaisse quelqu’un qui fraude peut en effet être plus élevée. Une question alternative est l’évaluation du nombre de personnes que l’on soupçonne de frauder. Ces questions ont été posées avec les questions d’opinion. Elles peuvent être utiles pour vérifier l’impact que peut avoir sur son propre comportement ce type d’opinions sur le comportement d’autrui. Si le répondant est un professionnel et qu’on lui demande son

Résumé

W

35

avis en tant qu’expert dans son secteur ou sa profession, cela pourrait aussi fournir une estimation de la fraude réellement commise.9

Une tentative d’explication de l’étendue de la fraude Sur la base d’un échantillon plus large, une différentiation de toutes ces variables pourrait être faite en fonction du groupe socio-professionnel, des revenus, de la région, du sexe. Un rapport de ces estimations point par point n’aurait aucun sens ici. Sur la base de l’échantillon restreint, il est en effet possible de faire une première recherche sur les facteurs qui déterminent le comportement frauduleux. Nous comparons dans le tableau ci-dessous, par le biais d’une analyse probit, un certain nombre de facteurs qui déterminent la probabilité de demander du travail au noir, d’offrir du travail au noir et de commettre une fraude fiscale (définie ici comme une déclaration fiscale qui n’est pas complétée correctement). Le sexe, la région, le groupe socio-économique, les revenus, le comportement des autres et la propre moralité sont retenus comme éventuels facteurs explicatifs.

Tableau 5.

Analyse probit.

Paramètre

Variable

Intercept Sexe Région Groupe socio-économique

Masculin Francophone Indépendants Allocataires Inactifs Oui

Connaît quelqu’un (demande) Connaît quelqu’un (offre) Connaît quelqu’un (fraude fiscale) Revenus (1) Moralité (2)

Demande de travail au noir

Offre de travail au noir

Fraude fiscale

-0,3613 -0,0331 -0,0896 0,8488**

-2,1466*** 0,6068** 0,424* 0,1701

0,5952* 0,0162 -0,2762 0,1999

-0,4383** -0,4079 0,8468***

-0,7391*** 0,3395

0,3426* 0,0457

1,1406** -0,0152 Difficile Totalement d’accord Relativement d’accord Pas d’accord

-0,3393* -0,5279* 0,261 -0,1914

0,1508 -0,652* -0,5856* -0,23

0,0667 0,3207 -0,0158 -0,1203

Remarque : *, **, et *** indiquent un niveau de pertinence de respectivement 10%, 5% et 1%. 1 Est-ce que l’on s’en sort à la fin du mois ? 2 Les impôts sont trop élevés ? Source : Calculs réalisés sur la base des données SUBLEC.

9.

Le HIVA a appliqué ce type de méthode dans le secteur de la construction. Pacolet J. & Baeyens K. (2007).

36

X

Résumé

Connaître quelqu’un d’autre qui achète aussi des biens ou des services au noir ou qui travaille au noir semble avoir une influence significative sur son propre comportement. Cela est confirmé par une étude expérimentale concernant la fraude fiscale et la fraude aux allocations.10 Cela peut être un point de départ important pour la politique, à savoir casser cette spirale du ‘l’autre le fait bien aussi’ ou ‘tout le monde le fait’, donc ‘moi aussi’. Le sexe n’a aucun impact sur la consommation de biens ou de services au noir, mais bien sur l’offre de travail au noir. Les hommes sont en effet nettement plus représentés dans l’offre de travail au noir. L’absence de l’influence du sexe sur la demande peut s’expliquer par le fait que de nombreux achats sont en fait des achats familiaux, de sorte que les hommes et les femmes prennent/subissent ensemble la décision de faire réaliser certaines activités au noir. Bénéficier d’une allocation a un impact négatif sur la demande et sur l’offre de travail au noir. C’est l’inverse de ce qui est généralement attendu ou de la définition attribuée au travail au noir, à savoir une stratégie de survie pour les allocataires afin de compléter leurs revenus ou de consommer à moindre prix. Les indépendants semblent en effet être plus demandeurs de travail au noir mais on ne constate pas d’effet significatif sur l’offre, ni sur la fraude fiscale. Ces deux types d’effets vont à l’encontre des opinions largement répandues et font que l’on a encore besoin de ce type d’enquêtes pour tirer des conclusions avec un degré de certitude plus élevé. De même, les revenus et les opinions sur la fraude (comme proxy pour la moralité fiscale de la personne concernée) ont fourni peu, voire pas de résultats significatifs sur la base de cet échantillon. Ainsi, la demande de travail au noir serait moins grande selon que l’on s’en sort difficilement avec ses revenus, ce qui va de nouveau à l’encontre de la stratégie de survie.

Déterminants de la fraude et lutte contre la fraude Un triangle de forces est régulièrement utilisé pour expliquer l’étendue de la fraude mais ces forces peuvent en même temps servir de point de départ dans la lutte contre la fraude : il s’agit de la moralité fiscale (d’une personne, d’un pays), de la pression fiscale (ou de l’avantage de se soustraire à cette pression) et du contrôle (risque d’être pris et amende). Dans les triangles ci-dessous, nous synthétisons l’importance relative que les répondants accordent à ces trois facteurs, tant dans l’explication de l’offre de travail au noir que dans leur jugement concernant le facteur qui peut le plus fortement influencer la lutte contre la fraude. Pour les causes, on mentionne presque exclusivement la pression fiscale, donc l’avantage de frauder. Pour le jugement de la politique, on accorde toutefois aussi de l’attention facteur de contrôle. La politique même peut ainsi en déduire qu’il ne faut pas tout attendre de la seule diminution de la pression fiscale, mais que la population s’attend aussi à ce qu’il y ait un système de contrôle adéquat. Les deux graphiques montrent que l’on accorde moins d’importance à la moralité fiscale mais si nous voyons l’impact de l’effet de démonstration, agir sur cette moralité fiscale est sans doute aussi un moyen qui permet de réduire la fraude.

10. Lefebvre, Pestieau, Riedl, Villeval, 2011.

Résumé

W

37

Cela correspond donc bien à une autre observation faite dans cette étude, à savoir que le répondant mentionne la ‘moralité fiscale’ comme une raison (16%) de ne pas frauder.

Moralité, comportement

10,9%

11,4% Taxation, 77,7% régulation, ‘red tape’

Controle, risque d’être pris, mise en application

Belgique

Source : Calculs réalisés sur la base des données SUBLEC.

Figure 1. Quelles sont, selon vous, les raisons d’accomplir un travail non déclaré ? (n = 246)

Moralité, comportement

6,8%

Taxation, régulation, ‘red tape’

35,1%

58,1% Belgique

Controle, risque d’être pris, mise en application

Source : Calculs réalisés sur la base des données SUBLEC.

Figure 2. Quelles mesures sont, selon vous, les plus efficaces dans la lutte contre le travail non déclaré ? (n = 246)

Quelques observations marquantes comme hypothèse pour la poursuite de l’étude Un très grand nombre de détails sont présents dans le questionnaire où, pour des raisons d’échelle, on peut en conclure qu’il s’agit de pistes d’étude intéressantes plutôt que de conclusions définitives. Une observation étonnante est par exemple qu’en Flandre, il est relativement plus souvent mentionné que le travail au noir émane des entreprises tandis que les répondants francophones parlent davantage du circuit informel des amis, des collègues,

38

X

Résumé

des connaissances et de la famille. Cela peut fournir une morphologie intéressante du travail au noir dans le cadre d’une enquête à grande échelle. Une autre question est également pertinente pour les autorités qui souhaitent lutter contre la fraude et qui reçoivent parfois comme contre-argument que de nombreuses activités disparaîtraient si l’on devait réprimer le travail au noir : 2/3 des répondants auraient acheté le bien ou le service sur le marché régulier si c’est le seul endroit où il était proposé, et un quart des répondants se tournerait vers les activités de bricolage. Le questionnaire a révélé que le pourcentage de tâches ménagères réalisées au noir avant l’introduction des titres services représente un pourcentage plus élevé que présumé jusqu’ici. À la question de savoir ce qui inciterait le répondant à acheter les biens et les services dans le circuit officiel, la moitié répond qu’ils seraient convaincus par les garanties contre les défauts et les vices. Cela peut constituer un point de départ pour une campagne d’information contre le travail noir. Les réponses des allocataires sur l’adéquation de l’allocation donnent une image interpellante d’un État providence sous pression : 58% des bénéficiaires d’un revenu de remplacement le trouvent trop faible ou beaucoup trop faible, et 35% le trouvent juste suffisant par rapport à leur revenu précédent. Mais, comme mentionné plus haut, cela n’a manifestement pas incité ces répondants à proportionnellement travailler ou consommer davantage au noir. Nous trouvons donc aussi un rapport étonnamment positif entre l’offre de travail au noir et la demande de travail au noir. Mais pas l’inverse. Cela est sans doute à nouveau logique : celui qui admet travailler au noir admettra aussi acheter des objets au noir, mais pas l’inverse. La combinaison des deux questions pourrait être utile pour imputer certaines données ou corriger l’étendue du travail au noir. On pourrait faire la même chose au moyen de questions d’opinion : un pourcentage relativement grand de nos répondants avait une estimation élevée de la fraude dans le reste de la population. Vu que l’opinion sur le comportement frauduleux d’autrui semble aussi avoir un impact sur son propre comportement frauduleux, la réponse sur le propre comportement frauduleux pourrait être corrigée avec ces informations. Voilà du travail méthodologique en perspective si nous avions suffisamment d’éléments sous la forme d’une enquête à grande échelle.

Bien-fondé de la politique et comment agir à l’avenir Nous mentionnons ci-dessous quelques-unes des nombreuses tentatives d’observations que nous avons pu faire lors de l’interprétation des premiers résultats Nous ne voulons pas grossir les différences en fonction du groupe socio-professionnel ainsi que de la région. L’étude pilote sur les phénomènes de fraude en Belgique doit être considérée comme provisoire, parfois contre-intuitive, parfois ambitieuse face à des évidences ou des opinions existantes, et précieuse surtout comme hypothèse pour la poursuite de l’étude. À divers endroits, le bienfondé de la politique anti-fraude a pu être pointé du doigt, notamment en ce qui concerne le nombre de groupes à risque où la constatation a été faite qu’il en existe beaucoup moins que ce que l’on croit habituellement, l’impact de l’effet de démonstration, les attentes par rapport à ces politiques et l’impact qu’elles peuvent avoir. Dans le rapport, nous avons, pour

Résumé

W

39

certains de nos commentaires, mis en garde contre le fait qu’ils risquent non seulement d’être expérimentaux mais aussi spéculatifs. Cela doit être pris en compte dans le débat. Des exemples sont : quels groupes présentent plus ou moins de risques de fraude, qu’est-ce qui détermine la fraude (pression fiscale, manque de contrôle, comportement des autres, propre moralité fiscale) et quelle est la meilleure manière de lutter contre la fraude (pression fiscale, fermeté dans la politique de contrôle et de sanction, campagnes de sensibilisation). C’est justement avec ce type d’études que l’on peut se prononcer à ce sujet. En formulant une définition large et exhaustive de la fraude sociale et fiscale (seule l’évasion fiscale n’a pas été évoquée – mais elle peut sans doute être déduite des registres dont disposent les administrations), cette enquête est pertinente pour les divers pouvoirs publics, administrations et groupes cibles. C’est également là l’intérêt d’une enquête auprès de la population. Elle doit être représentative dans ces divers aspects et permettre une analyse détaillée ultérieure. Il s’agissait d’un questionnaire relativement exhaustif, relativement intensif (mais nous connaissons des enquêtes beaucoup plus intensives) mais pas assez extensif. L’échelle était trop petite. La méthode proposée est coûteuse en temps et en argent. L’utilisation peut donc être de longue durée, comme cela est le cas aussi avec d’autres enquêtes auprès de la population comme l’enquête sur le budget des ménages, le SILC, l’enquête sur les forces de travail.11 Cela peut nous fournir des chiffres concernant le fondement de notre économie et les aspects cachés du comportement économique, de sorte que l’enquête doit être réalisée sur une base régulière mais certainement pas annuelle. Organiser l’enquête avec un certain intervalle de temps nous semble donc aussi recommandé mais il faut bien une première fois. Sur la base de cette étude pilote, il nous semble que le bon moment est arrivé pour réaliser enfin ce premier déploiement complet ‘full blown’ et commencer le véritable travail sur le terrain.

Bibliographie ECB, Household Finance and Consumption Network (2008a), Reducing non-response bias. ECB, Household Finance and Consumption Network (2008b), Imputation and data-editing. European Commission (2007), Special Eurobarometer 284. Undeclared Work in the European Union, Brussels, 90 p. Lefebvre M., Pestieau P., Riedl A. & Villeval M.C. (2011), ‘Tax Evasion, Welfare Fraud, and “The Broken Windows” effect: An Experiment in Belgium, France and the Netherlands’, IZA Discussion Paper No. 5609, 49 p. Pacolet J. & Baeyens K. (2007), Deloyale concurrentie in de bouwsector. Een terreinverkenning van mechanismen van sociale fraude, hun omvang en hun gevolgen voor de sector, HIVA-KU Leuven, Leuven, 149 p. Pacolet J. & De Wispelaere (2009a), Naar een observatorium ondergrondse economie. Een haalbaarheidsstudie, Acco, Leuven, 166 p. Pacolet J. & De Wispelaere (2009b), ‘The underground economy: designing an appropriate survey methodology to reveal sensitive behaviour (social and fiscal fraud)’, HIVA-KU Leuven, mimeo.

11. Une autre enquête vient également d’être lancée en 2011 sur un autre domaine social important, les avoirs. Il s’agit d’une enquête auprès de la population menée en Belgique et dans d’autres pays, sous la direction de la BCE. Voir références ci-dessus.

40

X

Résumé

Pacolet J. & Merckx S. (2008), SUBLEC: designing a survey methodology for fiscal en social fraud in Belgium: recent international comparative evidence and conclusions for Belgium, HIVA-KU Leuven, Leuven, mimeo. Pacolet J., Perelman S., Pestieau P., Baeyens K. & De Wispelaere F. (2009), Zwartwerk in België. Een indicator van omvang en evolutie – Travail au noir en Belgique. Un indicateur concernant l’étendue et l’évolution, Acco, Leuven, 195 p. Van Teijlingen, E. R. & Hundley V. (2001), ‘The importance of pilot studies’, Social Research Update, University of Surrey, winter, issue 35, p. 1-4.

Foreword

‘Tussen droom en daad staan wetten en praktische bezwaren’ (‘Between dream and reality there are laws and practical objections’) is a well known quote of the Flemish writer Willem Elsschot. The dreams and ambitions of the FPS Social Security, PPS Science Policy and the researchers to organize a ‘full-blown’ survey on the size and scope of undeclared work and income, social and fiscal fraud has only be realized partially in this project because of the limited size of the sample. This makes detailed analysis and interpretation premature. But we do have the feeling that the instrument for obtaining this information and the methodology has been improved. The first analysis seduces to tentative and perhaps also speculative conclusions. But we should reverse them in research hypotheses for such a full-blown rollout of the survey in the future. Since the survey turned out a pilot study on methodology and potential use of the results, this report does not provide the final answers but, on the contrary, opens many new hypotheses and questions that could have been answered by this type of research. But many times we could not resist in making tentative and probably speculative interpretations. They are meant as a challenging teaser for further work on this unfinished project. But perhaps also as an inspiring trigger for further policy already now. Since it opens, not with complete certainty, but will that ever be possible on the underground economy that wants to stay uncovered, but enough convincing to think about further actions and targets. This report also includes suggestions for further improvement of the methodology and the survey instrument, including new hypotheses to be verified. The questionnaire was not only inspired by obtaining a broad and exhaustive overview of all kind of forms of social and fiscal fraud, its occurrence, its determinants, and its volume. Besides exhaustiveness and accurateness, it was also intended to create a sound database for empirical verification of the theory about compliance and non–compliance. At the same time it reflects in many items that the problem of fraud and the fight against it is an actual topic. The project could be organized within an agreeable context of mutual understanding of the research teams at the KU Leuven (HIVA), the ULg (CREPP) and ULB (METICES). It benefited from the financial support of the PPS Science Policy and the FPS Social Security within the context of the AGORA-programme and the financing of the field work by the FPS Social Security. We especially owe gratitude to Aziz Naji of BELSPO and Didier Verbeke and Koen Vleminckx of the FPS Social Security for their stimulating support and leading engagement to further develop the research on the undeclared economy in Belgium. The CBSS, the Sectoral Committee of Social Security of the Privacy Commission and SMALS made this

42

X

Foreword

survey possible by providing us their principal agreement and practical help and support in defining and contacting the sample population. Here Chris Brys of the CBSS was our guide on this road between scientific dreams and laws and practical obstacles. A steering committee of the representative stakeholders in the fight against fraud further accompanied us with comments and feedback during the preparatory work. The survey departments of respectively HIVA and CREPP, their enthusiastic collaborators and interviewers made the roll-out of the survey less of an adventure but more of a learning tour between our respondents. It is finally the willingness of the respondents to answer that gives this report content and body. From the individual stories and points of view of those respondents, the real life, we hopefully reconstructed somewhat of a realistic although preliminary picture of this real economy with this report. Errors and misunderstandings and too speculative jumps to conclusions are the sole responsibility of the researchers. They hope to share their enthusiasm in the search for a scientific view on these phenomena in the societal debate on fraud and the ways how to reduce it.

Abbreviations

ADSEI BELSPO CBSS CREPP DGSIE EC EU GDP HIVA INR LFS METICES NACE NBB NSSO NSSOLPA OECD OLAF SILC SNA SUBLEC UNECE VAT

Algemene Directie Statistiek en Economische Informatie Belgian Science Policy Office Crossroads Bank for Social Security Center of Research in Public Economics and Population Economics Direction Générale Statistique et Information Economique European Commission European Union Gross Domestic Product Research Institute for Work and Society Instituut voor de Nationale Rekeningen Labour Force Survey Migrations, Espaces, Travail, Institutions, Citoyenneté, Epistémologie, Santé Nomenclature générale des Activités économiques dans les Communautés Européennes National Bank of Belgium National Social Security Office National Social Security Office for the Local and Provincial Administrations Organisation for Economic Co-operation and Development Office européen de Lutte Anti-Fraude Statistics on Income and Living Conditions System of National Accounts Survey Black Economy United Nations Economic Commission for Europe Value added tax

Introduction

From 2008 until 2011, three research institutes – HIVA (KU Leuven), CREPP (ULg) and METICES-TEF (ULB) – were involved in the SUBLEC (survey on the black economy) project. One of the main pillars of this project was the organization of a large-scale survey on social and fiscal fraud or the so-called underground economy in Belgium. This project was financed by the Belgian Science Policy and the Federal Public Service Social Security. The last years the survey method was promoted by the European Commission as one of the promising methods to reveal the level and structure of undeclared activities. Revealing the covered and reporting on socially less or not accepted activities is however methodologically difficult. The Eurobarometer recently organized on this issue demonstrated promising but also disappointing results. International literature indicates that face-to-face interviewing is appropriate for a survey that deals with social and fiscal fraud. The methodology the SUBLEC project uses to estimate the extent of this phenomenon, is the direct methodology of asking the citizens their demand for and supply of undeclared work and their (un)declared income by means of a face-to-face interview. It was oriented to the total population living in Belgium. The aim of this survey was to describe and to understand forms and causes of undeclared activities, to measure the size and the structure of undeclared activities. Large definitions of fraud were used, covering undeclared work as well as other undeclared income but also benefit fraud. To obtain more information about the definition, the methodology, the questions and the special risk groups, we have performed a literature review and organized an international seminar on surveys of social and fiscal fraud. On the basis of this we concluded how to organize this kind of survey in the best way. It became a face-to-face interview of individuals from 18 to 75 years old living in Belgium, to measure the underground economy. Irregular migrants were not included in the survey.

Chapter 1

Scope of the survey and applied methodology

1.

The definition of the underground economy

Terminology is not always clear and many terms can be related to the ‘underground economy’.1 First we can refer to the 1993 System of National Accounts (SNA) where the ‘non-observed economy’ is defined. “The non-observed economy refers to all productive activities that may not be captured in the basic data sources used for national accounts compilation” (UNECE, 2003, 2008). The ‘non-observed economy’ consists of (OECD, 2002): – Underground production; – Illegal production; – Informal production; – Household production for own final use; – Production missed due to deficiencies in data collection programme. The underground economy (OECD, 2002; UNECE, 2008): “Production activities that are legal but deliberately concealed from public authorities: – to avoid the payment of income, value added or other taxes; – to avoid the payment of social security contributions; – to avoid having to meet certain legal standards such as minimum wages, maximum hours, safety or health standards, etc.; – to avoid complying with certain administrative procedures, such as completing statistical questionnaires or other administrative forms”. The definition used by the 1993 SNA for achieving ‘exhaustiveness’ in the national accounts is defined very strictly. Illegal production is not included in the underground economy however, as “the borderline between underground and illegal production is not entirely clear” and is interpreted in a strict sense (OECD, 2002). For the calculation of the exhaustiveness

1.

Van Eck (cited in OECD, 2002) listed the following words: alternate, autonomous, black, cash, clandestine, concealed, counter, dual, grey, hidden, invisible, irregular, marginal, moonlight, occult, other, parallel, peripheral, secondary, shadow, submerged, subterranean, twilight, unexposed, unofficial, untaxed, underwater.

48

X

Scope of the survey and applied methodology

of the Belgian national accounts the term underground economy should contain the illegal and the black economy (INR, 2006). The figure of Kazemier (2003) below delivers an insight in the relationship between the underground economy, the illegal economy, the informal economy and exhaustiveness (figure 1.1).

Illegal economy Exhaustiveness

Informal economy

Underground economy

Activities that contribute to GNP Source:

Kazemier, 2003.

Figure 1.1. Relationship between the illegal economy, the informal economy, the underground economy and exhaustiveness.

The underground economy can be interpreted in a larger sense than only its part in the calculation of the national accounts (see also figure 1.1). The different aspects of fiscal and social fraud and avoidance are much larger than the limited content of the underground economy used by the 1993 SNA. “Non-productive activities do not contribute to national income. This also means that non-observed not-productive activities (for example social benefit fraud) do not affect the quality of national accounts” (Renooy, Ivarsson, van der Wusten-Gritsai and Meijer, 2004). Fiscal fraud also includes (non-productive) activities that have no influence on the GDP but however leads to fewer revenues for the state. The overestimation of professional expenses by employees (negative impact on the revenues coming from the personal income taxes), the non-declaration of income from capital or the capital itself (negative impact on the revenues coming from the withholding tax on income derived from securities or the inheritance taxes), black capital legalized by a tax amnesty, ... are examples of activities which have no impact on the national accounts (NBB, 2010). The definition of the underground economy by the 1993 SNA bears close resemblance to the definition of undeclared work used by the European Commission (1998): “any paid activities that are lawful as regards their nature but not declared to the public authorities, bearing in mind that differences in the regulatory system of Member States must be taken into account. Applying this definition, criminal activities would be excluded, as would work which does not have to be declared”. Illegal and criminal activities are also excluded in this definition. For the financing of the welfare state and the public expenditures in general, a broad definition of the underground economy is needed (Pacolet and Verbeke, 2007). Limiting the

Scope of the survey and applied methodology

W

49

underground economy or undeclared work to only ‘paid productive activities which are lawful’ and thereby excluding other kinds of tax evasion and tax avoidance is too narrow. Social fraud, but also fiscal fraud and tax avoidance should be a concern for policy makers. The narrow definition of the European Commission can be used to describe undeclared work, but the fight against fiscal and social fraud should be enlarged in the societal debate from benefit fraud, undeclared or under-declared work and income and tax avoidance (Pacolet, 2007a; 2007b). This broad definition can also contribute to the belief that law enforcement is assured for all segments of fraud and illegal activities. The EU introduced in its employment policy the attention on undeclared work, with the risk to forget other undeclared income (Pacolet and Verbeke, 2007). It excluded the illegal activities. All this is understandable from the point of view of an employment policy. It is not from a global view on a fight against fraud and the scientific analysis of it. Experience on the field, scientific studies on the phenomena, reveal needs for larger definitions. Broadening the scope of the debate is necessary because all fraud contributes to less revenue for the state or more expenditure. Broad definitions should also be used in policies to guarantee equal treatment. It can avoid underestimation of the problem, and could reveal a lot of synergies in studying and controlling fraud. Figure 1.2 situates the scope of the questionnaire. Questions about fiscal and social contribution and benefit fraud but also avoidance are asked. But also income which is not or less taxed, is integrated in the questionnaire. The purchase of counterfeit products is an example of a question which asks something about illegal activities.

Avoidance

Charge not or less taxes because

Fiscal contributions

Cannot fully tax

Undeclared income

– To form – Relocate fiscal basis – Spread in time

Moveable property

Social contributions

Grey zone

Undeclared work

– To form – Relocate fiscal basis – Spread in time

Exemption and subsidy

Social benefits

Evasion

Benefit fraud and misuse

Grey zone

Domestic help, fiscal amesty

Fiscal underestimation

Don’t want fully tax Cadastral income own house Savings account

A fixed deduction of certain occupations

Domestic help, service vouchers

A fixed deduction of certain occupations

Criminal and illegal activities

Source:

Pacolet and De Wispelaere, 2009a based on Pacolet & Geeroms.

Figure 1.2. Broad definition of fiscal and social fraud and avoidance.

By using a large definition of the underground economy it is the intention to obtain a detailed view of all the aspects of social and fiscal fraud and avoidance. This should result in

50

X

Scope of the survey and applied methodology

a picture of the amount of revenues lost by the Belgian government because of undeclared or under-declared work and income and not limited to those activities which have an influence on the GDP. Although a broad definition of fiscal fraud and avoidance is needed and although the borderline with illegal activities is sometimes thin, the used methodology focuses on social and fiscal fraud.

2.

The choice for a face-to-face survey “It’s time to acknowledge how little we really know about the unobserved economy despite forty years of effort to measure its size and growth” (Feige and Urban, 2008).

Measuring the extent of the underground economy is still a problematic issue. The various estimates of the underground economy in Belgium are a good example of this. Schneider (Feld and Schneider, 2010) estimates the size for 2007 at 18.3% of the GDP, while the Belgian national accounts (NBB, 2010; Pacolet, Perelman, Pestieau, Baeyens and De Wispelaere, 2009) assume a size of 3.8% of the GDP. In absolute figures this means a difference of € 48.4 billion (€ 61.3 billion – € 12.9 billion).2 The Eurobarometer survey about undeclared work organized in 2007 even results in a size of 0.6% of the GDP (or € 2 billion) (= demand for undeclared work) (EC, 2007b). Differences in Belgian regions are also calculated by means of macroeconomic models. The size would be higher in the Brussels-Capital Region than in the Walloon and Flemish Regions (Herwartz, Schneider and Tafenau, 2010). The provinces in the Flemish Region would have a smaller ‘shadow economy’ than the provinces in the Walloon Region (Ibid.). Variations in the extent of the underground economy can be explained by the different measuring methodologies used. Direct and indirect methods at micro and macro level can be used. Indirect methods compare macroeconomic aggregates (e.g. discrepancy method, currency demand approach). Closely related is the economic approach based on the construction of an econometric model (DYMIMIC/MIMIC, electricity consumption) (Schneider and Enste, 2000; Ciccarone, Marchetti and Pavlovaite, 2009). The use of macroeconomic models to estimate the underground economy has been criticized many times already. Most of the comments are: no view at specific sectors or socio-economic categories, no precise definition, too simplistic assumptions, no stable results, overestimation of the underground economy (EC, 2007a; OESO, 2002). Direct methods are “based on contact with or observation of persons and/or firms” (Pedersen, 2003). Well-designed surveys and audits of social and fiscal inspection services can provide a detailed view of the underground economy. Surveys allow to get more information from the whole population, from specific economic categories (employees, self-employed persons, unemployed persons, ...), from employers in

2.

GDP in 2007 for Belgium: 334,948 million €.

Scope of the survey and applied methodology

W

51

specific sectors,3 from social and fiscal inspection services. Contrary to indirect methods, one can assume that the results from surveys underestimate the underground economy because persons will hesitate to answer completely honestly on socially less appreciated behavior. On the 25th of February 2008, the three research institutes involved in this project organized a ‘kick-off’ seminar with the Federal Public Service Social Security and the Belgian Science Policy (See annex I for programme). The research team confronted its research strategy with international experts and the expectations of national experts of the controlling authorities. The methodological possibilities for the organization of the sample, questionnaire and roll-out were discussed with the ambition to learn from recent experiences. The conclusions of this seminar have been used as input for the further organization of the survey in Belgium. A literature review of foreign and Belgian surveys dealing with the whole or an aspect of the underground economy was made by HIVA in preparation of the kick-off seminar (Pacolet and Merckx, 2008). It was the intention to obtain more information about the methodology, the questions and the specific risk groups. Other interesting studies, remarks and recommendations were added to this literature review after the seminar. The conclusion was that the faceto-face method would be the best method to ask respondents about social and fiscal fraud. Most surveys about the underground economy have been carried out by face-to-face interviews (Pacolet and Merckx, 2008). Also the special Eurobarometer survey (No. 284) on undeclared work used this research method (EC, 2007b). A preceding European ‘feasibility study on a direct survey about undeclared work’ also found that “the most frequently applied method of data collection in past surveys on undeclared work was face-to-face interviewing” (TNS, Rockwool Foundation Research Unit and Regioplan, 2006). The European feasibility study concludes “from the point of view of representativeness, in the majority of countries the face-to-face methodology is the preferable option for a study on undeclared work” (Ibid., 2006, p. 89). Kazemier and van Eck (1990) carried out different sorts of surveys to measure the underground economy in the Netherlands (postal, face-to-face and telephone surveys) and already concluded that face-to-face interviews obtained the best results. A recently executed pilot study in the Netherlands postulates that the face-to-face interview is still the best method to ask people on their underground activities (de Heij and Kazemier, 2007, p. 12).4 From the literature review about the design of an appropriate survey methodology (Pacolet and De Wispelaere, 2009) also other elements for applying a face-to-face survey appeared (e.g. a lower risk of missing data). The presence of an interviewer can produce positive as well as negative effects. The fact that the respondent has to report his or her answer to the interviewer would increase the chance of a socially desirable response and misreporting (Tourangeau and Yan, 2007; Rockwood, Sangster and Dilleman,

3.

4.

In Belgium for instance, we obtained very promising results with a survey among the employers in the construction sector (Pacolet & Baeyens, 2007). The survey revealed opinions of professionals in the sector on the size and form of undeclared work and social fraud in their sectors, without asking directly if they themselves were engaged in such activities. Two methods were used in the pilot study. A face-to-face survey and a mixed mode survey (an internet survey combined with a postal and a telephone survey).

52

X

Scope of the survey and applied methodology

1997). On the other hand comprehension problems can be solved by the interviewer and the willingness of the respondent to formulate accurate answers will be higher (Holbrook, Green and Krosnick, 2003).

3. 3.1

Development of the questionnaire Scope

The SUBLEC project had the ambition to organize a survey about the underground economy. This was not only limited to undeclared work but also integrated social benefit fraud and different aspects of fiscal fraud. The title of the questionnaire was “A profound survey on declared and undeclared income and work”. Specific questions about undeclared work, social benefit fraud and fiscal fraud (not restricted to income from labor but also income from capital or the capital itself) were thus included in the questionnaire. The presence of the Eurobarometer survey with results for Belgium, and comparative information for 26 other countries was an opportunity. For that reason we maintained similar questions from that survey. For other issues, we compared our results with other population surveys.

3.2

Strategy

How to formulate the questionnaire and how to decide what design is appropriate? These are two of the most important questions one could ask when a survey is dealing about a sensitive issue like undeclared income and work. Therefore a literature review was performed by HIVA (Pacolet and De Wispelaere, 2009b). It helped us with the design of the survey and the formulation of the questions. We have opted to formulate the sensitive questions about social and fiscal fraud gradually because it would be inadvisable to formulate the threatening questions at the beginning of the survey or to formulate them very suddenly and without any introduction (De Pelsmacker and Van Kenhove, 1999). Formulating so-called ‘loading’ questions to introduce the threatening questions, make them more easy to answer or give them a less threatening character. It seemed advisable to ask some introduction questions (e.g. demographic characteristics) and opinion questions first, followed by questions on the demand for underground activities and to ask for the supplied underground activities in the final section (de Heij and Kazemier, 2007). The demand for undeclared work would also be more socially accepted than the supply of undeclared work. Indirect questions where the respondent is asked for his opinion about the way most people behave are also less threatening than direct questions where the respondent has to report his own behavior (Fisher, 1993). Finally some closing/debriefing questions were added to the questionnaire. These questions were aimed at measuring the sensitivity of certain questions. The questionnaire was divided in 7 sections: – Socio-demographic information;

Scope of the survey and applied methodology

– – – – – –

–

W

53

Socio-economic questions; Opinions about social and fiscal fraud; Demand for undeclared work – household expenditure; Fiscal fraud; Supply of undeclared work; Specific questions for – Employees, – Self-employed persons, – Benefit recipients, – Non-working people; Closing questions for the respondent.

The issue of declared and undeclared activities was introduced straightforwardly in the introduction letter and in the questions. An alternative could have been to limit the opinion questions and the questions to third persons. The questions were posed in a direct way to the respondent. Euphemistic terms were not used to assure questions were well understood by the respondents. Nevertheless it was possible to remove the threat partially by asking the respondent about the fraud he or she committed in the past (during the last 12 months) and not about his/her current ‘fraud behavior’. To guarantee the comparability with the Special Eurobarometer on undeclared work, we have maintained some of the questions from the survey. In the definition of undeclared work of the Eurobarometer one used the terms ‘niet-aangegeven arbeid/travail non-declaré’ in the Dutch and French translation and not the familiar terms ‘zwartwerk/ travail au noir’. Also in the SUBLEC-survey the term ‘niet-aangegeven arbeid’ is used. Remark that the SUBLEC survey interviewed respondents between 18 and 75 years old while the respondents from the Eurobarometer were 15 years and older.

3.3

Communication of the questions

We have printed the response categories of some questions on a response card. The use of a response card gives the respondent enough time to think about his/her answer. On the back of the response card the number of the question was printed. For specific opinion questions we have used a new method since we wanted to avoid response order effects. The order in which the respective items are presented on the list may greatly influence the obtained reports (Schwarz and Oyserman, 2001). Response order may influence responses through primacy effects (selecting the first category) or recency effects (selecting the last category). Primacy effects are more likely with response options presented visually (in a self-administered survey or by use of a show card) (Martin, 2006). Scientific literature suggests that researchers have to rotate the order of responses (Holbrook, Green, Krosnick, 2003). Since this was impossible with a face-to-face interview without computer assistance, a so-called ‘response disc’ was used (questions C9, C10, C11, C12, C24) (figure 1.3). In a circle (so with no beginning and no ending) all the possible responses where listed.

Scope of the survey and applied methodology

Ir (il regu le l g al iere en i ) mm

Rechth e sociale bbenden o p (leeflo bijstand on OC MW, ...)

ven t actie re nie ...) Ande rouw, (huisv

en

z lo

k er W

ig ra nt en

n

Zel f hoo standig fdb ero en in ep

ente

X

n in

Zelfstandige bijberoep

Invaliden n

e rd

en

ikt

ee

nem De

elti

jds

erk ew

en erd ne sio pen Ge

Pr ar ima be ire ids on

Voltijdse werknemers

ers

ch

Source:

ge s

n sio

en

ep

gg

u Br

Stud

54

Own figure.

Figure 1.3. Example of a response disc.

4.

Sample selection

To obtain our sample it was necessary to submit a data request to the Sectoral Committee of Social Security and Health within the Privacy Commission. We not only needed the names and addresses of the selected persons in the sample but also some of their socio-economic characteristics (region, age, sex, socio-economic category) since it was our intention to compare respondents with non-respondents at an aggregate level, based on those socioeconomic characteristics. We concluded to execute a face-to-face interview with individuals from 18 to 75 years old to measure the underground economy. Irregular migrants were not included in the survey, because they are not registered by the Crossroads Bank for Social Security (CBSS). The sample was taken from this central administrative database to allow, at an aggregate level, the confrontation of information from the survey with the total population. The gross sample was initially set at 12 900 individuals and the target to reach was 4 300 individuals net. Because of budget limitations we had to reduce our gross sample to 5 202 individuals. This gross sample was split evenly over French-speaking and Dutch-speaking respondents. Normally the Brussels-Capital Region could include both language communities. In reality, due to the limited sample size, the Dutch-speaking part turned out to be absent from the selection. Our target was still a response rate of 33% or 1 734 individuals. The sample was structured by socio-economic category (employee, self-employed, pensioner, unemployed, ...) since it was our ambition to obtain a detailed picture of four categories: employees, self-employed persons, social benefit recipients and non-active people. The

Scope of the survey and applied methodology

W

55

CBSS asked the different social departments to make the sampling. This made it possible to receive the most recent situation of the socio-economic categories. Table 1.1 shows the population for Belgium, the Flemish Region and the Walloon Region together with the Brussels-Capital Region. The share of the socio-economic categories in the total population and our total gross sample are also listed. We defined an overrepresentation of the self-employed persons in the gross sample. One can also observe an overrepresentation of the pensioners. It should be noted that the total share of pensioners in the Belgian population will be higher than observable in table 1.1 and will thus be closer to the defined gross sample. Our sample was defined until the age of 75 years and not 74 years as defined in table 1.1. In 2010 more than 17% of the Belgian population is estimated to be at an age of 65 years and older (Federal Planning Bureau, 2008). The total share of pensioners is even larger. The respondents were chosen by means of a stratified sample method that grouped them over a limited number of municipalities. This was done from a cost saving point of view and for the convenience of the interviewers.5 The sample was split in two equal halves. The sampling procedure was performed separately for these two ‘regions’. In principle, municipalities were selected with probability proportional to size, without replacement. So municipalities could be selected only once. Big municipalities were put in a separate stratum, and selected by default. The cluster size for these big cities was larger than the standard cluster size. The municipalities in the BrusselsCapital Region, other than Brussels itself, were put in a separate stratum. The questionnaire was developed by the two research groups (one Dutch-speaking person and two French-speaking persons with a good knowledge of the other language) and validated for content and comprehension by the third team. This guaranteed proper understanding and phrasing in both languages.

5. 5.1

Roll-out Pretest

A pretest allows to detect understanding and interpretation problems in the survey design (Krosnick, 1999). Missing relevant response categories could also be detected (de Leeuw, Hox and Huisman, 2003). Most of the time, the pretesting procedure includes a small number of interviews followed by a debriefing session. A small pretest was done by HIVA (Dutch questionnaire) and CREPP (French questionnaire) in June 2010. 3 face-to-face interviews (employee, self-employed and benefits recipient) were performed by the two research institutes. The interviewer who did the interviews had to fill-in an evaluation form of the questionnaire. It was the intention to detect possible mistakes or incompletenesses in the questionnaire, the briefing document, the response cards and the response discs. This made it possible to make some corrections just before the final launch of the survey.

5.

The sampling and clustering procedure was defined by dr. Karel Van den Bosch, Antwerp University.

Source:

2.0% 1.4% 1.8% 5.8%

152,025 103,258 135,552 441,120 100.0%

1.1% 8.7% 1.8% 3.2% 1.4% 12.6%

84,658 658,589 134,736 246,159 109,786 959,394

7,604,325

9.7% 7.7% 2.5% 0.8%

735,576 584,836 192,473 61,979

3,297,738

73,784 81,849 188,228

59,442

31,754 365,348 48,745 112,149 47,098 375,123

406,007 219,547 71,311 23,478

1,020,164 173,711

Annual reports social departments, FPS Economy, SMEs, Self-employed and Energy.

Total (population 18-75)

34.9% 4.6%

2,656,308 347,876

4,306,587

29,474 53,703 252,892

92,583

52,904 293,241 85,991 134,010 62,688 584,271

329,569 365,289 121,162 38,501

1,636,144 174,165

2,601

34 35 163

65

35 296 42 86 22 413

193 229 73 24

789 102

2,601

10 17 159

74

42 174 54 74 22 467

114 278 90 29

923 74

5,202

44 52 322

139

77 470 96 160 44 880

307 507 163 53

1,712 176

Total gross sample

100.0%

0.8% 1.0% 6.2%

2.7%

1.5% 9.0% 1.8% 3.1% 0.8% 16.9%

5.9% 9.7% 3.1% 1.0%

32.9% 3.4%

% share of total

X

Workers private sector Workers public sector within NSSOLPA Workers public sector within NSSO Self-employed worker main activity Self-employed worker second activity Self-employed activities after age of pension Helpers Unemployed person Temporary unemployed person Permanent incapacity Primary incapacity Pension worker + self-employed worker until 74 years old Pension public servant until 74 years old Social reintegration allowance Allowance disabled person Housewife/husband (until 64 years old)

Population % share of Population Population Gross sample Gross sample Belgium total Walloon Region- Flemish Walloon Region- Flemish (2007-2008) Brussels-Capital Region Brussels-Capital Region Region Region

Table 1.1. Gross sample for Belgium, Flemish Region, Walloon Region and Brussels-Capital Region.

56 Scope of the survey and applied methodology

Scope of the survey and applied methodology

5.2

W

57

Briefing of the interviewers

At the end of August 2010 a briefing of the interviewers was organized for the Dutch-speaking interviewers by HIVA and for the French-speaking interviewers by CREPP. All the interviewers received a briefing document. This document contained an explanation of the project and working method, some practical forms and guidelines but it mainly explained the different questions and possible difficulties. Since the fiscal and social legislation contains terms which are not commonly used by a great part of the respondents and neither by the interviewers, it was necessary to explain these terms. After all, a good understanding of the question is the basis for receiving a correct answer.

5.3

Method of contacting the sample

It was the intention to have the research institutes themselves contact the whole sample. Therefore we needed the names and addresses from the persons in the sample. Because of privacy reasons this was not possible according to the Sectoral Committee of Social Security and Health within the Privacy Commission. Addresses were selected by the BCSS. Invitations from the research institutes were sent to those selected persons by a third party, Smals,6 and the persons willing to participate in the project had to send back a letter of consent confirming their willingness to participate in the research. This letter of agreement was directly sent back to the research institutes (HIVA for the Flemish Region and CREPP for the Walloon Region and the Brussels-Capital Region). The introduction letter formulated by the research institutes and sent by Smals contained the description and goals of the survey, the research institutes involved, the voluntary character of participation, the importance of participation, and the anonymity and confidentiality of the answers given by the respondent. It was also clearly underlined that participation in the survey and the answer formulated would in no way influence their rights and that also on individual questions the answer could be refused. Three types of envelopes were used for sending the introduction letters. The Dutch letter for the Flemish Region was sent in an envelope with the HIVA logo on it. The French letter for the Walloon Region was sent in an envelope with the new ULg logo on it. The French and Dutch letter for the sample in the Brussels-Capital Region was printed recto verso and was sent in an envelope with the old ULg logo on it. The introduction letter was sent in two waves (one on the 15th of July 2010 and one on the 18th of August 2010) because of the difference in delivery of the selected addresses by the different social departments. On the 13th of September 2010 a reminder letter was sent to the whole sample. Since the response rate in the first mailing turned out so low (only 2%), the reminder envelope announced with a sticker the rewarding of the participants with a voucher of € 10. The reminder and perhaps also the announced voucher could double the number of participants.

6.

Smals is a non-profit organization which supports the IT systems of the social security and other public services at federal level.

58

X

Scope of the survey and applied methodology

Since it was our strategy to announce and talk straightforwardly about declared and undeclared activities, it could not be excluded that the willingness to participate, and certainly for those engaged in undeclared activities, would be reduced substantially. Our original design of contacting directly the selected persons was explicitly aimed at increasing the response rate and avoiding biases. It was not allowed for privacy reasons. That is why the design and understanding of this survey needed to be reoriented. Since the scale of the sample became much smaller, we decided to limit the detail of the analysis and interpretation and to consider the study more as a pilot study to assess the feasibility of the methodology and the potential results.

5.4

Face-to-face interviews

Between September and October 2010 the 246 face-to-face interviews were performed.

5.5

Debriefing

At the end of November 2010 a debriefing was organized by HIVA (for the Dutch-speaking interviewers) and CREPP (for the French-speaking interviewers). We asked the interviewers about the first reactions of the respondents, about possible problems with parts of the questionnaire or specific questions, about their impression whether respondents were giving truthful answers.

6.

Data weighting method

Given the differences in the response rate, we decided to continue our analysis on the basis of weighted data. Weights are defined taking into account three variables: language, socio-economic category and sex. This was necessary due to the unbalance of the degree of participation in those three variables. No data were immediately available to combine the three variables language, sex and socio-economic category at a population level. It was neither our intention to generalize the survey to the whole population in the pilot study. The gross sample gives a good impression of the different categories. There is however a small overrepresentation of the pensioners and self-employed persons. 4 major socio-economic categories are defined (employees, self-employed persons, social benefit recipients, nonactive people). Verification has been done when the category ‘social benefit recipients’ was divided in more specific categories. Here only marginal differences appeared. The net response was compared with the initial gross sample. This allowed to correct an over- or underrepresentation in the net sample (e.g. the overrepresentation of the pensioners). We observe a higher weight for the self-employed persons. The data in chapter 2 and 3 about the characteristics of the respondents and non-respondents and the socio-demographic and socio-economic information of the respondents are unweighted. Data about the demand for undeclared work (chapter 4), the supply of undeclared work (chapter 5), and opinions (chapter 6) are weighted.

Scope of the survey and applied methodology

W

59

The low response rate made it impossible to analyze detailed questions. For some questions we made a distinction between the French-speaking and the Dutch-speaking respondents. In the Brussels-Capital Region only French-speaking respondents answered the questionnaire, so this distinction between Dutch-speaking and French-speaking is in fact a distinction between the Flemish-region and the Walloon & Brussels-Capital region. It was our ambition to analyze three important socio-economic groups, namely employees, self-employed persons and social benefit recipients. Due to the low response rate it was impossible to analyze the answers of the self-employed persons (only 15 persons in the response group). We have tackled however the answers on the specific questions for employees and social benefit recipients.

Table 1.2. Weights. Categories Language Sex

Socio-economic category

Dutch Dutch Dutch Dutch Dutch Dutch Dutch Dutch French French French French French French French French ALL

Employee Self-employed Social benefit recipient Other Employee Self-employed Social benefit recipient Other Employee Self-employed Social benefit recipient Other Employee Self-employed Social benefit recipient Other ALL

Source:

Male Male Male Male Female Female Female Female Male Male Male Male Female Female Female Female ALL

CBSS and SUBLEC data.

Net sample

Gross sample

n

%

n

%

29 5 24 7 18 3 23 4 19 5 33 6 38 2 24 6 246

0.118 0.020 0.098 0.028 0.073 0.012 0.093 0.016 0.077 0.020 0.134 0.024 0.154 0.008 0.098 0.024 1.00

600 268 531 36 511 146 358 123 577 251 511 45 507 135 454 118 5,171

0.116 0.052 0.103 0.007 0.099 0.028 0.069 0.024 0.112 0.049 0.099 0.009 0.098 0.026 0.088 0.023 1.000

Weights 0.98 2.55 1.05 0.24 1.35 2.32 0.74 1.46 1.44 2.39 0.74 0.36 0.63 3.21 0.90 0.94 1.00

Chapter 2

Characteristics of respondents and non-respondents

An important aspect for the survey on social and fiscal fraud is the non-response. It is important to verify if there were certain groups of the population who systematically did not respond to the questionnaire. So we want to measure if non-response correlates with age, sex, the statute of the respondent or the activity sector in which the respondent is working. The decision to participate in the survey is an individual choice. Nevertheless some general characteristics can be observed when the list of respondents is compared to the list of nonrespondents.

1.

The different characteristics

We provided a list of respondents to the Crossroads Bank for Social Security (CBSS). The CBSS added the following characteristics for every person in the sample (for respondents as well as for non-respondents) – Participated: 1 (Yes) or 0 (No); – Region: Flemish Region, Walloon Region, Brussels-Capital Region; – Sex: 1 (Man) or 2 (Woman); – Age group: 15-19, 20-24, 25-29, 30-34, 35-39, 40-44, 45-49, 50-54, 55-59, 60-64, 65-69, 70-74, 75-79, 80-84, 85-89, unknown; – Official socio-economic classification: employee, self-employed person – helper, unemployed person, temporary unemployed person, primary incapacity, permanent incapacity, pensioner, handicapped person, social reintegration income and housewife/house husband; – NACE: The classification of economic categories (2 digit).

2.

First analysis

The total sample counted 5,171 persons.1 It was our intention to divide the sample in two equal classes. This is the reason why the Flemish Region represents 50% of the sample, as well as the Walloon Region together with the Brussels-Capital Region.

1.

Not 5,202 persons because only 13 persons who receive a social reintegration allowance were contacted (and not 44 persons). The persons who receive a social reintegration allowance were first contacted by a social worker to explain the aim of the survey.

62

X

Characteristics of respondents and non-respondents

246 persons have been questioned, what implies a response rate of 4.8%. The response rate differs per region. The response rate in the Walloon Region (5.6%) was higher than in the Brussels Capital Region (3.5%) and the Flemish Region (4.4%). Perhaps the distribution of socio-economic categories or age categories over the regions can explain this distinction. Further in this report we will examine these assumptions. The share of the Walloon Region in the class of respondents (43.9%) is higher than its share in the total sample (37.1%). The results are statistically significant (p = 0.0474). The size of the sample did not allow further correction by the region (especially a large city like Brussels). The following tables all have the same structure: the first line for each category is the gross number, the second line the percentage of the row total and the third line is the percentage of the column total.

Table 2.1. Respondents and non-respondents by region (n = 5,171). Frequency Row Pct Col. Pct Flemish Region

Walloon Region

Brussels-Capital Region

Total

Source:

Non-respondent

Respondent

2,459 95.57% 49.93% 1,810 94.37% 36.75% 656 96.47% 13.32%

114 4.43% 46.34% 108 5.63% 43.90% 24 3.53% 9.76%

4,925 95.24%

246 4.76%

Total

Ratio column response/nonresponse

2,573 49.76% 1,918

0.93

37.09% 680

1.19

13.15%

0.73

5,171 100.00%

Own calculation based on data from the CBSS.

Table 2.2. Respondents and non-respondents by sex (n = 5,171). Frequency Row Pct Col pct Men

Women

Total

Source:

Non-respondent

Respondent

Total

Ratio column response/non response

2,696 95.64% 54.74% 2,229 94.77% 45.26%

123 4.36% 50.00% 123 5.23% 50.00%

54.52% 2,352

0.9

45.48%

1.1

4,925 95.24%

246 4.76%

5,171 100.00%

Own calculations based on data from the CBSS.

2,819

Characteristics of respondents and non-respondents

W

63

Our sample contained more men (54.5%) than women (45.5%) (table 2.2). Nevertheless the same number of male and female respondents has answered our questionnaire. The response rate for women (5.2%) was a little higher than the one for men (4.4%). The results are not statistically significant (p = 0.1450).

Table 2.3. Respondents and non-respondents by socio-economic category (n = 5,171). Frequency Row Pct Col Pct

Non-respondent

Respondent

Employee

2,089 95.17% 42.42% 771 96.38% 15.65% 457 97.23% 9.28% 96 100.00% 1.95% 945 92.74% 19.19% 42 95.45% 0.85% 154 96.25% 3.13% 51 98.08% 1.04%

106 4.83% 43.09% 29 3.63% 11.79% 13 2.77% 5.28% 0 0.00% 0.00% 74 7.26% 30.08% 2 4.55% 0.81% 6 3.75% 2.44% 1 1.92% 0.41%

6 46.15% 0.12% 314 97.52% 6.38%

7 53.85% 2.85% 8 2.48% 3.25%

0.25% 322

23.75

6.23%

0.51

4,925 95.24%

246 4.76%

5,171 100.00%

Self-employed person

Unemployed person

Temporary unemployed person Pensioner

Primary incapacity

Permanent incapacity

Handicapped person

Social reintegration allowance Housewife/house husband Total

Source:

Total

Ratio column response/nonresponse

2,195 42.45% 800

1.02

15.47% 470

0.75

9.09% 96

0.57

1.86% 1,019

0

19.71% 44

1.57

0.85% 160

0.95

3.09% 52

0.78

1.01%

0.39

13

Own calculations based on data from the CBSS.

In table 2.3 we focus on the distribution over the socio-economic categories. The results are statistically significant (p < 0.0001). It could be possible that some categories are more willing to participate in a survey on social and fiscal fraud (or in general in a survey) than others.

64

X

Characteristics of respondents and non-respondents

Pensioners have the highest response rate with 7.3%, which is 2.5% point higher than the total response rate. This corresponds to 30.1% of the interviewed persons. This share is much higher than its share in the total sample (19.7%) and even the double of the share in the total population (14.6%). 106 employees have answered the questionnaire. This means a response rate of 4.8% or 43.1% of the interviewed persons. The response rate for persons who receive a social benefit is rather low. This seems to be especially the case for unemployed persons and temporary unemployed persons. The response rate for self-employed persons (3.6%) is below the total average response rate. It is also lower than the response rate for employees but it is higher than the response rate for the social benefit recipients like unemployed persons. Table 2.4. Respondents and non-respondents by age category (n = 5 169). Frequency Row Pct Col Pct 15-24

25-34

35-44

45-54

55-64

65-74

75-89

Total

*

Non-respondent

Respondent

300 98.04% 6.09% 887 95.17% 18.01% 963 96.40% 19.56% 977 95.69% 19.84% 929 94.89% 18.87% 845 92.96% 17.16% 23 100.00% 0.47%

6 1.96% 2.45% 45 4.83% 18.37% 36 3.60% 14.69% 44 4.31% 17.96% 50 5.11% 20.41% 64 7.04% 26.12% 0 0.00 0.00

4,924 95.26%

245 4.74%

Total

Ratio column response/nonresponse

306 5.92% 932

0.40

18.03% 999

1.02

19.33% 1,021

0.75

19.75% 979

0.91

18.94% 909

1.08

17.59% 23

1.52

0.44%

0.0

5,169* 100.00%

From two persons in the sample we do not know the age category.

Source:

Own calculations based on data from the CBSS.

Age may also have an influence on the response rate (table 2.4). We find that the results are statistically significant (p = 0.0021). We can also observe the impact of pensioners on our survey when we divide the non-respondents and respondents by age category. The persons between 64 and 74 years old in the sample have a response rate of 7%. This age category represents 26.1% of the respondents. The persons between 15 and 24 years old have the lowest response rate (2%). The age category 35-44 also participated less in the survey.

Characteristics of respondents and non-respondents

W

65

Table 2.5. Respondents and non-respondents by activity of the employees (NACE) (n = 2 195). Frequency Row Pct Col Pct Specialized construction works (NACE 43) Wholesale trade services, except of motor vehicles and motorcycles (NACE 46) Retail trade services, except of motor vehicles and motorcycles (NACE 47) Food and beverage serving services (NACE 56) Employment services (NACE 78) Public administration and defence services; compulsory social security services (NACE 84) Education services (NACE 85)

Human health services (NACE 86) Residential care services (NACE 87) Social work services without accommodation (NACE 88)

Source:

Nonrespondent

Respondent

79 97.53% 3.78% 105 98.13% 5.03% 154 96.86% 7.37% 56 91.80% 2.68% 128 94.81% 6.13% 153 91.07% 7.32%

2 2.47% 1.89% 2 1.87% 1.89% 5 3.14% 4.72% 5 8.20% 4.72% 7 5.19% 6.60% 15 8.93% 14.15%

160 93.02% 7.66%

12 6.98% 11.32%

115 89.84% 5.51% 80 98.77% 3.83% 81 95.29% 3.88%

13 10.16% 12.26% 1 1.23% 0.94% 4 4.71% 3.77%

Total

Ratio column response/nonresponse

81 3.69% 107

0.50

4.87% 159

0.38

7.24% 61

0.64

2.78% 135

1.76

6.15% 168

1.08

7.65%

1.93

172 7.84%

1.48

128 5.83% 2.23 81 3.69% 85

0.25

3.87%

0.97

Own calculations based on data from the CBSS.

The sector (by NACE)2 where the employee is employed may have an influence on the response rate. The employees have a response rate of 4.8%. The next table only shows the sectors with a relatively high number of employees. Employees active in the human health services (NACE 86) (10.2%) and the public administration (NACE 84) (8.9%) have the highest response rate. Also employees engaged in the education services (NACE 85) (7%) have a higher response rate than the mean for the employees (4.8%). It is also interesting

2.

The Statistical Classification of Economic Activities in the European Community (NACE) is a European industry standard classification system.

66

X

Characteristics of respondents and non-respondents

to look at the considered fraud sensitive sectors such as the construction and the hotel and catering industry. Employees working in these sectors are possibly less willing to participate in the questionnaire. A response rate of 8.2% for the employees employed in the food and beverage serving services (NACE 56) is much higher than the mean. The response rate in the sector of specialized construction works (NACE 43) is rather low (2.5%). The response rate of 4.8% was much lower than the response rate we initially had in mind. People from the Walloon Region were most inclined to answer the questionnaire. The socioeconomic category may have played the most important role in the decision to participate. 7.3% of the contacted pensioners were willing to respond to the questionnaire. We observe a small response rate for the unemployed persons (2.8%). In line with the high response rate of the pensioners, 7% of the contacted persons between 65 and 74 years old have participated.

Table 2.6. Response rate: a summary. Response rate Total

4.76%

Region – Flemish Region – Walloon Region – Brussels-Capital Region

4.43% 5.63% 3.53%

Sex – Man – Woman

4.36% 5.23%

Socio-economic category – Employee – Self-employed person – Pensioner – Unemployed person

4.83% 3.63% 7.26% 2.77%

Age category – 65-74 years

7.04%

Chi-Square (prob.)

0.0474**

0.1450

< 0.0001***

0.0021***

Note: *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. Source:

Own calculations based on data from the CBSS.

For each of the crosstabs we have checked the significance. The most significant variables are the ‘socio-economic category’ and the ‘region’. The variable ‘sex’ is not significant. The socio-economic category is also a major dimension for further analyses.

3.

Logistic Regression

By a logistic regression it is possible to verify if some of the independent variables (sex, region, age and socio-economic category) had an impact on the decision to respond on our

Characteristics of respondents and non-respondents

W

67

questionnaire. The dependent variable (response) is a dummy variable (0 = non response; 1 = response). References = man, Brussels-Capital Region, 45-54 years old, employee. The Walloon region (p = 0.0386) has a positive influence on the response rate. The social categories unemployed (p = 0.0050) and housewife (p = 0.0182) have a negative influence on the response rate.

Table 2.7. Logistic Regression (n = 5,171). Parameter

Estimate

Pr > ChiSq

Intercept Women Flemish Region Walloon Region 15-24 25-34 35-44 55-64 65-89 Primary or permanent incapacity Handicapped person Unemployed Self-employed Housewife Pensioner

-3.5066 0.2343 0.2036 0.4807 -0.8573 0.0989 -0.2081 0.1101 0.1548 -0.3644 -1.0800 -0.8345 -0.3541 -0.8963 0.1890

< .0001 0.0784* 0.3770 0.0386** 0.0524* 0.6507 0.3651 0.6229 0.5955 0.3352 0.2879 0.0050*** 0.1013 0.0182** 0.4587

Note: *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. Source:

Own calculations based on data from the CBSS.

Finally we could also argue that women seem to be overrepresented in our survey (p = 0.0784) while self-employed people are underrepresented (p = 0.1013). These results show the overrepresentation of some categories in the respondents. This is why we applied a weight for particular categories to our sample of respondents.

4.

The impact of a monetary incentive

The use of a monetary incentive is very common in (face-to-face) surveys. This can create intended but also unintended effects. The most important intended effect is a higher response rate (Singer, 2002; Davern, Rockwood and Campbell, 2003; Willimack, Schuman, Pennell and Lepkowski, 1995; Ryu, Couper and Marans, 2005; Castiglioni, Pforr and Krieger, 2008). The fact that a monetary incentive may attract a specific group of respondents can be an important unintended effect. Compared to internal motivators, external motivators (e.g. monetary incentive) may have a negative impact on the data quality. Also the size of the

68

X

Characteristics of respondents and non-respondents

monetary incentive will possibly have an effect on the number of respondents but of course also on the total cost of the research. In August 2010, given the very low response rate, we decided to introduce a financial incentive. The goal was to significantly increase the number of people willing to participate in the survey. The introduction of a voucher of € 10 was decided and it was clearly indicated on the envelope of the reminder letter. The table below shows the number of people who decided to participate in the survey before and after the introduction of the voucher. Note the greater increase for the French part, compared to the Dutch speaking part. It could have caused the higher response rate of the Walloon region compared to the Dutch-speaking population.

Table 2.8. Response rate before and after the monetary incentive. # respondents (before) Total – Dutch speaking – French speaking Source:

# respondents (after)

# respondents (total)

114

132

246

56 58

57 75

113 133

Own calculations based on SUBLEC data.

However, given the low number of respondents, it is impossible to draw conclusions about the impact of the voucher. Moreover, given the period the letters were sent (school holidays), it is possible that respondents in the group ‘after voucher’ initially did not see the first letter. Besides, some respondents refused the voucher and gave it to the interviewers. So it’s difficult to draw conclusions about the impact of the voucher even if the number of respondents more than doubled after the reminder.

Chapter 3

Socio-demographic and socio-economic information

1.

Socio-demographic information

The socio-demographic information of the respondents is unweighted. It will give an impression of the characteristics of the respondents. 52% male and 48% female respondents answered the questionnaire. The average age of the respondents is 51. Almost 10% of the respondents was not born in Belgium and 6.1% does not have the Belgian nationality. There is also a huge difference between the Dutch-speaking and the French-speaking respondents concerning the country of origin and nationality. Almost 75% of the respondents have an education level between 3 (upper secondary education) and 5 (first stage of tertiary education). 37% of the respondents are single; this percentage is lower for the Dutchspeaking than for the French-speaking respondents. The average size of the household is higher for the Dutch-speaking respondents then for the French-speaking respondents. This has an impact on the number of people who are receiving an income (from labor or social benefits).

Table 3.1. Summary of socio-demographic data (unweighted data) (n = 246). Belgium (n = 246)

Dutch-speaking (n = 114)

French-speaking (n = 132)

Sex – Man – Woman

52.0% 48.0%

57.5% 42.5%

47.4% 52.6%

Age – Mean

51

52

51

Native country – Belgium – Other country

90.2% 9.7%

97.4% 2.6%

84.2% 15.8%

Nationality – Belgian – Other nationality

93.9% 6.1%

97.3% 2.7%

91.0% 9.0%

70

X

Socio-demographic and socio-economic information

Belgium (n = 246)

Dutch-speaking (n = 114)

French-speaking (n = 132)

2.0%

0.9%

3.0%

4.5%

4.5%

4.5%

16.7%

19.6%

14.3%

24.1%

29.5%

19.6%

23.7%

21.4%

25.6%

25.3%

20.5%

29.3%

3.7%

3.6%

3.8%

63.0% 37.0%

70.8% 29.2%

56.4% 43.6%

Composition household (divided by total number of households) – Younger than 15 years old – Between 15 and 18 years old – Older then 18 years old

0.5 0.1 2.0

0.6 0.1 2.2

0.5 0.1 1.8

Receiving an income (labor, social benefit) – Mean

1.8

2.0

1.6

Education (ISCED) – Level 0 – Pre-primary education – Level 1 – Primary education or first stage of basic education – Level 2 – Lower secondary or second stage of basic education – Level 3 – (Upper) secondary education – Level 4 – Post-secondary nontertiary education – Level 5 – First stage of tertiary education – Level 6 – Second stage of tertiary education Civil state – Couple – Single

Source:

Own calculations based on data SUBLEC.

When we look at the social status (unweighted), most of the respondents were employees (42.3%) or social benefit recipients (42.3%). Some of them indicated a second social status, indicating mixed situations or careers. Due to the lack of respondents we have only made a more detailed analysis for the employees and the benefit recipients.

Table 3.2. Social status of the respondents (unweighted data) (n = 246).

Employee Self employed Benefits recipient Not active Not applicable Total Source:

Firstly

N

42.3% 6.1% 42.3% 9.4%

104 15 104 23

5.3% 7.3% 2.4% 0.8% 84.5%

13 18 6 2 207

100.0%

246

100.0%

246

Own calculations based on SUBLEC data.

Secondly

N

Socio-demographic and socio-economic information

W

71

Socio-economic information

2.

The socio-economic situation at micro (for an individual) or at macro (for a country) level may have an influence on the decision to commit social or fiscal fraud. Financial necessity is often called as one of the aspects for doing undeclared work or not declaring income. But also the (inter)national economic conditions are sometimes considered as one of the explaining variables (e.g. impact of the financial crisis) (Pfau-Effinger, 2009; Schneider, 2010). For this reason we have made a detailed view of the financial situation of the respondents and asked for their opinion about the Belgian economy. The impact of the financial situation on the decision to ask or supply undeclared work is verified further in this report by a probit analysis. A longitudinal survey on undeclared income and work would make it possible to verify the impact of the economic and financial crisis. The socio-economic questions are described for Belgium but also for the Dutch and Frenchspeaking citizens. The French-speaking respondents are more negative about the economic conditions in Belgium. One third of these respondents consider the Belgian economy in the last 12 months as ‘much worse’ than before. The Dutch-speaking respondents have a more positive view about the development of the Belgian economy. In general few respondents were very optimistic (response: ‘much better’). 34.1%

35,0%

30.3%

30,0%

27.6%

27.2%

25,0%

23.5% 20.5%

20,0%

20.9% 18.9% 14.4%

15,0% 10,0% 5,0%

34.0%

33.1%

7.7% 1.7% 2.1% 2.5%

0.0% 0.5% 1.0%

0,0% Much better

A little better

Remained the same

Belgium Source:

Dutch-speaking

A little worse

Much worse

Don't know

French-speaking

Own calculations based on SUBLEC data.

Figure 3.1. B1. Opinion about the development of the Belgian economy during the last 12 months. (n = 246)

43% of the respondents have difficulties to get by on their monthly official income. More French-speaking respondents suffer difficulties (49%) than Dutch-speaking respondents (36%). The percentage of respondents (25.5%) in a difficult or very difficult situation is somewhat higher than the results in the SILC (weighted data) (17.1%).1

1.

SILC= Survey on Income and Living Conditions

72

X

Socio-demographic and socio-economic information

29.4%

30,0%

25.3% 24.9% 24.5% 24.1%

25,0% 20,0%

18.9% 17.2% 15.6%

18.1% 16.1% 14.1%

15,0%

18.9%

12.1% 10,0%

9.0% 8.2% 7.5%

9.4% 6.7%

5,0% 0,0% Very difficult

Difficult

Rather difficult

Belgium Source:

Rather easily

Dutch-speaking

Easily

Very easily

French-speaking

Own calculations based on SUBLEC data.

Figure 3.2. B2. Is it possible to live on your monthly official income? (n = 246)

Most of the respondents with difficulties to get by on their monthly official income have to save on costs (54.6%) or have to use savings (23.9%). 8.2% would engage in informal activities but still the double would rebound on do-it-yourself activities. Remarkable is that those answering that they have difficulties to live on their income, still answer it is ‘sufficient’ (6% + 7.3%) if asked in another way.

Limit expenditures

54.6%

Use savings

23.9%

Overtime

6.1%

Take an additional job as an employee

2.8% 3.7%

Be self-employed next to my work as employee Do informal activities, such as doing odd jobs, etc... in other households

8.2%

Do his/her own household work, tinkering

15.5%

Borrow from the bank, friends or family

10.2%

My available monthly income (official) is sufficient to make ends meet The avalaible monthly income (official) of the household in which I live is enough to make ends meet

6.0% 7.3% 17.2%

Other Don't know

1.6%

0,0% Source:

10,0%

20,0%

30,0%

40,0%

50,0%

60,0%

Own calculations based on SUBLEC data.

Figure 3.3. B3. What do you have to do to get by until the end of the month? Only for categories with difficulties (B2 – 1 to 3) – Belgium (n = 105) – (multiple answers possible).

Socio-demographic and socio-economic information

W

73

66% of the Dutch-speaking respondents state that their current financial situation makes it possible to put some money aside. This is clearly more than for the French-speaking respondents (48%).2 70,0% 60,0%

66.0% 56.9% 51.9%

50,0%

47.6% 42.9%

40,0% 34.0% 30,0% 20,0% 10,0% 0,0%

0.2% Yes

No Belgium

Source:

Dutch-speaking

0.5% 0.0% Don't know

French-speaking

Own calculations based on SUBLEC data.

Figure 3.4. B4. Does your current financial situation make it possible to put some money aside? (n = 246)

On average 2/3 of the respondents has no loan or credit when a possible mortgage is excluded. The situation for the French-speaking respondents is similar to the situation for the Dutch-speaking respondents. 68.6%

67.6%

70,0%

66.7%

60,0% 50,0% 40,0% 32.4%

33.3%

31.4%

30,0% 20,0% 10,0% 0,0% Yes

No Belgium

Source:

Dutch-speaking

French-speaking

Own calculations on data SUBLEC.

Figure 3.5. B5. Do you have a loan or credit at this moment? (Mortgage excluded) (n = 246)

2.

Remark that the saving rate can vary per region.

74

X

Socio-demographic and socio-economic information

We further consider three dimensions which have possible effects on fraud (figure 3.6); tax morality, benefit of evading taxes and regulations, ands costs of fraud being detected and punished. In our socio-economic questions we have discussed two of those dimensions in a specific opinion question (B6, see table 3.3), namely the dimension ‘taxes, regulation, red tape’ and the dimension ‘morality, behavior’. Later in the opinion questions this triangle will come back. Then all of the three dimensions are discussed.

Morality, behavior

Taxes, regulation, ‘red tape’ Source:

Inspection, risk of being caught, enforcement

Pacolet and De Wispelaere, 2009a.

Figure 3.6. Dimensions of social and fiscal fraud.

In this part we get some more insight in the opinion of the respondents regarding the government, the legislation and the tax burden. This may have an influence on the morality and/or the behavior of the respondents (see figure 3.6; Torgler, 2007). A great part of the respondents agree with the idea that the tax burden in Belgium is too high. Even 48% of the respondents totally agree with this idea. The opinions about the fairness of our tax system differ but incline more to a negative view. 50% of the respondents totally or rather disagree with the opinion that our tax system is a fair system against 30% who totally or rather agree. Almost 80% of the respondents fully or rather agree that our social and fiscal legislation is too complicated. The judgments about the spending of the tax revenues differ substantially between the Dutch- and French-speaking respondents. The Dutch-speaking respondents are more convinced of the good spending of the tax revenues than the French-speaking respondents. The French-speaking respondents also disagree more with the idea that they receive enough back in exchange for the paid taxes. They are also less convinced of the fact that the government is acting correctly for its citizens. This reveals an unexpected higher mistrust in the French-speaking population towards the state and its spending. Each of those opinions or attitudes deserves a further analysis, but we limit the description because of the pilot-character of the survey.

1.1% 0.0% 2.2%

6) The Belgian government acts correct for its citizens Belgium Dutch-speaking French-speaking

Own calculations based on SUBLEC data.

5.3% 7.5% 3.2%

5) I get enough back in exchange for the taxes I pay Belgium Dutch-speaking French-speaking

22.7% 28.4% 17.1%

35.5% 39.3% 31.8%

20.7% 28.8% 12.8%

35.1% 36.6% 33.5%

27.4% 26.3% 28.4%

31.9% 26.8% 37.0%

Rather agree

32.8% 37.5% 28.2%

17.5% 16.8% 18.3%

19.5% 23.4% 15.7%

10.7% 11.5% 9.8%

17.1% 17.8% 16.4%

10.6% 11.9% 9.2%

Neither agree or disagree

27.4% 24.6% 30.1%

21.2% 19.3% 23.0%

35.4% 30.3% 40.2%

6.5% 5.6% 7.4%

32.1% 31.7% 32.4%

8.5% 9.8% 7.3%

Rather disagree

14.9% 8.6% 21.1%

19.3% 17.2% 21.5%

22.3% 15.8% 28.6%

1.8% 0.0% 3.5%

18.2% 17.5% 18.9%

0.4% 0.0% 0.7%

Totally disagree

1.1% 0.9% 1.4%

1.1% 0.0% 2.2%

0.7% 0.0% 1.4%

1.5% 0.0% 3.0%

3.0% 2.0% 3.8%

0.8% 0.6% 1.0%

Don’t know

W

Source:

1.5% 1.7% 1.3%

44.5% 46.2% 42.7%

2.3% 4.7% 0.0%

47.8% 50.8% 44.8%

4) The Belgian government is spending the tax revenues good in general Belgium Dutch-speaking French-speaking

3) The fiscal and social legislation is too complicated in Belgium Belgium Dutch-speaking French-speaking

2) The Belgian tax system is a fair system Belgium Dutch-speaking French-speaking

1) The tax burden is much too high in Belgium Belgium Dutch-speaking French-speaking

Totally agree

Table 3.3. B6. To what extent do you agree/disagree with following statements? (n = 246)

Socio-demographic and socio-economic information

75

Chapter 4

Demand for undeclared work

1.

General

38.8% of the respondents admit they have bought a good or service which embodied undeclared work during the last 12 months. This percentage is higher for the Dutch-speaking respondents (42.1%) than for the French-speaking respondents (35.5%). Compared to the Eurobarometer survey (18%), the demand for undeclared goods and services is much higher in the SUBLEC survey (38.8%). The percentage of respondents who have bought an undeclared service (35.2%) is much higher than the percentage of respondents who have bought an undeclared good (14.1%). This was also the case in the Eurobarometer survey (15% vs. 8%). It seems reasonable, given the ‘intangible’ aspect of services, making it more difficult to control? This observation could imply that a more and more service-oriented economy would become more sensitive to fraud. Of course there can be a great difference between the amount spent on these undeclared services and goods. On average € 1,553 was paid for the most expensive undeclared goods or services with a standard deviation of € 3,693. The average amount for Belgium in the Eurobarometer was € 1,050. When we translate these results to the total Belgian economy, 1.9% of the Belgian GDP could have been spent on the purchase of undeclared goods and services. This is more than the results obtained from the Eurobarometer survey (Belgium: 0.6% of the GDP and EU27: 0.5% of the GDP). It is also possible to estimate the share of the undeclared goods and services in the total household budget. For 2009 we find a total expense of € 14,651 per person and € 34,441 per household (Household Budget Survey, see website ADSEI/DGSIE). This implies a share of 4.1% of the expenses per person and 1.7% of the expenses of the household. The answers to our survey probably refer to the total household spending, and less to individual expenditure. This can also explain the insignificant impact of sex on this spending: the results reveal the household spending, and not the individual spending.1

1.

In the fully engaged roll out of the survey the unit of data collection needs to be cleared out: at individual or household level.

78

X

Demand for undeclared work

Table 4.1. Demand for undeclared goods and services. Yes

No

Don’t know

Refusal

In the last twelve months, have you acquired any services of which you had a good reason to assume that they embodied undeclared work, i.e. that the income was not completely reported to tax or social security institutions? (n = 246) Belgium 35.2% 64.8% 0% 0% Dutch-speaking 39.5% 60.5% 0% 0% French-speaking 31.0% 69.1% 0% 0% Eurobarometer Belgium 15% 82% 2% 1% Eurobarometer EU 9% 84% 5% 2% In the last twelve months, have you acquired any goods of which you had a good reason to assume that they embodied undeclared work, i.e. that the income was not completely reported to tax or social security institutions? (n = 246) Belgium 14.1% 84.2% 1.7% 0% Dutch-speaking 15.2% 82.9% 1.9% 0% French-speaking 13.0% 85.4% 1.6% 0% Eurobarometer Belgium 8% 89% 2% 1% Eurobarometer EU 6% 86% 6% 2% In the last twelve months, have you acquired any goods or services of which you had a good reason to assume that they embodied undeclared work, i.e. that the income was not completely reported to tax or social security institutions? (n = 246) Belgium 38.8% 59.4% 1.7% 0.0% Dutch-speaking 42.1% 56.0% 1.9% 0.0% French-speaking 35.5% 62.8% 1.6% 0.0% Eurobarometer Belgium 18% Eurobarometer EU 11% Mean

Std dvp

Thinking about the most important good or service coming from undeclared work you acquired, could you please indicate roughly how much money you spent on it in the last twelve months?(in €) (n = 79) Belgium 1,553 3,693 Dutch-speaking 1,632 4,456 French-speaking 1,461 2,814 Eurobarometer Belgium 1,050 Eurobarometer EU 1,028 % of GDP* Belgium Eurobarometer Belgium Eurobarometer EU *

1.9% 0.6% 0.5%

GDP/capita Belgium 2010 = € 32,400; GDP/capita Belgium 2007 = € 31,500; GDP/capita EU27 2007 = € 25,000 (Eurostat).

Source:

Own calculations based on SUBLEC data; EC, 2007b.

Within the context of this project a ‘side project’ was organized to reveal the behavior on fiscal and benefit fraud by way of laboratory experiments executed in Belgium (the Walloon Region and the Flemish Region) and two other European countries: France and the Netherlands (Lefebvre, Pestieau, Riedl and Villeval, 2011). Students were recruited from

Demand for undeclared work

W

79

undergraduate classes in economics and business only. Only Dutch-speaking students participated in the Maastricht sessions, Flemish students in the Leuven sessions, Walloon students in the Liège sessions and French students in Lyon. In the closing part of the session, some questions about demand for and supply of undeclared work were asked. The results from these questions are compared with the SUBLEC results. Compared with the SUBLEC results, the results obtained in the experiments with students in Belgium are rather close, but nevertheless reversed. Walloon students (44.7%) have acquired more undeclared goods or services than their Flemish (39.3%) counterparts. This is contrary to our results. The figures are lower in France (28.6%) and in the Netherlands (34.4%).

Table 4.2. Demand for undeclared goods or services by students, 2010.*

*

Yes

No

Don’t know

Flanders (n = 135) Wallonia (n = 114) France (n = 133) Netherlands (n = 131)

39.3% 44.7% 28.6% 34.4%

46.7% 43.0% 54.1% 60.3%

14.1% 12.3% 17.3% 5.3%

Total (n = 513)

36.5%

51.3%

12.3%

In the last twelve months, have you acquired any services or goods of which you had a good reason to assume that they embodied undeclared work?

Source:

Data obtained from M. Lefebvre (CREPP-ULg).

49.5% 45.8% 53.9%

Friends, colleagues or acquaintances

Relatives

Neighbours

4.5% 0.0%

9.8%

0.8% 1.4% 0.0% 15.4% 12.1% 19.3%

Other private persons or households

22.6%

Firms or businesses

33.3%

10.3% 4.7% 5.6% 3.6%

Other

2.5% 1.9% 3.2%

Don't know 0,0%

10,0% Belgium

Source:

20,0%

30,0%

Dutch-speaking

40,0%

50,0%

60,0%

French-speaking

Own calculations on data SUBLEC.

Figure 4.1. D5. Among the following, could you please indicate from whom you bought this good or service? (n = 98)

80

X

Demand for undeclared work

Coming back to the SUBLEC results, the respondents who bought an undeclared good or service principally got this from friends, colleagues or acquaintances (49.5%) (figure 4.1). Also in the Eurobarometer most of the goods or services were delivered by friends, colleagues or acquaintances (Belgium = 34%; EU27 = 33%) (EC, 2007b). Perhaps the fact that more services are acquired in this formal way, could rely better on an informal network of relatives and friends. Firms, as in the Eurobarometer for Belgium, are the second supplier of undeclared goods or services. The figures also reveal a much more informal way of organizing the use of undeclared goods or services in the French-speaking part of Belgium compared to the Dutch-speaking part of Belgium. It could reveal that undeclared work is much more formalized and embedded in firm activities in Flanders, but of course this also is only a hypothesis which needs further research.2

Lower price

54.7%

Faster service

15.1%

Better quality

10.0%

In order to help someone who is in need of money

9.6%

It was a favour amongst friends/colleagues It was a favour amongst relatives

13.7% 3.0%

Good/Service is not/hardly available on the regular market

10.2%

Other Don't know

16.2% 1.0%

0,0% Source:

10,0%

20,0%

30,0%

40,0%

50,0%

60,0%

Own calculations on data SUBLEC.

Figure 4.2. D6. From the following, what made you buy it undeclared instead of buying it on the regular market? – Belgium – (n = 82) (multiple answers possible).

The main reason for buying an undeclared service or good is the lower price (54.7%) (figure 4.2). Also in the Eurobarometer this was the most important reason for buying an undeclared good or service (Belgium = 56%; EU27 = 66%) (EC, 2007b). On the demand side, the cost-saving strategy will probably be the most important reason for buying goods and services on the irregular market (see also Pfau-Effinger, 2009). On most of the delivered

2.

Has it become more ‘systematic’ in Flanders, and did it remain a more ‘informal’ mutual support system in Wallonia? Is a different typology of undeclared work possible?

Demand for undeclared work

W

81

goods and services one has to pay a VAT rate of 21%. On top of other direct taxes and social security contributions, this means the costumer will save a lot of money when he buys his goods or services on the irregular market, in the hypothesis at least that the price benefit is really transferred to him and not ‘cashed’ by the provider. The advantage is probably split up between both of them. This cost-saving strategy of the customer (and/or provider) will however have a negative impact on the revenues of the state, and on the social protection of the provider and indirectly also of the user, but these ‘external effects’ of their fraud behavior are probably not visible to them. The figure below shows that if this good or service had only been available on the regular market, more than 65% of people would have bought it from the regular market. In the Eurobarometer we find a percentage of 52% for Belgium (EC, 2007b). This intention would have important consequences on possible revenues if the irregular market became totally regulated. Almost two thirds of the money consumed in the irregular market would then have been spent in the regular market. So a third of the money would never have been spent on the regular market for that good or service but it remains available for other spending (or savings). So why allow undeclared work: the fact that another third of the consumption would be replaced by ‘prosumption’ (‘the job would have been done by myself’) would even imply there is no welfare loss. The activities enter again in that other (often not calculated) part of our national income: home production of services.3 But undeclared activities risk in any case not being counted in official GDP either. Self provision seems to be more evident for domestic services as housekeeping, gardening, so this kind of question could probably also tell us something on the behavior with regard to service vouchers when they would become less generous or be abolished. Would it really result in more undeclared work or just in do-it-yourself?

65.8%

I would have bought it from the regular market I would have postponed the acquisition of this service or good

11.9%

The job would have been done by myself or another member of the household I would have renounced from purchasing this service or good

Other

0,0% Source:

27.0%

4.3%

2.3%

10,0%

20,0%

30,0%

40,0%

50,0%

60,0%

70,0%

Own calculations on the basis of SUBLEC data.

Figure 4.3. D7. From the following, what would you have done if this good or service had only been available on the regular market? – Belgium – (n = 85) (multiple answers possible).

3.

In the national account guidelines ‘self production of goods’ is included in GDP, ‘self production of services’ not.

82

X

Demand for undeclared work

The main reason for not buying products or services on the black market is the non-possibility of recourse in the event of failure or error (figure 4.4). More than 50% of respondents mention it. A high enough personal income was a reason for not buying undeclared goods or services for only 19%. This contrasts with the ‘saving-strategy’ applied by the buyers of undeclared goods or services. Finally, the figures incline to prove that the risk of getting caught or being controlled is not important in the decision of buying declared products or services. The high frequency of the category ‘personal moral objections’ leads to think that our respondents hold the values of honesty at the top rank. It is of course a typical question that invites for socially desirable answers. I have no recourse in the event of failure or error

51.0%

My income is high enough so I can pay everything with an invoice

19.0%

I am aware of the negative impact on state revenues

13.5%

Personal moral objections

29.4% 24.8%

I just did not think about it 5.2%

Too many risks to be controlled Excessive fines if you get caught

7.3%

Other

13.5%

Don't know

0.6%

0,0% Source:

10,0%

20,0%

30,0%

40,0%

50,0%

60,0%

Own calculations based on SUBLEC data.

Figure 4.4. D8. What are the factors that led you to not buy products or services on the black market? – Belgium – (n = 148) (multiple answers possible).

2.

Most important categories of expenditures

In the Eurobarometer, most of the acquired goods or services coming from undeclared work in Belgium were delivered as household services (31%) (EC, 2007b). This percentage is much higher than in the EU27 (17%). We have integrated some questions about household expenditures and more specifically about the possible expenses for housing, car and domestic help in our questionnaire.

2.1

Construction and housing

First, we tackle construction works. The construction industry can be considered as a fraud sensitive sector. In their business survey about ‘the unfair competition in the construction

Demand for undeclared work

W

83

industry’, Pacolet and Baeyens (2007) find that 12% of all proposals between construction firms are unfair, and an estimated loss of work of 25-31%. But also in the annual reports of the Belgian social inspection services we can find that an important percentage of the infringements are committed in the construction industry (SIOD, 2009; TSW, 2010; SI, 2010). A great part of the construction works are delivered to families. On average 75% of the respondents have an own house. 62% of the Dutch-speaking owners have a mortgage. This percentage is much lower for the French-speaking owners of a house (40.8%). More than half of the owners have done some changes and renovations on their house during the last 12 months. 62.6% of the Dutch-speaking people did changes and renovations in the last 12 months, in comparison with only 45.4% in the French part. When there are no (substantial) renovations there is probably less chance that there is a new or ongoing mortgage, what seems to be the case.

Table 4.3. Housing. Yes

No

Do you own a home (co-owner, tenant for life)? (n = 246) Belgium 75.3% 24.7% Dutch-speaking 76.8% 23.2% French-speaking 73.8% 26.2% Do you have a current mortgage for the repayment of this housing? (n = 186) Belgium 51.4% 48.6% Dutch-speaking 62.0% 38.0% French-speaking 40.8% 59.2% Is your home a new building built in the last twelve months? (n = 186) Belgium 4.2% 95.8% Dutch-speaking 4.7% 95.3% French-speaking 3.8% 96.3% Has your housing undergone changes and renovations over the last twelve months? (n = 186) Belgium 54.0% 46.0% Dutch-speaking 62.6% 37.4% French-speaking 45.4% 54.6% Source:

Own calculations on data SUBLEC.

More than a third (36.7%) of the respondents were entitled to subsidies from the government for the construction works done in their house. There is a potential for further research to verify a possible link between subsidies or fiscal stimuli and undeclared work, what is sometimes suggested as a strategy to reduce fraud. 32,9% of the respondents answered the work had not been (entirely) invoiced. For these 33%, out of € 1,000, € 351 had not been invoiced on average. In most cases it was the respondent who proposed not to invoice or just to invoice a portion of the work (44,4%). We observe differences in the extent to which both regions are evading taxes for the delivered construction works. The construction works were more frequently entirely invoiced for the Dutch-speaking respondents (72.2% vs. 60.5%). Almost two thirds (63%) of the Dutch-

84

Demand for undeclared work

X

Table 4.4. Construction works. Have you been entitled to subsidies from the government for work done in your home? (n = 105)

Belgium Dutch-speaking French-speaking

Yes

No

36.7% 37.7% 35.4%

63.3% 62.3% 64.6%

Has the work been charged? (n = 102)

Belgium Dutch-speaking French-speaking

Yes

Yes, but charged partially

No, entirely without invoice

67.1% 72.2% 60.5%

13.8% 10.3% 18.3%

19.1% 17.5% 21.2%

Out of € 1 000 that you paid, how much is not been charged? (n = 30)

Belgium

Mean

Std dvp

351.66

386.23

Who proposed not to charge or charge just a part of the work? (n = 30) Myself Belgium Source:

The person who performed the work In consultation

44.38%

22.74%

28.46%

4.42%

Own calculations based on SUBLEC data.

The introduction of a general reduction of VAT, which would allow me to pay less

36.1%

A tax reduction for customers who use a licensed contractor

27.8%

A more restrictive control policy: increased penalties and/or higher risk to be discovered

15.4%

You are entitled to government incentives that reduce the total cost You are not satisfied with the quality of work performed

33.9%

6.4%

You want a guarantee in case of faults

Other

Don't know

0,0% Source:

A third

50.8%

3.3%

9.7%

10,0%

20,0%

30,0%

40,0%

50,0%

60,0%

Own calculations on data SUBLEC.

Figure 4.5. D17. What would induce you to charge all the work next time? – Belgium – (n = 30) (multiple answers possible).

Demand for undeclared work

W

85

speaking respondents who were committing tax fraud have entirely paid the work without an invoice. The division is more equal for the French-speaking tax evaders (for 46% a part of the work was invoiced; for 54% no invoice was made). Figure 4.5 shows the reasons which would induce respondents to charge all the work next time they would do construction works. For more than half of the respondents the guarantee in case of faults is the most important reason (50.8%). This ‘fear’ could be used in campaigns to warn people about undeclared work.4 The economic reasons are cited quite regularly (VAT reduction: 36.1%; tax reduction: 27.8% or incentives reducing the total cost: 33.9%). For construction works which have been invoiced partially, the customer still has a guarantee. However, almost 20% will not have this guarantee.

Car repairs

2.2

Car repairs are possibly sensitive for tax evasion and undeclared work too (although here guarantees for quality and safety are essential too, so ‘rational consumers’ would avoid fraud here). For car repairs, the percentage of people who have not paid the entire repair with an invoice is around 10%. The fraud percentage is indeed lower here. The respondents seem to answer in a rational way. Within these 10%, in most cases the person who performed the work or a third person (e.g. the owner of the garage) proposed not to invoice the work or to invoice just a part of it (51%). One fourth of the tax avoiding car owners asks not to invoice the work or to only invoice a part of it. Table 4.5. Car repairs. Do you own a car? (n = 246)

Belgium Dutch-speaking French-speaking

Yes

No

76.2% 82.4% 70.2%

23.8% 17.6% 29.8%

Have the latest maintenance works on your car have been invoiced? (n = 185)

Belgium Dutch-speaking French-speaking

Yes

Yes, but partially invoiced

No, entirely without invoice

Don’t know

88.9% 90.1% 87.3%

0.5% 0.0% 1.1%

9.8% 8.9% 10.9%

0.9% 1.0% 0.8%

Who proposed not to invoice the work or to invoice just a part of it? (n = 22)

Belgium Source:

4.

Myself

The person who performed the work

In consultation

A third

Don’t know

24.9%

30.1%

19.6%

20.9%

4.5%

Own calculations based on SUBLEC data.

A campaign of ‘bad examples’ or ‘horror stories’ in this context (and probably they are already well known) could be used.

86

X

Demand for undeclared work

Domestic work and service vouchers

2.3

Car repair and maintenance and perhaps also construction works seem to be less controllable ‘de visu’, so people should mistrust fraudulent providers. Domestic help is better observable and controllable. Should fraud be more massive? Undeclared work delivered for domestic work is a widespread problem. In Belgium a household can use ‘service vouchers’ for the following household activities5: – In the home of the user: house cleaning, doing the laundry and ironing, sewing, cooking; – Outside the home of the user: doing groceries, transportation of persons with reduced mobility, ironing in an ironing shop. A household has to pay € 7.5 for one service voucher or one hour of domestic work. 30% of the price is tax-deductible up to a total amount of purchased vouchers of € 2,560 (2011). So the actual price will be of € 5.25. The recognized companies receive € 21.1. The difference between the price for the consumer and the amount for a recognized company, € 13.6, is paid by the federal government. The system of the service vouchers is a huge success in Belgium. In 2010 there were 760,702 users, 97.2 million purchased service vouchers, 2,664 recognized companies and more than 100,000 employees (Pacolet, De Wispelaere and De Coninck, 2011). One of the main goals of the service voucher is turning undeclared domestic work into regular jobs. In earlier studies (IDEA Consult, 2009), between 10 to 20% of the voucher users admitted to have made appeal to an undeclared household help in the past. 1 to 3% of the employees employed with service vouchers declare to have worked as an undeclared household help in the past. The authors concluded nevertheless that the service voucher was a successful instrument in the fight against fraud. Reading those figures we could hardly come to that conclusion. What does our SUBLEC-survey reveal? 37.3% of the respondents indicate they have a housekeeper for doing the cleaning of their house (table 4.6). Of this percentage, 75.5% uses service vouchers (or 28.1% of the respondents). This is disproportionate to the figures we know about the use of service vouchers. On average 9% of the Belgian population older than 20 years and 12% of the Belgian households is using service vouchers (RVA-ONEM, 2011). It should however be noted that our response group is mainly composed out of persons in the higher income sectors (education, health care, administration) and retired persons. Both can use more service vouchers. Tabel 4.6. Housekeeper. Yes

No

Do you employ a housekeeper for the maintenance of your home? (n = 246) Belgium 37.3% 62.4% Dutch-speaking 35.5% 64.5% French-speaking 39.2% 60.3% Do you have an insurance for accidents at work for your private home help? (n = 21) Belgium 65.7% 33.1% Source:

5.

Own calculations based on SUBLEC data.

Art. 1, 2°, Royal Decree of 12 December 2001 ‘betreffende de dienstencheques’.

Don’t know 0.3% 0% 0.5% 1.2%

Demand for undeclared work

W

87

Figure 4.6 shows a very high use of service vouchers for those using housekeeping help (75.5%). Despite the widespread use of vouchers for domestic help, still 20.4% of the users employ an undeclared household help (or 7.6% of the respondents). It would imply, for a complete ‘laundering’ of the sector that it could grow further with about 25%, without taking into account new users of home help, if these figures are reliable. A third of these users have no insurance for accidents at work for their undeclared domestic worker. Compared with other evaluation studies of the service voucher, this would imply that a huge part of domestic work still remains undeclared. Also completely different from the studies of IDEA Consult is the answer with regard to the demand for undeclared help before the use of service vouchers. In the SUBLEC-survey 48.2% answered that they asked for undeclared work (figure 4.7). In IDEA Consult studies this is only 10 to 20%. There is no obvious reason why our respondents should be more honest in answering these questions, except perhaps that the methodology is indeed better (face-to-face instead of a telephone survey). That would imply that it is a much more effective way of transforming undeclared work in declared. We already criticized the IDEA Consult conclusions on their own evidence before. Our new evidence remains to be confirmed since it would imply a substantial amount of undeclared work in this field, while we had more the impression that completely new users were created. It does not change our conclusion that it is impossible to sustainably transform undeclared work in declared work by subsidizing it massively. The survey would even make us wonder whether it is appropriate to subsidize two (or ten)6 users to regulate 1 user.

3.

Probit analysis

In this part, the goal is to determine which variables explain the demand for goods and/ or services not reported. By a probit regression, it is possible to verify if some of the independent variables (sex (A1), language (taal), financial need – income (B2), knowing other people buying undeclared goods and/or services (C6), socio-economic variables (F1a) and one about morality (B6_1)) had an impact on the decision to demand for undeclared goods and/or services. The dependent variable (demand) is a dummy variable (0 = no demand; 1 = demand). The financial need/income variable is built by combining the first 3 answers to the question B2 (difficult to make ends meet) and the last 3 responses to the same question (easy to make ends meet). The results are summarized in the following table 4.7. (References = woman, Flemish, employee, knowing nobody, high income and neither agree or disagree).

6.

‘Two’ is when 48% of the work was done undeclared before, ‘ten’ would be the case if the previously undeclared activities would be only 10%, as sometimes revealed in the evaluation reports of the system. In the questionnaire it should be made clear to which time period it is related.

88

X

80,0%

Demand for undeclared work

75.5%

70,0% 60,0% 50,0% 40,0% 30,0% 20.4%

20,0% 10,0% 1.0%

1.6%

1.5%

Other

Don't know

0,0% Household help paid with service vouchers

Source:

Private household help (declared income)

Undeclared private household help (undeclared income)

Own calculations based on SUBLEC data.

Figure 4.6. D21. Household help – Belgium – (n = 89) (one possible answer).

60,0% 51.9% 50,0%

48.2%

40,0%

30,0%

20,0%

10,0%

0,0% Yes Source:

No

Own calculations based on SUBLEC data.

Figure 4.7. D24. In the past, did you use a private household help who came to work without being declared? – Only for the current users of service vouchers – Belgium – (n = 68).

Demand for undeclared work

W

89

Table 4.7. Probit analysis demand for undeclared work. Parameter

Variable

Intercept Sex (A1)

Man

Region (taal)

French-speaking

Socio-economic category (F1a)

Self-employed Benefits recipient Inactive

Know someone (C6)

Yes

Income (B2)

Difficult

Morality (B6_1)

Totally agree Rather agree Disagree

Pr > Khi2

Estimation

Std error

-0.3613 -0.0331 (-0.012) -0.0896 (-0.033) 0.8488 (0.328) -0.4383 (-0.155) -0.4079 (-0.135) 0.8468 (0.266) -0.3393 (-0.121) -0.5279 (-0.188) 0.261 (-0.092) -0.1914 (-0.66)

0.3613 0.1804

0.3172 0.8543

0.1816

0.6217

0.3974

0.0327**

0.1953

0.0248**

0.3397

0.2298

0.2477

0.0006***

0.1879

0.0710*

0.2873

0.0661*

0.2954

0.3768

0.3929

0.6261

Note: *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. Marginal effects7 between brackets. Source:

Own calculations based on SUBLEC data.

Sex and Region do not have an influence on the demand. The absence of an influence on the demand could be explained by the fact that information on the demand is probably given for expenditures at the household level (housing, car, and domestic help). Information on the gender of the respondent is of less relevance then. This will be confirmed later on by the answers on the supply of undeclared work. This is given at individual level. The benefit recipients appear to use fewer goods and services not reported (p = 0.0248). It is opposite to the hypothesis that undeclared work is a way to overcome financial distress. This negative sign appears even combined with the ‘income’ variable. The table shows a different result for the fact of being self-employed (p = 0.0327). It seems to have a positive influence on the demand. Other results may seem surprising: agreeing totally with ‘The tax burden is much too high in Belgium’ (p = 0.0661) and meeting revenue difficulties (p = 0.0710) would lead respondents to consume less undeclared goods and services. A last but important result is that if respondents know someone who uses such goods or services, they tend to buy them too (p = 0.0006). This is also relevant for policy makers. It would imply that preventive actions, avoiding such bad examples, could create a vicious circle of good conduct.

7.

The impact of the explanatory variables on the probability of asking undeclared goods or services is expressed as marginal effect.

90

X

Demand for undeclared work

Compared to employees, the likelihood of asking undeclared goods or services is 15.5% lower among social benefit recipients and is 32.8% higher among self-employed persons (see table 4.7 – marginal effects). Knowing someone who asks for undeclared goods or services increases the likelihood by 26.6% compared to knowing no one. Suffering difficulties to get by on the monthly income reduces the likelihood of asking undeclared work by 12.1%. It suggests that basic goods are bought on the formal market.

4.

Conclusion

The characteristics of the people who are asking for undeclared services and goods were verified. This allowed us to make a kind of profile of the demand side of undeclared work. We have looked at the impact of the sex, the region and the social statute. More men than women and more Dutch-speaking than French-speaking respondents seem to ask for undeclared goods or services. For both variables the results are nevertheless not statistically significant. This was also the case in the probit analysis. Most noticeable is the high demand for undeclared work by self-employed persons (72.5%). This category lays 33.7% point higher than the average (72.5% vs. 38.8%). The probit analysis confirms the positive influence of the self-employed persons on the demand for undeclared work (p = 0.0327). Benefit recipients and inactive people are underrepresented in the group of people asking for undeclared goods or services. Either they have a lower spending capacity, or they produce some goods and services themselves since they have more leisure time. The negative sign of the income variable (here more difficulties to get by) confirms this. It challenges the literature that says that undeclared work is useful as a surviving strategy. As we doubted from the beginning, we discovered a cluster of ‘poor but honest’ persons. The results are statistically significant (p < 0.001). In the probit analysis we indeed find a negative influence on the demand for undeclared work for the benefits recipients (p = 0.0248). Table 4.8. Summary demand for undeclared work. Yes

No

Don’t know

38.8%

59.4%

1.7%

41.5% 35.7%

56.3% 63.2%

2.25% 1.13%

Region – Dutch-speaking – French-speaking

42.1% 35.5%

56.0% 62.8%

1.9% 1.6%

Socio-economic group – Employee – Self employed – Benefits recipient – Not active

41.6% 72.5% 23.3% 25.4%

56.2% 27.5% 74.6% 73.1%

2.2% 0.0% 2.0% 1.6%

Total Sex – Man – Woman

0.4836

0.5504

< 0.001***

Note: *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. Source:

Chi-Square (prob.)

Own calculations based on SUBLEC data.

Chapter 5

Supply of undeclared work

1. 1.1

General Frequency

14.1% admits to have done undeclared work during the last 12 months (figure 5.1). This percentage is much higher than the percentages we observe in the Eurobarometer survey about undeclared work. In that survey, a percentage of 6% was found for Belgium and a percentage of 5% for the EU27. In our SUBLEC-survey 17.2% of the French-speaking respondents carried out undeclared activities during the last 12 months. Only 10.9% of the Dutch-speaking respondents carried out undeclared activities. In Flanders, it seems that there are more people that ask for undeclared work than people that supply it. Perhaps the fact that more formal providers (firms) offer this undeclared work, could explain the difference. This needs further research. For instance does the worker consider his activities in the firm as ‘undeclared’ work?

1.2

Intensity and volume

Essential for calculations of the total volume of the underground economy is, besides the frequency of undeclared work, also the volume of activities provided. Only a minority of the undeclared workers (17.2%) performed this activity only once (figure 5.2). Most of the undeclared workers did it a few times (45.6%) or with a certain regularity. These results are similar to the results of the Eurobarometer for Belgium. To know the volume of undeclared work, we multiplied the average number of weeks and hours carried out by undeclared workers by the average remuneration for this undeclared activity (table 5.1). The undeclared workers who were paid only in cash received an average pay of € 15.5 per hour. This average wage of € 15.5 per hour is similar to the results from the Eurobarometer. In total the undeclared workers received an average pay of € 1,354 during the last 12 months. As a control question, we also asked for the total amount they received for this undeclared activity. The amount of € 1,332 is in accordance with the calculations we made. We observe a much higher difference between both in the Eurobarometer. An average of 0.6% of the GDP is lost via undeclared activities. This is much less than the 1.9% of the GDP calculated for the demand for undeclared goods and services.

92

X

Supply of undeclared work

According to the 2009 Household Budget Survey, Belgians had an average disposable income of € 17,441 per person. This implies that the average amount received for carrying out undeclared work is only 1.1% of this individual disposable income.

100% 90% 80% 70% 60%

85.9%

82.8%

89.1%

93%

50%

92%

40% 30% 20% 10%

14.1%

10.9%

Belgium

Dutchspeaking

17.2% 6%

5%

Eurobarometer Belgium

Eurobarometer EU27

0%

Yes Source:

Frenchspeaking No

Refusal

Don't know

Own calculations on data SUBLEC; EC, 2007b.

Figure 5.1. F3. Did you yourself carry out any undeclared activities in the last twelve months for which you were paid in money or in kind? Herewith we mean activities which were not or not fully reported to the tax or social security authorities and where the person who acquired the good or service was aware of this. (n = 246)

Eurobarometer Belgium

SUBLEC

0%

13%

49%

17.2%

45.6%

20%

40% Just once

Bron:

38%

A few times

37.2%

60%

80%

100%

With certain regularity

Own calculations based on SUBLEC data; EC, 2007b.

Figure 5.2. F5. Thinking about the most significant undeclared work you just mentioned, did you carry out this activity only once or a few times or do you carry it out with certain regularity? (n = 35)

Supply of undeclared work

93

W

Table 5.1. Volume of undeclared work (F6-F11) (n = 26).

Weeks Hours per week Amount per hour (in €) Total (in €)* Total (in €)** % of GDP***

SUBLEC

Eurobarometer Belgium

Eurobarometer EU27

11.1 7.9 15.5 1,354 1,332 0.6%

10.9 14.5 14.6 2,308 1,000 0.2-0.4%

10.5 18.9 16.6 3,294 1,118.5 0.2-0.7%

* Weeks, hours and amount per hour. ** Control question. *** GDP/capita Belgium 2010 = € 32,400; GDP/capita Belgium 2007 = € 31,500; GDP/capita EU27 2007 = € 25,000 (Eurostat). Source:

Own calculations based on SUBLEC data; EC, 2007b.

Almost the majority of the undeclared workers have carried out these activities for friends, colleagues or acquaintances (43.9%). Only a fifth of the undeclared workers have done this for firms or businesses. The figures indicate the undeclared work is mainly a personal or household service. We already found this informal way of organizing undeclared work in the demand for undeclared work (especially for the French-speaking part) (figure 4.1). These figures almost do not differ from these of the Eurobarometer for Belgium. 45% answered ‘friends, colleagues or acquaintances’ and 24% ‘firms or businesses’.

43.9%

Friends, colleagues or acquaintances

45% 19.4%

Relatives

12% 12.3%

Neighbours

24.5%

Other private persons or households

17% 20.1%

Firms or businesses

24% 15.7%

Other

2% 0%

10%

20% SUBLEC

Source:

30%

40%

50%

Eurobarometer Belgium

Own calculations based on SUBLEC data; EC, 2007b.

Figure 5.3. F13. Among the following, would you please indicate for whom you carried out this activity? (n = 35)

One of the main reasons for carrying out undeclared work is that both parties take advantage of it (52.1%). The customer does not pay VAT (and (or other) taxes), which reduces

94

X

Supply of undeclared work

the total amount of the bill. The supplier of undeclared work will not pay social and fiscal contributions, which makes the gross income equal to the net income. The financial necessity is also an important reason (17.0%), which does not mean that it is a surviving strategy. Both answers can be linked to the main reason why people are buying undeclared goods or services, namely ‘the lower price’ (figure 4.2). Knowing the main reasons for doing undeclared work is important when one wants to tackle undeclared work. The financial aspect (lower price for the customer, no social and fiscal contributions for the provider) seems to be the main reason for asking and offering undeclared work. It is however not possible to link the saving-strategy of the consumer directly with a financial necessity. The probit analysis shows that a low income has a negative influence on the demand for undeclared.

The person(s) who acquired it insisted on the nondeclaration Bureaucracy/red tape to carry out a regular economic activity is too complicated You could not find a regular job

16.5% 7.0% 3.6%

Both parties benefited from it

52.1%

Taxes and/or social security contributions are too high

13.9%

It is just a seasonal work and so it is not worth to declare it

16.3%

Working undeclared is common practice in your region/sector of a activity so there is no real alternative

9.9%

The State does not do anything for you, so why should you pay taxes

9.9%

Financial necessity Other 0,0% Source:

17.0% 12.1% 10,0%

20,0%

30,0%

40,0%

50,0%

60,0%

Own calculations on data SUBLEC.

Figure 5.4. F14. Among the following, what were the reasons for doing this activity in an undeclared way? – Belgium – (n = 35) (multiple answers possible).

For a quarter (25.0%) a sufficiently high income (from labor or a social benefit) is playing an important role in the decision not to perform undeclared work. For policy makers this observation could inform us about the potential increase of undeclared work when benefits are cut. Also personal moral objections prevent people from carrying out undeclared work (20.4%). The third most important reason is a lack of time for doing undeclared work. The lack of social protection seems of relevance for some 10% of the sample. We also observe differences between the French- and Dutch-speaking respondents. Personal moral objections seem to count more for the French-speaking part. A sufficiently high income and lack of time are two reasons which play a more important role for the Dutchspeaking part.

Supply of undeclared work

2.9% 3.1% 2.7%

Personal moral objections

17.1%

20.4% 24.0% 25.0%

A sufficient high income to get by on it The financial consequences are to high when catched

95

10.2% 10.9% 9.5%

Lack of social protection against illness and/or accidents Not building up a pension

W

21.2%

28.7%

2.5% 2.5% 2.4%

1.2% It is not generally accepted 0.0% 2.5% The chance of being catched is too high

3.3% 1.7% 5.1%

Lack of time off for doing undeclared work Lack of technical skills for offering undeclared work Avoiding the risk of losing a social benefit

12.4% 2.0% 1.9% 2.1% 2.9% 2.1% 3.7%

14.4% 14.5% 14.3%

Other

0,0%

5,0% Belgium

Source:

15.1% 17.7%

10,0%

15,0%

Dutch-speaking

20,0%

25,0%

30,0%

French-speaking

Own calculations based on SUBLEC data.

Figure 5.5. F15. Among the following, what were the reasons for doing no undeclared work in the last twelve months? Firstly – (n = 211).

2.

Specific socio-economic categories

Due to the lack of respondents only the specific questions for employees and benefit recipients are tackled. This was not possible for self-employed persons.

2.1

Employees

For achieving ‘exhaustiveness’ in the national accounts the total sum of wages also includes the corrections made for undeclared labor. In 2003 there was an increase of 0.7% of the total wages in the national accounts for undeclared labor (Pacolet, Perelman, Pestieau, Baeyens & De Wispelaere, 2009). Initially it was our ambition to estimate the amount of the undeclared wages by sector. This good practice, the ‘bottom up’ approach, is also applied in the national accounts. The lack of respondents made it impossible to implement this strategy in our pilot study. Most of the employees (weighted) are active as non-manual workers in the private sector (48.2%). Most of the employees also have a permanent contract (93.2%) and a full-time job (58.3%).

96

X

Supply of undeclared work

Table 5.2. Employees – context (G2-G4) – Belgium – (n = 116). Professional status – Private sector – manual (‘blue-collar’) – Private sector – non-manual (‘white collar’) – Public sector – permanent position – Public sector – temporary position – Don’t know

16.7% 48.2% 19.2% 15.4% 0.5%

Type of contract – Permanent – Temporary

93.2% 6.8%

Type of job – Full-time – Part-time

58.3% 41.7%

Source:

Own calculations based on SUBLEC data.

We have compared the weekly working time defined in the labor contract with the actual situation. An average of 32.2 working hours a week is defined in the labor contracts. In reality the employees are working 34.4 hours a week or 2 hours and 11 minutes more than defined in the labor contract. The question is whether these overtime hours are (un)declared (exception higher management). In the Labour Force Survey, in Belgium organized by the General Direction of Statistics and Economic Information (ADSEI/DGSIE), this question is also asked. In the past we always emphasized the importance of this survey and the possibilities to add questions about undeclared work (Pacolet & De Wispelaere, 2009a).1

Table 5.3. Weekly working time (contract versus actual situation) (G5-G5b) – Belgium – (n = 110).

Contract Actually Difference Source:

Mean

Std Dev

32.2 34.4 2.2 (2 hours 12 minutes)

9.6 11.3

Own calculations based on SUBLEC data.

The employees receive an average gross monthly income of € 2,530. Their average net monthly income amounts to € 1,598. An estimation of the evaded wage could be obtained by comparing the gross income of this survey with the tax declaration, but if we would limit ourselves to that use, better population surveys on income exist for the moment (SILC). A study of DULBEA compared the Household Budget Survey with the tax declaration and calculated a loss of fiscal revenues (after taxes) of € 20 billion or 6% of GDP (Diallo, Karakaya, Meulders, Plasman, 2010). But there again certain limits can be formulated since income concepts are not always comparable.

1.

See also our report on using the LFS for describing paid an unpaid (or perhaps undeclared) overtime (Pacolet & Coudron).

Supply of undeclared work

W

97

Taking the difference between the contractual and actual working time and the gross income into account, a yearly amount of € 2,247 is not declared to the tax administration. This is 7.4% of the average yearly gross income. But we do not have the confirmation in our survey that the overtime is not declared. Part-time work and overtime are risk factors for undeclared work that should be further investigated. Table 5.4. Gross and net income (G6-G7), in €, Belgium.

Gross income (n = 96) Net income (n = 111) Source:

Mean

Std Dev

2,530.26 1,598.30

1,362.88 663.25

Own calculations based on SUBLEC data.

17.0%

Company car

41.0%

Reimbursement home to work movement

39.3%

Meal voucher 26.8%

GSM 12.8%

PC–internet

40.4%

Hospitalization insurance 7.9%

Profit taking–share option

14.9%

Fuel card

28.6%

Group assurance

23.6%

Sport check-gift vouchereco-voucher 2.6%

Crèche

12.4%

Payment of expenses 4.4%

Bonus schemes

7.9%

Benefit rate

31.0%

Thirteenth month, fourteenth month

71.5%

Endof-year-bonus

91.1%

Holiday pay 1.4%

Sale or production bonus

6.9%

Extra income from working overtime

1.4%

Extra income from commission Tips

0.0% 16.6%

Others 1.4%

None Don't know

0.6%

Refusal

0.0%

0,0%

Source:

10,0%

20,0%

30,0%

40,0%

50,0%

60,0%

70,0%

80,0%

90,0% 100,0%

Own calculations based on SUBLEC data.

Figure 5.6. G8. Which allowances, bonuses or benefits are attributed to you?- Belgium – (n = 116).

In addition to their income, employees can also receive allowances, bonuses or benefits (figure 5.6). 91.1% of the employees receive a holiday pay. Most of the employees (71.5%) also

98

X

Supply of undeclared work

have an end-of-year bonus. Fringe benefits such as commuting reimbursement (41.0%), hospital insurance (40.4%), meal vouchers (393%) are also very popular. Only 2% of the employees are receiving ‘cash in hand’ remunerations (figure 5.7). The ’cash in hand’ is paid for all or part of the regular salary or the remuneration for extra work or overtime hours. This share was much higher (6%) in the Eurobarometer for Belgium.

100% 80% 60% 94%

98%

89%

40% 20% 0%

2%

6%

5%

SUBLEC

Eurobarometer Belgium

Eurobarometer EU27

Yes Source:

No

Refusal

Don't know

Own calculations based on SUBLEC data; EC, 2007b.

Figure 5.7. G13. Sometimes employers prefer to pay all or part of the regular salary or the remuneration for extra work or overtime hours cash-in-hand and without declaring it to tax or social security authorities. Did your employer pay all or part of your income in the last 12 months in this way? (n = 115)

90,0%

84.9%

80,0% 70,0% 60,0% 50,0% 40,0% 30,0% 20,0%

15.1%

10,0% 0,0% Yes Source:

No

Own calculations on data SUBLEC.

Figure 5.8. G17. Did you receive a reimbursement of expenses during the last twelve months which can be considered as tax evasion? – Belgium – (n = 114).

Supply of undeclared work

W

99

15% of the respondents admit to have received a reimbursement of expenses which can be considered as an evaded tax (figure 5.8).

Social benefit recipients

2.2

Almost 60% of the social benefit recipients were retired. Due to the lack of respondents it was impossible to make an individual analysis for each category of social benefit recipients.

Table 5.5. I1. Category social benefit recipient – Belgium – (n = 104). Unemployed Retired Early retired Permanent incapacity Primary incapacity Disabled Social reintegration income Other Source:

15.0% 57.5% 4.5% 8.0% 0.8% 2.1% 3.8% 8.3%

Own calculations based on SUBLEC data.

Most of the social benefit recipients consider their benefit as just enough or too low compared to their last income. This confirms a growing concern about the adequacy of social benefits. For us, it reflects a ‘shocking’ picture of a welfare state in decline.

Much too low

28.5%

Too low

29.8%

Just enough

35.2%

Too high

1.0%

Much too high

0.0%

Don’t know 0,0%

Source:

5.5% 5,0%

10,0%

15,0%

20,0%

25,0%

30,0%

35,0%

40,0%

Own calculations based on SUBLEC data.

Figure 5.9. I4. What do you think of the sum of your social benefit compared to your last received income? – Belgium – (n = 89)

100

X

Supply of undeclared work

Almost a fifth (19.2%) of the social benefit recipients is not aware of their rights and duties (according to their own opinion). This implies that the risk for doing something that is not in accordance with the law is higher. It could also illustrate the risk of non-uptake of all kind of benefits. According to their knowledge, 5.6% of the social benefit recipients did not respect, the conditions of payment of the social benefit. Only a small percentage of the benefit recipients declare to have made a wrong statement about their family situation (2.0%) or to have cumulated their social benefit unlawfully with another social benefit (1.6%). 4.3% of the social benefit recipients have cumulated their social benefit with undeclared work.

Table 5.6. Social benefit fraud, Belgium. I5. Are you aware of your rights and duties as a benefit recipient? (n = 89) Yes 75.9% No 19.2% Don’t know 4.9% I6. Do you respect the conditions for the payment of your social benefit? (n = 91) Yes 94.4% No 5.6% I7. Did you make the right statements about your family situation? (n = 91) Yes 98.0% No 2.0% I8. Did you cumulate your social benefit with other social benefits unlawfully during the last twelve months? (n = 91) Yes 1.6% No 97.6% I9. Did you cumulate your social benefit with undeclared work during the last twelve months? (n = 91) Yes 4.3% No 95.7% Source:

3.

Own calculations based on SUBLEC data.

Coherence between the supply of and the demand for undeclared work

We observe (except with regard to the volume) a coherence between the supply of undeclared work and the demand for undeclared work. 70% of the respondents who are carrying out undeclared work also ask for undeclared work. 65% of the respondents who are only working at the regular market did not ask for undeclared work. Only 6% of the respondents who are not demanding undeclared work are carrying out undeclared work. 25% of the respondents who are demanding undeclared work are also carrying out undeclared work. It seems that carrying out undeclared work delivers a great impulse to ask also for undeclared goods or services. The results are statistically significant (p = 0.0001). For undeclared goods and services, it could be an indicator to correct the (missing?) answers on the supply of undeclared work.

Supply of undeclared work

W

101

Table 5.7. Cross table demand and supply of undeclared work (n = 246).

Supply of undeclared work

Row Pct Col Pct Yes 70.2% 25.5% 33.7% 74.5% 38.8%

Yes No Total

Source:

4.

Demand for undeclared work No Don’t know 27.0% 2.8% 6.4% 23.0% 64.8% 1.6% 93.6% 77.0% 59.4% 1.7%

Total 14.1% 85.9% 100.0%

Own calculations based on SUBLEC data.

Probit analysis

In this part, the goal is to determine which variables explain the supply of undeclared work. By means of a probit regression, it is possible to verify if some of the independent variables (sex (A1), language (taal), financial need – income (B2), knowing other people carrying out undeclared work (C3), socio-economic variables (F1a) and one about morality (B6_1)) had an impact on the decision to supply undeclared goods and/or services. The dependent variable (supply) is a dummy variable (0 = non supply; 1 = supply). Table 5.8. Probit analysis supply of undeclared work. Parameter Intercept Sex

Variable

Region

French-speaking

Socio-economic category

Self-employed

Man

Benefit recipient Inactive Know someone (C3)

Yes

Income (B2)

Difficult

Morality (B6_1)

Totally agree Rather agree Disagree

Estimation -2.1466 0.6068 (0.085) 0.424 (0.058) 0.1701 (0.026) -0.7391 (-0.098) 0.3395 (0.058) 1.1406 (0.109) 0.1508 (0.021) -0.652 (-0.091) -0.5856 (-0.072) -0.23 (-0.028)

Std error 0.5601 0.2443

Pr > Khi2 0.0001*** 0.013**

0.2478

0.087*

0.4241

0.6884

0.2845

0.0094***

0.359

0.3443

0.4464

0.0106**

0.252

0.5495

0.3433

0.0575*

0.3397

0.0847*

0.4726

0.6265

Note: *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. Marginal effects between brackets. Source:

Own calculations based on SUBLEC data.

102

X

Supply of undeclared work

The results are summarized in table 5.8. (References = woman, Flemish, employee, knowing nobody, high income and neither agree or disagree). Sex and Region have an influence on the supply now. Being a man (p = 0.013) or speaking French (p = 0.087) has a positive influence on the supply of undeclared work. Here the respondent probably refers to his own behavior in his answer, while in the question regarding the demand, it could refer to the household behavior. The benefit recipients appear to produce fewer goods and services not reported (p = 0.0094). This is also expected (See our previous overview Pacolet & Geeroms (1995), of reasons for doing undeclared work). Some theories tell us that the reasons for having a job and a regular income also make people active in the underground economy. Those that do not have these skills are unsuccessful in both economies. The table doesn’t show a significant result for the fact of being self-employed (p = 0.6884). The fact that there is a positive relation between economic hardship (income – difficulties to survive) could confirm the survival economy. But the relation is not significant. We need a larger sample and better models perhaps to make it clearer. Significant is that when the respondent is a benefit recipient, he engages less in undeclared work. Is their income sufficient? The majority answered however that it was not! (see figure 5.9) Nevertheless they engage less in undeclared work. Other results may seem surprising: totally agreeing with ‘The tax burden is much too high in Belgium’ (p = 0.0575) or rather agreeing (p = 0.0847) would lead respondents to produce less undeclared goods and services. A last but important result, if respondents know anyone who offers such goods or services, they tend to do it too (p = 0.0106). It confirms us in our plea for prevention and persistent action for strong control and fines. Being a man increases the likelihood of carrying out undeclared work by 8.5%. Compared to employees, the likelihood of carrying out undeclared work is reduced by 9.8% for social benefit recipients. Knowing someone who carries out undeclared work increases the likelihood by 10.9% compared with knowing no one.

5.

Conclusion

As was done for the demand for undeclared work, the characteristics of people offering undeclared work are verified. More men than women tend to deliver undeclared work (18.2% vs. 9.2%). These results are statistically significant (p = 0.0445). The probit analysis confirms this. Also in the Eurobarometer, this ratio was 2 vs. 1 (EU27 = men: 6% and women: 3%) (EC, 2007b). The French-speaking part of Belgium provides more undeclared work than the Dutch-speaking part (17.2% vs. 10.9%). These results are however not statistically significant (p = 0.1517). However, in the probit analysis the French-speaking region has a positive influence on the supply of undeclared work (p = 0.087). Not-active persons are most involved in undeclared work (24.4%). The social benefit recipients are underrepresented as providers of undeclared work (5.4%). The involvement of employees and self-employed persons is more of less the same (18% and 19%). The results are statistically significant (p = 0.0267). The probit analysis only results in a negative influence of the social benefit recipients (p = 0.0094)

Supply of undeclared work

W

103

on the supply of undeclared work. Self-employed persons do not stick out as offering more undeclared work; the benefit recipients are even offering less undeclared work. Perhaps, as explained above, they have less capacities or opportunities to do so.

Table 5.9. Summary supply of undeclared work (n = 246). Yes

No

Total

14.1%

85.9%

Sex – Man – Woman

18.2% 9.2%

81.8% 90.8%

Region – Dutch-speaking – French-speaking

10.9% 17.2%

89.1% 82.8%

Socio-economic group – Employee – Self employed – Benefits recipient – Not active

18.0% 19.1% 5.4% 25.4%

82.0% 81.9% 94.6% 74.6%

0.0445**

0.1517

0.0267**

Note: *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. Source:

Chi-Square (prob.)

Own calculations based on SUBLEC data.

Chapter 6

Other forms of fiscal fraud

Undeclared work is only one aspect of the definition of fiscal fraud. In this part different aspects of fiscal fraud are tackled, strongly focusing on the personal income taxes. The SUBLEC-questions want to reveal or quantify some of those phenomena.

1.

General

In table 6.1 we summarize some of the results on questions about the behavior of the respondents with regard to certain fraud related aspects. The hotel and catering industry is known as a fraud sensitive sector. Figures of the social inspection services are proving this (Pacolet, Perelman, Pestieau, Baeyens and De Wispelaere, 2009). In the future the hotel and catering industry has to install a cash register. At this moment 65% of the respondents do not ask a VAT receipt when they go to a restaurant. The French-speaking part is asking this receipt more often than the Dutch-speaking part (71.3% vs. 54.5%). The Belgian government introduced a tax amnesty (DLU/EBA) in 2004. Less than 1% used this tax amnesty, 99% did not use it. It seems as if our respondents were the ‘we are the 99%’.1 7.5% of the Belgian respondents indicate they have a bank account abroad. 33.7% of the respondents confirm they didn’t report the foreign bank account in the tax declaration. This means about 5% has reported a foreign bank account. This comes close to the percentage in the tax declaration (see further). So on top of that is another 33.7%, leading to a total of 7.5%. It is possible to compare this figure with the tax declaration because it is obligatory to report this foreign account in the tax declaration. In 2007 a total of 235,000 foreign bank accounts were reported. Compared to 2006 (104,000 persons) and 2004 (52,000 persons) this is a strong increase.2 In 2007 there were 6,675,663 registered taxpayers, of whom 3.5% reported having an account abroad (FPS Finance, 2010).3 This would imply that the number in the tax declaration has a potential to increase further. Perhaps the answers in the declaration and in our survey are in both cases an under-declaration.

1. 2. 3.

One of the slogans of the ‘Occupy Wall Street’ in 2011. D. Van der Maelen, 2008, ‘Dubbel zoveel Belgen hebben buitenlandse rekening aangegeven’. In reality one taxpayer can have more foreign accounts. So the percentage of taxpayers who declared their bank account abroad will be lower.

106

X

Other forms of fiscal fraud

Table 6.1. General questions. Yes

No

Don’t know

Refusal

E3. When you go to a restaurant, do you ask a cash register slip (VAT receipt)? (n = 246) Belgium 35.1% 62.9% 1.8% 0.3% Dutch-speaking 28.7% 71.3% 0.0% 0.0% French-speaking 41.4% 54.5% 3.6% 0.6% E4. Have you used the tax amnesty (EBA/DLU – one-off declaration)? (n = 246) Belgium 0.8% 98.5% 0.7% Dutch-speaking 0.8% 98.3% 0.9% French-speaking 0.8% 98.7% 0.5%

0.0% 0.0% 0.0%

E5. Do you have a bank account abroad? (n = 246) Belgium 7.5% 91.5% Dutch-speaking 6.1% 93.9% French-speaking 9.0% 89.1%

0.0% 0.0% 0.0%

1.0% 0.0% 1.9%

E6. Does the tax authority know you have a bank account abroad? (n = 20) Belgium 66.3% 33.7% 0% E7. Do you own a second home abroad? (n = 246) Belgium 7.1% 92.6% Dutch-speaking 3.2% 96.9% French-speaking 11.1% 88.3%

0.3% 0.0% 0.6%

0.0% 0.0% 0.0%

E8. Does the tax authority know you have a second home abroad? (n = 18) Belgium 75.0% 25.0% 0.0% Source:

Own calculations based on SUBLEC data.

Belgium

7.2%

Dutch-speaking

8.5%

5.8%

French-speaking

10.5%

Source:

78.9%

16.4%

0% One

75.9%

13.5%

20% Two

Three

72.9%

40% Four

60% Five

Six

80% Seven

Eight

100% Nine

Ten

Own calculations on data SUBLEC.

Figure 6.1. E9. On a scale from “1”to “10”, can you indicate whether your most recent tax declaration has been properly completed? “1” means that you have not completely filled correctly and “10” means that you fully satisfied. (n = 246)

Other forms of fiscal fraud

W

107

Also the ownership of a second home abroad can be verified on the basis of the tax declaration. 7.1% of the Belgian respondents in our survey have a second home abroad but 25% of these houses are not known by the tax authorities. So about 5% of the respondents reported a second home abroad in their tax declaration. In 2009, only 48,102 of the 6,798,564 taxpayers reported a second home abroad (or 0.7%) (FPS Finance, 2010).

2. 2.1

Frequency and volume of several other forms of fiscal fraud Frequency

Figure 6.1 shows that more than 24% of the respondents do not fill in their tax declaration completely correctly. 21% of the Dutch-speaking respondents admit this, compared to 27% of the French-speaking respondents. None of the Dutch-speaking respondents gave a score of less than 7. These figures seem to suggest the existence of a large group of small tax evaders and a very small group of big tax evaders and the rest assumes, pretends they comply completely with the rules. The fight against counterfeiting has become a priority in Belgium in recent years (see also the awareness campaign).4 6.7% of the Belgian respondents bought a counterfeit product in the past 12 months. The percentage of people who committed fiscal fraud by not declaring capital income amounts to 3.5%. Only 0.3% did not report the received real estate income. 5.5% evaded inheritance taxes in the course of the last 5 years. Finally 1.9% evaded registration fees. Further detail in the questionnaire should have been introduced here (for instance between financial wealth and real estate). Table 6.2. Different aspects of fiscal fraud. Yes

No

Don’t know

E1. Have you bought counterfeit items in the past twelve months? (n = 246) Belgium 6.7% 92.9% 1.1% Dutch-speaking 6.0% 92.4% 1.6% French-speaking 7.4% 92.0% 0.6%

Refusal 0% 0% 0%

E11. Over the last twelve months, have you received any income from capital revenues that you haven’t declared to the tax administration despite the fact that it was not exempted? (n = 246) Belgium 3.5% 95.6% 0.9% 0% Dutch-speaking 4.4% 95.0% 0.6% 0% French-speaking 2.7% 96.2% 1.2% 0% E13. Over the last twelve months, have you received any real estate income that you haven’t declared to the tax administration despite the fact that it was not exempted? (n = 246) Belgium 0.3% 99.3% 0.4% 0% Dutch-speaking 0.0% 100.0% 0.0% 0% French-speaking 0.6% 98.7% 0.7% 0%

4.

See www.contrefacon.be

108

X

Other forms of fiscal fraud

Yes

No

Not applicable5

E15. In the course of the last five years, have you inherited a capital or a property that you have not declared to the fiscal administration and for which you consequently have not paid inheritance taxes? (n = 246) Belgium 5.5% 89.2% 5.3% Dutch-speaking 2.7% 87.8% 9.6% French-speaking 8.3% 90.7% 1.1% E16. In the course of the last twelve months, have you avoided registration fees when buying a home? (n = 246) Belgium 1.9% 89.0% 9.1% Dutch-speaking 2.6% 82.6% 14.9% French-speaking 1.2% 95.5% 3.4% Source:

2.2

Own calculations based on SUBLEC data.

Volume

In general the evaded amount seems to be rather small (table 6.3). 2.3% of the respondents did not report all their incomes in the tax declaration. Remarkable is the high percentage of the inheritance that was not declared to the tax administration by the fraudulent respondents (33.2%). We have to be careful about the interpretation due to the lack of respondents (especially for the capital income, the real estate income, the inheritance and the registration tax).6 The relatively high percentage of fraud in inheritance tax, although the fiscal treatment is moderate for most of the population (see Pacolet & Strengs, 2011), illustrates perhaps the resistance of the population against such a tax, considered as a tax on already taxed income.

Table 6.3. Volume of fiscal fraud. Mean

Std dvp

E10. Can you tell us what percentage of your income you have hidden in your tax declaration? (n = 225) Belgium 2.3% 11.4 Dutch-speaking 3.3% 16.0 French-speaking 1.4% 4.9 E12. What is the percentage of your capital income that you have not declared to tax authorities despite the fact that they were not exempted? (n = 18) Belgium 3.0% 6.5

5. 6.

During the debriefing with the interviewers and also during the data analysis we had the impression that some of the respondents answered negatively to these questions while these were not applicable for them. These possibilities occur only rarely in a life time so the chance it would appear in the sample is small. By expanding the time span over which the respondent has to report such phenomena increases again the chance it would appear. Since these are rare events, the chance the respondent remembers it is probably high enough.

Other forms of fiscal fraud

Mean

W

109

Std dvp

E14. What is the percentage of your real estate income that you have not declared to tax authorities despite the fact that they were not exempted? (n = 13) Belgium 1.8% 7.2 E16. What is the percentage of your inheritance that has not been declared to tax institutions? (n = 15) Belgium 33.2% 40.2 E18. What is the percentage of the purchase value of the home for which you have not paid a registration tax? (n = 8) Belgium 5.1% 6.8 Source:

3.

Own calculations based on SUBLEC data.

Probit analysis

In this part, the goal is to determine which variables could explain the other forms of fiscal fraud. By means of a probit regression, it is possible to verify again if some of the chosen independent variables (sex (A1), language (taal), financial need – income (B2), knowing other people committing fiscal fraud (C17, C19, C21)), socio-economic variables (f1a) and one about morality (B6_1)) had an impact on the decision to engage also in this kind of fraud. The dependent variable (fiscal fraud) is a dummy variable (0 = non fraud; 1 = fraud). The dependent variable is built on the E9 question. If the respondent didn’t answer 10 to the question: “On a scale from “1” to “10”, can you indicate how your most recent tax declaration has been properly completed? “1” means that you have not completely filled correctly and “10” means that you fully satisfied”, we consider the presence of fiscal fraud. Otherwise, the respondent is seen as ‘honest’. Other variables on fiscal fraud could also be used, but we did not consider them further because of the limited size of the sample. The results are summarized in table 6.4. (References = woman, Flemish, employee, knowing nobody, high income and neither agree or disagree). The results are less significant. This could be related to the definition of the dependent variable. An alternative could be one or another summary index of all kinds of fiscal fraud. But since these forms are so heterogeneous, with different determinants, the analysis would again not be justified. A larger sample is needed to further proceed with this kind of analysis. Concretely, we cannot conclude to the existence of influent variables on fiscal fraud. Sex, Region, Income, Morality or Knowing someone who commits fraud have no influence on the fraud. The only significant relation is the relation with the socio-economic category of benefit recipients. This category would tend to cheat more than others on their tax declaration (p = 0.0932), what is a surprising factor. Their benefit is known. In the future their declaration will probably become even automatic. So our respondents indicate here that they had less or more often ‘forgotten’ to declare something.

110

X

Other forms of fiscal fraud

Table 6.4. Probit analysis fiscal fraud. Parameter

Variable

Intercept

Estimation

Std error

Pr > Khi2

0.5952

0.3346

0.0752*

0.0162 (0.005)

0.1861

0.9305

-0.2762 (-0.081)

0.1916

0.1495

Self-employed

0.1999 (0.055)

0.4009

0.6179

Benefits recipient

0.3426 (0.099)

0.2041

0.0932*

Inactive

0.0457 (0.013)

0.318

0.8858

-0.0152 (-0.005)

0.195

0.9379

Sex

Man

Region

French-speaking

Socio-economic category

Know someone (C17, C19, C21)

Yes

Income (B2)

Difficult

0.0667 (0.019)

0.1961

0.7337

Morality (B6_1)

Totally agree

0.3207 (0.094)

0.2913

0.271

Rather agree

-0.0158 (-0.005)

0.2942

0.9571

Disagree

-0.1203 (-0.037)

0.3897

0.7575

Note: *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. Marginal effects between brackets. Source:

4.

Own calculations based on SUBLEC data.

Conclusion

According to our response group for Belgium, on average 2.3% of the income is not declared in the tax declaration and almost a fourth of respondents do not fill in their tax declaration completely correctly. Some differences show up between the Dutch-speaking and Frenchspeaking part of Belgium. More French-speaking respondents committed (admitted) income fraud than Dutch-speaking respondents. The volume of the income fraud was however higher for the Dutch-speaking respondents. The results for the three variables (sex, region and socio-economic category) are not statistically significant. This is also the case in the probit analysis.

Other forms of fiscal fraud

W

111

Table 6.5. Summary fiscal fraud (Tax declaration). Yes

No

Chi-Square (prob.)

– Man

22.9

77.1

0.6156

– Woman

25.6

74.4

– Dutch-speaking

21.1

78.9

– French-speaking

27.1

72.9

28.5

71.5

– Self employed

21.8

78.2

– Benefits recipient

17.3

82.7

– Not active

39.0

61.0

Sex

Region 0.2692

Socio-economic group – Employee

Note: *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. Source:

Own calculations based on SUBLEC data.

0.1507

Chapter 7

Opinion questions

In this section of the report, all the results about opinion questions are discussed. As previously, when there is no distinction between Dutch and French-speaking respondents, this is due to a small number of answers. For several questions which were also present in the Eurobarometer, we compared the results to our results.

1. 1.1

Opinion on undeclared work In the SUBLEC-sample

The respondents suppose that almost 4 out of 10 persons are (sometimes) carrying out undeclared work (table 7.1). Above, we found that to the question whether the respondent him/herself has carried out undeclared work, only 14% answers positively. On average € 280 of the € 1,000 of labor income would be not declared to the tax or social security authorities. The respondents also suppose that 4 out of 10 persons are purchasing undeclared good or services, worth € 231 of the € 1,000 of disposable income. An interesting point is the fact that almost 8 out of 10 respondents know people who have purchased products or services coming from undeclared work (79.2%) or who have carried out undeclared work (78.5%). Observing the above-mentioned potential impact of the opinion on fraudulent behavior of others on the own behavior, it could be an important strategy to correct the opinion (if mistaken) or to contain the damaging impact. It should be interesting to verify the following assumption: people who tend to report a high percentage of people who purchased an undeclared good or service are also purchasing undeclared goods or services themselves (table 7.2). The assumption can also be made for the supply of undeclared work. The borderline of a small or high percentage is the average (demand for undeclared work = 42% and supply of undeclared work = 38%). 59% of the respondents who bought an undeclared good or service reported a high percentage when answering the opinion question. 61% of the respondents who did not buy an undeclared good or service reported a low percentage when answering the opinion question. These results are statistically significant (p = 0.003). With regard to the supply of undeclared

114

X

Opinion questions

work, we observe that 68.4% of the respondents who carried out undeclared work reported a high percentage when answering the opinion question. This is statistically significant too (p = 0.0107). There could be an influence from the behavior and the fiscal morals of the people the respondent knows on the respondent’s behavior. But there could also be some projection of the respondent’s own behavior on that of others. Instead of asking the respondent directly if he/she carries out the activity we ask if he/she knows someone who is doing this. This strategy we found warranted in a ‘firm’ survey (in the construction industry) because here you ask experts in the sector about what is happening in the sector, without asking them directly what they are doing themselves (Pacolet & Baeyens, 2007).

Table 7.1. Opinion about undeclared work. Mean

Std dvp

You think about 100 people, how many work without declaring their income or part of their income to tax or social security authorities in Belgium? (n = 240) Belgium 38.3% 20.7 Dutch-speaking 37.3% 21.1 French-speaking 39.2% 20.3 In your opinion, for € 1,000 in labor income, how much is not reported to tax or social security authorities in Belgium? (n = 217) Belgium 279.2 202.8 Dutch-speaking 267.6 200.5 French-speaking 291.4 205.2 You think about 100 people, how many regularly purchase a service or product that comes from undeclared work? (n = 229) Belgium 41.6% 24.6 Dutch-speaking 41.3% 25.3 French-speaking 41.9% 24.1 On disposable income of € 1,000, according to you, how much is used to purchase goods or services from undeclared work? (n = 212) Belgium 231.2 203.1 Dutch-speaking 210.6 210.6 French-speaking 253.2 253.2 Yes

No

Don’t know

Refusal

Do you personally know people who work without declaring their income or part of their revenues to tax or social security authorities in Belgium? (n = 246) Belgium 78.5% 21.5% 0.0% 0.0% Dutch-speaking 78.9% 21.1% 0.0% 0.0% French-speaking 78.1% 21.9% 0.0% 0.0% Do you know people who, according to you, purchase products or services coming from undeclared work, that is to say that the income from it has not been fully communicated to tax or social security authorities? (n = 246) Belgium 79.2% 20.0% 0.5% 0.3% Dutch-speaking 82.0% 17.8% 0.2% 0.0% French-speaking 76.5% 22.2% 0.7% 0.6% Source:

Own calculations based on SUBLEC data.

Opinion questions

W

115

Table 7.2. Relationship between opinion and reality on demand and supply of undeclared work – Belgium – (n = 246). Low percentage

High percentage

Chi-Square (prob.)

Demand for undeclared work (border opinion = 42%) Yes 41.0% No 61.0%

59.0% 39.0%

0.003***

Supply of undeclared work (border opinion = 38%) Yes 31.6% No 55.3%

68.4% 44.7%

0.0107***

Note: *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. Source:

1.2

Own calculations based on SUBLEC data.

Comparing the SUBLEC sample with the experimental group of students

The above results are again compared with the results obtained in the experiments with students in Belgium (see Lefebvre, Pestieau, Riedl and Villeval, 2011). Walloon (67.5%) and Flemish (65.9%) students know less people who purchased products or services coming from undeclared work than the SUBLEC respondents. In France, these figures are even lower (54.1%). However, there are a lot of students who answered ‘Don’t know’, which is quite different from the SUBLEC survey. The opinion results in the experiments about the supply of undeclared work are close to the SUBLEC results. This time, more Walloon students (79.2%) know people who work without declaring the income or part of revenues than their Flemish (76.3%) counterparts. The figures are much lower in France (65.4%) and in the Netherlands (67.9%).

Table 7.3. Students who know people who have acquired (demand) or carried out (supply) undeclared work, 2010. Yes

No

Don’t know

Demand

Supply

Demand

Supply

Demand

Supply

Wallonia (n = 114) Flanders (n = 135) France (n = 133) Netherlands (n = 131)

67.5% 65.9% 54.1% 62.6%

79.8% 76.3% 65.4% 67.9%

17.5% 20.0% 30.1% 29.8%

12.3% 12.6% 23.3% 28.2%

14.9% 14.1% 15.8% 7.6%

7.9% 11.1% 11.3% 3.8%

Total (n = 513)

62.4%

72.1%

24.6%

19.3%

13.1%

8.6%

Source:

1.3

Data obtained from M. Lefebvre (CREPP-ULg).

Comparing SUBLEC with the Eurobarometer

Figure 7.1 shows huge differences between the results of the Eurobarometer survey and the SUBLEC survey. Many of the respondents in the Eurobarometer were thinking that less than 50% of the population is working in the underground economy. In the SUBLEC survey we

116

Opinion questions

X

observe a different opinion. Many of the respondents (35%) think that more than half of the population is doing undeclared work. The differences between the French and Dutch-speaking part are quite small. In the Eurobarometer for as well Belgium as the EU27, this is only 10%. 40% 37% 35% 35%

33%

30%

25% 21% 19% 18%

20%

15%

10%

13% 11% 8%

13%

12% 8%

9% 7%

Less than 5%

From 5 to 10%

8% 3% 2% 2%

0% 0%1% From 10 to 20%

Eurobarometer EU

Source:

13% 10%

4%

5%

0%

19%

13% 10%

10% 9% 8%

7% 6%

5%

21% 19% 18%

19% 16% 17% 16%

From 20 to 30%

From 30 tot 40%

Eurobarometer Belgium

From 40 tot 50%

Belgium

From 50 to 100%

Dutch-speaking

Don't know

1% 0% 1% 0% 0% Refusal

French-speaking

Own calculations based on SUBLEC data; EC, 2007b.

Figure 7.1. C1. You think about 100 people, how many work without declaring their income or part of their income to tax or social security authorities in Belgium? (n = 240)

100,0% 90,0% 78.9% 78.5% 78.1%

80,0% 70,0% 60,0%

56.0%

55.0%

50,0% 40,0%

42.0%

38.0%

30,0%

21.1% 21.5% 21.9%

20,0% 10,0%

4.0% 1.0%

0,0% Yes Eurobarometer EU

Source:

No Eurobarometer Belgium

Don't know Belgium

Dutch-speaking

3.0%1.0% Refusal French-speaking

Own calculations based on SUBLEC data; EC, 2007b.

Figure 7.2. C3. Do you personally know people who work without declaring the income or part of their revenues to tax or social security authorities in Belgium? (n = 246)

Opinion questions

W

117

The same goes for the probability that the respondent knows someone who committed fraud. The SUBLEC respondents, in comparison with the Eurobarometer, know much more people who work without declaring their income or part of their revenues to tax or social security authorities in Belgium (figure 7.2). In the SUBLEC survey some 78% of the respondents knew someone. The Eurobarometer reveals 56% for Belgium and 38% for the EU27. It is reasonable that when you know more people that commit fraud, you also guess that the amount of people doing it is higher. The answers are consistent. The perceived risk of being discovered for doing undeclared work in Belgium is considered as low (figure 7.3). The Dutch-speaking part assumes the risk of being caught as smaller than the French-speaking part. Perhaps the expected risk of being caught will differ between undeclared and regular workers. In table 7.4 we observe no coherence between a small perceived risk of being discovered and doing undeclared work. The results are not statistically significant (p = 0.4769). 45,0%

43.1% 40.7% 38.4%

40,0% 35,0% 30,0%

26.9%

25,0%

22.7%

20,0% 15,0%

20.9%

18.4%

15.3% 14.3% 13.3%

17.5% 14.2%

10,0% 6.6% 5,0%

4.5% 2.3%

0.3% 0.6% 0.0%

0,0%

Very high

Rather high

Neither high nor low

Belgium Source:

Dutch-speaking

Rather low

Very low

Don't know

French-speaking

Own calculations based on SUBLEC data; EC, 2007b.

Figure 7.3. C7. People who work without declaring the income risk that tax or social institutions find out and issue supplementary tax bills and perhaps fines. In your opinion, the risk of being discovered in Belgium is ...? (n = 246)

Table 7.4.

Expected risk of being caught and the decision to do undeclared work – Belgium – (n = 246).

Undeclared work No undeclared work

Not low

Low*

Chi-Square (prob.)

47.1% 40.6%

52.9% 59.4%

0.4769

Note: *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. * Answer: rather low, very low. Source:

Own calculations on data SUBLEC.

118

X

Opinion questions

Most of the respondents (87.6% for Belgium in the SUBLEC survey) think that, on top of the reimbursement, they have to pay a fine. These results of SUBLEC are in line with the results of the Eurobarometer SUBLEC. More than 5% of the respondents estimate they could go to prison, which is not unrealistic since this kind of penalization indeed rests on fiscal fraud.

87.6% 86.6% 85.7%

90,0%

78.0%

80,0% 70,0%

60.0%

60,0% 50,0% 40,0% 30,0% 20,0%

18.0% 14.0%

10,0%

13.0%

9.8% 8.1% 6.4%

7.0%

5.3% 7.9% 4.0% 2.7%

4.0%

2.0%

0,0% Normal taxation and/or normal social security contributions, but no fine

Normal taxation and/or normal social security contributions, plus a fine

Eurobarometer EU

Source:

Imprisonment

Eurobarometer Belgium

Belgium

Don't know

Refusal

Dutch-speaking

French-speaking

Own calculations based on SUBLEC data; EC, 2007b.

Figure 7.4. C8. In your opinion, what punishment is to be expected if the authorities find out that someone has had an income from work of € 1,600 per month which was not declared to tax or social security authorities? (n = 246)

As we have explained before, for several opinion questions (C9-C10-C11-C12 and C24), the respondents had to use the ‘response disc’ in order to avoid response order effects. Of the respondents, the unemployed persons are most likely to carry out undeclared work. The self-employed persons (main and second activity) and the illegal immigrants are also considered as ‘sensitive’ to work in the underground economy. In the Eurobarometer survey these were the three categories which the respondents considered as most likely to carry out undeclared work too (EC, 2007b). One can observe some differences in opinion between the Dutch- and French-speaking respondents. Almost half (48.8%) of the French-speaking respondents consider the unemployed persons as the most likely category to carry out undeclared work compared to only a fourth (25.0%) of the Dutch-speaking respondents. Selfemployed persons are considered more likely to do undeclared work by the Dutch-speaking respondents (20.2%) than by the French-speaking respondents (11.8%). Perhaps we rediscover a different ‘culture’ of undeclared work between the two regions here, one more

Opinion questions

W

119

‘informal’ (of the unemployed) in the French-speaking part and a more organized, formal, by formal firms, in the Dutch-speaking part.

Unemployed 16.0%

Self-employed (main activity)

11.8% 9.5%

Self-employed (second activity)

4.8%

The full-time workers

Permanent incapacity

14.2%

0.5% 0.0% 1.0% 5.3% 6.4% 4.3% 0.4% 0.0% 0.7% 1.3% 2.7% 0.0% 4.7% 7.5%

Students

1.9%

18.0% 15.0% 21.0%

Illegal immigrants (illegal) 3.1% 3.9% 2.3%

Recipients of social assistance (CPAS, …)

Other inactive (housewives, etc.)

0.4% 0.0% 0.7%

Refusal

0.7% 0.9% 0.6%

0,0%

10,0% Belgium

Source:

20.2%

0.8% 1.7% 0.0%

The part-time-workers

Primary incapacity

48.8%

2.3% 2.6% 2.1%

Early retired

Retirees

37.0%

25.0%

20,0% Dutch-speaking

30,0%

40,0%

50,0%

French-speaking

Own calculations based on SUBLEC data.

Figure 7.5. C9. In your opinion, who are more likely to carry out undeclared work? Specify the 4 groups. In first? (n = 246)

1.4

Opinions integrated in the fraud triangle

For more than 3 out of 10 (33.6%) respondents the most important reasons to carry out undeclared work are the high taxes and social security contributions. Furthermore a financial need is considered as one of the main reasons to carry out undeclared work (figure 7.6). For the French-speaking respondents as well as for the Dutch-speaking respondents, these reasons are assumed as most decisive. In our triangle of dimensions of undeclared work the opinions are clustered in the three dimensions (figure 7.7). The dimension ‘taxes, regulation, red tape’ is more decisive than the dimensions ‘inspection, risk of being caught, enforcement’ and ‘morality, behavior’.

120

X

Opinion questions

1.9% 0.8% 3.0%

Bureaucracy/red tape to carry out a regular economic activity is too complicated

8.4%

Lack of control by authorities

Sanctions are too weak In certain sectors or regions, there is no real alternative

5.5%

11.5%

1.8% 0.6% 2.9% 1.5% 0.6% 2.4% 9.2% 6.6% 11.8%

Salaries in the regular businesses are too low

Lack of regular jobs on the labor market

1.5%

4.8% 8.1% 26.7% 25.2% 28.2%

The financial need (income or allowance too low) The state does not do anything for the people, so why should they pay taxes? Nobody would buy those goods or services on the regular market

0.8% 0.0% 1.6% 0.4% 0.8% 0.0% 33.6%

Taxes and/or social security contributions are too high

28.9% 3.6%

It is generally accepted by society

1.0%

2.3% 4.1%

0.6%

3.2% 2.8% 3.7%

Other

Refusal

6.3%

1.5% 1.1% 1.9%

It occupies their free time

Recipients insist on non-reporting

0.3% 0.0% 0.6%

0,0%

5,0%

10,0% Belgium

Source:

38.2%

15,0%

20,0%

Dutch-speaking

25,0%

30,0%

35,0%

40,0%

French-speaking

Own calculations on data SUBLEC.

Figure 7.6. C10. What are in your opinion the reasons that motivate people to work without declaring all or part of their income? In first? (n = 246)

The opinions of the respondents about the main reason for not carrying out undeclared work are divided (figure 7.8). The fact that the undeclared worker is not socially protected when he/she is ill or has an accident is considered as the main reason for not carrying out undeclared work by more than 20% of the respondents. But also a personal moral reticence (16.0%), especially for the French-speaking respondents (20.9%), can be decisive in the decision. Finally the fact one comes around with his/her income or allowance (17.6%) can be a reason to stay away of the irregular market. The risk of losing welfare benefits is the fourth reason for not carrying out undeclared work (for Flanders even the third). Combined with the first reason (no legal protection) it illustrates that having and maintaining entitlements to social security and welfare prevents people from engaging in undeclared work. It questions those studies that link highly developed welfare states to high risks of undeclared work all the time. Sometimes the opposite is observed (See also Pacolet & Marchal (2003) for further debate).

Opinion questions

W

121

Morality, behavior

10,9%

11,4% Taxes, regulation, 77,7% red tape Source:

Inspection, risk of being caught, enforcement

België

Own calculations on data SUBLEC.

Figure 7.7. C10. What are in your opinion the reasons that motivate people to work without declaring all or part of their income? In first? (n = 246) – clustered in the three dimensions of undeclared work.

For the social benefit recipients we have checked how many of them answered ‘risk of losing welfare benefits’ as first answer. 19% of the social benefit recipients gave this as first answer (against 12% of the total population).

20.6% 20.7% 20.4%

No legal protection for illness and/or accident No constitution of pension rights

8.9%

10.4% 11.9% 16.0%

11.1%

Personal moral reticence

20.9% 17.6% 16.6% 18.5%

Income or allowance is high enough to come around 8.2% 9.4%

If one gets caught, the consequences are too heavy Not generally accepted by society

7.1% 0.0% 0.0% 0.0%

7.0% 7.7% 6.3%

The risk of getting caught is too high 3.2% 3.0% 3.4%

Lack of free time to pursue activities unreported

3.7%

Lack of technical expertise to offer work not stated

2.1%

5.2% 12.7% 12.0% 11.2%

Risk of losing welfare benefits 1.4% 1.7% 1.2%

Other 0,0%

Source:

5,0%

10,0%

15,0%

20,0%

25,0%

Own calculations based on SUBLEC data.

Figure 7.8. C11. What are in your opinion the reasons for not doing undeclared work? In first? (n = 246)

122

X

Opinion questions

When asked about the most effective policy to tackle undeclared work, the respondents mainly focus on the dimension ‘taxes, regulation, red tape’ (e.g. reduction of social and fiscal charges on labor) (figure 7.9). Also the ‘dimension of control’ is assumed as crucial in the fight against undeclared work. We again observe small regional differences. The Frenchspeaking respondents focus more on the dimensions ‘taxes, regulation, red tape’ and ‘morality, behavior’ compared to the Dutch-speaking respondents. The dimension ‘inspection, risk of being caught, enforcement’ is considered as more important by the Dutch-speaking respondents than by the French-speaking respondents. Comparing figure 7.7 and 7.9 the contradiction becomes clear that much more influence is attached to morality and less to control, while for the policies to combat fraud, more is expected from control and less from morality influencing policies. Perhaps the two strategies should be followed.

Morality, behavior

6.8%

Taxes, 58.1% regulation, red tape

Source:

35.1% Belgium

Inspection, risk of being caught, enforcement

Own calculations based on SUBLEC data.

Figure 7.9. C12. What are in your opinion the most effective policy measures to fight against undeclared work? Clustered in the three dimensions of undeclared work. In first? (n = 246)

2.

Social benefit fraud

Table 7.5 shows the opinion on social benefit fraud. One thinks that 30% of the social benefit recipients are not respecting the conditions of payment. 5 out of 10 persons know a social benefits recipient who is not respecting the conditions of payment. Such a revealed mistrust of the public opinion (since that is indeed what we are analyzing here, an opinion poll) risks of course to be devastating for the trust in the welfare state (see Pacolet, 2003, on Undeclared work and the welfare state).

Opinion questions

W

123

Table 7.5. Opinions on social benefit fraud. Mean

Std dvp

In your opinion, out of 100 people receiving a social allowance in Belgium, how many do not respect the conditions which are leading to the payment? (n = 246) Belgium 30.1% 20.9% Dutch-speaking 28.9% 20.3% French-speaking 31.3% 21.4% Yes

No

Don’t know

Refusal

Do you personally know people who receive their social allowance without respecting the conditions which are leading to the payment? (n = 246) Belgium 52.1% 47.3% 0.6% 0.0% Dutch-speaking 47.4% 52.6% 0.0% 0.0% French-speaking 56.8% 42.1% 0.0% 0.0% Source:

3.

Own calculations based on SUBLEC data.

Fiscal fraud

The respondents expect that almost 38% of the population is not reporting all their revenues to the Belgian tax authorities (table 7.6). But not every income is expected to be evaded at the same level. 40% of the population is expected to withhold (a part of) the inheritance. Also 4 out of 10 respondents know someone who did not declare the entire inheritance. 32% of the persons who bought a house are expected to commit fraud by not paying registration fees on the total price (part of the price paid under the counter). There is nevertheless a great difference in opinion between the Dutch-speaking respondents (38.1%) and the French-speaking respondents (25.7%). Also 48.4% of the Dutch-speaking respondents knows someone who evaded registration fees compared to only 32.1% of the French-speaking respondents. A third of the respondents know someone who did not report his capital income entirely (Belgians are known for their discretion about their income and probably also tax declaration). They expect that 34.6% of the persons who are receiving a capital income, is withholding this (partly/entirely) for the tax authorities. Real estate income is expected to be evaded the least (22.9%). For capital income as well as for real estate income, French-speaking respondents expect a higher percentage of persons who are evading taxes compared to the Dutch-speaking respondents. The respondents assume the dimension ‘taxes, regulation, red tape’ (56.3%) and the dimension ‘inspection, risk of being caught, enforcement’ (40.5%) as most effective in the fight against fiscal fraud (figure 7.10). These results are comparable to the opinion question about the most effective policy to tackle undeclared work, but there seems to be less consistency with the mentioned arguments that explain undeclared work. In the public opinion there is seemingly not a direct relation between the reasons to engage in undeclared work, and the ways to prevent it.

124

X

Opinion questions

Table 7.6. Opinions on fiscal fraud. Mean

Std dvp

Out of 100 people, how many have willfully failed to report income to the tax authorities? (n = 246) Belgium 37.7% 25.1% Dutch-speaking 37.0% 23.3% French-speaking 38.3% 27.7% In your opinion, out of 100 people receiving non exempted capital income, how many do not report this fully to the Belgian tax authorities? (n = 246) Belgium 34.6% 26.5% Dutch-speaking 31.1% 26.8% French-speaking 38.3% 25.8% In your opinion, out of 100 people receiving real estate income, how many do no report this fully to the Belgian tax authorities? (n = 246) Belgium 22.9% 22.0% Dutch-speaking 19.8% 17.4% French-speaking 26.5% 25.3% In your opinion, out of 100 people receiving an inheritance (capital and real estate), how many do not report this fully to the Belgian tax authorities? (n = 246) Belgium 40.2% 28.7% Dutch-speaking 42.0% 30.2% French-speaking 38.5% 27.3% In your opinion, out of 100 people who buy a house, how many evade registration fees? (n = 246) Belgium 31.9% 26.8% Dutch-speaking 38.1% 28.0% French-speaking 25.7% 24.5% Yes

No

Don’t know

Refusal

Do you personally know people who received capital income without declaring this fully to the tax authorities in Belgium? (n = 246) Belgium 33.6% 66.0% 0.4% 0.0% Dutch-speaking 31.0% 68.9% 0.2% 0.0% French-speaking 26.3% 63.1% 0.6% 0.0% Do you personally know people who received real estate income without declaring this fully to the tax authorities in Belgium? (n = 246) Belgium 27.3% 70.9% 1.8% 0.0% Dutch-speaking 25.6% 74.4% 0.0% 0.0% French-speaking 29.0% 67.4% 3.6% 0.0% Do you personally know people who received an inheritance without declaring this fully to the tax authorities in Belgium? (n = 246) Belgium 41.3% 57.4% 0.9% 0.4% Dutch-speaking 43.5% 55.9% 0.6% 0.0% French-speaking 39.1% 59.0% 1.2% 0.7% Do you personally know people who evaded registration fees by the purchase of a house? (n=246) Belgium 40.3% 59.1% 0.6% 0.4% Dutch-speaking 48.4% 51.6% 0.0% 0.0% French-speaking 32.1% 66.7% 1.1% 0.7% Source:

Own calculations based on SUBLEC data.

Opinion questions

W

125

Morality, behavior

3.2%

Taxes, regulation, red tape Source:

40.5%

56.3% Belgium

Inspection, risk of being caught, enforcement

Own calculations on data SUBLEC.

Figure 7.10. C24. What are in your opinion the most effective policy measures to fight against fiscal fraud? Clustered in the three dimensions of undeclared work. In first? (n = 246)

4.

General

In the next paragraph some questions are reported on traditional items where the compliance behavior of a population is tested.1 The behavior is judged on the degree of acceptability (0 = totally unacceptable and 10 = total acceptable). All behavior is related to undeclared work, social benefit fraud or fiscal fraud except for one. The behavior ‘use of public transport without a valid ticket’ (acceptance of 2.8) can be compared to all the other behaviors related to the underground economy or fraud. An employee who is carrying out undeclared work after his regular hours (4.5) and using undeclared work for the renovation of the bathroom (4.5) are the most accepted behaviors. Dutch-speaking respondents assume the use of public transport without a valid ticket (2.2) less acceptable than hiding income in the tax declaration (3.2), which is a remarkable result. There is a strong resistance against social benefit recipients who are combining their social benefit with undeclared work (1.5) or receiving the social benefit without entitlement (1.6).

1.

See also recent results from Denmark (Rockwool Foundation Research Unit, ‘Lying and social security fraud are viewed more leniently today than previously’, June 2011). “Failing to buy a ticket when using public transport” has an acceptance of 1.9. “Claiming social security payments to which you are not entitled” and “Cheating on the tax declaration, if one has the opportunity” have an acceptance of 1.7 and 1.8.

126

X

Opinion questions

3.1 3.2 3.1

Hide income in the tax declaration 2.4 2.3 2.6

Avoid taxes by hiding income in official ‘tax havens’ 1.5 1.3 1.8

Carry out undeclared work full time and qualify for unemployment benefit at the same time

2.6 2.5 2.6

Reporting ill, without being it

2.4 2.4 2.4

Illegal immigrants are put to work by an employer in Belgium 1.7 1.7 1.7

Workers are not reported to the NSSO by their employer

4.5 4.3 4.8

Work after hours without being declared

4.5 4.3 4.7

To renovate his bathroom using undeclared work

2.8 2.2

Use public transport without a valid ticket

3.4 1.6 1.5 1.7

Receive a social benefit without entitlement

1.0 Source:

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.00

Own calculations on data SUBLEC.

Figure 7.11. C25. For each of these behaviors, can you say how you find them acceptable or unacceptable? Please use the following scale: “1” means that you find this behavior “totally unacceptable” and “10” as you are “totally acceptable”. (n = 246)

5.

Conclusion

It can be concluded that the fiscal fraud by the population is expected to occur more frequently than social benefit fraud. It is in line with the prioritizing that we sometimes had put in our research on those issues. At the same time this survey illustrates the research strategy that all fraud should be brought in the picture, and be tackled. The occurrence of fiscal fraud and undeclared work are expected almost at the same frequency. In the fight against undeclared work and fiscal fraud there is a strong focus on two dimensions of our ‘triangle of fraud dimensions’, namely the dimensions ‘tax’ and ‘control’. But since ‘morality’ is also considered as an important determinant of fraudulent behavior, it should perhaps be given as much attention in policy making.

Chapter 8

Lessons learnt from the survey

First we will discuss the problems we had during the roll out of the survey. We will focus on our data request by the Sectoral Committee of Social Security and Health within the Privacy Commission. Afterwards we will tackle the low response rate and describe possible actions to avoid this in the future. The methodology of interviewing will be discussed also. Finally we assess the sensitivity of some questions for the respondents themselves.

1.

Influence of the data request on the results

Initially it was our ambition to personally contact the whole sample for the organization of a face-to-face interview. Therefore we needed the names and addresses from the persons in the sample. The gross sample was initially set at 12,900 individuals and the target to reach was 4,300 individuals net. It is the scale of a population survey as the SILC (Survey on Income and Living Conditions) and the household budget survey, but much smaller than for instance the LFS (Labour Force Study) or the health survey. The first data request (June 2009) was however not accepted by the Sectoral Committee of Social Security and Health within the Privacy Commission. Since a postal survey (self-administrated questionnaire) is the rule and the sample has to be contacted by the CBSS, the intention to contact the sample by the research team and to execute a face-to-face interview was not accepted despite recent precedents in other research. In a reaction of the research team to the Sectoral Committee of Social Security and Health (August 2009) the need for a face-to-face interview and the direct contact to the sample by the research team was further justified. Because of a smaller budget we also had to reduce our gross sample to 5,202 individuals. Our target was still a response rate of 33% or 1,734 individuals. A new data request was submitted in November 2009. Because of the new imposed procedure, this response rate was not obtainable any more, what would result in a smaller net sample. Due to the timing of the project, it was not possible to introduce a new request for a larger gross sample. In January 2010 the deliberation of the data request was published by the Sectoral Committee of Social Security and Health. The request for the organization of the survey by a faceto-face interview was accepted by the Sectoral Committee but not our proposition about

128

X

Lessons learnt from the survey

the way of contacting the sample. The research team was obliged to follow the following procedure: an external organization (Smals) had to contact the respondents in the sample. Only when the respondent agreed to participate and sent the response form back to one of the research institutes (HIVA for the Flemish Region and CREPP for the Walloon Region and Brussels-Capital Region) it was possible to contact the respondent. This decision had consequences on the organization of the survey but probably also on the response rate and the representativeness of the respondents. We were aware of the fact that this decision would have a negative impact on the response rate. It was however not possible to raise the number of gross respondents in the accepted data request. A new data request was the only possibility, which was not longer possible bearing in mind the timing of the project. The fact the sample had to send a letter to the research team if one wanted to participate would misrepresent the original sample. We suppose only persons who are really motivated to participate and persons who declare all or almost all their income will participate at the survey. The group of persons who are regularly committing fraud will not answer the letter despite the fact that anonymity and confidentiality is guaranteed, ... except when they plan to fraud the survey. An agreement was however not a 100% guarantee for participation. Finally, 246 persons have been questioned (response rate of 4.76%). It was our intention to obtain a response rate of 33% or 1,734 respondents. A direct contact with this sample group via the interviewers, would have delivered a much higher response rate. The interviewer can personally convince his persons to participate, which is more effective than a letter. Lessons learnt: – A letter sent to the sample with the question whether they want to participate is not an appropriate working method. We still prefer a direct method where the interviewer is immediately contacting the person by a personal contact. That way it is possible to convince the participants personally of the importance of the questionnaire. At the same time an appointment can be made for the interview.

2.

Methodology of interviewing

The item non response was still a problem despite the use of a face-to-face interview and the routing on the paper questionnaire. Despite the briefing of the interviewers and the briefing document some misinterpretations of the questionnaire were observed. Computer technology is more and more used to collect data from the respondents. The use of CAPI (Computer-Assisted Personal Interviewing) instead of a paper questionnaire will reduce the non-response and possible mistakes of the interviewers Lessons learnt: – The use of CAPI must be taken into consideration.

Lessons learnt from the survey

3.

W

129

The use of a face-to-face interview

Sometimes it is argued that questioning people on sensitive issues as income or wealth is unfeasible and from that reason sometimes omitted. It is our experience that you sometimes not only have to ‘sell’ a survey to the financer of the project, but also to the colleagues working on it, to the interviewer and finally in this case also to the respondent. Surveys on sensitive issues need to be sold to the respondent: ’Boost interviewers’ confidence about their ability to ‘sell the survey’ was one of the recommendations of the Household Finance and Consumption Network of the European Central Bank organizing now (2011), also in Belgium, a ‘Household Finance and Consumption Survey’, asking e.g. detailed information on this very income and wealth. This survey design illustrates the feasibility but also the efforts that needs to be put in it. In the design of this recent Euro-area wide survey, we discover a confirmation of many steps taken in our own survey on undeclared work. Proper sampling, oversampling of groups at risk, choice for face-to-face interviews, use of CAPI as suggested here for a ‘full blown’ survey, the importance of (first) contacting the respondents (sending an introduction letter, handing over a package of information on the survey at the door, use of incentives, and not in the least not less than four efforts to contact the respondent at different times to minimize non-response, and the maximum of collected detailed information. To improve the data quality also ‘data-editing’ and ‘imputation of missing data’ is recommended. One of the benefits of this study was the detailed questionnaire. The low response rate however made it impossible to analyze all the questions in detail. Only by means of a faceto-face interview it is possible to go through this detailed questionnaire. A complicated issue like fiscal and social fraud needs support. By a postal or web survey this is less possible or too complicated. The respondents indicate they were comfortable to answer the questions despite of the presence of an interviewer. The interviewers did not have the impression respondents were lying about certain answers, even though the chance of a social desirable answer is high in a faceto-face interview. The chance the respondent is giving unrealistic answers is perhaps also smaller in the presence of an interviewer. Postal and web surveys cannot guarantee this. The cost for the organization of face-to-face interviews is perhaps the most negative aspect. When the goal is to analyze the undeclared incomes and activities in detail, one needs a large enough gross sample. In our survey the small response rate made this impossible from the start despite of the level of respondents who committed fiscal or social fraud. The initial gross sample of 12,900 persons could have fulfilled this ambition. A detailed questionnaire only has an advantage when there are enough respondents.

4.

Sensitivity of the questions

We have added some closing/debriefing questions to the questionnaire. This should be useful to detect sensitive questions in the survey (Where there any questions in this interview that you felt uncomfortable answering?) (Bradburn, Sudman, Blair and Stocking, 1978).

130

X

Lessons learnt from the survey

The honesty of the respondents is possibly rather high when we observe the answers given on the questions to what extent they felt (un)comfortable to answer the questions about the demand for and the supply of undeclared work.

19.9%

Not at all uncomfortable

72.8%

29.0%

Little uncomfortable

17.8%

43.4%

Average uncomfortable

8.2%

7.7%

Very uncomfortable

1.2% 20,0%

0,0%

40,0%

60,0%

Others Source:

80,0%

100,0%

Yourself

Own calculations based on SUBLEC data.

Figure 8.1. K1. Extent the respondent felt comfortable/uncomfortable: Demand for undeclared work (n = 246)

18.0% Not at all uncomfortable

81.0%

22.6%

Little uncomfortable

13.6%

44.6%

Average uncomfortable

3.8%

14.8%

Very uncomfortable

1.6%

0,0%

20,0%

40,0% Others

Source:

60,0%

80,0%

100,0%

Yourself

Own calculations based on SUBLEC data.

Figure 8.2. K2. Extent the respondent felt comfortable/uncomfortable: Supply of undeclared work (n = 246)

Lessons learnt from the survey

W

131

Almost 73% of the respondents were not at all uncomfortable to answer the question about the demand for undeclared work (figure 8.1). They think most of the other people will feel on average uncomfortable (43.4%). The percentage of respondents who were not at all uncomfortable is even higher for the question about the supply of undeclared work (81%), what seems reasonable (figure 8.2). Although most of the respondents think people will feel more uncomfortable to answer this question than the question about the demand for undeclared work. In literature we find that questions about the supply of undeclared work should be more sensitive than questions about the demand for undeclared work (de Heij and Kazemier, 2007). At the debriefing, the interviewers also remarked that they had the impression respondents were very open to talk about social and fiscal fraud and their income. The questions were not observed as threatening to the respondents.

Chapter 9

Summary and conclusions

The pilot study on declared and undeclared work and income was organized in the summer of 2010. It was the end of a long preparatory road of assessing the desirability and feasibility of a direct methodology to describe the size, structure and determinants of undeclared work and income. The preparatory road included stock taking of international experience on this direct methodology, identifying the scope of the survey and identifying the optimal methodology in as well sampling and defining the questionnaire, preparing and discussing the needed agreements, not in the least to guarantee privacy rules and the financing of a larger scale survey and finally informing and mobilizing the stakeholders. This report can now present the temporary endpoint of this road. For many reasons the roll-out was different than originally planned so that instead of a large scale population survey it had to be reduced to a nevertheless also already reasonable size, a decent pilot study on the methodology of organizing a survey on undeclared work and income in Belgium. It ended up with a large definition of supply of and demand for undeclared activities, social and fiscal fraud and benefit fraud and a broad overview of the characteristics and possible determinants, within an almost complete sample of the Belgian population between 18 and 75, differentiated along social-economic categories. This project had the ambition to organize a survey about the underground economy. This was not limited to only undeclared work but also integrated social benefit fraud and different other forms of fiscal fraud. Due to the small number of respondents this study has to be considered as a pilot study. This report is partial, tentative and sometimes maybe even speculative. It is certainly not to be considered as definitive. It has been a validation of the feasibility of such a survey, the instrument, and the relevance. For some elements we could verify some of the answers with the results in the Special Eurobarometer No. 284 from 2007, for other elements, further external validation was warranted but we refrained from it because of the size of the sample. The survey on undeclared work and income was organized to inform policy makers on the size of these phenomena, and provide recommendations for policy makers to better fight undeclared work and income. Because of the size of the survey the observations that can be made now are rather recommendations on how to proceed further, and not the definitive answers (as far as research ever provides definitive answers) on size, structure and how to fight it better. The observations are to be read as tentative, sometime counterintuitive, sometimes challenging existing evidence and opinion, and worth to be used as hypotheses for further verification

134

X

Summary and conclusions

and research. Many times however the relevance of the collected information for the debate on undeclared work and income has been given. Within this large questionnaire and population sample sub-questionnaires and subsamples were defined, so that one SUBLEC survey embedded many partial surveys on all the considered phenomena of social and fiscal fraud. Below, we describe the strengths and weaknesses of the design and roll-out of our survey, some tentative results and finally some illustrations of the use in the policy debate, and a final recommendation to continue along the road we took.

1.

Strengths and weaknesses of the present pilot study

We presented here a quasi detailed analysis as if we had a full scale database. This was worthwhile since in could reveal some of the inconsistencies in the questions or the answers, but at the same time it already revealed the relevance of the collected information. At the same time we limited our reporting since going in further detail did not make sense since a larger sample was needed for analyzing along those lines of detail,.

2.

Some summary of first tentative results

38.8% of the Belgian respondents bought an undeclared good or service during the last 12 months. This percentage of demand for undeclared work is much higher than in the Eurobarometer for Belgium and for the EU27. But not only the percentage of people who are asking for undeclared work is important, the amount of this undeclared work is of a great importance too. During the last 12 months an average amount of € 1,553 was spent on the most expensive undeclared goods or services. This amount is higher than the results in the Eurobarometer (Belgium: € 1,050 and EU27: € 1,028).

Table 9.1. Size of undeclared work. SUBLEC

Eurobarometer: Belgium

Eurobarometer: EU27

Demand for undeclared work: services/goods General – Services – Goods Average amount (€) % GDP

38.8% 35.2% 14.1% 1,553 1.9%

15% 8% 1,050 0.6%

11% 9% 6% 1,028 0.5%

Supply of undeclared work: services/goods General Average amount (€) % GDP

14.1% 1,332 0.6%

6% 1,000 0.2%

5% 1,119 0.2%

Source:

Own calculations based on SUBLEC data; EC, 2007.

Summary and conclusions

W

135

Also the percentage of supply of undeclared work (14.1%) is higher in the SUBLEC survey than in the Eurobarometer for Belgium (6%) and for the EU27 (5%). The average amount received for the undeclared work during the last 12 months is € 1,332, which is somewhat higher than the results in the Eurobarometer (Belgium: € 1,000 and EU27: € 1,119). The frequency and volume of undeclared work can be translated in a percentage of the GDP. 1.9% of the GDP is spent on undeclared goods and services and 0.6% of the GDP is lost by carrying out undeclared work. Normally the demand for and the supply of undeclared work should be equal to each other. Questions about the demand for undeclared work and the extent of it can be considered as less sensitive than the supply of undeclared work. The volume of the supply of undeclared work will be an underestimation. Probably this will also be the case for the demand for undeclared work. These figures are also higher than the results from the Eurobarometer. On the basis of a probit analysis it was possible to verify which independent variables had an influence on the demand for and the supply of undeclared work and on fiscal fraud. None of the independent variables are significant for all the three dependent variables.

Table 9.2. Probit analysis. Parameter

Variable

Demand for undeclared work Pr > Khi2

Fiscal fraud

Estimation

Pr > Khi2

Estimation

-0.3613

0.3172

-2.1466

0.0001***

0.5952

0.0752*

0.6068

0.013**

0.0162

0.9305

Estimation Intercept

Supply of undeclared work

Pr > Khi2

Sex

Man

-0.0331

0.8543

Region

French-speaking

-0.0896

0.6217

0.424

0.087*

-0.2762

0.1495

Socio-economic category

Self-employed

0.8488

0.0327**

0.1701

0.6884

0.1999

0.6179

Benefits recipient

-0.4383

0.0248**

-0.7391

0.0094***

0.3426

0.0932*

Inactive

-0.4079

0.2298

0.3395

0.3443

0.0457

0.8858

1.1406

0.0106** -0.0152

0.9379

Know someone (demand)

Yes

0.8468

0.0006***

Know someone (supply) Know someone (fiscal fraud) Income (1)

Difficult

-0.3393

0.0710*

0.5495

0.0667

0.7337

Morality (2)

Totally agree

-0.5279

0.0661*

-0.652

0.0575*

0.3207

0.271

Rather agree

0.261

0.3768

-0.5856

0.0847*

-0.0158

0.9571

-0.1914

0.6261

-0.23

0.6265

-0.1203

0.7575

Disagree

0.1508

Note: *, **, and *** indicate significance at the 10%, 5% and 1% level, respectively. 1 Get by on their monthly income? 2 The tax burden is too high in Belgium? Source:

Own calculations based on SUBLEC data.

Knowing persons who are asking for or carrying out undeclared work has a positive influence on doing this also. The impact of information about others’ behavior on the decision

136

X

Summary and conclusions

to commit fraud was also recently observed by laboratory experiments about tax evasion and welfare fraud (Lefebvre, Pestieau, Riedl and Villeval, 2011). The sex has no influence on the decision to buy undeclared goods or services. This is not the case for the supply of undeclared work, where men are more likely to do undeclared work. Being a benefit recipient has a negative influence on the decision to ask for undeclared work or to carry out undeclared work. Self-employed persons appear to ask for undeclared goods and services more fresuently. Those results alone need a further research since they challenge the usual opinions on those matters to a large degree. It was our intention to draw a detailed picture of the underground economy in Belgium. Due to the lack of respondents this was not fully possible. Nevertheless we have obtained indications of the size of different aspects of the underground economy. In the table below we have listed the frequency, the volume and the number of people who know someone committed this fraud for each of the different aspects.

Table 9.3. Summary of the Belgian underground economy. Frequency (% of people) Demand for undeclared work Supply of undeclared work Envelope wage Social benefit fraud Benefit cumulated with undeclared work Tax return not completely correct Capital revenues Real estate income Inheritance Registration fee *

Know someone 79.2% 78.5% 52.1%

2.3%* 3.0% 1.8% 33.2% 5.1%

33.6% 27.3% 41.3% 40.3%

Asked to all respondents.

Source:

3.

38.8% 14.1% 2.0% 5.6% 4.3% 24.1% 3.5% 0.3% 5.5% 1.9%

Volume (% of total amount)

Own calculations based on SUBLEC data.

Relevance for policy making and how to continue

The observations are to be read as tentative, sometimes counterintuitive, sometimes challenging existing evidence and opinion, and worth to be used as hypothesis for further verification and research. Many times however the relevance of the collected information for the societal debate on undeclared work and income has been given. The comments we made on some of the observations are sometimes not even tentative but risk being speculative, and should be read as such. Examples are about the groups more or less at risk for fraud, about the determinants for fraud (fiscal pressure, lack of control, bad examples, morality), about the appropriate ways to tackle it (fiscal pressure, perseverance in control since it could have a deterrence effect, awareness campaigns).

Summary and conclusions

W

137

By using a large and exhaustive definition of social and fiscal fraud (only tax avoidance is not included since it can be collected, but has not been done up until now, from registered information) the survey can be used for many policy makers and subcategories of the population. A population survey needs a certain size to guarantee representativeness and allow more detailed analyses. This one was exhaustive and moderately intensive (we know more intensive population surveys), so that it is time consuming and costly. It can provide figures about the underground structure of our economy and economic behavior, so that it takes place on a regular, but not permanent basis. There are even good reasons to repeat it with a certain time gap.

Annex 1

Programme SUBLEC Kick-off seminar– 25 February 2008

Morning Session Chaired by Jozef Pacolet (HIVA-Onderzoeksinstituut voor Arbeid en Samenleving, Katholieke Universiteit Leuven) 08.30

Welcome and coffee

Part I Welcome

09.00 09.10

Welcome: Tom Auwers/Koen Vleminckx: Federal Public Service Social Security and Aziz Naji: Science Policy Didier Verbeke, Federal Public Service Social Security, Ambition of SUBLEC

Part II Population oriented surveys

9.20 10.00

10.30

11.00

11.30 11.45 12.00 12.30

Friedrich Schneider, Recent experience in the use of surveys measuring and understanding the underground economy Overview of surveys on social and fiscal fraud and its relevance for statistical and control issues: Jozef Pacolet and Sigrid Merckx, Onderzoeksinstituut voor Arbeid en Samenleving, Katholieke Universiteit Leuven Belgian experience with direct surveys and questions for the new survey: Sergio Perelman and Pierre Pestieau, CREPP Centre de Recherche en Economie Publique et de la Population ULg How to include goups at risk: clandestine workers, illegal migration, in the sample and how to reach them, E. Krzeslo, Centre de sociologie du Travail, de l’Emploi et de la formation, ULB Coffee Debate The Experience of the Rockwool Foundation and its relevance for the fiscal adminstratio. Søren Pedersen, Ministry of Taxation and before Rockwool Informal economy: comparing a mixed mode and face to face methodlogy, José Gouweleeuw, Harold Eding, Statistics Netherlands (Alternatives or in-

140

13.00

X

Programme SUBLEC Kick-off seminar– 25 February 2008

terest audience: Brugt Kazemier, Ron (H.N.) de Heij, Statistics Netherlands, Kees de Wit, Ministerie van Sociale Zaken en werkgelegenheid, directie Financieel-economische zaken Debate

13.15 Lunch Afternoon session Chaired by Sergio Perelman, CREPP Centre de Recherche en Economie Publique et de la Population ULg 14.00 14.30 15.00 15.30 15.45

Lessons from the Eurobarometer 2007: Arnold Riedmann, Research Manager TNS Infratest Sozialforschung Lessons from the Eurobarometer 2007: Sergio Perelman. Some additional analysis for Belgium Lessons from the Eurobarometer 2007: relevance for Europe, Guido Vanderseypen, European Commission Debate Coffee

Part III Firms oriented surveys

16.00 16.30 17.00

Use of business surveys in detecting fraud: Jan Hanousek, CERGE-EI, Charles University, Prague Feasibility of surveys in construction in Belgium: Jozef Pacolet, Onderzoeksinstituut voor Arbeid en Samenleving, Katholieke Universiteit Leuven Debate

Part IV Some conclusions to go forward in Belgium

17.15

Didier Verbeke, Jozef Pacolet, Sergio Perelman and Estelle Krzeslo

Annex 2

Steering committee

Aseglio Michel De Dyn Hans De Troyer Marianne De Wispelaere Frederic Dubois Luc Dubois Véronique Fegatilli Ermano Gratia Marianne Heirman Jean-Claude Kermarrec Gaël Krzeslo Estelle Laureyns Patrick Le Bussy Yves Mathues Hugo Naji Aziz Pacolet Jozef Perelman Sergio Renaud Luc Serlippens Monique Smet Martine Vanhuysse Sven Verbeke Didier Verdonck Tom Verlinden Eric Willems Marc

FOD WASO NBB METICES, ULB HIVA, KU Leuven RIZIV SIOD CREPP, ULg FOD SZ FOD SZ/SIOD SIOD METICES, ULB Kabinet Staatssecretaris Devlies RVA Kabinet Staatssecretaris Clerfayt POD Wetenschapsbeleid HIVA, KU Leuven CREPP, Ulg RSVZ RSZ FOD Financiën Kabinet Minister Laruelle FOD SZ RIZIV RSZ Inspectie Werk en Sociale Economie

References

Blom A. G., de Leeuw E. D. and Hox J. J. (2010), ‘Interviewer Effect on Nonresponse in the European Social Survey’, Discussion Paper Series, 202-2010, Mannheim Research Institute for the Economics of Aging, Mannheim, 50 p. Bradburn N. M., Sudman S., Blair E. and Stocking C. (1978), ‘Question threat and response bias’, Public Opinion Quarterly, 42, p. 221-234. Castiglioni L., Pforr K. and Krieger U. (2008), ‘The Effect of Incentives on Response Rates and Panel Attrition: Results of a Controlled Experiment’, Survey Research Methods, Vol. 2, No. 3, p. 151-158. Ciccarone G., Marchetti E. and Pavlovaite I. (2009), Study on indirect measurement methods for undeclared work in the EU, GHK, Birmingham, 107 p. Davern M., Rockwood T. H., Sherrod R. and Campbell S. (2003), ‘Prepaid Monetary Incentives and Data Quality in Face-to-face Interviews’, Public Opinion Quarterly, Vol. 67, p. 139-147. de Heij R. and Kazemier B. (2007), ‘Undeclared work, tax evasion and avoidance in The Netherlands’, Paper prepared for the international conference on Undeclared Work, Tax Evasion and Avoidance: A momentum for change in Belgium and Europe, 21-22 June 2007, Brussels, Belgium, 12 p. De Pelsmacker P. and Van Kenhove P. (1999), Marktonderzoek. Methoden en toepassingen, Garant, 737 p. Diallo H., Karakaya G., Meulders D. and Plasman R. (2010), Raming van de belasting fraude in België, Working paper N° 10-07.RR, DULBEA-ULB, Brussel, 55 p. ECB, Household Finance and Consumption Network (2008a), Reducing non-response bias. ECB, Household Finance and Consumption Network (2008b), Imputation and data-editing. European Commission (1998), Communication of the Commission on undeclared work, COM(1998) 219, 23 p. European Commission (2007a), Communication from the Commission to the Council, the European Parliament, the European Economic and Social Committee and the Committee of the Regions. Stepping up the fight against undeclared work, COM (2007) 628, Brussels, 11 p. European Commission (2007b), Special Eurobarometer 284. Undeclared Work in the European Union, Brussels, 90 p. Fegatilli E. (2009), ‘Undeclared work in the European Union – What can we learn from an European survey?’, Working Paper 2009-02, CREPP-ULg, Liège, 20 p. Feige L.F. and Urban I. (2008), ‘Measuring underground (unobserved, non-observed, unrecorded) economies in transition countries: Can we trust GDP?’, Journal of Comparative Economics, Vol. 36, No. 2, p. 287-306. Feld L.P. and Schneider F. (2010), ‘Survey on the shadow economy and undeclared earnings in OECD countries’, German Economic Review, Vol. 11, No. 2, p. 109-149. Fisher R. J. (1993), ‘Social desirability bias and the validity of indirect questioning’, The Journal of Consumer Research, 20, (2), p. 303-315. FOD Financiën (2010), Jaarverslag 2009. Tabellen en grafieken, Brussel, 116 p. Herwartz H., Schneider F. and Tafenau E. (2010), ‘One share fits all? Regional variations in the extent of shadow economy in Europe’, paper presented at workshop ‘Shadow Economy, Tax Policy, and

144

X

References

the Labour Markets in an International Comparison: Options for Economic Policy’, 15-16 April 2010, University of Potsdam. Holbrook A. L., Green M. C. and Krosnick J. A. (2003), ‘Telephone versus face-to-face interviewing of national probability samples with long questionnaires: comparisons of respondent satisficing and social desirability response bias’, Public Opinion Quarterly Volume, 67, p. 79-125. Household Finance and Consumption Network, (October 2008), Household Finance and Consumption Survey: Modalities for Implementation, European Central Bank, mimeo. Household Finance and Consumption Network, (October 2008), Imputation and data-editing, European Central Bank, mimeo. Household Finance and Consumption Network, (October 2008), Reducing non-response bias, European Central Bank, mimeo. IDEA Consult (2010), Evaluatie van het stelsel van de dienstencheques voor buurtdiensten en –banen 2009, IDEA Consult, Brussel. INR (2006), De berekeningsmethode voor het bruto binnenlands product en het bruto nationaal inkomen volgens het ESR 1995, NBB, Brussel, 549 p. Jobber D., Saunders J. and Mitchell V-W. (2004), ‘Prepaid monetary incentive effects on mail survey response’, Journal of Business Research, Vol. 57, p. 21-25. Kazemier B. and van Eck R. (1990), The supply of hidden labour in the Netherlands: a model, Central Bureau of Statistics, Voorburg/Heerlen, Nr. NA-041, 35 p. Kazemier B. (2003), ‘De zwarte economie: een overzicht van methoden en ramingen’, in J. Pacolet and A. Marchal (red.), Zwartwerk en fraude: een bedreiging voor de verzorgingsstaat in België en Europa, BTSZ, Brussel, p. 913-955. Krosnick J. A. (1999), ‘Survey research’, Annual Review of Psychology, 50, p. 537-567. Lefebvre M., Pestieau P., Riedl A and Villeval M.C. (2011), ‘Tax Evasion, Welfare Fraud, and “The Broken Windows” effect: An Experiment in Belgium, France and the Netherlands’, IZA Discussion Paper No. 5609, 49 p. Martin E. (2006), Survey questionnaire construction, US Census bureau, Washington DC, 13 p. NBB (2010), ‘De zwarte economie in de Belgische nationale rekeningen. Cijfers 2007 en evaluatie van de hogere ramingen gesuggereerd door andere studies’, NBB, Brussel, 17 p. OECD (2002), Measuring the Non-Observed Economy. A handbook, OECD Publications, Paris, 250 p. OLAF (2006), Deterring Fraud by Informing the Public, EC, Brussels. Pacolet J. (2007a), ‘Undeclared work and the fight against it in Belgium’, European Employment Observatory Review: Spring 2007, National report for Belgium, http://www.eu-employment-observatory. net/resources/reviews/NationalArticles/BelgiumUDW2007.pdf Pacolet J. (2007b), Consultancy on the synthesis of recent information on undeclared work in the EU, in preparation of the communication from the European Commission. Pacolet J. and Baeyens K. (2007), Deloyale concurrentie in de bouwsector. Een terreinverkenning van mechanismen van sociale fraude, hun omvang en hun gevolgen voor de sector, HIVA-KU Leuven, Leuven, 149 p. Pacolet J. and De Wispelaere (2009a), Naar een observatorium ondergrondse economie. Een haalbaarheidsstudie, Acco, Leuven, 166 p. Pacolet J. and De Wispelaere (2009b), ‘The underground economy: designing an appropriate survey methodology to reveal sensitive behaviour (social and fiscal fraud)’, HIVA-KU Leuven, mimeo Pacolet J., De Wispelaere F. and De Coninck A. (2011), De dienstencheque in Vlaanderen. Tot uw dienst of ten dienste van de zorg?, Steunpunt WVG, Leuven. Pacolet J. and Merckx S. (2008), SUBLEC: designing a survey methodology for fiscal en social fraud in Belgium: recent international comparative evidence and conclusions for Belgium, HIVA-KU Leuven, Leuven, mimeo Pacolet J., Perelman S., Pestieau P., Baeyens K. and De Wispelaere F. (2009), Zwartwerk in België. Een indicator van omvang en evolutie – Travail au noir en Belgique. Un indicateur concernant l’étendue et l’évolution, Acco, Leuven, 195 p.

References

W

145

Pacolet J. and Strengs T. (2011), De kost van fiscale en parafiscale uitgaven en ontwijking in België, HIVA-KU Leuven, Leuven, 148 p. Pacolet J. and Verbeke D. (2007), Satellite Accounting for Undeclared Work, Paper presented at the Seminar on undeclared work of DG EMPL December 2007, Brussels. Pedersen S. (2003), The shadow economy in Germany, Great Britain and Scandinavia. A measurement based on questionnaire surveys, The Rockwool Foundation Research Unit, Copenhapen. Pfau-Effinger (2009), ‘Varieties of Undeclared Work in European Societies’, British Journal of Industrial Relations, Vol. 47, Nr. 1, p. 79-99. Renooy P., Ivarsson S., van der Wusten-Gritsai O. and Meijer E. (2004), Undeclared work in an enlarged Union. An analysis of undeclared work: an in-depth study of specific items, European Commission, 236 p. Rockwood T. H., Sangster R.L. and Dilleman D. A. (1997), ‘The effect of response categories on questionnaire answers: context and mode effects’, Sociological Methods Research, 26, p. 118-140. Ryu E., Couper M. P. and Marans R. W. (2005), ‘Survey Incentives: Cash vs. In-kind; Face-to-face vs. Mail; Response Rate vs. Nonresponse Error’, International Journal of Public Opinion Research, Vol. 18, No. 1, p. 89-106. RVA-ONEM (2011), Jaarverslag 2010 – Rapport annuel 2010, Brussel-Bruxelles. Schneider F. and Enste D.H. (2000), ‘Shadow Economies: Size, Causes, and Consequences’, Journal of Economic Literature, Vol. XXXVIII, p. 77–114. Schneider F. (2010), ‘The influence of the economic crisis on the shadow economy in Germany, Greece and the other OECD-countries in 2010: What can we do?’, 19 p. Schwarz N. and Oyserman D. (2001), ‘Asking questions about behavior: cognition, communication, and questionnaire construction’, American Journal of Evaluation, 22, p. 127-160. Singer E. (2002), ‘The Use of Incentives to Reduce Nonresponse in Household Surveys’, Survey Methodology Program, No. 051, University of Michigan, 34 p. TNS Infratest Sozialforschung, Rockwool Foundation Research Unit, Regioplan Beleidsonderzoek (2006), Feasibility Study on a direct survey about undeclared work, 156 p. Torgler B. (2007), Tax Compliance and Tax Morale: A Theoretical and Empirical Analysis, Edward Elgar, Cheltenham, 307 p. Tourangeau R. and Yan T. (2007), ‘Sensitive questions in surveys’, Psychological Bulletin, 133, (5), p. 859-883. UNECE (2003), Non-Observed Economy in National Accounts. Survey of National Practices, United Nations Publication, Geneva, 256 p. UNECE (2008), Non-Observed Economy in National Account. Survey of National Practices, United Nations Publication, Geneva, 340 p. Van Teijlingen E. R. and Hundley V. (2001), ‘The importance of pilot studies’, Social Reserach Update, University of Surrey, winter, issue 35, p.1-4. Willimack D. K., Schuman H., Pennell B-E. and Lepkowski J. M. (1995), ‘Effects of a Prepaid Nonmonetary Incentive on Response Rates and Response Quality in a Face-to-face Survey’, Public Opinion Quarterly Volume, vol. 59, p. 78-92.

Jozef Pacolet & Frederic De Wispelaere

Naar een observatorium ondergrondse economie Studies over sociale en fiscale fraude

de publieke opinie hebben baat bij het zichtbaar worden van de omvang en de nadelige gevolgen van het zwartwerk en van de strijd ertegen. ‘Exhaustiviteit’ of het verwerven van het globale beeld en de mobilisatie van alle betrokken instanties zijn hierbij essentieel. De auteurs doen suggesties voor de organisatie van een observatorium voor ondergrondse economie. De uiteindelijke keuze van de vorm zal bij de overheid liggen. Belangrijker dan de vorm is echter het feit dat het gecreëerd zal worden. In opdracht van het Federaal Wetenschapsbeleid en de Federale Overheidsdienst Sociale Zekerheid wordt in deze publicatie de haalbaarheid en de wenselijkheid van een informatie- en analysecentrum voor de ondergrondse economie nagegaan De auteurs gaan na welke informatie over sociale en fiscale fraude er thans beschikbaar is. Ze vergaarden hiertoe informatie uit zowel theoretische bronnen als de praktijk van de talrijke instanties die in België verantwoordelijkheid dragen voor de strijd tegen sociale en fiscale fraude. Talloze binnenlandse en buitenlandse voorbeelden, als ook de opinies ter zake van belanghebbenden, tonen aan dat een observatorium wenselijk, haalbaar, maar vooral ook noodzakelijk is. Sociale inspecties, fiscale controlediensten, het politioneel en gerechtelijk apparaat, de sociaaleconomische partners, de onderzoekswereld, de beleidsmakers en

Dit onderzoek kadert in een reeks van wetenschappelijke studies omtrent de sociale en fiscale fraude die door het Federaal Wetenschapsbeleid de jongste jaren werd gelanceerd. JOZEF PACOLET is professor in de economische wetenschappen. Hij is hoofd van de onderzoeksgroep Verzorgingsstaat en Wonen en lid van de directie van het HIVA-KU Leuven. FREDERIC DE WISPELAERE is licentiaat in de handelswetenschappen en master-namaster in het bedrijfsrecht. Hij is onderzoeker bij de onderzoeksgroep Verzorgingsstaat en Wonen van het HIVA-KU Leuven. ISBN 978 90 334 7595 5 // 2009 // 168 blz. // 30,00 euro Uitgeverij Acco Blijde Inkomststraat 22, 3000 Leuven tel. 016/62 80 00 fax 016/62 80 01 e-mail: [email protected] www.uitgeverijacco.be

Jozef Pacolet, Sergio Perelman, Pierre Pestieau, Kathleen Bayens & Frederic De Wispelaere

Zwartwerk in België. Een indicator van omvang en evolutie Studies over sociale en fiscale fraude

coexister des estimations très divergentes en la matière. Dit onderzoek kadert in een reeks van wetenschappelijke studies omtrent de sociale en fiscale fraude die door het Federaal Wetenschapsbeleid de jongste jaren werd gelanceerd.

De inschatting van de omvang van het zwartwerk verbetert voortdurend maar is toch nog steeds gebrekkig. Het statistisch apparaat is niet aangepast aan de hedendaagse behoefte om dit in beeld te brengen. De opdracht is ook niet eenvoudig: het is het zichtbaar maken van wat zich per definitie wenst te verbergen. Cette publication a donc pour objectif l’identifi cation des moyens disponibles pour améliorer les estimations de l’étendue de la fraude. Suite à la demande de la Politique Scientifique et du Service Public Fédéral Emploi, Travail et Concertation Sociale, cette publication rassemble des informations, aux niveaux international et national, portant sur l’ampleur du travail non déclaré, de la fraude sociale et de la fraude fiscale. Une multitude de méthodes ont été répertoriées. En les conciliant et en les combinant nous obtenons une image plus positive de notre pays, qui a laissé pendant longtemps

JOZEF PACOLET is professor in de economische wetenschappen. Hij is hoofd van de Onderzoeksgroep Verzorgingsstaat en Wonen en lid van de directie van het HIVA – KU Leuven. SERGIO PERELMAN est professeur en sciences économiques et directeur du Center of Research in Public Economics and Population Economics (CREPP – Université de Liège). PIERRE PESTIEAU est professeur en sciences économiques. Il est président du CREPP. KATLEEN BAEYENS is doctor in de toegepaste economische wetenschappen en was als onderzoeker verbonden aan de Onderzoeksgroep Verzorgingsstaat en Wonen van het HIVA – KU Leuven. FREDERIC DE WISPELAERE is licentiaat in de handelswetenschappen en master-na-master in het bedrijfsrecht. Hij is onderzoeker bij de Onderzoeksgroep Verzorgingsstaat en Wonen van het HIVA – KU Leuven. ISBN 978 90 334 7594 8 // 2009 // 200 blz. // 30,00 euro Uitgeverij Acco Blijde Inkomststraat 22, 3000 Leuven tel. 016/62 80 00 fax 016/62 80 01 e-mail: [email protected] www.uitgeverijacco.be

Gedrukt en gebonden bij Acco, Leuven