L
----
P -
Tijdschrift - L
ISSN 0772-7674
voor
economie en aiS.xayeiiaa -
Driemaandelijks publikatie van d e Faculteit der Economische en Toegepaste Economische Wetenschappen van de Katholieke Universiteit Leuven, uitgegeven m e t medewerking van Ekonomika. vereniging van afgestudeerden van de faculteit. uitgegeven m e t d e steun van h e t Ministerie van de Vlaamse Gemeenschap Departement Onderwijs.
REDACTIE: Hoofdredacteur: Prof. Dr. F. Abraham Kernredactie: G. De Bruyne, M. Dekimpe, J. D'Haene, D. Heremans, J. Konings, C. Lefebvre, F! Sercu, M. Vandebroek, C. Van Hulle, R. Veugelers, H. Gaeremynck en H. ÜewacnIer. Redactieraad: Deze raad omvat naast de hoofdredacteur en de leden van de kernredactie eveneens: professoren F! Beghin (R.U.Gent), H. Daems (K.U. Leuven), R. De Bondt (K.U.Leuven), M. Dombrecht, (Nat. Bank, Brussel), S. Proost (K.U.Leuven), E. Schokkaert (K.U.Leuven), F! Van Cayseele (K.U.Leuven) W. Vanthielen (L.U.C. Limburg), J. Vuchelen (V.U. Brussel). Redactiesecretariaat: A. Ronsmans Tijdschrift voor Economie en Management Naamsestraat 69, 3 0 0 0 LEUVEN Tel. 016132.66.88 - Fax. 016132.66.10 - 3 2 . 6 6 . 9 8 E-mail:
[email protected] Verantwoordelijke uitgever: F. Abraham, Beukenlaan 3 4 , 3 0 0 1 Heverlee De Redactie beoordeelt de kwaliteit van de bijdragen maar kan niet verantwoordelijk gesteld worden voor de inhoud. The Editorial Board judges the quality of the contributions published but can not be held responsible for their content. ABONNEMENTSVOORWAARDEN: Volledige jaargang: 4 afleveringen (ca. 5 0 0 blz.) 2 . 0 0 0 BF als steunabonnement* 1 . 2 0 0 BF als institutioneel abonnement (bibliotheken, instellingen, bedrijven) 8 0 0 B F als persoonlijk abonnement (vooraf te betalen via persoonlijke rekening) 5 0 0 B F als student 1 . 5 0 0 BF of $ 4 0 als buitenlands abonnement 3 5 0 B F als los nummer Abonnementen en bestellingen worden uitsluitend vereffend op PR. 000-0112553-33 Tijdschrift voor Economie en Management, Leuven.
KATHOLIEKE UNIVERSITEIT T E LEUVEN FACULTEIT DER ECONOMISCHE E N TOEGEPASTE ECONOMISCHE WETENSCHAPPEN
economie en
INHOUD Advertentie RiskIReturn Relationship Conditional on Market Movements on the Brussels Stock Exchange J. CROMBEZ AND R. VANDER VENNET
163
Earnings Management and Institutional Differences: Literature Review and Discussion H. VANDERBAUWHEDE AND M. WILLEKENS
189
Temporary Workers and their Transition Probabilities t o Different Labour Market States: the Belgian Experience A. PEETERS
213
Wages and Firm Profitability in the Manufacturing Sector in Côte d'lvoire M. GOEDHUYS 2 3 3 Advertentie Roofdieren in de markt?
R. DE BONDT
253
Eindverhandelingen
267
Boekbesprekingen
273
JUNI 2000 DRIEMAANDELIJKS
JAARGANG X tv
postacademische vorming 7
' li
l"\,,
a,,
,g
",V:,:"
Overheid en economische activiteit: Recht en Beleid
--%P
q&;
,&33&*;%3
Eerz znztzatzef vari het Iriatiluut voor de Overlzr~dz J nz het I~zstrtuutvoor r,*$ *,%.J: &--$L *"%c2 %4, A~lr~z~r~r~trntref Reclzt ei1 de Drenst Postric~1~len7isthe Vorrrring Kader en doelstelling. De iol van de overheid in de econoinische activiteit IS het laatste decciiiituiii sterk geevolueerd en iicenit anno 2000 veisiliillende, uiteeillopende voiincn aai1 Een aantal belangr~jlteoorzaken van deze evolutie 2y;h-p&> vloeien vooit uit het steikcr managen van liet overheidsoptreden, liet eificieriter ws%>**4-:>Q f aanwenden van de overlieidsiniddilen ei1 het dcmonopolieeieii vrii tal van overheidstalten De opleiding heeft tot doel de cursisten wegwijs te make11 in de ae,-*,-&i;. regclgeviiig die Iiet opmdeii vali de overiieid in Iiet econoiniscb leven vonn geen
.-q& >ah
l$,s$;:
p ,
ai
aa
DuurLijd & d a t a De opleiding voorziet 111 een programina van 45 u u , gespieid over 15 sessies op donderdagavond 19110,26110,9111,2311 l , 30111,7112,14112, 21/12/2000, 1111, 1811,2511. 112, 812, 1512, 221212001
Prijs 55 000 b i - Lokatie Leuven - Faculty Club Doelgroep. Ainbtcnuen van de Federale en Vlaamse overheid, provinciale en lokde
besturen, overheidsbedrijven, intercominu~iales.magistraten, advocaten en bednjtsjunsten Het aantal deelnemer< is bepcrkt tot maximuni 30, de volgoide van inschiijven is hierbij bepaleiid Modules I Revoegdheid van de Oveiheid inzake het uitoefenen van een ecoilomisclie activiteit, 2 Uitvoering in eigen belieei, uitbesteding aan deiden, privatisering, veizelfstandiging, outsouicing, 3 Ecoiiomisch handelen door de @$&*~;$$~ Overheid financieel management, 4 Hct (rntei)geineentelijk initiatief geineentebedillven en intercommunales, 5 Autonome oveiheidsbedrijven en publiekreclitclijl
:<*;;8-S inbtruincntaiiuiii "&>a$ c .->g," i
$z:;:&L$g-;:e: ;:g*$v>2e:
<-":-i .A*,
Info, inschrijfing en folder Iiittitifl~rr vooi rfe Oi~elhrlrl,Van Everisticrat 21, 3000 Lez~vrrz,tel 016/32 32 70, fcu 0/6/32 32 67, in@ \OL X11Iruvf17 LI( he, li I I J i L ~ I I ~ < ' L IY11 I Li( ~ Y / ~ ~11117 OIZ Ddgeh~hsged~lualiseeideiniormatie ovei postdc'idernisdie vomiingsinitiatievcn a a l de K U Leuven vincit i s 11, dc databank Open Vorming op www.dpav.kuiieuvei1.;ac:.B,e
$&.,2%*-Xe, 9
"'d"?$*
-q"'" -b
-$
i3*<&,)$&
3<
$$$b %%
~
Tijdschrift voor Economie en Management Vol. XLv, 2,2000
RisWReturn Relationship Conditional on Market Movements on the Brussels Stock Exchange by J. CROMBEZ*, R. VANDER VENNET*
I. INTRODUCTION Thirty years after the introduction of the Capita1Asset Pricing Model (CAPM) by Sharpe (1964) and Lintner (1965), the finance profession is again confronted with a paradox. At a time when beta - the risk measure used to explain expected stock return in the CAPM - has gained worldwide acceptance as the primary risk measure for asset selection, numerous reports of its death are being launched. The pathbreaking contribution of Fama and French (1992) has shown that beta alone is unable to explain the cross-section of US asset returns, especially in recent times. Miles and Timmermann (1996) report similar evidence for UK stock returns. As a result, the asset pricing research is shifting to a quest for more comprehensive multi-factor models that may provide a better explanation of asset returns. Others try to rehabilitate beta mainly by refining the empirica1 content of the tests. A useful approach appears to be the exploration of the inherently conditional nature of the market model in order to capture the timevarying and mean-reverting properties of the data. In this respect, the analysis of the betalreturn behavior in bul1 versus bear market conditions has intuitive appeal because of the hypothesized relation bemeen business cycle conditions and the evolution of the stock market. Hence, the question remains to what extent beta, as a measure of Faculteit Eco~iomieen Bedrijfskunde, Universiteit Gent, Gent. Comments by an anonymous referee are gratefully acknowledged.
systernatic risk, remains a usefurrl tool for portfolio selection in a conditional setting. This paper loolts at the evidence for tliie Brussels Stock Exchange. The approach migkit be interesting for at least two reasons. By al1 accounts, Brussels l-ias a smal1 stock market and it should be useful to examine whether well-docuinented anoinalies in large and broadly covered stock markets can also be detected in smaller financial centers. Moreover, in view of the increasing international diversification of financial asset portfolios, investors may be interested in the behavBor of risk and return on a sinall European stock market. Froin a more practica1 point of view the question Is whether the easy-to-calculate and easy-to-interpret betas are still useful in a portfolio selectbon environment and whether the findings on conditional beta pricing for the U S eyuity market are appiicabie to different markets, i.c. ei~eBrussels Stock Exchange. We first examine the overall performance of beta in the explanation of Belgian stock returns over the period 1990-1996. Our results largely confirm the Fama and French (1992) conclusion that beta is ainable to explain the cross-section of average stock returns. We then investigate whether beta can stil1 serve as a useful risk rneasure by analyzing the relationship between returns and beta conditional on the marlitet movement. The results suggest that there is a fairly close association between beta and stock returns in both upward and downward moving markets. The analysis is robust for various specifications of beta and various definitions of up and down markets. Hence, when making asset allocation decisions on the Brussels Stock Exchange, investors may still want to incliade betas in their risk management practice as a measure of exposure to downside risk or as an indicator of upward potential. The rest of the paper BS organized as follows. Section Ijl reviews the relevant literature on beta tests. Section Ijl1 justifies the use of betas conditional on the market evolution. The calculation of the betas and the rnarket conditions are explained in section HV. Section V concludes. Ijl. OVERVIEW OF THE EMPIRICAL LITEMTURE The CAPM is undoubtedly the most successful model to link the risk and expected return of capital assets. Ht is not only a standard tool fsr investment analysis but the model is als0 widely used for capita1 budgeting purposes. According to the CAPM, the differences in risk premiums across assets are due to differences in the systematic risk of
the assets. This risk is called beta and measures the sensitivity of the return of an asset relative to movements in the rnarket return. Given the risk-free rate, the CAPM predicts that t-he expected return of an asset is an upward-sloping linear function of its beta (see Figure 1). A similar relationship is the outcome of the zero-beta variant of the iaiodel developed by Black (1972). S h e classic support of the CAPM is the study of Black, Sensen and Schsles (1972). For all the stocks on the NYSE covering the years 1931-65 they regress average rnonthly returns on beta for 10 portfolios ranked by pre-estimated betas. For tlmis sample period they find that the data are broadly consistent with Black's version of the C m M .
FIGURE 1 The risk-~etumrelatiolzship accordi~zgto the CAPM
In another stiady for the period 1926-1968,Fama and McBet-Bi(19'73) do not reject t-he i~ypothesisof a posiaive linear relation between average return and beta, which again tends I s corroborate the CAPM. From the 1980s onwards, a number of authors have reportecl evidence contradicting the CAPM and suggested that other factors might be instrumental in explaining stock returns. The first anomaly was reported by Banz (1981) who showed that the cross-sectional variation in average returns is better explained by size (value of inarket equity) than by beta over the period 1936-1975. More specificaligr, observed returns to stocks of smal1 firms were found to be substantially higher than the average return to large stocks after adjusting for risk. Otber
anomalies to the CAPM were found by Stattman (1980) and Rosenberg, Reid and Lanstein (1985) who report a positive relationship between US stock returns and the ratio of book to market value. Chan, Hamao, Lakonishok (1991) confirm this finding for Japanese stocks. Basu (1983) concluded that the earnings-price ratio helps explain the cross-section of stock returns, next to beta. Bhandari (1988) finds a positive relationship between company leverage and average return. The most damaging set of counterevidence to the CAPM was produced by Fama and French ((1992), FF) covering the period 19631990. When they estimate a model with only beta they find a slope coefficient of -0.15 with a standard error of 0.46, implying that the relation between beta and return is flat at best. Thus, the results indicate that beta has no ability to explain the cross-sectional variation in average returns for a wide collection of stocks listed on the NYSE, AMEX and NASDAQ. Moreover, the size variable is significant with or without including beta. And when the authors include the bookto-market ratio they find that this variable is even more powerful than size. FF attribute the shift in conclusions concerning the validity of the single-beta asset pricing model to differences in sample periods. When they rerun their regressions for 1941-1965 they find a positive slope coefficient for beta. It thus seems that the positive association between return and beta, as predicted by the CAPM, has vanished in recent times (see also Chan and Lakonishok (1993)). The finding of the anomalies led to the construction of enhanced asset pricing models. One avenue is to construct a multi-factor model of asset returns to account for other influences than systematic risk. This approach leads to the construction of APT models (arbitrage pricing theory), initiated by Ross (1976), which allow for priced factors that are orthogonal to the market return, or multi-factor versions of Merton's (1973) intertemporal CAPM. Fama and French (1993, 1996) show that many of the CAPM anomalies are captured by a three-factor model in which excess portfolio returns are explained by their sensitivity to (1) the excess return on a broad market portfolio, (2) the difference between the return on a portfolio of small stocks and the return on a portfolio of large stocks (SMB, small minus big), and (3) the difference between the return on a portfolio of high-bookto-market stocks and the return on a portfolio of low-book-to-market stocks (HML, high minus low). They suggest that the book-tomarket factor may proxy for relative distress. Jagannathan and Wang (1993) estimate a model with a value-weighted market portfolio, human
capita1 (proxied by the growth rate of per capita labor income) and which allows for time-varying betas. Over the period 1963-1990, the model explains a substantial portion of the cross-sectional variation in average stock returns. Some researchers have tried to link the behavior of stock returns directly to underlying macroeconomic factors, extending the approach initiated by Chen, Wol1 and Ross (1986). Finally, a number of authors ascribe the evidence against the CAPM to irrational pricing and inefficient capita1 markets (Lakonishok, Shleifer and Vishny (1994); MacKinlay (1995), Haugen and Baker (1996)). Confronted with this persuasive case against the CAPM, a number of studies have tried to challenge the findings of the so-called anomalies on methodological grounds. Kothari, Shanken and Sloan (1995) argue that even with the F F results, a wide range of economically plausible risk premiums cannot be rejected statisticaiiy. The view that the data are too noisy to firmly invalidate the CAPM is als0 espoused by Black (1993). H e notes that the size coefficient is not statistically different from zero in the 1981-1990 period, which suggests that the size effect may have disappeared after it was discovered by Banz (1981). Black (1993) concludes that most of the F F results may be attributable to data mining. Breen and Korajczyk (1993) argue that the bookko-market results may be driven by survival bias in the Compustat datafiles (see als0 Kothari, Shanken and Sloan (1995)). However, Chan, Jegadeesh and Lakonishok (1995) and Fama and French (1996) show that the selection biases are probably not very large. This brief survey suggests that the traditional CAPM-beta at least lacks efficiency and completeness as a measure of risk or even that there is no risk-return tradeoff when only beta is used as a risk measure. The persistent finding that other variables have superior power in explaining the cross-section of stock returns tends to corroborate the latter conclusion. Unfortunately, there are no firm theoretica1 guidelines as to what variables to include in a multi-factor asset pricing model. Furthermore, while it is not disputed tbat observed return premia miglit be explained by a factor model, a related issue is whether the factors can accurately represent economically relevant aggregate risk. Daniel and Titman (1997) argue that the return premia associated with portfolios formed on tbe basis of the factors identified by Fama and French (1993) cannot be interpreted as a compensation for factor risk, such as distress.
III. THE CONDITIONAL NATURE OF THE BETA-RETURN RELATIONSHIP Even if some of the recent evidence suggests that there is no compensation for beta, it does not imply that beta is completely useless for investment decisions. As argued by Chan and Lakonishok (1993), as long as beta behaves as a stable indicator of exposure to movements in the market, investors should still consider the sensitivity of stocks to the market factor. Risk is commonly defined as the chance that ex post realized returns differ from the expected ones. Hence, the dispersion or variance of return outcomes around their average values is generally used as a measure of risk. However, positive surprises (i.e., higher realized returns than expected) can hardly be considered to be unfavorable. In fact, risk-averse investors are primarib interested in minimizing downside risk. For that purpose they need a risk measure that reflects the exposure of financial assets to substantial drops in the market as a whole. Standard theory suggest that high-beta stocks tend to experience larger gains or losses than low-beta stocks relative to the market. FIGURE 2 The risk-return relation according to the CAPM conditional on market movements.
The question then becomes whether beta is an accurate measure of downside risk, regardless of the long-term relationship between beta and return. Conversely, when the market is bullish, investors would like their portfolios to perform accordingly. Again, the question is whether beta can be used as an indicator of upward potential. Figure 2 shows what the risk-return relation would look like if beta acts both as a measure of downside risk and upward potential. In order to test this hypothesis, a number of studies have conditioned the relation between beta and returns on the direction of market movements. The market model predicts an overall positive relationship between beta and expected returns. However, realized returns may substantially differ from expected returns. In fact, for high beta portfolios to have higher risk, there must be instances where realized returns for high beta portfolios wil1 be lower than those for low beta portfolios. This will especially be the case when the market exhibits a downward trend so that R, becomes negative. This reasoning suggests a segmentation of the relationship between beta and realized returns. It will be positive when the market moves upward and negative in periods of downward market movements. Gonsequently, a regression of realized returns on beta will be biased towards finding a flat relationship and cannoe provide a crucial test of the market model. Pettengill, Sundaram and Mathur (1995) test for a systematic conditional relationship between average realized returns and beta. The purpose is not to test market efficiency but ascertain whether or not the market beta has information value and is priced accordingly. A positive slope on beta is expected in periods with positive excess market returns, a negative relationship in periods with depressed market returns. Over the entire period 1926-1990, and in various subperiods, they find a significant conditional association between beta and returns. Bhardwaj and Brooks (1993) classify months as either bull or bear if the market return is higher or lower than the median return over the 1926-88 period. They find that systematic risks in bul1 and bear periods are statistically different. These results build on the conclusions of Fabozzi and Francis ((1977), (1979)) who already analyzed changes in risk (beta) and abnormal return (alpha) during bull and bear markets in the 1960s. The findings of the recent studies are in accordance with previous research which has documented that equity risk premiums vary with market conditions and business cycles, thus rejecting the constant risk assumption (see Ferson and Harvey (1991)).
Differential return premia between upward aiid downward markets have been attributed to varying systematic risk over the business cycle (Schwert (1989)) or transaction cost differences (Karpoff (1987)). IV. TESTING AND RESULTS A. The unconditional beta-return relationship
We test the unconditional and conditional relationship between beta and return on the Brussels Stock Exchange over the period 19901996. To be included in the sample domestic stocks must have a listing on the Belgian Stock market and be traded over the period January 1985 to December 1996. This condition implies that al1 included stocks had a publicly known continuous price during this period. The 64 stocks that fulfill this requirement represent 85% of the total market capitalization. Portfolios are formed by a monthly ranking of the sample based on the estimated historica1 betas. Portfolio betas are then calculated as an equally weighted mean of the individual betas. Betas are calculated over rolling time windows using monthly return observatioi~sfrom January 1985 to November 1996. For example, the individual 60-month betas for January 1990 are estimated by ordinary least squares using data from January 1985 to December 1989. Al1 stocks are ranked by their estimated beta in December 1989 and divided in four quartiles, whereby the equally weighted mean of each quartile is the portfolio beta (Q1 is the low beta portfolio LBP, Q4 the high beta portfolio HBP). The associated return is the percentage change in the portfolio return index (including dividends) between the last trading day of December 1989 and the last trading day of January 1990. Subseque~itly,the starting and ending months of the 60-month interval are rolled forward one month, etc. Consequently, the actual beta-return reiationship is tested over the entire 1990-1996 period. Betas are obtained from the market model:
where R, is the return on the i-th portfolio including dividends, Pi measures the sensitivity of portfolio returns relative to broad market movements, R, is the total return on the market index and ei is the error term. We estimate betas over two different time intervals and relative to two market indices. The first index is the widely used BEL-20 mar-
ket index. It contains the 20 blue chip Belgian companies with the largest market capitalization and highest trading volumes. The second market proxy is an equally weighted index of the 64 stocks in the sample (EWMI). The reason for the use of two alternative equity market indicators is the wel1 known problem of choosing the appropriate proxy for the true market portfolio that would minimize the estimation errors. Since the BEL-20, by construction, contains only the 20 most actively traded stocks, we also perform the estimations relative to the EWMI as a robustness check. Betas are calculated over a period of 24 months and 60 months. This produces four types of betas with a different history and relative to a different market portfolio. The two estimation windows for beta allow a distinction between the impact of the short and long-term sensitivityof stocks to market movements. The use of two market indices nlay partly reflect the return behavior of smal1 versus large stocks. The BEL20 index is a capitalization-weighted index, while smal1 stocks are better represented in the EWMI. First we examine the overall beta-average return relationship on the Brussels Stock Exchai~geby performiilg a broad test of the market model where the pi are obtained from equation 1for al1 the portfolios and b is the risk premium associated wit11 taking one unit of systematic risk in the Belgian stock market. We als0 look at the differences in return produced by the high beta portfolio (HBP) versus the low beta portfolio (LBP). TABLE 1 Estirnated coeficients of the market model (equation 2) for the four types ofbetn, 1990-1996. The significance ofthe estimated coefficients is nssessed with t-values. The probabiliiy of an estimated zero riskpremium (b) is given in the last column. a
P~~,B~Lzo
0.72
P24,EWMI P6o,nr-~zii
P60,EWM1
0.18
b
tial
t@)
RZ
p-value (b)
0.18
1.06
0.25
0.02%
0.80
0.64
0.31
0.92
0.25%
0.36
0.79
0.19
0.80
0.19%
0.43
0.076
0.79
0.08
0.002%
0.94
The overall test of the validity of the market model on the Brussels Stock Exchange for the period 1990-2996 is performed by estimating equation 2 for the full sample of portfolios. Stocks are ranked montlily by their beta over the period January 1985 to November 1996 for the 60-month betas and over the period January 1988 to November 1996 for the 24month betas. Betas are calculated relative to the BEL20 index or the EWMI. Quartile portfolios are formed based on the beta ranking. For each portfolio, equally weighted monthly returns are calculated. This yields a sample of 336 observations for each beta. The regression coefficients are listed in Table 1. For all beta types, the estimated slope coefficients b are not statistically different from zero. None of the t-statistics reveal a level of significance close to conventional levels. The four test values to examine wbether the nul1 hypothesis of b = 1holds are: 20.88; 9.47; 3.88 and 17.14 with a theoretica1 value of 1.97 at the 5% significance level. We are forced to conclude that the existence of a positive overall relationship between beta and return must be rejected for a representative sample of domestic stocks listed in the Belgian stock market. This is in line with the findings of Pettengil et al. (1995) who test the significance of the risk premium for US data for the period 1971-1990 and find a premium of 0.05% (t-value of 1.3), whicb is both statistically and economically insignificant. Mthough the estimated coefficients are insignificant, it can be observed that the values for the intercept and the dope vary with the use of different betas. The shifts in the regression lines are mainly due to differences in the variance of the quartile beta portfolios for the different beta specifications. This finding underlines that unconditional tests of the market model may be highly sensitive to the measurement of beta. A s o the lX2 vaPues for the different specifications differ markedly. Nevertheless, the conclusion is that the single-beta model fails to capture the behavior of returns on the Belgian stock market over the period 1990-96. Another way to assess the overall relation between beta and return is to look at the average return of stock portfolios with different betas. Again, stocks are pre-ranked by their beta relative to the BEL20 index or the EWMI. Each month equally weighted average returns are calculated for the four quartile portfolios (Q1 is low beta, Q4 high beta). The monthly returns for the four quartiles and the four types of beta are presented in Table 2. The last column tests the hypothesis that the realized return on the HBP and the LBP are significantly different. Based on the t-values the hypothesis that the returns on the
HBP and the EBI? are equal cannot be rejected. None of the t-statistics for the four beta types is significant. The Same conclusion holds for a painvise comparisoil of al1 quartile portfolio returns.
Average qa~artileportfolio returns over fl?eperiod 1990-1996. The portfolios are formed by ranking the stocks moiztlzly according to their beta. The retunz is the equally weighted meun of tize retzirnsproduced by ench portfolio based on a oize-rnonth holdingperiod (in %). Next to the average inonthly return for each quartile poi-tfolio, the result of the t-te$t that ineasures the differente in return between guartile 1 and quartile 4 is given in the last column.
These results imply that there is no positive linear relation between beta and portfolio stock returns on the Brussels Stock Exchange over the period 1990-96. Moreover, Table 2 suggests that the relationship between beta and return is not even monotonic for any of the beta types. For the 60-month beta calculated relative to the BEL20 index, a measure widely used by Belgian portfolio managers, both Q2 and Q3 seem to produce at least as high returns as Q4. B. The conditional beta-return relationship 1. C o n d i t i o n e d o n t h e marlcet e v o l u t i o n
In the second part the tests of the market model are conditioned on the evolution of the market index. We use two broad definitions of up and down markets. The first, statistica1 method is based on the series of monthly returns of the equally weighted market index. The weak criterion defines markets as upward or downward depending on whether the monthly market return is larger or smaller than zero. This produces 46 up (R, > 0) and 38 down market months (R, 4 0). In the strongest case, we define upward markets as those months in which the EWMI-return exceeds the average value of al1 positive market returns. Downward markets are those in which the return on the market index falls below the average of all negative market returns. This division produces a broad neutra1 interval in which the market is clas-
sified as neither bullish nor bearish. The application of this procedure gives 18 months in which the market can be classified as substantially upward moving and 13 substantial bear months, out of 84 market returns in the sample. In order to check for the robustness of this classification, we allow the neutral inteilral to vary. Both the positive and the negative bounds are lowered with 0.5 and 0.75 times the standard deviation of the positive and negative market returns, respectively. Table 3 lists the intervalszd the resulting classification of u p and downaarkets. In the table, Rn, is the average of positive market returns, R , is the average of negative market returns, while o, and o, are the standard deviations of the positive and negative market returns, respectively.
Classification o f ~ and q down markets on the Brussels Stock Exchange over the peiiod 1990-1996 based on the evolutioiz of returns on the equally weighted inarket index.
The performance of beta as an overall risk measure is evaluated in upward and downward markets by estimating the following equation
where 6 = 1in upward markets and 6 = 0 in downward markets (as in Table 3). This regression uses al1 available data but partitions the estimated coefficients according to the market evolution. In up markets y, is the relevant coefficient, y, in down markets. The hypothesized sign of y, is positive since we expect portfolios with higher betas to outperform low beta portfolios in bul1 market conditions.
TABLE 4 Estimated coeficients for the maïket model, 1990-96, for vatying definitions of upward und dowìzward inarkets and for the fouï ypes of beta. The t-values of tlze estimated coefficients are reported below the coejjïcieìzt estimates.
TABLE 5 Quartile portfolio retur-rtsin upward and downward markets, 1990-96. PanelA Average quartile portfolio returns in upward and downward markets for the four types of defined betas. The portfolio returns are the equally weighted returns produced by each portfolio based on a one-month holdingperiod.
This would imply that in upward markets the systematic risk is rewarded and y, is the reward (risk premium) per unit of risk. Hence the testable hypothesis is H,: y,=O versus H,: y, > O. Since y, is estimated in periods of downward market conditions, the expected sign of this coefficient is negative. The nul1 hypothesis to be tested is H,: y2=0 versus H,: y, i O. The equation isestimated for al1 definitions of up and down markets and for the four betas(see Table 4). Applying this test with 24 and 60-month betas allows US to examine the predictive power of the short versus long-term market sensitivity. The performance of beta as an overall risk measure is als0 evaluated by investigating the return pattern of the quartile portfolios (see Table 5). In order to assess the power of beta to serve as a useful measure of market risk, we test whether the average returns produced by the HBP and the LBP are significantly different. The hypothesis is that the HBP wil1 produce significantly higher returns than the LBP in upward markets and substantially lower returns in bear market conditions. TABLE 5 Quartzle portfolio returns in upward and downward markets, 1990-96. Panel B Percentage spread between the returns on the HBP and the LBP The dfference in meun monthly return between the high beta and the low beta portfolio w given as well as the value for the t-test measuring its statzstical significance (between brackets). *; **; *** indicate significance at the 10%, 5%, 1% level.
Al1 estimated coefficients y, and y, have the expected sign and are highly significant over al1 four definitions of up and down markets and for the four specifications of beta. The adjusted R 2values indicate that the conditional estimations provide superior explanatory power over the unconditional test of the market model. The significant positive coefficients on y, suggest that portfolios with higher betas are indeed associated with substantially higher returns in upward markets. This result holds for both beta estiination intervals and both market proxies. Similarly, the y, estimates indicate that beta is a consistent measure of downside risk, since high beta portfolios experience larger negative returns than low beta portfolios in downward markets. The conclusion is that there is a strong conditional relation between beta and return on the Brussels Stock Exchange over the period 1990-96.Again, this is in iine with the findings for the US data (Pettengil et al., 1995). In the period 1971-1990they find a premium of 3.92% (t-value = 8.23) in up markets and -3.09% (t-value = -7.46) in down markets. Thus, beta is found to behave as a consistent indicator of both upward potential and downside risk. The estimated slope coefficients and the R 2 statistics tend to increase as the classification of up and down markets becomes more pronounced. This suggests that the conditional betaireturn relationship becomes stronger for more pronounced stock market evolutions. The estimated intercepts are not significant, except for the estimations with p, in the most extreme classification of up and down markets. This result is probably due to the fact that only very substantial positive and negative return months are included in this sample. The results of the nested regression model thus suggest that investors may be able to improve the riskireturn mix of their portfolios by conditioning the portfolio beta on the expected market movement. When the rnarket is expected to rise, returns can be improved by investing in high-beta stocks. When the investor anticipates a market decline the downward risk can be minirnized by shifting to low-beta portjolios. Since we do not use an explicit model of market expectations, we are unable to test this prediction directly. We can only examine the pattern of realized returns of beta-ranked portfolios in different market conditions. Panel A of table 5 reports the average returns for the quartile portfolios formed on pre-ranked betas for the four definitions of upward and downward markets. Panel B reports the spreads between the return of the HBP and the LBP of panel A and the associated tstatistics. The results in panel A show that the evolution of returns on
the quartile portfolios ranked on different betas is mainly monotonic. In the 16 beta-return rankings in down markets, 4 reversals occur, but only behveen Q1 and Q2, i.e. the portfolios with the lowest betas. In upward markets, one reversal occurs between Q1 and Q2 and two between the higher-beta portfolios Q3 and Q4. It is striking that five of the seven return reversals occur for the P,,,, BEL2". Apparently, this widely used risk measure is unable to capture the systematic risk of stocks in a consistent way. Panel B shows the differences between Q1 and Q4 returns and their significance. For the broad definitions of upward markets, the return differences are generally significant shows a relatively poor (bottom half of panel B). Again, the P,, performance in distinguishing the return potential of high versus lowbeta portfolios. For the strong specification of upward markets, the significance is weaker. In these cases, the number of observarions is probably too smal1 to allow a strong dichotomization. The significance also becomes weaker when we move from the second broadest (0.750) to the widest definition of market movements (R, > 0). This finding results from the smoothening influence of the close-tozero market returns. In the case of downward markets the results are generally weaker. The differences between returns for the HBP and the LBP are only significant for a limited number of cases and for broad definitions of down markets. These results suggest that the discrimination based on portfolio returns may work better in upward markets than in downward markets, although it has to be stressed that the analysis uses ex post realized average portfolio returns. Combined with the very strong regression results based on the whole sample, the above findings imply that beta may still sewe as a useful indicator of conditional market risk.
,,,,,
2. C o n d i t i o n e d o n t h e r e t u r n of a r i s k - f r e e investment A second way to distinguish bul1 and bear market conditions is based on the comparison of stock market returns with the return on a riskfree investment. We define periods of relative stock market attractiveness as those cases where the risk premium earned in the stock market exceeds the term spread on risk free investments. Excess stock market return can be defined as the return on the EWMI minus the return on Belgian three-month treasury bills. The term spread is the difference between the 10-year government bond yield and the trea-
sury bil1 return. Consequently, the criterion for classifying markets as up or down reduces to a comparison of the monthly stock market return with the government bond yield. This procedure has intuitive appeal to investors, because it confi-onts the expected stock market performance with the alternative yield on risk free bonds. When tlie stock market return exceeds the yield on bonds, investors have an incentive to invest in stocks. On the other hand, when risk free bonds are expected to yield more than stocks, investors will prefer the riskless investment. Moreover, this classification procedure als0 captures the influence of business cycle conditions and inflation expectations on stock markets. Fama and French (1989) find that the term spread tracks a time-varying premium in stock returns similar to that in long-term bond returns. Moreover, the term spread is related to variation in expecteci returns in response to short-term variation in business cycle conditions. This classification procedure yields 35 cases where the monthly realized risk premium on the stock market exceeds the rislrLess yielld spread and 49 cases where the stock market premium is lower than the term spread. Again, we estimate equation 3 with 6= l when R,-R,> R,-R, and 6=0 in the reverse cases. The results appear in Table 6. The results are similar to those in Table 5. As expected, the y, coefficient is positive and significant for al1 four definitions of beta, while y, carries a consistent negative sign. Furthermore, the alphas are al1 insignificant and the estimated conditional relationship explains roughly half of the observed variation in realized returns. These findings show that beta is a solid measure of upward potential in periods where the stock market outperforms the bond market. Similarly, beta serves as a useful indicator of downside risk in the stock market in periods where the expected yields on bonds will induce investors to invest in risk-free securities. Hence, investors may improve the performance of their portfolios by conditioning the asset selection on the anticipated market movement based on expected business cycle conditions.
TABLE 6 Estimated coefjîcients for the rnarket model, 1990-96 for vaqing sstates of stock nzarket nttractiveness based o n a conzparison of tlze stock rnarkelpretnizmz with the yield spr~nd.T-statistics for the estinzated coeflicients are @ven below the coeficient estiinates.
Similar to the previous exercise, we report the returns on betaranked quartile portfolios and we test whether the differences between the realized returns on the HBP and the EBP are significantly different. Table 7 contains the basic findings. The results generally behave as exgected since the monthly returns increase in absolute value as the portfolio risk becomes higher, moving from Q1 to Q4. The numbers largely confirm the findings in Table 6. Panel B suggests that returns can be improved significantly, especially by focusing on shortterm betas in those cases where the stock market outperforms the bond market. The 60-month beta calculated relative to the BEL20 again performs poorly. This is an important finding because this measure is often treated as the preferred risk parameter in the asset management practice. TABLE 7 Portfolio returns in vaïying states of stock market attractiveness Panel A Average quartile portfolio returns in upward and downwai-d markets classified o n the basis ofyield spreads, 1990-1996. The portfolio return is the equally weighted meun of the returns produced by each portfolio based o n one-month holdingperiods and is given for al1 defined beta measures.
TABLE 7 Portfolio returns in valyiizg states of stock market attractiveness Panel B Percentage spread between returns ort the HBP nrzd the LBI) The diflerenre in rizean nzonthly return betweerz the high beta and the low beta portfolio is given for al1 defirzed beta nzeasures as wel1 as tize statistical significarzce ofthe dijference and its t-value (in brackets), *., ""., " e " .. signijïcant at 10%, S%, 1%
3. V a r i a n c e p r o p e r t i e s of t h e b e t a p o r t f o l i o s We examine the theoretica1 expectation that the variance of highbeta portfolios is larger than that of low-beta portfolios by performing a variance ratio test. The variance of the returns on beta-ranked portfolios for the various definitions of market conditions is calculated relative to the mean portfolio return. The difference between the dispersion of returns on the HBP versus the LBP can be tested since the ratio of the variances is F-distributed. Table 9 shows that the variante ratio test is almost consistently significant in declining markets. Apparently the volatility of returns seems to be more pronounced in adverse market conditions. In upward markets the test is only significant for the more narrow definitions of market swings. As in previous tests, the results based on the P,, B,L2, are always insignificant. Thus, we find evidence that HBP effectively displays a higher variante than LBP. However, the calculation of the variances may be biased because the mean returns in upward markets are higher in absolute value than the mean returns in downward markets. Therefore we also report the results of a variance ratio test over the whole 1990-96 period. In this case the variance is calculated as the mean of the distance between the individual return on the portfolio and the average EWMI return. Based on the test values in Table 10, the hypothesis of equal variances for HBP and LBP can be rejected, except for the ,P, BELZO specification. Thus, the variance ratio test is not biased by asymmetric sample means in upward and downward markets.
TABLE 8 F-values for the hypothesis H ,e?c = u2,?, The reported nulnbers are the values for the F-test measuring the difference in vanance for the low beta and the high beta portfolio. **, ***: significant at 5%, 1
TABLE 9 F-values for the hypothesis H, d ,,,,,=,'G for the entire sample penod. The F-test is done cornparing the vanance ofthe low beta and the high beta portfolio for the period 1990-1996. The significante is given by itsp-value
4. T h e extreme values Finally, by way of illustration, we show the behavior of beta-ranked portfolio returns in the 10 most extreme up and down markets. Here, extreme up and down months are simply the 10 largest absolute market increases and the 10 biggest declines, measured by the variation in the return on the EWMI. The return behavior of high-beta and lowbeta portfolios are illustrated using 24-month betas relative to both the BEL20 index and the EWMI. Figure 3 and Figure 4 show the results. FIGURE 3 Return on the Low beta portfolio and the High betaportfolio in the 10 most extreme up nzarkets, betas are calculated using a 24 month horizon.
FIGURE 4 Return on the Low beta porlfolio and the High beta portfolio in he 10 most extreme down markets, betas are calculated using a 24 month horizon.
On the X-axis, the up and down markets are ranked according to the absolute value of the market variation in that month. Figure 3 shows that there are two cases of substantial up markets (July 1993 and January 1996) where the WBP is outperformed by the LBP. In al1 other cases, the HBP tends to produce higher returns. In down markets, Figure 4 shows only one reversal (September 1991). In the other months the WBPs exhibit inferior returns relative to the LBP. Visual inspection learns that the returns on the HBP are roughly proportionate to the magnitude of the associated market movements.
V. CONCLUSION
Recent empirica1 evidence has raised serious doubts about the ability of market model betas to explain the cross-section of stock returns. Yet, the inherently conditional nature of the beta-return relationship suggests that it may be useful to link the risk-return tradeoff to the underlying market evolution. The question is whether beta remains a consistent risk indicator over varying market conditions. This paper analyzes the unconditional and the conditional relationship between stock returns and beta on the Brussels Stock Exchange over the period 1990-1996. Consistent with findings for other countries, notably
Fama and French (1992), we find that traditional betas are unable to explain the cross-section of observed stock returns over the sample period. This conclusion is robust for varying definitions of beta, using different estimation windows and different market indices. This would imply that beta is an incomplete measure of market risk and that its use as a systematic risk measure for asset selection purposes may add little value. However, when we test a nested regression model conditioned on the occurrence of upward and downward markets, the beta factor turns out to be a strong and consistent indicator of both upward potential in bul1 markets and downside risk in bear markets. We find that high beta portfolios produce significantly higher returns than low beta portf o l i o ~in up markets and substantially lower returns in down markets. Moreover, these findings are robust for various definitions of beta and different specifications of upward and downward markets. The analysis of ex post realized returns on beta-ranked portfolios produces similar conclusions, although the significance is generally weaker. We also look at the difference in the variance of the two extreme portfolios Q1 and Q4. Theoretically, the variance of the returns on the high beta portfolio should be larger because Q4 is expected to yield both higher returns in upward markets and lower returns in downward markets. Finally, we present the results for the 10 most extreme down and up market months, similar to Chan and Lakonishok (1993) and Grundy and Malkiel (1996). Overall, the results suggest that investors can improve the performance of their portfolios by using betas conditional on the market evolution in their asset selection practice (styling and risk exposure). Hence, the power of beta as a reliable asset allocation t001 depends on the ability of investors to anticipate future stock market movements, e.g. based on macroeconomic fundamentals. Another practical implication is that investors should still be able to take or avoid systematic risk based on betas in their portfolio allocation. This is important because previous research suggested that beta was no longer useful in this kind of decisions. Betting on low beta stocks still means taking a lower exposure to downside risk. However, the sesults show that following a low-beta strategy does not necessarily lead to lower returns, as suggested by conventional portfolio theory. Hence, a betabased strategy can be useful in limiting the risk exposure (practitioners often use the tracking error) without losing return opportunities. An important research question remains how other variables that
have been found to explain cross-sectional asset returns behave in a conditional setting. APPENDIX Calculation of the bourids and variances for upward and downward markets 84
C Rmi X Fi
R- =
i='
and R,: =
84
x
Fi i =l 2
G,
" CFi-(Ri-Rn)'
1
=
CF,
x 84
'=l
Rmi X (1 - F!)
,
X ( 1 - Fi) ik1
-
2
XL
'=l
i=1
I= 1
1
=-
1
with R n = 8 4 . x F I . R ~
Ik1
OP
with Fi = 1 if R,; < O ; Fi= O otherwise.
C (I - Fi) ik1
"
C ( l - F i ) . ( ~ t - % ) ~ with '=l
l , R P = ~ . ~ ( I - F ~ ) . F I ~ C ( 1 - Fi) '=' ik1
REFERENCES Banz, R.W., 1981, The Relationship Between Return and Market Value of Common Stocks, Journal of Financial Economics, March, 3-18. Basu, S., 1983, The Relationship Between Earnings Yield, Market Value, and Return for NYSE Common Stocks, Journal of Financial Economics, 12,126-156. Bhandari, L.C., 1988, DebtIEquity Ratio and Expected Common Stock Returns: Empirica1 Evidence, Jo~~rnal of Finaizce, June, 507-528. Bhardwaj, R.K. and Brooks, L.D., 1993, Dual Betas from Bull and Bear Markets: Reversal of the Size Effect, Journal of Financial Research, Winter, 269-283. Black, F., 1972, Capital Market Equilibrium with Restricted Borrowing, Journal of Business, July, 444-455. Black, F., 1993, Beta and Return, Journnl of Porffolio Management, Fall, 8-18. Black, E, Jensen, M.C. and Scholes, M., 1972, The Capital Asset Pricing Model: Some Empirical Tests, in Jensen, M., ed., Studies in the Tlieory of Capital Markets, (New York, Praeger), 79-121. Breen, W.J. and Korajczyk, R.A., 1993, On Selection Biases in Book-to-Market Based Tests of Asset Pricing Models, Northwestern Universi& Workingpaper 167. Chan, L.K.C. and Lakonishok, J., 1993, Are the Reports of Beta's Death Premature?, Journa1 of Portfolio Managemenl, Summer, 51-62. Clian, L.K.C., Hamao, Y. and Lakonishok, J., 1991, Fundamental and Stock Returns in Japan, Journal of Finance, December, 1739-1764. Chan, L.K.C., Jegadeesh, N. and Lakonishok, J., 1995, Issues in Evaluating the Performance of Value Versus Glamour Stocks: Selection Bias, Risk Adjustment and Data Snooping, Journal of Financial Economics, 38,269-296. Chen, N., Roll, R. and Ross, S., 1986, Economic Forces and the Stock Market, Jo~lrnaloj' Business, 59, 383-404.
Daniel, K. aiid Titinan, S.; 1997. Evidcnce oii the Characteristics of Cross Sectional Varia!ion in Stock Returns, Jolir.irni oJ'Firiance. March, 1-33. i'ahozzi, EJ. anci Fraiicis, J.C., 1977. Stability Tests for Aiphns 2nd Bctas over Bull and Bear Market Conditioiis, .Tour?inl c![Fiiia:ice, Sel?tcmbeï. 10513-1099. Wina. E.F. and McBetb, J.D., 1973, Risic, Return aild Ecluilibriuiil: Empricial Tests, .Torli.nul ofPoliticn1 Ecorzornj~,Mabr-Juiie. 607-636. Wnia. E.F. aild French, K.R., 1989, Business Conditioiis anci Expected Returi~so11 Stocks and Boncls, 5oz~nzfi1ofFinnricin1 Econoi.iiics, 25, Novcinber, 23-49. Fania, E.F. ancl Frciicli, K.K., 1997, Tlic Cross-Sectio11oi Expectcd Stock Returils, Jou~nal of Finnizce, Jurie, 477-465. Faii~a,E.E and French, IC.R., 1993, Coinmoi~Risk Factors in the Ret~irnso11 Bonds ;iiiiI Stoclcs, Joilninl of Fii~a:icinlEcorroirzics, February, 3-56. Fama, E.F. aiid Freiicti, K.R., 1996, Multiiactoi- Explanations of Asset Priciiig Anomaljes, .Tor~ïiiulof Finniice, March, 55-84. Faina, E.17.and Frcnch, lC.R., 1996, She CAPM is Wantcd, Dead or A.live,.To~tn~alofliii~nirce, December, 1947-1958. Ferson, W.E. and Harvey, C.R., L991; The Variation o T Econoinic Risk Prciniums, Joi!ïnnl oflJolitical E C O I Z O M 99, ~ .385-415. ~, Francis, J.C. aild Fabozzi, EJ.. 1979, The Effects of Cilanging Ivfacroecononiic Coilditions on ihe Parameters of tlie Single index Markei Model, .Toz~rnaio/'Firluncial nrrd Qzlniititntii~eAaalysis, June, 551-360. Grundy, K. and Mallciel, B.G., 1996, Reports oSBcta's Yeath Have Been Grcatly Enaggerated, .To~!rrralqf.Pol;rlolio iMLlizugenient. Spring, 36-44. Haugen, R.A. ancl Baker, N.L., 1996, Gommoi~alityin thc Determinailts of Expected Stock Returns, .Toi~innlofFinnncin1 Econoniics, 41,40 1-439. Jagannathan, R. and Wang, Z., 1996, The Conditional CAPM and tlie Cross-Scction of Expected Returns, J ~ L L U oJ'Finnrice, IUI Evfarch, 3-53. Karpoff, J.M., 1987, The Relation Betwecn Price Changes aiid Trading Voluine: a Survey, Jozirizni ofFiriatrcia1 nr~dQl~antitaliveAizniysis, 22, 109-126. ICothari, S.P., Shanken, J. and Sloan, R.G., 1995, Anothcr Look at the Cross-Section of Expected Stock Returils, .Toiri.~inlq[Fiiziiizce, March, 185-224. Eakonisliok, J., Shleifcr, A. and Vishny, R.W., 1994, Contraria11 Investn~eilt,Entrapolation, and Risk, .Tor~r.t~al of Finunce, 49, 1541-1578. Lintner, J., 1965, The Valuation of Risk Assets and lhc Sclectiori of Rislcy Investme~ilsin Stock Portfolios and Capital Budgets, Review oJ-'Econornics nrzcl Statislics, February, 13-37. MacICinlay, A.C., 1995, Multifactor Models Do Not Explain Deviations from the CAPM, Jour.na1 oJ'Fir~~rizciul Ecoizo17zics,35, 3-78. Merton, R.C., 1973, An Iiiterternporal Capital Asset Pricing Model, Econometrictl, 41, 867-887. Miles, D. a i ~ dTiinmerinann, A., 1996, Variati011in Expected Stock Returns: Evidc~icco11 the Pricirig of Equities from a Cross-Scction oT WIC Cornpanies, Ecorzor?iica,63,369-382. Pettengill, G.N., Sui~daram,S. and Mathur, I., 1995, The Conditional Relatioii Behvceri Beta aiid Returns, Joz~nzulo f Fiiiancirrl and Qiluntiti~riveiliicilysis, March, 101- i 10. Koscnberg, B., Rcid, IC. and Lanstein, R., 1985, Persuasive Evidc~iceof Markct IiiefSicicncy, .Jobii-~~(zl of Poi./fi>liijMonogement, 11, 9-1 7. Ross, S.A., 1976, Thc Arbitrage Theory of Capitai Assct Pricing,.101~uiol of Econoiiric TJleoi);, 13,341-350. Schwert, G.W., 1989, Why Does Stock Markcl TY'ol~iiiiity Change over Siiile, .To~~iïiul (>J' Finunce, 44, 1115-1153. Sharpe, W. F., 1964, Capita1 Asset Priccs: a Thcory of Markct Equilibrium Under Coiiditions of Risk, .Ìoui.tinl of Firzaizce, Novembcr, 237-268. Stattnian, D,, 1980. Boek Values a ~ i dStock Returns, Tlze Cliicngo&IBU1:n.ioi!~nnloJ'Selecied l'apers, *, 75-45.
Tijdschrift voor Economie en Management Vol. XLV, 2,2000
Earnings Management and Institutional Diffwuences: Literatare Review and Discussion by H. VANDER BAUWHEDE" a n d M. WILLEKENLI*
This paper provides (1) a review of the empiricali earnings management literature, (2) a discussion of the institutional differences between Anglo-Saxon and non Anglo-Saxon countries and (3) an assessment of the impact of institutional differences on earnings management. As (accounting) earnings can be considered an important summary statistic of a firin's financial performance, one can question whether managers do not "manage9' those earnings. Ht is elear that financial, investment and operational decisions can Pnfluence earnings. However, accounting decisions too can be used to manage earnings in a particular direction, for GAAP leave some discretion to managers in reporting the financial position and operating results of their organizatisn. Examples of accounting decisions that can influence earnings are accrual decisions, accountinggroceduire choices aizdclianges, timing of adoptiolm of a mamdated accou~itingchange, and the l i k . Earnings management has been researched in the literahre both analytically and empiricaiily.Analytica1 models (see, for example, Lambert (19846, Demksi et al. (1984), Verreechic! (1986), Dye (1988), Triaeman and Titman (1988), Fudenberg end Tirole (1995), Evans and Sridhar (1996)) study conditions under which earnings management
' Departrnent of Applied Economics, K.U Lcuvcn, Leuven. We gratetully acknowledge the anonymous reviewer for valuable coininents on a previous version of this paper.
can occur, whereas empirical studies document instances of earnings management (Schipper (1989)). In this paper, we limit ourselves to a review of the empirical earnings management literature. The review provides some evidence that managers have incentives to manage earnings, that they do actually engage in earnings management and that there are factors that constrain their ability to manage earnings. Some studies report also on the consequences of actual or assumed earnings management. Most empirical studies were performed in Anglo-Saxon countries. Since there exist institutional differences between Anglo-Saxon and non Anglo-Saxon countries, we argue that the factors that create incentives for and constraints on earnings management may be different for these environments, and some results of Anglo-Saxon studies may not hold in non Anglo- Saxon countries. The remainder of the paper is organized as follows. The second section presents a literature review. The third section discusses the institutional differences between Anglo-Saxon and non Anglo-Saxon countries. An assessment of their viable impact on factors that create incentives and constraints on earnings management is provided in section four. We conclude with a summary.
11. LITERATURE REVIEW Empirica1 studies mainly report on the incentives to manage earnings. Some recent empirical studies also examine factors that might constrain earnings management. The following paragraphs give an overview of (1)the incentives to manage earnings, (2) the factors that might constrain and (3) the consequences of earnings management. A. Incentives Incentives to manage earnings may stem from the existence of explicit and implicit contracts, a firm's relations with capita1 markets, the need for external financing, the politica1 and regulatory process or several specific circumstances. A firm may also be induced to engage in a specific type of earnings management, i.e. income smoothing.
1. Explicit and implicit contracts A first category of incentives examined stems from the contracting literature. In the early literature it is often argued that earnings management is induced by the existence of explicit contracts, for example bonus plans and debt covenants. The recent literature, however, also focuses on inaplicit contracts, as a source of earnings management incentives. The main concern that "drives" these studies, is that, although the flexibility provided by GAAP can improve efficient contracting, managers might use this flexibility to act opportunistically. A first well established hypothesis in the empirical financial accounting literature is that bonus plans based on accounting earnings might induce managers to manage (manipulate) earnings through their accounting procedure and accrual decisions, in order to increase their cash compensation. More specifically, early studies test the bonus plan hypothesis, which states that "Ceteris paribus, managers of firms with bonus plans are more likely to choose accounting procedures that shift reported earnings from future periods to the current period" (Watts and Zimmerman (1986), p.208). Later studies examined related and more refined hypotheses using different measures of earnings management and using other datasets. Examples of studies on the bonus plan hypothesis include Healy (1985), McNichols and Wilsoii (1988), Gaver et al. (1995), Holthausen et al. (1995) and Dechow et al. (1996). Although the early literature finds evidence that is generally consistent with the bonus plan hypothesis (Watts and ZimInerman (19901, p.138), more recent studies document mixed results (that is, results differ when other datasets and measures of earning management are used). This comes as no surprise, as Watts and Zimmermaii ((1990), p.139) argue that the early tests are not very powerful, because they rely on simplifications of the (contracting) theory. Another well established hypothesis in the contracting literature is that the (1) existence of andior (2) closeness to debt covenants might induce managers to manipulate accounting earnings. Early empirica1 studies ihat test the covenant based hypothesis use leverage as a proxy for (1) the existence of andl or (2) closeness to debt covenants. In fact, they test what is called the debtiequity hypothesis. More specifically, it is hypothesized that "Ceteris paribus, the larger a firm's debtlequity ratio, the more likely the firm's manager is to select accounting procedures that shift reported earnings from future periods to the current period." (Watts and Zimmerman (1986), p.216). Later studies
provide "direct" evidence of the covenant based hypothesis. DeFond and Jiainbalvo (1994) and Sweeney (19941, for example, contribute to the literature because they do not use a proxy, but base their study of the covenant based hypothesis on a sample of firms that actually reported to haveviolated debt covenants. Other studies include: Dechow et al. (1996) and DeAngelo (1994). Except for the DeAngelo (1994) study, both the early accounting choice studies, which use proxies, and more recent research, which circumvents the use of such proxies, document (in general) evidence consistent with the covenant based hygothesis. The above evidence suggests that explicit contracts have an impact on earnings management throiigh accounting decisions. Bowen et al. (1995) and Kasanen et al. (1996) examine whether implicit contracts (tl-iai is, for example, implicit contracts between the firm and its customers, suppliers, short-term creditors, employees, capita1 providers and other stakeholders) do also influence managers' accounting decisions. Both studies find evidence which is consistent with implicit contracts inducing earnings management. Moreover, Bowen et al. (1995) find that implicit contracts can explain cross-sectional variante in firms' accounting procedure choices in addition to traditional variables, for example, the proxies for the explicit contracts mentioned above, that is bonus plans and debt covenants. However, the literature on implicit contracting incentives is rather new and Limited, and so it seems too early to draw general conclusions. 2. C a p i t a l m a r k e t s a n d n e e d f o r e x t e r n a l f i n a n c i n g
The contracting perspective is not the only way to look at management's accounting decisions. An alternative is to address those decisions from a capita1 market perspective (see, for example, Healy and Palepu (1993)). In general, research findings provide empirica1 evidence that a firm's relation with capita1markets can create incentives to influence earnings. Some of those studies expect firms to manage earnings opportunistically (Shivakumar (1998), Dechow et al. (1996), Rangan (1998) and Teoh et al. (1998)). Other studies consider that earnings may be managed in order to communicate private information to investors (Subramanyam (1996), Neill et al. (1995)). Subramanyam (1996), for example, finds supportive evidence of the hypothesis that managers use discretionary accruals to communicate information about future profitability, that is the economic value of the
firm. Also the findings of Neill et al. (1995) suggest that accounting method choice can signa1 firm value. Dechow et al. (1996) test whether managers manipulate earnings to influence investor's perception of firm value in order to be able (1) to raise additional financing on more favorable terms or (2) to sell their stockholdings for a higher price. They find that the need for external financing seems to be an important motive to explain managers engaging in earnings manipulation, whereas insider trading seems to be less important. Some recent studies find that earnings are managed prior to or around initia1 public offerings (Friedlan (1994), Aharony et al. (1993), Neill et al. (1995)) or seasoned equity offerings (Shivakumar (1998), Rangan (1998), Teoh et al. (1998)). 3. P o l i t i c a l a n d r e g u l a t o r y p r o c e s s The politica1 and regulatory process through, for example, taxes, rate regulation and investigations by regulatory agencies may als0 create incentives to manage earnings (see, for example, Watts and Zimmerman (1986)). A hypothesis often tested in accounting choice studies is thepolitical cast hypothesisl size hypothesis, which states that "Ceteris paribus, the larger the firm, the more likely the manager is to choose accounting procedures that defer reported earnings from current to future periods" (Watts and Zimmerman (1986), p.235). Note that formulating the hypothesized impact of the politica1 and regulatory process this way, assumes that larger firms are more political sensitive than smaller firms. Recent research on the impact of the political and regulatory process on earnings management includes Jones (1991), Guenther (1994), Bowen et al. (1995), Hunt et al. (1996), Key (1997) and Han and Wang (1998). In general, one can conclude that prior research provides evidence consistent with earnings management induced by the politica1 or regulatory processes. 4. Specific circumstances
There exist also some studies that examine whether specific circumstances induce earnings management. Liberty and immerm man (1986), for example, study the impact of labor union contract negotiations on managers' accounting decisions, but do, however, not find evidence of earnings management. And DeAngelo (1988) examined whether proxy contests create incentives for managers to influence earnings and finds evidence consistent with this hypothesis. In particular, it
seems that, during the election carnpaign, managers use their accounting discretion to create a favorable picture of their performance. Finally, a recent study (Burgstahler and Dichev (1997)) finds evidence consistent with the hypothesis that earnings are managed to avoid earnings decreases and losses.
5. Bncome s m o o t h i n g Some studies examine a specific type of earnings managemenit, i.e. income smoothing. The hypothesis is that managers might manage earnings to reduce the variability of reported earnings or to align reporteed earnings with expected earnings. Explanations for income smoothing inclerde job protection and avoidance of shareholder Pnterference, tax avoidance, irnproving terrns of trade aild pursuing a fixed dividend pay-out ratio. Studies include Eckel(1981), McNichols and Wilson (1988), Hunt et al. (19961, Subramanyam (1996), DeFond and Park (1997), Youi~g(1998). The evidence on the ineoine smoothing hypothesis is however mixed.
From the above review it is clear that untiB now empirica1 research has mainly focused on tlie incentives to manage earnings. Recent research also examinesfactors that vnight constrain eariiings management. These include prior accounting decisions, ownership structure, strength of the internal controls, internal governance (especially the existence of an audit committee and some characteristics of the board of directors, such as, for example, board size, board composition and separation of the fernctions of Chairman of the Board of Directors and @hief Executive Officer) and quality of the external audit. The impact of prior accounting decisions on earriings management was studied by Depond and Jiambalivo (1991). They found that earniamgs "manipulation" is more likely in firrns where the rernaining number of income-increasing G U P possibilities is smalicr. Moreover, Sweeney (1994), argues that the accounting flexibility available to managers, which is iniluenced by their prior accounting decisions, is an important determinant of managers9accoernting responses to debt covenant violation. The impact of ownership structure on earnings management was examined by DelFsnd and Siarnbalvo (1991), Warfield et al. (1995) and Rajgopal and Venkatachalam (2998). Research results suggest a negative
relation between (1) managerial ownership (Warfield et al. (1995)) or institutional ownership (Rajgopal and Venkatachalam (1998)) and (2) the magnitude of accounting accrual adjustments ( = a measure for the extent of earnings management). Earnings "manipulation" was, however, found more likely in firms with diffuse ownership (DeFond and Jiambalvo (1991)). Further, results indicate that earnings overstatements are less likely in firms that have an audit committee (DeFond and Jiambalvo (1991)). However, no evidence was found that stronger internal controls reduce the likelihood of earnings overstatement (DeFond and Jiambalvo (1991)). Studies on the impact of the characteristics of the board of directors on earnings management (Dechow et al. (1996), Peasnell et al. (1998) and Beasley et al. (1996)) find that at least some of those characteristics are related to earnings management. Further, several studies examine whether a higher quality audit constrains (opportunistic) earnings management more than a lower quality audit. The impact of audit quality on both GAAP as wel1 as nonGAAP earnings management was tested. Evidence was found of a relation between audit quality and earnings management within GAAP. In particular, Becker et al. (1998) found that discretionary accruals for clients of non-big6 auditors are higher than for clients of big6 auditors. Francis et al. (1997) found that the amounts of discretionary accruals of big6 client firms are lower than those of non-big6 client firms. The evidence on the relation between audit quality and nonGAAP earnings management is, however, mixed. Dechow et al. (1996), do not find that the use of a Big6 auditor differs between firms that were subject to enforcement actions by the SEC (that is, firms that are "suspectedn to have engaged in non-GAAP earnings management) and the control firms. Higher audit quality was, however, found to reduce the likelihood of errors and irregularities (DeFond and Jiambalvo (1991)) and to increase the likelihood of auditor-client disagreements over income-increasing accounting methods (DeFond and Jiambalvo (1993)). Finally, Shivakurnar (1998) finds some evidence which, he argues, is consistent with auditors limiting managerial discretion over accounting procedures.
Soine studies do not only examine the factors that induce or constrain earnings management, but study also the consequences of actual or assurned earnings management. Thls research is, at the hnoment, limited t s the consequences of earriings management induced by a firrn's relation with capital markets. In particular studies examine whether firms (or firm managers) succeed in their attempt I s influence their relations with a specific stakeholder, i.e. investors. They provide also some evidence on the long run consequeiices of earnings management. Several studies examine tlie stock priee effect of changes in accounting procedures. Early studies try to discriminate befween kwo csmpeting hypotheses, i.e. the mechanistic (hilctional fixation) hypotlaesis and no-eikcts hypothesis (see Watts aiid Zimmerman (1986)). The fint hypothesis is hased oai the assumption that inlvestors are functionally fixated, i.e. interpret the earnings number the same way regardless of the accounting procedures used to calculate them (Wjtts and Zimmerman (1986), p.16C). This implies that investors can be foo8ed and that stock prices wil1 react to announcements of changes in accounting procedures, even when those changes have no cash flsw effects. The no-effecks hypothesis, a joint hypothesis of the efficient rnarket hypotbesis, the C M M and the assumptions of zero inforrnatioii, contracting and ~ransactionscosis, by contrast claims that the rnarket can see through the effects of accounting changes. Consequently, stock prices should onHy react to announcements of changes in accounting procedures when there are cash flow sffects. Studies include tbose on voluntary changes in inventory procedures (LW0 vs. FIFO) by for example Sunder (1973 and 19759, Riclis (19823, Biddle and Eindahl(l982). Those studies however failed ts discriminate between the two competing hypotheses (see Watts and Ziinmerman (1986)). The early studies asseamed that tax was the only possible cash flow effect. Later studies dropped some assumptions of the no-effeets hypothesis, i.e. the assumptions of zero infor~nationand contracting costs. Consequently, changes in accounting procedures could have cash flow effects, other than taxes. In particular, those changes could irifluence contracting and politica1 costs. This irnplies that accounting changes without tax effects couid have stock price effects, even when Invesiors are not fu~ictionallyfixated. Those studies did how-
ever not try to discriminate between the functional fixation and noeffects hypotheses. They ratkner wanted to explain accounting procedures (Watts aild Limmerman (19861, p.110). For an overview of studies, see Watts and %immerman ('(1986), chapter 12). Except for the study of Neill et al. (19951, recent sieidies focus on the implications of possible accruals n~anagementaround an IPO or seasoned equity offering. Shivakurnar (19981, for example, finds that, when firms manage earnings prior to an equity offering, investors already recognize this at the equity offering announcement date, which suggests that investors cannor be "fooled" by managing earnings. He finds also a decrease in tlle stock price response to subsequenl earnings announcements, which he attributes to earnings management. The evidence of Dechow et al. (1996) suggests however that firms managing earrnings do initia%Iysucceed in their attempt to influence investors' perceptioni of firrn value. They a%soreport that their evidence suggests that, when earnings management is revealed, increases in the costs of capital follow. Rangan (6998) and Teoh et al. (1998) find that the POOP (stock aind income) performance of seasoned equity offerings in the post-offering aainouneement period, can be explained by earnings management around or prior to the equity offering. Neiiil et al. (1995) find ihat both offering values and kindergricing are related t0 accounting mebhod choice. He suggests thak this is not the result of opportunistfc earnings management, bilt of accounting rneths d choice being a credible signa1 to investors of firm value. Resuits as to whether investors are fooled in the short term are thus mixed. In the long run, however, the rnarket seems to see through earnings manageme~lt.
III. INSTITUTIONAL DIFFERENCES BETWEEN ANGLO-SMON AMD NON ANGLO-SMON COUNTRIIES From our review of tlie literature (FBower (19971, Bal1 et al. (1997), Nobes and Parkcr (1 9951, flexander and Netbes (19941, Joos and Lang j1994), FEE (19971, Paisey (1991) and Nobes (1984)) it is elear that there exisr instilutional differences hetween countries aioalg varlous dimensions. Those are, for exarnple, dlfferences in legaH sgistems, in providzrs of finance (in particular, the importante of capita! markets), in ownership and corporate governance and in khe link b,ptweei~ tax and accounting (see, for example, Nobes (1984)). Such factors may rause the varioas internatioi~aldiiferences that can be observed in
accounting (Nobes (1984), p.3). Differences at the accounting level are, for example, a different source of demand for accounting (that is, different goals for financial reporting and different key users of fiiiancial statements), different conceptual frameworks (FEE (1997)) and accounting systems, different sources of accounting rules and degree of detail in which they are specified. It seems a logica1 consequence that those differences will, in turn, have an impact on the ability and the incentives to manage earnings. In the literature on institutional and accounting differences between countries one often classifies countries as either 'Anglo-Saxon' or 'continental European'. Tlie US and the UK are typical examples of Anglo-Saxon countries, whereas Germany, France and Belgium are typical examples of continental European countries. Some countries on the European continent have however characteristics which are closer to these of (typical) Anglo-Saxon countries. In the Netherlands, Sweden and Switzerland, for example, the importance of capital markets as providers of finance is very substaaitial, as is the case in Anglo-Saxon countries. To avoid a geographical association of countries with a certain category, and because countries are similar or different with respect to some institutional characteristic but perhaps not another, we believe that a dichotomous classification of Anglo-Saxon versus continental European countries is not appropriate. Further, recent developments in corporate financing and accounting have made the (traditional) institutional differences between Anglo-Saxon and continental European countries less pronounced, although important differences still remain. Therefore we first provide a discussion of various institutional dimensions along which countries can differ in subsection 111.1. and then continue with a discussion of recent developments in subsection 111.2.. A. Institutional and accounting characteristics of countries
1. Differences in legal systems, financing and ownership a . C o m m o n l a w v e r s u s c o d i f i e d law s y s t e m s A country's legal system can be characterized by a common law system or a codified law system. Anglo-Saxon countries typically operate under common law systems. In such a system there is only a limited amount of statute law as only genera1 principles are enacted by the legislator (Flower (1997), pp.38-39). The courts then interpret
those principles and build up a body of case law. Most non AngloSaxon countries operate under codified law systems, which are based on Roínan luw and where rules are specified in great detail. The main role of courts in those systems is to enforce the law. b. Large stock market capitalization with w i d e s p r e a d o w n e r s h i p versus lirnited stock market capitalization with concentrated ownership The major differences in ownership and providers of finance beween countries can be summarized as follows. Typical Anglo-Saxon countries have wel1 developed capital markets. Companies which are important players in the economy are listed on the stock exchange (see also Note 1) and in these cornpanies, shares are held by a vast number of individual shareholders (widespread ownersbip). In non hglo-Saxon countries, however, capital markets are tgrpically less developed, although there is a current trend towards increasing stock market capitalization (see subsection 111.2). Only a minority of eompanies is listed on the stock exchange. Funds are mainly provided by, for example, banks (this is the case, for example, in Germany) and other financial institutions, the state (for example, in France), family members or business contacts. Ownership is concentrated (that is, there are few shareholders) in both listed and non-listed firms. Some figures may illustrate the differences. Stock market capitalization as a percentage of GNP in 1995 was: 123% in the UK, 82 % in the US but only 38% in BePgium, 33% in France and 24% in Germany. The stock market capitalization in the Netherlands (74%), Sweden (77%) and Switzerland (131%) is however of comparable size as in the US and UK (Daems (1998), p.33). 2. Accounting differences
a . O r i g i n of demanid f o r f i n a i l c i a l s t a t e m e n t s s h a r e h o l d e r s v e r s u s governrnerat The above differences in legal systerns, providers of finance and firins' ownership structure have several implications on the accounting level. In particular, the differences in ownership structures have an impact on the demand for accounting information and the users of financial statements. The fact that in Anglo-Saxon countries funds are
obtained from a vast number of individual shareholders raised through the capita1market makes the financial statements an important means to communicate information to shareholders. Hence shareholders are the main users of published financial statements. Companies are als0 willing to provide this information to enhance their ability to raise additional funds througb the capital market. By contrast, in non AngloSaxon countries, where ownership is (even in listed firms) typically concentrated, the major providers of finance obtain information through direct contacts and internal financial reports. Hn Germany, for example, bankers have often a seat on the board of directors. Hn non Ang10-Saxon countries demand for financial information originated mainly from government, who needed this information in order to plan and control the economy or to raise taxes. As companies had little incentive to make accounting information public, governments made publication of financial statements mandatory (FIower (19971, p.41)1. b . P r o v i s i o n of i n f o r m a t i o n t o c a p i t a 1 m a r k e t s v e r s u s p r u d e n t a s s e s s m e n t of d i s t r i b u t a b l e p r o f i t andlor tax assessment
Differences in ownership structure and related differences in major users of financial statements parallel the traditional differences in the (traditional) purpose of financial reporting. In the UK and US, the important players in the economy are listed companies. The dominant objective of financial statements in those countries is then als0 to provide information to capital markets. In non Anglo-Saxon countries where third party protection (for example, creditor protection) is paramount and demand for financial statements stemmed mainly from the government, both prudent assessment of distributable profit (to protect third parties) and tax assessment (for government) have traditionally been important objectives of financial reporting. Tax and financial reporting is closely Linked, whereas in the US by contrast companies have to file separate financial statements for tax purposes. c. S u b s t a n c e o v e r f o r m a n d m a t c h i n g versus prudence and historica1 costs Differences in objectives of financial reporting resulted in different accounting frameworks, that is, "the fundamental concepts which lie behind the accounting law or regulation" (FEE (1997), p.3). The FEE ((1997), p.8) reports that, in Anglo-Saxon countries the most impor-
tant principles seem to be the substance over form and the matching principles. In countries where creditor protection is considered a particular important objective of financial reporting (like, for example, in Germany), special attention is given to a conservative application of the prudence and historica1csst principles. In those countries where tax assessment seems to be an important objective of financial reporting, it is difficult to identify any dominant concepts. d . Profession based a c c o u n t a n c y versus law based accountancy Finally, differences in legal systems as wel1 as different users of financial statements are related to different sources of accounting rules and the degree of detail in which they are specified. Nobes ((19841, p.16) reports that in code-law countries accounting rules are mainly to be found in company Baws, commercial codes and tax regulations. This means that accounting rules are law-based (law-based accountancy). These codes usually also specify detailed rules for accounting and financial reporting. In common-law countries, however, it is the accounting profession itself which effectively governed accounting practice, which resailted in detailed accounting standards (profession-based accountancy). Company law did not prescribe detailed rules on how companies should publish their financial statements2.
B. Recent institutional developments and their (viable) impact on accounting practice A fairly recent important evolution in several continental European countries is the trend towards a more hglo-Saxon way of corporate financing and accounting. Two factors have been and still are important in this context: ( l ) accounting harmonization and (2) increased risk financing in continental Europe. Since the seventies, EC harmonization programs have attempted to reduce differences between national accounting systems in EC countries (see for example the Fourth and Seventh Directive which deal with the annual accounts of individual companies and consolidated accounts, respectively). Mthough these directives have been implemented across member states of the European Union, many differences remained. In the UK, for example, where accounting was traditionally profession-based, several accounting rules were introduced into the law (see, for example, Alexander and Nobes (1994), p.84) but
this didn't change the key characteristics of tlie British accounting system. While the Fourth Directive provided for example the possibility to separate tax and financial accounting, tiie traditional link between them still seems to exist in several non Anglo-Saxon countries (FEE (1997)). Also, while the true and fair view principle, which is typical Angio-Saxon (UK) was adopted in the fourth directive, its practica1 impact remains limited in non Anglo-Saxon cotintries, as it is often interpreted as complying with the rules in these countires (FEE (19971, The cfforts of tbe International Accounting Standards Gomsnittee (IASC) to harmonize accounting and finaricial reporting across countries (IASC core standards p r ~ j e c t introduce )~ (some) hglo-Saxon thinking outside the Anglo-Saxon world. ThPs harmonization atkempt is principally fuelled by the globalization of capital markets and is encourage$ and supported by the Initernationa? Orgaanization of Securities Commissions (IOSCO) and the (US) Secusities and Exchange Comission (SEC) (see for example Zeff (1999), Cairns (1997)). Many exchanges (the London stock exchange and Easdaq, for example) a3ready accept that foreign conipanies prepare financial statements according to the International Accou~itingStandards (IASs) developed by the IASC. Notable exceptioms are khe New York Stock Exchange and Nasdaq. On those exchanges, foreign cornpanies can file financial statements prepared using HASs, but the SEC still requires that they reconcile earnings and shareholders' equity to US GAAP According to the Europeaan Cornmission member states can aPlow companies to use IASs in preparing their consolidated accounts. Whether preparing statements according to US G U P would fit withln European law is nol yet clear however (Batt (1998)). To avoid double work, national regulators in some EU countries already allow, intend or consider to allow multinational cornpanies andlor companies that are listed on foreign exchanges to prepare and file their "local" group accouiits according to IAS or US GAAP (see for example Batt (B998), Belgrado et al. (1998)). Belgian firms for exarnple which are listed on a foreign exchange or operate internationally are allowed to prepare their Belgian group accounts according to PASS or the foreign principle~involved (for example, US GAAP)^. Some argue that IASs are doininated by "the Anglo-Saxon approach to financial reporting" (Zeff (1999), p.10)~.Also, as the W S E and Wasdaq do not yet allow financial statements based on PASS,companies may choose to follow US Adoption of IASs or US G
by non Anglo-Saxon cornpanies definitely introduces Anglo-Saxon accounting thinking in continental Europe. Companies which (may) undergo this impact are (1)inultinational parent companies which have to issue group accounts, (2) their foreign subsidiaries which report to them and whose figures have to be consolidated in the parent's accounts, (3) European subsidiaries of Arnerican cornpanies, and (4) companies which are quoted on a foreign exchange. However, PASS or US GAAP do not affect al1 non hglo-Saxon companies. Companies which do not belong to an international group, andlor which do not have t0 issuc group accounts6,continue to report according ho their national GAAP, when not listed on a foreign exchange. This is stil a substantial part of the economy in many continental European coiantries. Further, any compaa~y's(wkether or nol it sperates internationally) individual accounts continue to be based on national G U P . As yet, there is no uniform accounting practice, although differences between typical Anglo-Saxon and non hglo-Saxon practice decreases. There is a secsnd trend which may be expected to lead t0 a fuerrther reduction in institutional and accounting differences. In continental Europe, more firrns are looking for risk financing and are findilig their miay to the exhanges. Tbe increase in HPOs over the last years on several continental European exchanges illustrates this trend. The ownership and financing of such companies is tlius changing significantly, and this will also have an accounting impact. As more firms go public, ownership of a larger set of firrns wil1 become less concentrated and investors become important users of financial statements. The dominant purpose of finaiicial reporting wil1 shift from creditor protection andlor tax assessment to information provision7. Cornpanies wil1be more willing to provide this accounting information, as (1)they depend $oa larger extent on outside investors for the financing of their operations and (2) the coinpetition for outside risk financing increases. As users and purposes of financial reporting change, it is reasonable to expect that accounting frameworks wil1 adapt to fit the context in which the accounts are used. Overall, hglo-Saxon institutiona1 characteristics are starting to apply in continental European csuntries.
IV DHSCUSSHON OF THE IMPACT OF INSTITUTFIONAL DIFFERENCES OW INCENTIVES FOR, OPPORTUNITIES 84;AND CONSTMINTS ON EARNlNGS MMAGEiVgENT Incen~ivesof and constraints on earnings inanagement are mainly affected by a firni's corporate firrancing, its ownership structure, and the accounting environment in which it operaies. Since there exist differences between hglo-Saxon and non Anglo-Saxoia companies, the incentives and constraints which were Ernestigatecl in Anglo-Saxon studies (see section 11) may not equally apply in continental Europe. However, tke trend towal-ds a more hglo-Saxen way of financing and accounting in continental Europe (as discussed in subsection 111.2) again makes a dichotomous assessment of earnings management differences between h g l o - S a o n and non Angio-Saxorm countries impossible. A better assessment, perhaps, would be to discuss differences between companies (irrespective of nationality) which have Anglo-saxon characteristics (for example, widespread ownership, financed $y capital markets, adopting US GAAP or IASS, ....) and those which have not. A. bnstitutional diflerences and incentives for earnings management
Our views with respect to incentives for earnings management are the following. First, we believe that earnings management induced by explicit contracts, such as management compensation contracts and debt covenants, might be less important for companies characterized by a concentrated ownership structure. With concentrated ownership,which is a typical non Anglo-Saxon characteristic, there are fewer conflict of interest and information asymmetry problems between owners and managers. Pt is a fact that bonus plans and debt covenants are less prevalent in non hglo-Saxon countries. In listed continental European firms with concentrated ownership, however, another type of conflict of interest may become important, namely between smal1 and large shareholders. Differences in systems of corporate governance also help explain the relatively smaller reliance on bonus plans and debt covenants. In fact, in continental European countries, providers of finance (and other stakeholders to the company) are more directly involved in corporate governance, whereas in Anglo-Saxon countries corporate governance relies more on external board members and on monitoring
by external providers of debt and equity capital (Bali (19991, p. 8). In Beigium and Germany for example, important providers of firrance have often a seat on the board of directorss. Second, tor non-listed eompanies, the incentives to manage earnings created by capita1 markets (in particular, eamiilgs management in order to communicate private information to investors, or earnings management in view of saising new funds on capital markets at more favorable terms) do not apply. Note that this is still the majority of companies in non Ang'lo-Saxon countries, although there is an increasing trend in risk financing. Third, we believe that the incentives to manage earnings created by the political and regulatory process are especially important in non Anglo-$axon firms, given the close relationship between financial reporting (in the Individual accounts) and tax reporting. Hf taxes are assessed based on profit figures reported in individual accounts, the incentive applies to al1 types of companies (internatianally operating os not, listed or iiot). Fourth, in contrast to explicit contracts, we expect that implicit contracts are especially important in creating earnings management incentives in firms with concentrated ownership. Concentrated ownership might create implicit contracts between a firm and its major shareholders, who often have a seat on the board of directors. Kasanen et al. (1996), Itor example, find that (for listed firms) implicit contracts on dividends with institutional shareholders can be important in cireating incentives to manage earnings. Fifth, we expect that income smoothing, a particular type of earnings management, may be important in non Anglo-Saxon countries. Typical explanations for income smoothing in the (hglo-Saxon) literature are job protection, maximization of compensation and avoidance of shareholder interference. These incentives may however be less important for firms characterized by concentrated ownership and which consequently have a low degree of separation between ownership (shareholders) and control (management). In those companies other incentives are however playing a major role. Tax avoidance and pursuing a fixed dividend pay out ratio (see also above) may induce firms to engage in this particular type of earnings management. The rationale is the close link between tax and financial reporting (individual accounts), the link between the reported earnings figure and dividend payments (see Bal1 (1997), p. 21,and the potential influence of the (majority) shareholders over those dividend payments. Note
that this is especially likely to hold for eariiings management in individual accounts, as both dividend payoibts and taxes are linked to this reported earnings number.
B. I~tstitutionaldifferences and constraints on earnings nlnnagement Opportunities to manage earnings (through accounting decisions) depend on l-iowflexible the accounting standards are according to which financial stateinents are established, together with Ihe degree of enforcement of those standards by the authorities. Bt is difficult to draw general conclusions about differences Bn flexibility and enforcement between Anglo-Saxon and non hglo-Saxon countries. %tis interesting to note that there exist large daffeirences, for example, between US and UK accounting standards (both Anglio-Saxon countries). The latter are considered to be less detailed and allow more RexibiPiQ (see for exxample Bal1 et al. (1997)). Further is earnings management through accounting decisions in the US restricted by the severe scrutiny of the Securities and Exchange Commission (SEC). In non Anglo-Saxon countries the law specifies accounting rules in great detail (Iaw-based accountancy). %tis however common knowledge that in some of these countries (for example, Germany) earnings that are caIculated a@cording to local are managed through provisions. This would be rather rare in the Anglo-Saxon world (see fsr example Nobes and Parker (1995), p.47-48). Note also that earnings cannst only be managed through accounting decisions, but also through investment decisions (for example asset disposals). Opportunities to manage earnings may also be different due to differences in ownership structure. Prior hglo-Saxon research reported that ownership structure acts as a constraint on earnings management. The direction of the impact of ownership on earnings management in a typical non Anglo-Saxon setting is not clear. Concentrated ownership and hence closer monitoring by shareliolders (see for example Burkart et al. (1997), p. 694) may constrain earnings manipulation. However, majority shareholders (which have internal information) may also have incentives to manage eariiings in order to transfer wealth between them and users of financial statements (smal1shareholders and government) who have no access to internal financial information. Therefore, it may wel1 be that concentrated ownership no longer functions as a constraint on earnings management.
Further, it is not clear whether audit quality will function as a constraint on earnings management in non Anglo-Saxon countries. From Anglo-Saxon studies there is evidence that audits performed by Big Five auditors coristrain earnings management more than audits performed by non Big Five auditors. In our view, there are two competing hypotheses with respect to the impact of audit quality on earnings management in non Anglo-Saxori csuntries. The first hypotbesis is that, since Big Five auditors are part of an international (American) group with standardized audit procedures, a simiiar audit quality level wil1 be provided by Big Five audit firms across the world. Therefore audit quality wil1 work equally wel1 as a constraint on earnings management in non hglo-Saxon countries. We believe that this hypothesis may hold to be true for non hglo-Saxon firms which are Iisted and/or are operating intemationahly. A competing hypothesis is the following. As stock markets are less developed in non AngPoS a o n countries and auditing is mandatoq for closely held companies which meet certain legal form and size criteria, audit demand has a different origin in non Angio-Saxon countries. Many firms, in particular closely held firms and firms which are not operating internationally, only demand auditing because it is mandatoy and less s s for agency or signalang reasons. Because of lack of voluntary audit demand, firms wil1 try to fulfill this requirement as cheaply as psssible and demand a (relatively lower) level of atadPt quality which Qust meets legal requirements. As a result many closely held firms and firms which are nol operating internationally appoint non Big Five auditors because they are typically cheaper9. Big Five auditors are currently putting effort in increasing their market share in this market. However, to be succesful in this effort Big Five auditors will have to be price competitive with local auditors. Avalid question is whether their service wil1 still be quality differentiated from the other suppliers in that market. Empirica1 evidence is needed to test which hypothesis holds to be truelo.
1P'Blis paper gresented (i)a review of the empirica1 earnings management literature, (2) a discussion of the institutional differences beWeen Anglo-Saxon and non Anglo-Saxon countries and (3) a critical assessment of the impact of institutional differences on earnings management.
Empirica1 earnings management studies, which are mainly based on Anglo-Saxon data, examine the incentives for, constraints on and consequences of earnings management. Explicit contracts (such as bonus plans and debt covenants) as well as implicit contracts were found to induce earnings management. Evidence also suggests that earnings are managed (1)to communicate private information to investors about future firm value, (2) to sell stock for a higher price or (3) to raise additional financing on more favorable terms. Further, the political and regulatory process, and some specific circumstances (such as labor union contract negotiations, proxy contests, and earnings decreases or losses) induce earnings management. The evidence on earnings smoothing is mixed. There is some evidence that prior accounting decisions, ownership structure, audit committees and internal governance (especially some characteristics of the board of directors) constrain earnings management. Audit quality was found to constrain within GAAP earnings management. Evidence on its impact on non-GAAP earnings management is, however, mixed. There is conflicting evidence on whether investors can be "fooled" by earnings management in the short run. Even if so, (1) the cost of capital increases or (2) stock performance declines once earnings management is revealed. Also the informativeness of accounting earnings rnay be reduced. As major differences exist between non Anglo-Saxon and AngloSaxon countries, it rnay well be that results of Anglo-Saxon studies do not hold in continental Europe. In particular, we argue that earnings management induced by external contracts or a firm's relation with capital markets rnay be less important, whereas incentives created by the political and regulatony process and implicit eontracts rnay be especially important. Also income smoothing is claimed to be important in continental Europe, but the incentives rnay be different. Further, we question whether in continental Europe, a firm's ownership structure and the quality of its auditor rnay constrain earnings management. Further research in continental Europe is needed to test these views. NOTES 1. Note however that also in Anglo-Saxon countries, publication of financial statements is mandatory for some types of firms (e.g. listed firms in the US). 2. Note though that in the US the SEC (Securities and Exchange Commission) has legal power to establish accounting principles, but has delegated this to the accounting pro-
3. 4. 5. 6. 7.
8.
9.
10.
fession. However, the profession cannot ignore the opiriion of the SEC (Baker, Rapaccioli and Solomon (1995)). Note that the European commission abandoned the idea of setting European Staildards and decided to support the IASC (Zeff (1999)). A notable example for Belgium are the 1997 consolidated financial statements of Petrofina. See Zeff (1999) for more detail on the IASC harmonization program, its core siandards project and its relation with IOSCO and SEC. The obligation to issue group accounts is subject to certain size criteria Note however that tax assessment is likely to remain a purpose of financial reporting i11 non-Anglo-Saxon countries (FEE (1997) p. 12). Information provision through financial reports may be biased because of the tax influence. But, as (1) tax is usually assessed on the basis of the profits in individual financial statements and (2) information needed by capita1 markets is normally provided by consolidated ones, îinancial transparency can be achieved by eliminating the tax effect in consolidated financial statements (FEE (1997) p. 12). Differences in contracting practices inay also be related to differences in legal settings. Leuz et al. (1998), for example, conclude that in the US accounting based payout restrictions are included in debt contracts (i.e. the debt covenants) whereas in Germany they are mainly mandated (i.e. restricted by law). Francis (1984), Palmrose (1986), Francis and Simon (1987), Chan et al. (1991), Francis and Stokes (1986) found for example evidence of a big-five price premium in (1) both the large and small auditee segmeilts of the (US, UK or Australian) audit market, or (2) the small auditee segment alone. Vander Bauwhede and Willekens (1998) for example could not find a difference in (the amount of) discretionary accruals, which is a measure of earnings management, between clients of big Five vs non-big Five audit firms in Belgium.
REFERENCES Aharony, J., Lin, C.-5. and Loeb, M.P., 1993, Initia1 Public Offeriilgs, Accounting Choices, and Earnings Management, ContemporaryAccountiizg Research 10, 61-81. Alexander, D. and Nobes, C., 1994, A European Introduction to Financial Accounting, (Prentice Hall International Limited, New York). Ball, R., Kothari, S.P. and Ashok Robin, 1997, The Effects of Institutional Factors on Properties of Accounting Earnings: International Evidence, Worlting paper. Baker, C.R., Rapaccioli, D. and Solomon M.B., 1995, United States of America Individual Accounts, in Ordeheide, D. and KPMG, eds., Transnational Accounting, (Macn~illan Press Ltd, Basingstoke), 2957-3095. Batt, M.-C., 1998, European Accounting Harmonisation, Accountaizc)~- International Edition 121, April, 70-71. Beasley, M.S.,1996, An Empirica1 Analysis of the Relation Between the Board of Director Composition and Financial Statement Fraud, The Accounting Review 71,443-465 Becker, C. L., DeFond, M. L., Jiambalvo, J. and Subramanyam, K.R.,1998, The Effect of Audit Quality on Earning Management, ContemporaiyAccoulztirzg Research 15, 1-24. Belgrado L, Haesaert L., Haspeslagh V., Hillewaert H., Put C. and Veldeman C?., 1998, Balanslezen voor niet-ingewijden, (Roularta Books, Roeselare). Biddle, G.C. and Lindahl, F.W., 1982, Stock Price Reactions to LIFO Adoptions: the Association Behveen Excess Returns and LIFO Tax Savings, Joun2al ofAccourzfing Research 20, 551-588. Bowen, R.M., DnCharme, L. and Shores, D,, 1995, Stakeholders' Implicit Claims and Accounting Method Choice, Joui?zal ofilccountilzg and Ecoizomics 20,255-295. Burkart, M., Gromb, D. and Panunzi, F.,1997, Large Shareholders, Monitoring, and the Value of the Firm, Tlze Quarterly Journal of Economics C X I I , 694-728.
Burgstaliler, D. and Dichev, I., 1997, Earnings Management to Avoid Earnings Decreases and Losses. Jourt~nloJ'A;fccoui~li~rg nrrd Econornics 24, 99-126. Cairns, D,, 1997, The Future Shape of Harmonization: a Reply, Ihe Europerui Accounting Rel.iebv 6, 305-348. Chan, P., Ezzamel, M. and Gwilliarn, D,, 1991, Determinants of Audit Fees for Quoted UK Companies, Working paper. Daeriis. I-I., 1998, De paradox van het Belgische kapitalisme, (Lanrioo, Tielt). DeAngelo, L. E., 1988, Managerial Competitioil, Information Costs and Corporate Governance: the Use of Accounting Perforr~lanceMeasures in Proxy Contests, Journnl of ilcco~~riting nnd Ecorzornics 10, 3-36. DeAngelo, L., 1988, Discussion of Evidcnce of Earnings Management from the Provision for Bad Debts, Jouuzc11ofAccoz11ztingResearcli 26 Supplement, 32-40. DeAngelo, H., DeAngelo, L. and Skinner, D.J. 1994, Accounting Choice in Troubled Companies, Jouwzal ofAccountirzg and Ecoizoirzics 17, 113-143. Dechow, P.M., Sloan, R.G. and Sweeney, A.P.. 1996, Causes and Consequences of Earnirigs Manipulatioil: an Analysis of Firms Sul?jcct to Enforcement Actions by ihe SEC, Corztenzpornry Acco~in1iizg Resenrclz 13, 1-36. DeFond, M.L. and Jiambalvo, J., 1991, Incidente and Circuinstances of Accounting Errors, The Accouizrirzg Review 66, 643-655. DeFond, M.L. and Jiambalvo, J., 1993, Factors Relaled to A~iditor-ClientDisagreements over Incorne-Increasing Accoontiiig Metliods, CorztenzpornryAcco~/rztir2gXesearcli9,415431. DeFond, M.L. and Jiambalvo, J., 1994, Debt Covenant Violation and Manipuiation of Accruals, Journa1 ofAccounting and Econoinics 17, 145-176. DeFoild, M.L. and Park, C.W., 1997, Smoothing Income in Anticipation of Future Earnings, Joilrrzal of Accounting arzd Economics 23,115-139. Demski, J., Patell, J. and Wolfson, M., 1984, Decentralized Choice of Monitoring Systems, The Accouilzing Review 59, 16-34. Dye, R., 1988, Earnings Management in an Overlapping Gcneraiions Model, Journul ofAccouizting Research 26, 195-235. Eckel, N., 1981, The Income Smoothing Hypothesis Revisited, Abacus 17, 28-40. Evans 111,J.H. and Sridhar, S.S. 1996, Multiple Control Systems, Accrual Accounting, and Earnings Management, Journcrl ofAccouirting Research 34,45-65. Fédération des Experts Comptables Européens, 1997, FEE Comparative Study on Conceplual Accounting Frameworks in Europe, (FEE, Bruxelles). Flower, J., 1997, Introductioil, in Flower, J. and Lefebvre, C., eds., Comparative Studies in Accounting Regulation in Europe, (Acco, Leuven), 21-48. Francis, J.R., 1984, The Effect of Audit Firm Size on Audit Prices, a Study of the Australian Market, Joz~rnalofAccol,nting and Econonzics 6, 3.33-51. Francis, J.R. and Simon, D.T., 1987, ATest of Audit Pricing in the Smal1 Clieiit Segment of the U.S. Audit Market, Tlze Acco~~ntiilg Review LXII, 145-157. Frzincis, J.K. and Stokes, D.J., 1986, Audit Prices, Product Differeritiatioi~,aiid Scale Economie~:Further Evidence from the Australian Market, Jonrnal ofAccounling Research 24,393-93. Francis, J.R., Maydew, E.L. and Sparks, H.C.,1997, Tlie Role of Big Six Auditors i11 tlie Credible Rcporting of Accruals, (Forthcoming) Auditing: u Joul-na1oJPraciice & Theory. Friedlan, J.M., 1994, Accounting Choices of Issuers of Inilial Public Offcrings, Coizteinpornry Accounli~zgResea~ch11, 1-31. Fudenberg, D. aiid Tirole, J., 1995, A Theory of Income and Dividend Smoothing Based on Inc~imbencyRents, Jozii.i~alofPolitical Econor7zy 103, 75-93. Gaver, J.J., Gaver, K.M. and Austin, J.R., 1995, Additional Evidcnce oii Bonus Plans and Income Management, Journal ofAccouizti~zgalzd Ecoizornics 19, 3-28. Guenther, D.A., 1994, Earnings Management in Response to Corporate Sax Rate Changes: Evidence form the 1986 Tax Reforms Act, The Accoun:ing Review 69,230-243.
Han, J.C.Y. and Wang, S.; 1998, Political Costs and Earnings Management of Oil Cornpanies During the 1990 Persian Gulf Crisis, The Accounting Rei~iew73, 103-117. Healy, P.M., 1985, The ESSect oí' Boii~isSchemes on Accounting Dccisioris, Joz{r7ral ofAccorliztirig and Ecorlomics 7, 85-107. IHealy, P.M. 2nd Palepu, K.G., 1993, The Effect of Firms' Financial Discloslire Strategies on Stoclc Prices, Accorcnling florizorzs 7, 1-11. Holthatisen, R.W., Larcker, D.F. arid Sloan, R.G., 1995, Anniial Bonus Sclieines and the. Manipulation of Earnirigs, Joilnzal of Accouritirlg and Ecoizomics 19, 29-74. I-Iuiit, A., Moyer, S.E. and Shcvliri, T., 1996, Managing Interacting Accounting Measures to Mect Multiple Ol>jective:a Stiidy of LIFO Firms, Jo~uzaloJ'Acco~~rrfir~g urld Ecorzor7rics 21, 339-374. Jolles, J., 1991, Earnings Management Duriiig Import Relief Investigations, Jorcïizal ofAccountir7g Resecrïclz 29, 193-228. Joos, P. aiid Lang, M., 1993, The Effects of Accouriting Diversity: Evidence from the European Union, Joz~mnlof'Accuuratir~gHeseurch 32 Supplement, 141-168. Kasanen, E,, Ihnnunen, J. and Niskanen, J., 1996; Dividend-Based Earnings Management: Eiilpirical Evidence from Finland, Jolola-llal ofArcourrtirzg nnd Economics 22, 283-312. Key, I<.G., 1997, Political Cost Incentives for Earnings Management in the Cable Televisioii Indirst~y,.Jol~ïrzalofAccour~linpa r ~ dEco~zomics23, 309-337. Lambert, R., 1984, Income Sinoothing as Rational Eqtiilibriun~Behavior, The Accoilntir~g Reviel,) 59, 604-618. Leuz, C., Deller, D. and Stubenrath, M., 1998, An International Comparison of Accounting-Based Paj~outRestrictions in the United States, United Fingdom and Germany,Accor~ntirzgand Bilsiness Research 28, 111-129. Liberiy, S.E. and Zimmermail, J.L., 1986, Labor Union Contract Negotiatioils and Accounting Choices, Tlze Accounting Review LXI, 600-620. McNichnls, M. and Wilson, P., 1988, Evidence of Earnings Management from the Provision for Bad Debts, Jortrtzal ofAccoitrlling Research 26 Supplement, 1-31. Neill, J.D., Pourciau, S.G. and Schaefer, S.F. 1995, , Accounting Method Choice and IPO Valuation, Accor~i?tirzgIfoïizor~s9, 68-80. Nohes, C., 1984, International Classification of Fiilancial Reporting, (Croom Helm, London). Nobes, C and Parker, R., 1995, Comparative International Accounting, (Prentice Hall International, New Uork). Paisey, C., 1991, E C Financial Reporting Harmonisation, The Mask of Comparability, (The Tnstitutc of Cllartered Accountants of Scotlandj. Palmrose, Z.-V.. 1986, Audit Fees and Auditor Size: Further Evidence, .Journa1 ofAccorlntiilg Reseurclz 24, 97-110. Pcasnell, K.V., Pope, P.F. and kóung, S., 1998, Outside Directors, Board Effectivencss, and Earniilgs Management, Working Paper, (University of Lancaster). Rajgopal, S. and Venkatachalam, M., 1998, The Role of Institutional Iilvestors in Corporate Goveri~ance:an Empirica1 Investigation, 'Working paper, (University of Washington/ Stanford University). Rangan, S., 1998, Eariliiigs Management and the Perfoi-mancc of Seasoned Equity O f k rings, Jo~lriznloJ'Finnr?cialEconor?zics 50, 101-122. ofAccor~nlii~g XeRicks, W., The Market's Response to the 1974 LIFO Adoplions,.Io~nzi~l search 20, 367-387. Schipper, K., 1989, Commentary on Earnings Management, Acco~lntingHoïizons 3 , 4 , i ) l 102. Shivakuinar, L., 1998, Market Reaction to Seasoned Equity Offei-ing Annou~icemeritsand Earnings Managemcnt, Working paper, (London Business School). Subramanyam, K.R, 1996, The Pricing of Discretionary AccruaIs, Journal ofAccorlrztingand Econontics 22, 249-281.
Sundcr, S., 1973, Relatioriship Between Accounting Changes and Stock Prices: Problems of Mcasurernent and Soine Einpirical Evidciice, Jo~cïiialofiiccou?ztingResecir.c:?11, supplerneili. 1-45. Surider. S., 1975, Stock Price and Risk Related to Changes in Inveritory Valuation,Arcoi~iztiïzgReview 50, 305-315. Cweeney, A.P.. 1994, Debt-Coveriani Vioiatioris and Managers' Accounting Responses, JOLITriul ofAccoiciitiiig ar~dEconomics 17, 281-308. Teoh, S. H., Welch, I. and Woiig, TJ., 1998, Earnings Managemeiiî aild the Uiiderperforinailce of Seasoned Equity Offerings, .iouïi~nlof Finnnciul Ecorzonzics 50, 63-99. Trueman, B. and Titman, S., 1988, An Explanatioii for Accounting Income Smootliing,Jolliiltzl ofAccountir~gResenïclz 26, Supplement, 127.139. Vander Bauwhede, H. aiid Willeltens, M., 1998, Earnings Management and Iiistitutional Differences: Belgiail Evidence on Audit Quality as a Coiislraint on Earnings Management, Oriderzoeksrapport ilr 9834, (KULeuven, DTEW). Verrecchia, R., 1986, Managerial Discretion in the Choice Ainong Financial Reporting Alternaîives, Jo~~ï-rial ofilccounting arzd Ecorzomics 8, 175-195. Warfield, T.D., Wild, J.J. and Wild, K.L., 1995, Managerial Ownership, Accounting Choices, and Informativeness of Earnings, Jour-1101ofdccoilntiïzg arzd Ecorrornics 20, 61-91. Watts, R.L. and Zirnmermarl, J.L., 1986, Posilive Accounting Tneory, (Prentice-Hall Inc., Englewood Cliffs, New Jersey). Watts, R.L. and Zimmerman, JL., 1990, Positive Accounting Theoq: a Ten-Year Perspective, TJ~eAccountingRei~iew 65, 131-156. Young, S., 1998, The Determinants of Managerial Accounting Policy Choice: Further Evidence for the UK, Accounting anrl Busirzess Research 28, 131-143. Zeff, S.A., 1999, The Coming Confrontation on International Accounting Standards, Working paper, (Rice University).
Tijdschrift voor Economie en Management Vol. XIV, 2,2000
Temporary VITorkers and their 7i.aimsPtion Probabilities $o Different Labour Market States: the Belgian Experienee by A. PEESERS'
More and more firnis make use of temporary contracts as a screening and selection instrument to find a suitable worker for a permanent position. h o n g others this phenomenon can be explained by the existence of hiring and firing costs and the fact that traditional selection procedures never yield perfect information on the potential productivity of a worker. h r the employers, temporary contracts have the advantage that (more) information on the actual worker's productivity can be collectecl by monitoring the worker on-the-job, without being committed to keep this worker indefinitely. If the employer is not satisfied about the worker, he can easily be fired and replaced by a new job seeker. In case of a positive evaluation, this worker can be offered a permanent contract after the expiration of the temporary contract. Because temporary contracts are thus less risky and more flexible from an employer7sperspective, it is expected that workers who can not give a signa1 of their prsductivity at the moment of hiring (because they have no (recent) work experience) are more likely to be offered a temporary rather than a permanent contract1. In contrast, workers who move from one job to another after some years
" Department of Economics, K.U.Leuven and IDEA Consult, Brussel. This paper is based on chapter five of my doctoral dissertation on "Labour Turnover Costs, Employment and Temporary Work". I am grateful to Prof. E Abraham, Prof. G. De Bruyne, Prof. P. De Grauwe, Wim Koevoets, Prof. J. Konings, Prof. E. Schokkaert and Prof. J. van Ours for useful comments on this chapter.
of work experience will have a higher probability to be hired on a permanent basis, because their previous job gives the new employer some information about their (minimum) productivity. The purpose of this paper is to find some empirica1support for this theoretical hypothesis. This paper will focus on Wo research questions. At first, we wil1 analyse whether job seekers with little or no work experience have indeed a higher probability to have a temporary rather than a permanent einployment contract. A logit model wil1 be used to investigate which characteristics make it more likely that someone is employed on a temporag7 basis. Secondly, we wil1 estimate a logit model for the probability to switch to a permanent position after the expiration of a temporary contract. Based on the human capital depreciation theory, it is expected that individuals from a risk group (e.g. long-term or elderly unemplsyed) wili not only have a smaller probabdity t0 have a perinanent contract when they flow into a new job, but that they are aIso less likely to move to a permaneiik job after the expiration of their temporary contract. Tbsting the latter hypothesis can provide some empirica1 evidence on whether temporany employment is a channel to get integrated in the labour market or whether it leads to the creation of a duaB labour market, where some wlnerable groups have high turnover and repeated spells of unemployment or inactivity. This paper is organised as follows. The next section summarizes the two theoretical hypotheses, which wil1 tested in this study. Sectio11 three describes the data, while the estimation results are presented in section four. Finally, seetion five concludes. II. THEBRETICAL BACICGROUND The purpose of this paper is to find empirical evidence for two theoretical hypotbeses. The first hypothesis, derived from the matching model in Peeters (1999)2 is the following: Hypothesis 1: A worker who can not signal his minimumproductivitpi level is more likelg,to be offered a temporary contract compared to a worker who is able lo give such a signal. In order to test this hypothesis, we need to find some variables which can indicate whether or nat someone belongs to the group of individuals who can give a signal of their minimum productivity.
Straightfonvard candidates are age, years of work experience and a variable indicating whether someone recently experienced spells of unemployment. Based on the theoretical model of Peeters (19991, it is expected that age and work experience have a negative effect on the probability of having a temporary rather than a permanent contract, while the u~aemploymenthistory has a positive effect. Age According to the matching model in Peeters (1999), a worker who can not signal his minimum productivity level is less likely to be offered a permanent contract compared to a worker who is able to give such a signal. Because young individuals typically have little os no work experience, it is thus expected that they are more likely to be employed with a temporary rather than a permanent contract. An additional explanation is the fact that temporary contracts are often part of an apprenticeship period or a government employment programme for schoolleavers. Work experience %ftemporary contracts provide a cheap way to screen applicants on their potential productivity 2nd are used for this reason, we expect employees wifh DO or little work experience to be offered (first) a temporary contract because these applicants can not signal their minimum productivity level by referring to previous employment experiences. Unemployment experience Based on the theoretical model developed in Peeters (19991, we expect that employees with recent unemployment experience are less likely to be employed on a permanent basis because these individuals can not refer to a current (or recent) employment situation to signal their productivity to a new employer. Note that also supply factors ean explain why the probability of being employed on a temporary basis could be positively correlated with the experience of unemployment. If unemployed job seekers know that firms use information on work experience (hence the labour market history) as a screening device for productivity and as a selection criterion, unemployed (especially long-term unemployed) might become aware of their smal1 chances to get a permanent job and might therefore focus their applications to temporary vacancies.
The second hypothesis, which wil1 be tested in this paper, is based on the human capita1 depreciation theory and can be formulated as follows: Hypothesis 2: An individualfrom a riskgroup3(e.g. long-tem orelderly unemployed) has a lower transition pro babi li^, fiom a tempora? to a perfnanent contract. According to the buman capita1 depreciation theory" individuals can loose their skills and work experience due to (long) periods of unemployment or inactivity. Therefore, it is expected that individuals faced with recent spells of unemployment wil1 not only have a smaller probability to have a permanent contract when they flow into a new job, bui that they are also less likely to move to a permanent job after the expiration of their temporary contract. Hence, the unemployment history is expected to have a negative effect on the transition probability to a permanent contract.
The data used in this study is taken from the Panel Survey of Belgian Households (PSBH). This annual survey (first conducted in 1992) amoiig a sample of around 3500 households in BePgium collects information on many socio-economic and demographic variables both at the household and individual level. As explained in the introduction, this paper focuses on two research questions. On the one hand, it wil1 be analysed which characheristics increase the probability to have a temporary rather than a permanent contract. On the other hand, we wil1 concentrate on the transition probabilities of temporary workers to different labour market states and examine which characteristics make it more likely that someone is offered a permanent position after the expiration of his temporary contract. Therefore, we have extracted three subsamples from our dataset, i.e. one cross-section and two time-series. At first, the cross-section sample is taken from the third wave5 (carried out in 1994) of the PSBH. This selection is motivated by the fact that information on the labour market history of an individual (e.g. unemployment experience) is not available in the earlier waves. Our sample consists of 2810 individuals who were employed at least 15 hourslweek in 1994 and who revealed their type of employment
contract. We distinguish two contract types, namely permanent and temporaiy contracts. Permanent contracts are contracts signed for an indefinite duration, while temporary contracts are short-term contracts of a fixed duration or for a specified task, apprenticeship contracts, temporary work (interim) contracts or contracts associated with a government employment programme. In the third wave, 2568 individuals (91.4%) had a permanent employment contract, while 242 individuals (8.6%) were employed on a temporary basis6. These proportions are in line with the Eurostat figrires on temporary employment in Belgium. Secondly, two time-series samples were collected for the longitudinal analysis on the transition probabilities of temporary workers. A first sample consists of al1 the temporary workers of wave 3 (242 individuals) and their labour market state one year later (in wave 4). A second sample contains al1 the temporary workers of the first wave (350 individuals) and information on their labour market state during the three consecutive years. Note that it was not possible to construct a four-years panel on relevant explanatory variables due to significant changes in the questions over time7. Therefore the information from the initia1year (respectively 1994 for the first subsample and 1992for the second subsample) was used for the explanatory variables.
IV. RESULTS In this section we wil1 first discuss the empirica1 results of the crosssectional analysis investigating the characteristics which make it more likely that someone is employed on a temporary basis. The purpose of this analysis is to test whether a worker who can not signa1 his productivity is more likely to be offered a temporary contract compared to a worker who is able to give such a signal. Aftenvards, we wil1 present the findings of the longitudinal analysis of the temporary workers. There it wil1 be analysed which characteristics increase the probability of moving from a temporary to a permanent position, with special attention to the effect of past unemployment experience.
A. Temporary versus permanent workers (hypothesis 1 )
In the following cross-sectional analysis we wil1 estimate the probability of having a temporary rather than a permanent job (given that someone works as an employee) based on the third wave of the PSBH (1994). The dependent variable temp takes the value 1if the employee has a temporary contract, while the value O corresponds to a permanent contract. Tabel 1reports the results of the binary logit-regressions estimated to investigate the characteristics of temporary workers versus permanent employees. Column 1to 5 refer to a basic specification (referred to as specification 1) with 15 explanatory variables" wwhile the last two columns give the results of an extended specification including interaction effects (called specification 2). Note that column 3 and 5 give the calculated marginal effect of each variable on respectively the odds ratio9 and the event probability (i.e. prob (temp=l)), based on specification I. From Table 1,it is clear that both specifications explain the dependent variable (temp) significantlybetter than the model with the intercept only10 and that most explanatory variables are significant at the 5% significante level''. We wil1 first discuss the effect of the variables which are most interesting from a theoretical point of view. As explained in section 11, three variables (i.e. age, work experience and unemployment history) wil1 allow US to test the theoretical hypothesis that workers who can not signal their minimum productivity level are less likely to be offered a permanent contract compared to workers wlio are able to give such a signal (see hypothesis 1in section 11). Age As expected, young persons are more likely to be employed with a temporary rather than a permanent contract. The marginal effect indicates that the odds12for employees less than 25 years to have a temporary contract are 2.5 times higher than for employees above the age of 25. Alternatively, we can say that an average young employee ( 25 years) is 3.8 percentage points more Lilrely to have a temporary contract compared to an older individual (2 25 years). A dummy variable for being less than 25 years was preferred above a continuous variable for age, because the latter is too highly correlated with the variable work experience. Furthermore, the dummy seemed to pick up most of the effect because the positive correlation between the de-
pendent variable and age seemed to be especially apparent at the lowest end of the age distribution. TABLE 1 Parameter estimates of a logit model for the yrobability of beiízg e?nplo)led witlz a tenzporaly coíztractl"
others sector (services) agriculîure
/
1,596*
2,927
4,934
0,020
0.094
1,503*
2,699
1
-1,103
-0,690
0,332
0,006
-0,019
-1,473
-0,788
e Work experierice Tlais variable, constructed as tlne differente between someone's currerit age alad the aga at which this person starred his first paid job, can be used as a proxy for (potential) work experience, because at teils us how many years ara individual is already on ihe iabour market sirace his first job. Temporary movements out-of-the-labour force are laowever not taken in10 account due to missing data. The estimation results of specification 1 show that tbis variabie has a very significant and negative effect, which is consistent wlth our theoretica1 predictien. The probabilily of being employed on a temporasy- basis is thus negatively correlated to the years of wolk experience14. More specificaPly, one more year- ofwork experience decreases the chaiaces of having a temporag. contract rather than a permanent contract by 0.14 percentage points. In specification 2, the variable work experience is however no longer significant due to the incorporatlon of interaction effects between age and education.
Uneinployment experience A very significant and positive coefficient indisates that employees who have experienced unemployment durlng the last five years have a higher probabiliky (P.e. their chances are 6.5 percentage points higher) of getting a temporary contract compared to workers who had no recent unemployment experience. This finding provides some empirBcal support for our theoretica] hypothesis that employees who can n s t refer to a curreait or recent employer are less likely to be employed on a permanent basis. Buk as mentioned in section Hl, also supply factors can be mentioned as an explanation. Hf sinemployed job seekers are aware of their smal1 chances Zo get a permanent contract, because past employrnent experience is used as a selection criterion, they might focus their applications to temporag. vacancies. Alternative@, unobservable characteristics '"linked to those who experienced unemgloyment might be mentioned as explanations for the significant and positive coefficient of uneinployment experience. Note tbat we also experin~entedwith a dummy variable for longterm unernploymcnt (i.e. tlnose who experienced an ~anemployment spel1 of rnii~imurni 2 inonths during the past five years) and a continuoris variable referring to the frequency of unemployment spells during the past five years. When these variabies were used separateiy insiead of the dummy for having experienced unemployment during the last five year-s, both of thern had a positive and significant effect on
the probability of being temporarily employed. But this signif:icance was lost when they were combineci with the unemployment dummy of specification 1or 2. From these results, we can conclude that job seekers who are young, unexperienced or unemployed are more likely to be offered a tempora1-y rather than a permanent contract16. Hence, oUr empirica1 analysis provides support for the theoretica1 hypothesis that workers who can not signa1 &heirproductivity level at the moment of haring (because they Iack (recent) work experience) have a lower probability to be employed with a permanent contract (see hypothesis 1 in section Ia). Let US now take a closer look at the effect of some other interesting explanatony variables. Recent start A very significant positive coefficient indicates that individuals who started their current job in 1994 or 1993 (i.e. maximum 2 years at the inoment of investigation) are more likely (the odds are 3.3 times higher) to have a teinporargr contract than those who are already more than two years in their current job. This result illustrates the importance of temporary contracts for the inflow into new jobs and thus also supports the hypothesis that temporary contiracts are frequently used as a screening device for productivity. @
Nationality A significant positive coefficient for the durairny foreign indicates that non-belgian workers are more likely (i.e. 2.5 percentage points) to have a temporar-y contract compared to employees with the Belgian nationaliS. This result may suggesi a selection strateggr of employers in favour of Belgian workers, bmt It is very likely that some unobserved characteristics explain (part of) tlie sigimificance of this variable. c Gender Female employees seem to be less likely to have a temporary contract coinpared to the male employees. Note that this effect is however not significant in specificatioi~1 and o d y significant at the 10% leve9 in specification 2. klthough the raw data shows that proportionally more wsmen are in temporary contracts cornpared to men (11.4%
of the female employees versus 6.4% of the male workers), our results suggest that tliis distribution can not be explained by gender, but is due to other (personal) characteristics taken into account in the regression estimation. e Educational attainment The estimates for the four educational duininies indicate that employees with lower education levels have a smaller chance to have a temporary contract relative to the reference category of university degrees17. This result may be explained by demand as wel1 as supply factors. Because the required tasks for lower educated workers are often less demanding and more straightfonvard, temporary contracts might be less necessary as a screening device for the productivity of these workers. In contrast, it might be more difficult to infer the productivity of university graduates based on their degree, hence firms may be more likely to offer higher educated workers (first) a temporary contract in order to screen their productivity. Another demandrelated explanation for the positive correlation between the educational level and the probability of getting (first) a temporary contract is that employees with a university degree are on average more costly (higher wages and thus als0 higher firing costs). Therefore, firms might be more reluctant to hire these workers immediately on a permanent basis. In addition it is also possible that supply factors are at work, which make it more likely that higher educated workers search (first) for a temporary job. As mentioned in Groot (1990), higher educated individuals might be less risk averse than the lower educated as they possess more wealth in the form of human capital. Finally, this positive correlation might also be explained by the fact that higher educated workers are often employed in jobs which are temporary in nature for long periods of time (e.g. teachers, public servants, assistants and junior professors at universities). As a consequence, higher educated workers are more likely to be employed with a temporary contract and probably also more likely to stay in a temporary position for several years due to their type of job and not (necessarily) because of screening reasons. Additional insights on the effect of education can be obtained if we include interaction effects between the educational dummies and age. From the estimation results of specification 2, we can derive two interesting findings. At first, we notice that the significance of the lowest educational groups ("primary" and "secondaryl") increases, while
the significante of "secondary2" and "tertiaryl" falls compared to specification 1. Hence, it are especially the lowest educated workers (with secondary school as the highest educational attainment) who have a significant lower probability of being temporarily employed compared to tlie university graduates. Secondly, two significant interaction effects18 indicate that higher educated worlters (i.e. those with a third level non-university or university degree) have a lower probability of having a temporary contract the higher their age. B. Transition of ternporay workers to different labour market states (Itypothesis 2)
This section focuses on the transition probabilities of temporary workers to different labetur market states. As explained in sectio11 III, Wo time-series samples were collected for this longitudinal analysis. The first sample consists of al1 the temporaiy workers of wave 3 (242 individuals) and their labour market state one year later (in wave 4). The second sample contains al1 the temporary workers of the first wave (350 individuals) and their liabour market situation during the three consecutive years (fronn wave 2 to wave 4). As mentioned before, it was not possible to construct a panel on relevant explanatory variables due to significant changes in the questions over time. Therefore, the data frsm the initia1 year (respectively 1994 for wave 3 and 1992 for wave 1)wil1 be used for the explanatoiy variables and new dependent variables wil1 be constructed indicating the labour market state in tke consecutive years. Six labour market states were distinguislied, iiamely employment with a permanent contract, being temporarily empioyed, employment with unknown contract type19, selfemployment, unemployment and inactivity2'. l. Observed transitions t o different labour market
states Table 2 presents the observed transitions for the 242 temporary employees of wave 3 (1994) after one year (in wave 4). These figures show that 26.5% of the temporary employees in 1994 found a permanent job one year later, whiie around 50% of them are still employed on a temporary basis. Nearly 5% is either employed with an unknown csntract type or self-employed and 9.1% moved to unemployment or inactivity after the expiration of their temporary contract. Unfortunately, the situation of 10% of the temporary workers is unknown be-
cause these individuals are not included anymore in the sample of wave 4 (due to attrition in the panel). Taking these missing values into account, we can state that minimum 114 of the temporary workers flow to a permanent job after one year. TABLE 2 Obseived transitions to different labour market states for the 242 temporaty workers of wave 3 (situation after I year)
labour market state in wave 4 permanent contract temporary contract employed, unknown contract !ype self-employed uneinployed inactive missing total
frequency 64 119 7 4 15 7 26 242
%
26.45 49.17 2.89 1.65 6.20 2.89 10.74 1O0
Let US now take a closer look at the observed transitions for the temporaryworkers of wave 1.The advantage of this second sample is that we can analyse the transitions over a four-year period. Table 3 gives the obsewed transition probabilities for the 350 temporary workers of wave 1in the consecutive years. After one year (in wave 2), 28.6% of them had a permanent employment contract, while 40% still wsrked as a temporary employee. About 5% were employed with an unknown contract type or self-employed and 12% became unemployed or moved out-of-the-labour force. Due to attrition in the panel, 15% of the temporary employees of wave 1were lost in the sample of wave 2. Note that the one-year transition probability of moving from a temporary to a permanent contract (28.6%) is very similar to the one calculated for the temporary workers from wave 3 (see Table 2). Furthermore, it is worthwhile mentioning that the probability to move to a permanent position does not increase significantly after a time span of one year. After two years, this probability has only risen to 32.6% and it equals 33.1% after three yearsZ1.
Observed transitzorzs to difj"eer-entIabour market states for. the 350 tenlporaiy workers of wave l wave 1 freq %
1
I
permanent contract teinporary contract ernpl., unknown type self-ernployed unemployed inactive missing
I
I
350
wave 2 fi-eq %, 100 28.6 100 139 39.7 1 1 5 4.3 2 0.6
1
/
1
wave 3 freq % 114 32.6 92 26.3 5 1.4 4 1.1
/
/
wave 4 freq % 110 33.1 67 19.1 14 4.0 B 2.3
/
/
2. Estinlating the transition probability to a p e r m a n e n t job
In this subsection we wil1 examiile which characteristics make it more likely that a temporary worker is offered a permanent position after the expiration of his temporary contract. The aim of this longitudinal analysis is to find out whether this transition probability is lower for individuals from a risk group, i.e. those who recently experienced unemployment (see hypothesis 2 in section 11). Given the fact that information on past unemployment is only available from the third wave onwards, we wil1 only present the estimation results concerning the longitudinal analysis of the temporary workers of wave 3. The dependent variable, switchperm, takes the value 1 if the temporary worker has a permanent employment contract after one year and O if the transition to a permanent position did not take placeZ2.Table 4 presents the binaiy logit estimates for this one-year transition probability to move to a permanent job. The diagnostics indicate a much lower explanatory power of the estimated model compared to the cross-section results discussed in the previous section, bul this can be explained by the much lower number of obsewations available for the regression estimation. Nevertheless some interesting explanatory variables are significantz - as indicated by the t-values.
TABLE 4 Logit estimales of fhe one-yeai. fl-ansiíio~tprobability of nzoving to a pernzarzerzt job folthe iel?iporn~y~1orker.sof wave 3'-'
Ui~employrnentexperience As is obvious from Table 4, those who did experience unemployment during the past five years are less liliely to switch from a ternporaiy to a permanent contract a f er one year. This result thus confirrns the theoretical hypothesis, derived frorn the huinai1 capital depreciation theory, that tempora~yworkers belonging to a risk group (e.g. long-term unemployed) have a lower transition probability to a permanent contract (see hypothesis 2 in section 11). It is however also possible that the negative effect of past unemployment experience on the transi-
tion probability is (partly) driven by unobserved heterogeneity. Note that this result could provide some evidence for the existence of a dilal labour rnarket in which soms individuals are first offered a temporaly contract, but tliis contract is relatively quickly replaced by a permanent olie, while another group of iiidividuals experiences long t ~move ~ ~ to spells of temporary einployment (without the p o ~ s i h i l i to a permanent position) or consecutive short spells of temporary employme~itand unemployment (or inactivity). But given the small nuinber of observations and low explanatory powcr of specification 1 aiid 2, we lieed to be careful with deriving this conclusion on the basis of the estimation results reported in Table 4. Furthermore, i% is worthwbile inentioning the interesting effect of the I'ollowing variables: Age Ternporary employees less than 25 years seem to be less likely to switch to a permanent job after one year (compared to tkose who are at least 25 years old). AS is obvious from specificatisn 2, the opposite effect is however found for those individuals below the age of 30 years. Hence, it can be concluded that the one-year transitioii probability to a permanent job increases with age, but this rise occurs mainly between the age of 25-30. Werk experience Surprisingly, Table 4 reports a significant but negative coefficient for this explanatoiy variable. A possible explanation is that this variable is capturing some unobserved "bad" characteristics of individuals. The €act that someone has already many years of work experienceZ6and is still ernployed on a temporary basis can be a signa1 of "bad" characteristics. Hence, firms wil1 be Iess IikePy to offer a permanent contract to this type of individ~ials.011the other hand, it is also possible that supply factors cxplain (partly) this result. L If individuals know they have "bad" characteristics, they mighl be more likely to stay in temporary positions either because they are aware of their smal1 cliances to switch to a permanent contract or because they prefer temporary (hence more Rexible) employment contracts. Finally, it it also possib'ic that the reason why individuals stay in a temporary job for a long time 3s related to the type of job. More evidence for this last explanation can be found i11the fact that a considerable number of ternporany employees work many years on a temporary basis for the same employer (high e
t e n ~ r e ) ' ~Furthermore, . many of these individuals are the ones with the highest educational degïee. Note that the negative effect of work experience is however not significant anyinore in specification 2 (when ehe age dummy refers to al1 individuals less than 30 years old). lid-xational level Positive and significant coefficients for two educational dummies show that lower educated workers have a higher chance t0 switch from a temporary t s a permanent contract after one year, relative to the reference category of university degrees2'. A possible explanation for this effect is that higher educated individuals are more liltely to be employed in jobs which are typically of a temporaiy nature for many years ~ ~specification . 2, only the group of individ(e.g. public ~ e r v a n t s )I[n uals wilh a third level non-university degree has a significant higher transition probability compared to the university graduates. Finally, we can mention that similar effects of the above discussed variables were found when estimating a logit model for the one-year, two-year and three-year transition probability of the temporary workers of wave 130.The only exception is the variable 'unemployment experience', which could not be analysed in this sample due to missing data on this variable in the first wave of the PSBH.
Based on the Panel Survey of Belgian Households, this paper examined which characteristics make it more lilcely that someone is employed on a temporary basis and which characteristics increase the transition probability from a temporary to a permanent contract. A binary logit model was estirnated to analyse the first research question. The most important findings from this analysis are that job seekers who are young, unexperienced or unemployed are more likely to have a temporary rather than a permanent employment contract. These results provide some empirica] evidence for the theoretica1 hypothesis that workers who can not signal their productivity level at the moment of hiring (because they lack (recent) work experience) have a lower probability to be employed with a permanent contract (see Peeters (1999)). Hence, in the absence of a productivity signal temporary contracts are used as a screening device for productivity.
Secondly, we focused on the transition probability of temporary workers to different labour market states. Due to attrition in the panel, 10 to 15 % of these workers were lost after one year. Taking these missing values into account, we can nevertheless state that minimum 1/4 of the temporary workers flow to a permanent job after one year. Furthermore, it was found that this probability does not increase significantly anymore aftenvards. The binary logit estimations of the oneyear transition probability indicated tliat, as expected, temporary workers who experienced unemployment during the past five years are less likely to switch from a temporary to a permanent contract. This result confirms the theoretical hypothesis, derived from the human capital depreciation theory, that temporary workers belonging to a risk group (e.g. long-term unemployed) have a lower transition probability to a permanent contract. NOTES 1. See theoretical model in Peeters (1999), Chapter 4. 2. See Chapter 4 "Screening and Temporary Jobs in a Matching Model" in Peeters (1999). 3. A risk group is defined as a group of individuals who often lacli (recent) work cxperience aiid are often faced with depreciation of their human capita1 and loss of informal contacts with employers due to long periods of uneinployment or inactivitjj. 4. See Layard, et al. (1991) for empirica1 evidence and among others Blanchard and Diamond (1994) and Pissarides (1992) for a theoretical model. 5. Each year of the PSBH-survey is referred to as a "wave". Wave l corresporids to the first survey conducted in 1992. 6. The majority of these temporary worlcers was employed with a contract of fixed duraiion (43.8%). In addition, 7% were in a training period, 4% in a government employment programma, 16.5% had an interim contract, 4% contract for a specified task aild 24% had another type or did not specifl it. 7. Due io the incorporation into a Europanel survey, the questions were for exan~plesignificantly changed between the first two waves and the third wave. 8. A description of the variables is provided in the Appendix. 9. The odds or the odds raiion is defined as the ratio of the probability that ihe event occurs to the probability of the nonevent. 10. The joint hypothesis that al1 coefficients except the intercept are O can be rejected, because the likelihood ratio statistics exceed iheir critical values. For specificatiori 1 and 2 this statistic is respcctively equal to 471.04 and 494.13, while the criiical value for the distribution (with a=0.05) is rcspectivelj~37.65 and 44.70. 11. The nul1 hypothesis that a coefficient is 0 can be rejected if the computed t-statistic exceeds thc critica1 value (in our case this critical value equals 1.96 if the significance level is 5% aild 1.64 at a significance level of 10%). 12. See note 9. 13. " denotes sienificance ai 5% level. "" deriotes sirrnificance at 10% level. Mareinal effects evaluated at tlie sample mean of the independent variables. 14. We also experimented wil11 anotlier variable, namely tenure (i.e. the years someoile is working in liis current job) instead of work experience. The paraineter estimate for this variable was also negative and very sigiiificant. Although this specification gave even better results, tlie variable tenure is less interesting from a theoretical point of view
15.
16.
17. 18. 19. 20. 21. 22. 23.
24. 25. 26.
27. 28. 29. 30.
because it is vcry trivia1 tliat thc longer you are workiiig i11 tlic same joh, tlic inorc likely you are to liavc a perrilaricnt coiitract. Iiiclividuals can liave cliaracteristics uriobservable to [he researchcr, hut whicii nlay 110netheless be obscrved by the fii-in aiid affect tlie probability of bcing offei-ed a ternporary job. Tliis problciil of uriohserved heterogcricity as wel1 as tlie initia1 conditions prohlcni (arisirig wlicn tlic start of the ohservation period does iiot coiiicide with tlie lubour inarket eritrancc of al1 iiidividuals iii thc s~iinple)can only br properly addi-cssecl with panel data (Arulaiiipalaiii, Bootli aiid Tayloi- (1998)). Ui11 as explaiiied before, wc were not ahlc to constriict a panel on I-clcvaiit cxplanalory variables d ~ i eto significaii( cliangcs in tlic questioiis over tiinc. I-lence, i1 was not possihlc to distinguisli bctwecn unohscrved Iieterogeneity and true state depcnclencc (e.g. a causal link bctwecn past unemploymerit expcricnce aiid thc prohability of liaviilg a tcinporary coiitract). Tlie positivc effect o i being young 01-haviiig (rece~itly)cspcrier~ced~ineiilployment«n tlie probability of beiilg teniporarily eniployed was also statislically very significant lor tlirce subgroups (i.c. groups which were more homogeneous with respect to age, teilure and starting periode of the currci~tjob). Tliese rcsulls are iiot shown here, hut cali be found in Peeters (1999), Cliapter 5. A study for Ihe Netlicrlands (Gorter ei al. (1998)) came iip with a similar result. Naniely "age"tertiary1" and "age.Yertiary?". For sornc individuals, we know that tliey are employcd hut it is iinkiiown iyhcthcr they stil! have a temporary contract «r wliether this contract was replaced hy a perinaileilt one. Inaclivity coiisists of tliose followiiig training or educalioii (as tticir inain activily), cloing hoine dutics or child care, on military service o r rclircrnent. Note tliat tlie proportion of inissingvalues for the situation o l the temporary workei-s fromwave 1doubles betweenwave 2 and wave 4 (from 14.9% in 199.3 to 28.9% in 1995) due to attritioo in the panel. Tlie value O correspoiids not only to individuals still worlcing on a lemporary basis, but also to tliose wlio moved to self-employinent, unknown contract type, uriemploymeiii or inactivity. It is worthwhile nientioning here that binary clioice inodels have tlie unpleasant feature tliat oinitted variables bias al1 estimated coefficients towards zero. Eveii if the oinitted variable is uncorrelated witli the iiicludcd one(s), tlie coefficient on the included oiie(s) wil1 be incoiisistent (Greene (1997)). 111contrast, the iieglect of oinitted variables in standard linear regression moclels does :lot affect thc rcniaining coefficieiit estimates and merely sliows up in ai1 increased residual variance, provided tlie omitted i-egressors are uncorrelated with tlie included regressors (Cramcr (1991)). 'Veiiotes sigiiificance at 5 % level, 'l:'denotes sigiiificancc at 10% level. 01- "willigiiess" (because some iildividuals prefer to s~ipplyoilly teinporary employineili). Remcinber that this variable is definecl as the differente betwcen soineone's curreiit age and the age at which this persoli startcd Iiis first paid job. It is thus a proxy for (poteiltial) worli expcrience, but neglects tcnlporary nlovenients out-of-tlie-labour forcc. hniong the group o l Lemporary workcrs of wave 3, tlie riveragc tenure i11 tlie currcrit job is 3.7 years. Tlie majority of these workcrs has tcnure between O and 5 years (76.5%), bul a sigriificaiit group (23.55%)indicates a tenure between h and 14 years. Note that Gorter e t al. (1998) also Iound tliat temporary workers witli a liigh educational level ai-c Icss lilzcly to have thcir teinporary coiitract convcrtcd into a permaricnt olie. See also Ihc discussion »n the elfect of cducation in seclion 1V.h. Results nut showii, see Peelers ((1999) Chapter 5).
REFERENCES Arularnpalain. \V., Booth. A. ancl Thylor, M., 1998, Uiiemployment Persistcrice, mimeo. Blaricliard, O. aiid Dianioiid, P., 1994, 'Kanting, Unemploymcnt Duration and Wagcs, lieifierv of Ecorioinic Stzcrfies 6 1 , 417-434 Cramer, J., 1991, Tlie Logit Model: ar1 Iiitroduction for Econoniisis, (Edward Ariiold, Loildc>n),110. Gorter, C., Moolcnaar, D. and RLISSO, C., 1998, Tcniporary Jobs aild Tciiiporary Workers: the Dutcli Espcricilcc, mimco. Grecnc, W., 1997, Ecoiiometric Alialysis, Chapter 19, Models with Discïete Dependent Variahles, 3" ed., (Prcntice Hall, Englewod Cliff), 1075. Groot, W., 1990, Elctcrogcneous Jobs and Rc-employineiit Probabilities, O.+i-ilfHr~//etirl o$ Ecorloinic,~uild Sturisrics 52, 3, 253-267 Lityard, R., Nickell, S. and Jackman, R., 1991, Unemploymenl: Macroeconoiliic Performance aild the Labour Markel, (Oxford Uiiiversity Press, Oxford), 618. Pcetcrs, A., 1999, Labour T~iri-ioverCosls, Einployineiit and Tcmporary Woi-k, Ph.D. dissertation IC.U.Lcuven, 207. Pissarides, C., 1992, Loss of Still Duriiig Uneniployincnt and the Persistcncc of Employnielit Shocks, Quctl-îedyJo~ol~lacrl o[Ecoizoi~.iics107, 4, 1371-1391
APPENDIX
Description of the variables Variable name Dumniy i,arinhles:
/
temp switchperm
(
I
I
/
I I 1 I I
1 (
young ( < 25 yr.) young ( x 30 yr.) toreign female married primary secondaryl secoiidary2 tertiaryl tertiary2 unskilled blue skilled blue-collar white-collar management others agriculture construction industry services rion-profil sector flanders brussels wallonia public sector uneinpl. past 5 yrs. receiit start many cmpl. contracts supervision education in 1993 vocational training scale net-income work experience
I Defiiiition I Employed with temporary contract
(
I
I
1 1 f
I /
I
Switch to a permanent contract after cxpiration temp. contract Less than 25 years old Less than 30 years old Nationality is non-belgian (EU or !ion-EU) Women Married 13igliest educational attainment is primary school Highest educational attainment is junior seconciary school Highest educational attainment is senior secondary school I-Iighesl educational attainment is tliird level noil-university degree I-Iighest educational attainment is university degree Professional category is u~lsltilledblue-collar worlcer Professional category is skilled blue-collar worker Professional category is wliite-collar worker Professional category is manageinent Professional category does not belong to the catcgories above Employed in the agricultural sector Employed in the construction sector Employed i11 the industrial sector Employed in the services sector Employed in tlie non-profit sector Living in Flanders Living in Brussels Living in wallonia Employcd in tlie pubiic scctor Min. 1 iincmploymcnt spel1 duriiig past 5 ycars Current job started in 1993 01- 1994 More than 1 employment contract Supervising or coordinating other workers Followccl any education (full-time or part-time) in 1993 Followed ever any vocaiional training
1
1 / I
I I
1
I I
I I I Monthly net-incomc (in BF) divided into 18 categories I Diffei-ence between the current age and tlie age at which the individual started his first job (measured in years)
Tijdschrift voor Economie en Management Vol. XLY, 2, 2000
Wages and Firm Profitability in the Manuhcturing Sector in Côte d91voire
Wage levels above the competitive wage rate, combined with high levels of unemployrnent, are frequentb observed in developing countries, where wages in larger and formal firms are fotanad to be substantially higher while formal sector jobs are rationed. %he question why wages are paid above the market clearing level Plas been an issue of policy importante in these economies, where unemployment and its social consequences are a major concern. A theoretica1 explanation was first developed by Stiglitz (1974) in ehe context of poor econornies. He stated that higher wages induce better nutrition of the labour force, thereby increasing physical energy and resulting in higher productivity levels. A set of efficiency wage models was subsequently developed providing an explanation for the obsewed wage structure. These models advance that higher wages enhance productivity of workers and hlierefore pay off in terins of worker's efforts. Alternative explanations, maianly developed in the context of developed country labour markets, view higher wages as the result of collective bargaining power, which enables workers in more profitable firms to bargain over a share of the reilts. Crucial to this explanation is that the firm enjoys some degree of monopoly power in the product market which allows it t s make surplus profits and to share this Departement Toegepaste Economische Wetenschappen, K.U.Lcuven, Leuven. I would like to tliailk Leo Sleuwaegen, David Audretsch, Lode Belage, Joep Konings and Reinhilde Veugelers for insightful cominents aiid suggestions. I am very grateful to Hans Eyssen for his helpful work on the employee-level data.
with employees. Surprisingly enough, while product markets in developing countries ai-e often cilaracterised by oligopolistic structures with a few firms occupying a domiilaiit position in their industrj, the effect of firnn profitability on wages has remailled largely uilexplored in studies on wage formation in developing countries'. This study analyses how alternative models of wage determination may explain observed wage differentials in the manirfacturing sector in1 Côte d'xvoire. More specifically, the f o c ~ iss on the relationship between farm profitability and the wage paid to individual employees. It wil1 be investigated whether higher wages are paid by more profitable firms with sorne degree of market power in the product inarket, when other relevant wage determinants are taken into account. The empirical analysis is baced on a unique data set gathered in the fsarnework of the 'Regional Program oii Enterprise Developinent' project executed in Côle d7Hvoirein 1995 and 1996. TIie data set contains personal data of '799 wage workers in the Ivorian manufacturing sector including the person's education, working experience, current occupation, gender and wage, as wel1 as information on some key characteristics of the respective employer such as firm size, age, profitability, sector of activity and capital intensity. 'The paper is organised as follows. Section I1 presents the basic insights of efficiency wage theory and union bargainhg models in explaining why firms pay above market clearing wages. Section 111analyses rhe implicatlons of these models on how wages are set for individual worliers working in firins wikh different strtictural characteristics anid presents it relevant empirical findings. Section IV describes the characteristics of the labour market in Cdte d'Ivoire. %nsection V tlie estirnating model is presenled and the data and variables described. A wage equatio~iis estimated and results discussed in section Vl. Section VII concludes.
IH. DETERMINANTS OF THE WAGE LEVEL: A THEORETIGAL PERSPECTIVE According to neo-classical theory, labour is hired up to a level where the marginal product equals the reservation wage, and markets clear. Mowever, wage levels above the reservation wage combined with high levels of unemployrnent are comrnonly obsewed. In order to provide an explanatioil for the observed wage structure, several models have
been developed, either based oil the hypothesis that firms pay efficiency wages or share rents witli their employees. Iii the view of efficiency wage models, firms set l~igherwages tlian those that would fellow from the market t s enhance the productivity of the worlters. Several lines of thought have been developed to give alternative explanations of why labour prodilictivity depends on real wages. According to the 'shirking model', workers may have an inforrnation advantage over their employers concerning their own pcrformance. If workii~gii~ivolvessorne degree of disutility b r the worker, the latter may have an incentive to cheat or shirk (Shapiro, Stiglitz (39841, Yellen (1984)). To improve tkie efficiency of their workers, firins raise wages to pay more than their cornpetitors. Being cauglit shirking and being fired therefore involves a cost to the worker. If in ey~iilibriuma11 firms have set wages abovc market clearing levels, the resultirng unemployment is a sufficient penalty for the worker to deter shirking2. In a sirnilar view, the labour turriover model states that firms may choose to gay wages above the market clearing level 10 diminish costly turnover of the labour force (Salop (1979)). Workers have less ineentive to quit the higher the wage tbey receive in their current farm. A third model, presented by Ackerlof ((1982), (1984) j, postulates that workers9feelings of loyalty to their firin increase with the extent ro which Sirms share their profits with them. A higher wage paid by the firm is regarded as a kift' from the firn, which the workers fee1 like reciprocating by a 'gift' of higher work norms and higher efficiency. From an adverse selection perspective, tbe wage differential paid by the firm is mainly based on selection instead of incentive grounds (Weiss (1980)). If labour quality is ~notdirectly observable, and individuals possess superior knowledge about their own qualifications, low wage paying firms wil1 be unable to attract high quality labour. Firms paying higher wages attract on average a pool of better qualified applicants. %f,due to asymmetrie information, the hiring of employees occurs through a random selection from the pool of applicants, firms paying higher wages may expect to have more qualified workers. Efficiency wage models are not rniitually exclusive. Firins may pay above coinpetitive wages with the intention to reduce shirking arrd quits, to attract a labour force of above average quality, and to increase worker's sense of loyalty. TFLiese models have in common that they advance that the firm deliberately chooses to pay a wage premium, as it enhances firm performance3.
An alternative explanation for the obseived wage differential stresses the role of labour unions, as wel1 as the position of the respective firm in the product market. Labour unions and other institutions give some degree of monopoly power to the workers taken collectively which grants them the opportunity to bargain for a share of the firm's profits. Therefore, an essential assumption for union-bargaining models is that there is some profit to bargain over. Firms operating in very competitive product markets with free entiy are unable to make surplus profits, and the position of unions is weakened. The theory of union wage bargaining essentially consists of two general classes of models: the 'right-to-manage' model and the 'efficientbargaining' model (Ulph and Ulph (1990)). In the 'right-to-manage' model the union and the firm bargain only over the wage level, while the firm retains the right to choose the number of workers it hires. Once the wage is determined, the firm sets the employment Level so as to maximise profits. "diragesare set and surplus profits split according to the relative bargaining power of both parties. Higher wages are bargained for by the union, at the expense of employment, which falls below the competitive level. In efficient bargaining models unions bargain with firms both over the level of employment and the wage level. Any wagelemployment combination that emerges from the bargaining procedure should be 'efficient' i.e. should maximise the gain to the firm from the bargain, subject to achieving a level of gain for the union. The union shares the firm's incentive to use labour in the most efficient way, by hiring labour up to the point where marginal product equals the reservation wage. The union then tries to get as much of the firm's profits as possible by bidding the wage up above the resewation wage. The union reduces the firm's profits by a fraction, leaving employment unaffected.
IIH. WAGE DIFFERENTHALS M D WORKER AND FIRM CmMGTIERISTIeS: EMPIRPCAL EVIDENGE Hf al1 firms were identical and labour markets competitive, efficiency wage models would suggest that in equilibrium al1 workers with identical characterisiics would earn the Same wage, which is higher than the market clearing wage for that type of worker. However, firms are characterised by a high degree of heterogeneity, resulting from different technological complexity and human capita1requirements, thereby leading to a range of efficiency wage levels.
Empirica1 evidence indeed shows that wages val7 systematically over sectors of activity and firms of different size. The positive relationship between wages and firm size can be explained by the hypothesis that firm size proxies various unobserved wage determinants (Schmidt and Zimmermann (1991), Mellow (1982)). Larger firms are characterised by a higher degree of labour division, resulting in disutility for the worker for which larger firms have to pay compensating differentials. Larger firms typically use more complex and capital intensive production technologies requiring superior labour quality. Monitoring may be more difficult in larger firms resulting in an increase in the efficiency wage to avoid shirking of the workforce4~Moreover, unions tend to be better organised in larger firms, so wages may be raised and profits shared with the employees, as predicted by the union bargaining models. Wages are also found to vary substantially across industries. Lawrence and kawrence (1985) and Dicltens and Matz (1987) found that more capital intensive industries tend to pay higher wages. In a similar way Campbell (1993) found that more capital intensive firms pay higher wages. Some industries are found to pay systematically higher wages than other industries even after controlling for individual worker characteristics which capture observed5 or unobserved6labour quality differences and after controlling for undesirable aspects of the working conditions such as shift work, overtime work, many weekly hours, unpleasant physical conditions (Mrueger, Summers (1988)). Another empirica1 regularity is the correlation of wages across different occupations within a firm: i£ one occupation in an industry or firm receives a relatively high wage, al1 other occupations also tend to be wel1 paid. This advances the hypothesis that wages are related to the overall profitability of the firm and the bargaining power of employees to secure a share of the profits. This hypothesis may provide an explanalion for the observed inter-industry wage differentials. Firms in more profitable or concentrated industries are able to pay higher wages to employees at al1 levels of the organisation. Consequently, unions are expected to bargain harder in these industries. Union density, as measured by the proportion of workers in an industry belonging to a union, is found to be correlated with the industry wage rate (Thaler (1989)). However, unionisation increases wages for both union and non-union workers. Moreover, the observed union density may not be an effective measure of the impact of unions on wages, as the mere threat of unionisation and collective action might
enhance firms to offer supra competitive wages In order to prevent unionisation (Diclcens f 1986)). In developing countries, product markets arc typicaily characterised by oligopolistic market structures, with one or a few often stateowned or foreign owned firms occupying a dominant positiori in their industry. 111 the next sections ir wil1 be analysed wlietber firm market power has a significant impact o11 the wage level when labour quality differences, working conditions and firm characteristics which affect worlter's productivity are taken into account. However, tlae analysis does not seek to perform a strict test to discriminate between tlie different models of wage deterinination. Tlie different ~nodelsrather provide a theoretica1 background against which the explanatory power of certain variables ina the analysis ean be understood.
IV. CHAMCTERHSTHCS OF THE LABOUR MARKET IN COTE D'IVQIRE The relationship between emgloyers and their employees is regulated in the 'Code du Travai17,which was adjusted in 1995. The main issues regulated in the 'Code7include minimum wages, employers' social security contributions for labour, restrictions on laying-off and restrictions on the geriod temporary worlters can be hired. About the determination of wages, the code generally advances equal pay for employees irrespective of their sex, age, asceiidance, race, religion, poIitical or religáous 'oeliefs, sscral origin and adherente to a labour union. The Code allows freely khe creation of labour unions. Until 1994 the UGTCI (Union GénéraPe des Travailleurs de la Côte d'Ivo8re) was the only Pabsur union which was linked t s the politieal party of tlie president. In 1995 some new labour unions eimerged, the most important one being 'DDignité'. Tiie Code foresees a process of wage determinaliion by an interprofessional collective bargaining agreement ('Co11 ventions Collectives'), in whicli unions, employers and the government are represcntcd. The labour force in C6tc d'lvoire is relatively wei! educated fellowing Afriean standards. At the encl of the 1970's education 'oecame one of the priority investmerit targets of the Government, which it remailred during tlie severe crisis in the 1980's. The labour market is âlso characterised by high foreign presence. As a result of fsrmer policies high proportions of skilled and unskilied worlters immigrated
from other West-African countries, Burkina Faso, Mali, Togo, Ghana, Senegal. Non-Africans, mainly French and Libanese, are largely represented in the upper segments of the labour market. Eyssen (1997) found colivincing evidence that West African iinmigrants are discriminated in tlie manufacturing sector in Côte d'Ivoire7: West-African irnmigrants earn less than Ivorians. Eyssen finds also empirica1 evidence suggesting tliat femaie workers earn less than their male colleagues. Schultz (1993) on the contrary does not find strong indications of gelider reiated wage differentials in CBte d7Hvoire.However, he does noi control for firm or job specific charaeteristics. In a study by Hoddinot (1993) on tlie wage deterininants in urban areas in Côte d91voireevidence is found supporting the efficiencywage model. This is done by analysing the relationship between unemployment and wages, which is found to be negative. h a m and kesueur (1995) also analyse efficiency wage payments by manufacturing firms in Côte d'Ivoiïe and found evidence supporting the hypothesis that manufacturii~gfirms trade off higher wages against more supervisisn. The effect of Girin market power on wages has however not been analysed. Nevertheless, the Ivorian economy is characterised by oligopolistic structures and high levels of concentration in many sectorss. The dominant position ssme firms oecupy in their industry is likely to lead to higher levels of profitability. The effect of marltet power and associated profitability on the wage level of the workers is analysed in the next section.
V. THE MODEL Following the inodel developed by Bughili (8993) the wage level of employees determimed tlnrough union bargaining in a 'riglit I s manage' setting can be eskimated by tlre followinig equation:
where W, is the wage level in firm i, MP, is the inarginal productivity of labour in firm i, nii is the absolute value of the inverse residual demand elasticity which is inversely related to the firm's profitability (P,) and its marltet power. Profitability in turn depends on the firms market share (S,) and on the degree of colli~sionin the industry (Ci).
In an 'Efficient Bargaining' model however, the union's utility level is also determined by the level of employment. Hence, describing union preferences by a quasi concave fuilction in wages (Wi) and eniployment, (L,), U, (W;,L; Wi4:), the wage level is determined by:
where h, is the absolute value of the elasticity of the wage with respect to employinent along the union's indifference curve, or,
In line with efficiency models of wage deterinination, productivity of the individual worker j working in firm i, MPii, depends, besides on his observable labour quality (Q,), also on his earned wage relative to the efficiency wage Wi".
Assuming firms are maxirnising profits paying wages close to the efficiency wage level, wages are determined by characteristics of the firm (Xi) which tend to affect the level of the efficiency wage. Substituting equation (3) in (2), the following estimating model is proposed:
In line with the model, the dependent variable, W,, is the logarithm of total earnings (LEARN). Total earnings are defined as the sum of the individual worker's base wage, the average allowance paid by the firm to the worlters for that occupation, and the average social benefits paid by the firm. Paid leave is also considered as social benefits9. To control for differences in labour quality, Qj, a set of variables is included capturing forma1 education and genera1 and job specific ex-
perience. The appendix illustrates the construction of the variables and presents some summary statistics. The variables for formal education are al1 binary variables. They refer to the employee's highest level of formal education attained and are therefore mutually exclusive. The reference group are workers without any formal education. For primary education a distinction is made between persons who started and never finished primary school (SPRIM) and those who finished primary education after six years of study (CEPE). For secondary education a distinction is made between individuals with a technical formation (CAP, OTSEC for lower secondary or equivalent and BACT for higher secondary) and those who attended classica1 secondary education (SSEC, BEPC for lower secondary or equivalent and BAC for higher secondary). Higher education after secondary school is either technica1 (SUPTEC) or academic (UNIVIV and UNIVFOR for individuals with an acadernic degree from an Ivorian respectively foreign university). Genera1 experience is captured by the individual's age (AGE and AGESQ) and former apprenticeship (APPR). Firm specific human capita1 is measured by seniority (TENURE). A second set of variables is related to the function within the firm and the requirements of the job. The reference group are production workers. Binary variables are included for managers (MANAGER), administrative personnel (ADMIN), salesmen (SALES), supervisors and foremen (SUPERV), and workers responsible for equipment maintenance (MAINTEN). A number of variables are included to test for negative wage discrimination against feinale (FEMALE) or WestAfrican workers from neighbouring countries (WESTAFR) and for positive discrimination of non-Africans (NONAFR) or relatives of the owner of the firm (RELAT). Workers' productivity is further determined by characteristics of the firm. Capita1intensity (LCAPINT2) and firm size measured as the logarithm of the number of employees (LEMPL) are expected to have a positive effect on wages. Similarly, the age and location of the firm, the formal and legal status, the sector of activity and the ownership structure are included. Firms operating under competitive pressure with low profitability, such as informal firms, may be expected to pay lower wages. Moreover, firms hire labour from fragmented labour markets, such that the competitive wage and hence the efficiency wage may differ for the different groups of firms. A set of binary variables is entered into the equation for firms located outside Abidjan (NABID), for informal firms (INFORMAL), and firms belonging to a group
(SUBSID). Firm age is measured in logarithmic terms (LFIRMAGE). A set of sectoral variables (WOOD, METAL, TEXTILES) is included, the reference group being firms active in agro-industries. The market power of the firm, proxied by firm profitability P, is estimated as sales minus wages, direct and indirect costs proportional to sales (PROFITS). In a related study Goedhuys (1999) found that firm profitability is strongly determined by firm market shares, S,, besides sectoral effects. The effect of market shares is therefore indirectly accounted for in the estimation through its effect on profitability. By the time the interviews were done there was virtually only one labour union active in the country. The unions relative preference for employment versus wage, U,, is therefore taken to be constant. Further effects of union activity is measured by a binary variable (UNIONACT) equalling one if there is a union active in the firm. Union activity may als0 be stronger in state owned firms (SOE). While the importance of certain firm characteristics on the determination of wages is expected to be significant, the included explanatory variables do not allow to test different theories or to support one theory over another. Firm characteristics such as location, informal status and ownership structure, refer to very different groups of firms, operating under different competitive regimes and in different labour markets. These variables therefore do not allow to disentangle effects of wage bargaining from efficiency wage payments. VI. ESTIMATION AND RESULTS Following Donvick (1990), firm profitability is instrumented to account for endogeneity. The estimation results are presented in Table 1. The adjusted R-square statistic indicates that the model explains about 63% of the variation in the dependent variable. The average monthly wage equals 90420 F.CFA which at the time the interviews were done had an equivalent of approximately 186 USD. The base wage for a production worker who has not received any kind of formal education and who is active in a formal firm in agro-industries in Abidjan, equals 10162 E.FCA or 21 USD. It should be noted however that one year prior to the interviews the F.CFA devaluated by 100%, thereby sensibly lowering the wage levels when expressed in foreign currency. The results show in the first place that returns to formal education are significant. For about 32% of the workers in the sample, the cer-
tificate of primary school was the highest level of forma1 education ever obtained. Primary education, where basic literacy and arithmetic skills are taught, tends to increase earnings by 13%1°, as indicated by the coefficients of the variables SPRIM and CEPE. The initia1 years of genera1 secondary education do not seem to be rewarded above the mere certificate of primary school, as the coefficients of the variables CEPE and SSEC are of comparable magnitude. However, the accomplishment of six years of genera1 secondary school and the obtention of the 'Baccalaureate' (BAC), a certificate that grants students access to higher education, increases wages by 86%. A technical formation at the level of secondary school seems particularly valuable and productivity raising in the manufacturing sector and tends to increase wages significantly above the wage earned by workers without any formal education. Especially at the first years of secondary schooling, a technical education seems to be more rewarding than a classica1 education. The certificate of professional ability (CAP) which is obtained after four years of technical training and education raises wages by 26%, while six years of secondary school and the obtention of the 'Baccalauréat Technique' raises wages by 63%. Higher education, which is received by approximately 7% of the workers in the sample, raises wages substantially. The wage rise equals 148% for individuals who received technical higher education (SUPTEC) and 127% respectively 288% for individuals who obtained an academic degree from an Ivorian respectively foreign university or institution. Genera1 professional experience, measured by the worker's age, is also rewarded by higher wages. However, the positive effect of ageing decreases for older workers as indicated by the negative effect of the variable AGESQ. The coefficients indicate that positive returns to ageing fa11 to zero at the age of 49 to become negative thereafter. The effect of having been an apprentice formerly does not seem to have a significant influence on the wage level. Firm specific experience on the contrary does translate int0 slightly higher wages. Measured at the sample mean values, an additional year of job seniority results in a 0.5% increase in the wage level. Wages are also strongly determined by the responsibilities related to certain functions in the organisation. At the sample mean, salespersons and maintenance workers have their wages raised by 24% above the wage level paid to production workers, while administrative personnel, supervisors, and especially managers, have wages at a 30%, 40% and 60% higher level, al1 else equal.
TABLE 1 Wage detennilzants in the Côte d'Ivoire nza~zufacturingsector Coeffient
Standard Error
Depeiideiit variable: LEARN constant SPRIM CEPE CAP 0.238 OTSEC BACT SSEC BEPC BAC SUPTEC UNIVIV UNIVFOR AGE AGESQ TENURE APPR MANAGER ADMIN SALES SUPERV MAINTEN RELAT DFEMALE NONAFR WESTAFR TEXTILES WOOD METAL LEMPL INFORMAL NABID SOE SUBSID LFIRMAGE UNIONACT PROFITS LCAPINT2
Note: Standard errors are estimatcd using White's consistent estimator (White, 1980); Significante levels: :li""' 1%;* " 5%;
':10%.
f ome discriminatory practices are indeed suggested by the data. Female workers seem t s earl1 up to 14% less than eqiaally skilled male employees doing the same job in a similar fism. The sarne holds for non-livorian African employees who make up 19% of the workers In the sample and wbo seem to earn 7% less than their Evorian cslleagues. The opposite holds f s non-Mrican ~ employees, and the effect is found to be large and significant. Being non-African entails a wage increase of 238%. A slightkj positive discriminatio~iis observed for relatives of the owner or nzanager of the firn, however, the effect is n s t significant. Besides the characteristics of ihe workes and tlne requirements of the jok wages seem to be deteimined lo a large extent by thn c6larasteristics of the firm wlieii-ethe ii~dividrialis ernployed. The positke firm size-wage relationship, which is obsemed in many empirica1 studies, is also obsewed in the 1voria-wmanufacturirig sector. Aceording tss the estimates, a firm with l.vice the average eanplayment Hevei, 92 employees instead of 46, ivould gay about 9 % higher wages to al1 employees. Firm size ma)? proxy i~nobservedwage deterrninants, such as higher work dlcintiPity in larger firms due tio higher Labeur division or a more complex technoiogy. Higher capital inlensiky and more intense union activity, which may als0 be correlated with f i r n size, are accounted for by the variables %CAPINT2 and UWIONACT Both variables have a pssitive coefficient, kut the effect is nob significant. As firm size inight encomgasc the effect of union activity, tlae eqiaatien was reestimaled without inslusion of the variable LEMPL, which resulted in a positive and significant coefficient oïthe variable UNIONACT, which increases wages by 9%. Sectoral variabies also have an important impact on the wage level of the individual employee. Firrns in wood woa-Eng (WOOD) and in metal working (METAL) lens4 io pay syshernatieahly higher wages as compared to f i m s 4n agro-industries, tbe referente groiip. In a similar way, firms ia the inforrnal sector (INFORMAL), which tend lo be of a smaller size but large in number and operating under high competitive pressupes in their product market, gay about 34% Iower wages than forma1 firms. About 85% of the wsrkers in the sample are employed by farms Located in Abbdjan. These ffïms aiso tend to pay 23% higher wagec to their employees than films Iocated in the regions of Boualié i11 the centre of tbe country, or in San Pedro in the west (NABID). This may be explained by the fact that outside the industrial core region of Abidjan local dernand for labour is relatively low while a,
labour is abuiidaiatly available. Moreover, a higher cost of living in Abidjan may increase the opportunity cost for working i11 Abidjan. Firms belonging to a group (SUBSID) pay 50% more to equally traiiied employees. State owned enterprises (SOE), most of which have a dominant position or are working with a soft budget consiraint, also pay relatively lower wages. Tliis may be due to the i'act tliat state owned enterprises can offer higher job security to tlieir workers which BS coinpensated for by a somewhat lower wage. While controlling for labour quality differences and firm characteristics, the performance of the firm, as measured by firm profitability (PROFITS), has a large positive and significant effect on wages. Estimated at hhe sample mean, the elasticity of the workess' wage with respect to the profitability of the firm equals 3.71, inadicating that a ten percent rise in rhe profit margin of the firm, ii-on~the average profbt margin of .230 to 253, results in a 37.1% rise in the wages paid to al1 workers. This evidence supports the hypothesis thar workers collectively profit from the market power of the fisrn, helped by labour Begislation, which foresees a wage nego~iationmechanisrn between unions, employers and government officials. As mentioi~edearlier, an additional effect of union activity (BIPIJIONACT) is notsignificant in the estimation, partly due to correlation with firm size, but als0 because the impact of unions on the determinatioii of wages is typically characterised by measurement difficulties, as the mere threat of unionisation may result in worliing conditions which are similar to the conditions which would result from effective union presente. ALso at the level of the individual werkers in a firm, adherence to a labour i~nion does not translate into higher wages tlaan those paid to non-affiliated workers, which is also in line witl-i the specifications of the Labour Code.
This paper anabses wage differefitials as they are observed i11the manufacturing sector in Côle d'lvoire aiid investigates how alternative n~odelsof \vage determination may explain the obseïved differentials. In line with neo-classical theory, the results suggest that wages are positively related to worker characteristics which tend to increase worker9sproductivity. Returns to formal education are found to be highly significant. Wages rise consistently with higher levels of formal education and with higher responsibilities related to certain func-
tions in the organisation. Also technical and firm-specific skills acquired through technical education, through ageing and seniority increase wages substantially. Negative discriminatoiy practices against non-Ivorian West African and female workers and positive wage discrimination towards non-Africans are also observed in the Ivorian manufacturing sector. Besides worker characteristics, firm characteristics also have a significant impact on the wage level paid to employees. As such, larger firms, forma1 firms, firms located in Abidjan and firms belonging to a group pay significantly higher wages. This may indicate that different levels of technological complexity, different degrees of labour division or labour quality are compensated for. In larger firins tl-ie moiiitoring of workers becomes indeed more complex while at the same time far-going labour division increases disutility related to the jobs. On the other liand, these firm characteristics may also define the niches in which the firins operate, and hence the intensity of competition firms are exposed to. That the product market structure has an effect on wages paid to employees is supported by the data. They strongly suggest that wages are related to the position of the firm in the market and the associated profitability. More profitable firms are found to pay higher wages. Inter-industry differentials also remain significant in the rnanufacturing sector in Côte d'Ivoire even when capita1 intensity and relevant job characteristics are accounted for. These findings suggest for collective bargaining practices by which workers are granted some degree of bargaining power. Collective wage determination processes are indeed advanced by labour legislation and seem to be implemented especially in those firms which have a dominant position in their product market. NOTES 1. With tlie exceptioii of Teal (1996) ~isingdata froin Ghana, most stuclies using developing country data argue against a rent sharing interprctation (see among otliers Mol1 (1993) for South Aí'rica and Morrisoii (1994) for Eciiador). 2. It is in the workcr's interest to shirk il'g > p (w-wo) N, where g is thc ainount thc employee can gain by shirlcing, irrespective o l whcther or not hc is cauglit shirking, p is tlic probability tliat sliirkiiig is detected, w is the wage earncd, wo is the opportunity wage, N exprcsses tlie long-term value of the rclationship, nieasured as thc number of periods the worker caii work in tlie firm if iiot cauglit shirking. In order to prevent shirkiilg, firn1 can set higlier wagcs (w) or iiicrease tlie probability of being caught shirlcing (p) tlirough more severe supervision. The smallest wage that can detcr chcatiiig is thc efficiency wage aiid is given by w = wo g / (Np). Higher levels ofunemploy-
+
3.
4. 5. 6.
7.
8. 9.
10.
ment decrease the opportunitywage wo and wil1 tend to lower tlie efficiency wage. Higher levels of supervision compeiisate for lower efficiency wages. Using this condition to distinguish betweeri efficiency wage payments and rent sliaring, Konings and Walsh (1994) find evidencc supporting efficiency wage paymciits in UK coilipanies on the basis of a positive effect ïrom wages o11 firrn perforn~aiiccas measured by market share, in low unionised firnis. In high unionised firins evidcnce is found tliat fii-ins paid iiivoluiltarily high wages wilh a detrimental effcct oii market shares. Recent cvidciice is provided hy Ringuedé (1998) using French data. H e finds that above a critical firm size wages rise with firm size, as coiisistent witli efficiency wage Lheory. As measurcd hy education, tenure ancl age (Krueger and Summers (1988). Dickens and Katz (1987)). Murphy and Topel (1987) argue that different sectors attract laboui- of different unobserved and unmcasured quality. This hypotliesis is jeopardised by thc observed uniformity of wage diiferentials across occupations and is rejccted by a iiumber of empirica1 studics which tend to analysc the effect of unobserved labour quality differences (Kriieger and Summers (19SS), Gibbons and Katz (1989), Blaclcburn and Neumarck (1993)). The efficiency wage hypotliesis provides an explanation for discrimination among workers witli diSSerent observable characteristics (Yellen (1984)). Firsi, if jobs are rationed, an employer can satisfy bis taste for discrimiiiation ai a zero cost. Sccond, the function relating the effort level to the wage may differ ainong groups, so that each group has its own corresponding efficiency labour cost. Following World Bank estiinates (1994) of the 100 largest firms, nine firms create 90% or more of total val~readded i11their respcctive sector of activity, while another 11 firms have a share between 50 and 90% of value added in their sector. The 'Code du Travail' provides two days of paid holiday per month of effective work. Workers having 15, respectively 20,25and 30 years of tenure in the firm have two, rcspeclively four, six and eight extra days of paid leave per year. In practice however, paid liolidays are provided by most firms in the forma1 sector, but by very few firms in the informal sector. For workers in Sirms which provide paid holidays, montlily wages are multiplied by 32/11 to account for this benefit. For dummy variables, the percent change in wages cquals tlie antilog of the estimated coefficieiit minus half tlie variance of the coefficient, minus one, as suggested by Kennedy (1981) and Halvorsen and Palmqiiist (1980).
REFERENCES Aclcerlof, G , , 1982, Labor Contracts as Partial Gift Exchailge, Qilarterly Joclrrzal of Ecorzo~ n i c s97, , 543-569. Ackerlof, C., 1984, Gift Exchange and the Eí'liciency Wage Thcory: Four Views, Americnrz Econornic Review, Papers and Proceediiigs, 74, 79-83. Azam, J.-P., Lesueur, J.-Y., 1995, Efficiency Wage and Supervision in the Ivorian Manufacturirig Sector, Worlting papcr, (CRDI, Clerrnont-Fcrrand). Blackburn, M., Neuniark, D,, 1993, Oniitted-Ability Bias and tlie Increasc in the Return to Schooling, Jolrrnal of Lnbor Ecoiznnzics, 11, 521-544. Bughin, J., 1993, Uiiion-Firm Efficieiit Bargaining and Test o l Oligopolistic Conduct, The Review of Economics a~zdSiniistics, 75, 563-567. Caniphcll, C. M., 1993, Do Firms Pay Efficiency Wagcs'? Evidence with Data at the Firm Lcvel, .lóumnl o j Labor Ecotzol?~ics,11, 442-470. Dickens, W.T., 1986, Wages, Employmcnt and the Threat of Collectivc Action by Workers, (University of California, Berkerley), niimco. Dickens, W., I
Eyssen, H., 1997. Are West-African Imniigraiits Discrirniiiated in Côte d'lvoire?, Center for Economic Studies, Discussioii Paper Series, 97.03. Gibboiis, R., Katz. L.. 1989, Does Unmcasured Ability Explain Inter-Iiidustry Wage Differences?, NBER Working Paper Series, 3182. Goedhuys, M., 1999, Iiidustrial Organisation in Dcveloping Countries, Evidcnce Srom Côte d'Ivoire, Doctoi-al dissertation; (Iiatholieke Universiteit Leuven, Departnicnt of Applied Ecoiiomics). Halvorseii, R., Palmquist, R., 1980, The Interpretation o f Dummy Variablcs in Semilogarithmic Equaïions, Arneric~nrzEconol7zic Review, 70, 474-475. Hoddinott, J., 1993, Wages and Unemploymeiit in Urhail Côtc ci'lvoire, (Centre î'or tlie Study of Mrican Economies), Working Paper Series, 93.3. Kennedy, 1-1, 1981, Estiination with Correctly Intei-preted Dummy Variables in Semilogarithinic Eq~iations,Arnericatz Econor7zic Rei~iew,71, 801. Konings, J., Walsh, P., 1994, Evidence of Efficiency Wagc Payment in UKFirrn Level Panel Data, Econonzic Joz~rnal;104, 542-555. Krueger, A., Suminers, L., 1988, Efficiency Wages and the Inter-Industry Wage Structurc, Econometrica, 56, 259-293. Lawrence, C., Lawreilce, R., 1985, Manufacturing Wage Dispersion: ai1 End Gamc Interpretation, Rmoking.~Papers on Econonzic Activitj~,47-106. Mellow, W., 1982, Employer Size and Wages, Tlle Review oJ'Econonrics nrrd Stntistics, 64, 495-501. Moll, P. G., 1993, Industry Wage-Differentials aild Efficiency Wages: a Dissenting Viewwith South Mrican Evidence, Jouvzal o f Developmeizt Econotnics, 41,213-246. Morrison, A. R., 1994, Are Institutions or Economic Reiits Responsible for Interindustry Wage Differeiltials, World Developnzent, 22,355-368. Murpliy, K., Topel, R., 1987, Uncmploymcnt, Risk and Earnii~gs:Testing for Equalizing Wage Differences i11 the Labor Market, in Lang, K., Leonard, J., eds., Unemployment and the Structure of Labor Markets, (Basil Blackwell, London). Ringuedé, S., 1998; An Efficiency Wage Model for Small Firins: Firm Size and Wages, Economics Letters, 59, 263-268. Salop, S., 1979, A Model of Natura1 Rate of Unemploymeilt, Anzericnr~Econonzic Review, 69, 117-125. Scliinidt, Ch. M., Zimmermaiin, K. F., 1991, Work Characteristics, Firin Size and Wages, The Revieiv of Ecoiro~nicsnnd Staiistics, 73, 705-715. Scliultz, T. P., 1993, Invcstments in the Schooling and I-Iealth of Woineii and Mcn: Quantities and Returns, Tlze Joz~rrzulof H~lmarzResoul.ces, 28, 694-734. Shapiro, C., Stiglitz, J.E., 1984, Equilibrium Unemployment as a Worker Discipline Device, Ainericnn Econotnic Review, 74, 433-444. Stiglitz, J. E,, 1974, Alternative Theories of Wage Determination and Unemployment in L.D.C.'s: the Labor Turnover Model, Quaterly Jounzal oJ'Ecotlornics,88, 194-227. Tlinlcr, R. EI., 1989, Anomalies: Intcrindustry Wage DiSfere~itials,.Io~~n~al qj'Econon7ic Persl~ectives,3, 181-193. Teal, E, 1996, Tlie Size and Sources of Ecoiioniic Rents in a Developing Couiitry Maiiufacluring Labour Market, T11c Ecorzonzic Jou~irnl,106, 963-976. Ulph, A., Ulph, D,, 1990, Uriion Bargaining: a Survey oí'Receiil Work, in Spasí'ord, D,, Tzaniiatos, L., Ctirrent Issues in Labour Ecoriomics, (MacMillan Educatioi~alLtd., Loiidon). Yellcii, J. L., 1984, Eí'l'iciency Wage Models of Unemploymcnt, Ainericail Ecorlonzic Review, I'crl,a:r nnd I'roceeditigs, 74, 200-205. Weiss, A.. 1980, Job Q~icuesand Layofí's iii Labour Markets with Flexible Wages, J o z ~ f i r ~ l oJ'Polilicri1Econonzy, 88, 526-538. White, H., 1980, A Heteroscedasticity Co~isistentCovuriaiice Matrix Estimator and a Direct Test for Hctcroscedasticity, Ecoi~onzetrica,48, 817-832. World Bank, 1904, Rcpublic of Côte d'lvoire Private Scctor Assessinent, Report ilo 12885IVC, (Tlie World Bank, Washington D.C.).
APPENDIX Construction of t h e variables VarialAes
Mcaii Value Dependent variabie LEARN 11.417 Humali capital variables SI'RIM 0.128 CEPE 0.110 CAP 0.051 OTSEC
0.036
BACT
0.080
SSEC
0.213
BEPC
0.074
BAC
0.021
SUPTEC UNIVIV
0.039 0.016
UNIVFOR
0.015
general experience AGE 35.17 AGESQ 1313 APPR 0.414 TENURE 7.953 job characieristics MANAGER 0.068 ADMIN 0.11 1 SALES 0.050 SUPEIIV 0.059 MAINTEN 0.021 RELAT FEMALE NONAFR WESTAFR
0.066 0.095 0.018 0.193
logaritlini of totai earnings per worker (base wagetallowanccs+benefits) = l for workers who started primary educatioii bilt ciidn't fiiiisli it; = l for workers who finished priinary cdiicatioii; = 1 for workers who hold tlie 'Ccrtificat d'Aptitude Professionelle' (lower teclinical secondary school; 4-years); = l lor workcrs wlio Iiold a dcgrec at lower technica1 secoiidary level, other thaii CAP; = I for workers who Iiold a 'Baccalauréat Technique'-degree (higher tcchnical secondary scliool; 6 years); = l for workers who staried higher secondary educatioii but didii't finish ii; = l for worlcers who hold the 'Brcvet $Etudes duPremier Cycle' (4 years of geileral secoiidary school); = l for workers wlio hold a 'Baccalauréat' (complete geiieral secondaiy school; 6 years); = 1 for workers who hold a degree at higher teclinical level; = l for workers who obtained a ~iniversityor 'Grande Ecole' clcgree in Côte d'Ivoire; = l for workers who ohtained a university or 'Grande Ecole' degree abroad; age ol tlie worker, in years (AGE)' = l lor workers wlio have been appreiitice; years worked with current eniployer; = l lor workers iii maiiagcment funclions; = l Lor workers iii adiniiiistrative fiinctioiis; = l for workcrs in sales lunctioiis; = l for workers lor supervisors aiid fel-emeii; = l f»r workers in cquipinciit mainleiianee; = l for workers wlio are relalives of thc owner; = l for female workers; = l Ior workcrs of iion-African cthnic origin; = l for workers of Wcst-Africaii iion-Xvorian origin;
Firm cliaracteristics TEXTILES WOOD METAL INFORMAL NABID SUBSID SOE LFIRMAGE LCAPINTZ
UNIONACT
0.144 0.310 0.265
= l for workers iii textiles arid clotliing; = l for worlcers in tlie wood seclor: = l for workcrs in the inetal workiiig sector: = l for workcrs iii the inloriiial seclor; = l for wol-kers cmployed iii firms located outside Abidjaii: = 1 for workcrs in firiiis belongiiig t« a group; =l for workers in stalt owned firms; age of the fii-m, i11 logarilhmic lerms; capital i~itcnsityrneasured as Llie cost olelectricity, water. lelephone and fuel per employee, in logarithmic tcrms total number of cinployces working in the firiii, in logarithmic terms: proiitability, iueas~iredas (sales-wages-direct costs-indirect costs) 1 sales = l for firms wliere tliere is uiiion activily
Tijdschrift voor Economie e n Manageinent Vol. XLY 2.2000
RooMieren in de markt? door R. DE BONDT:':
I. INLEIDING
Eén van de zeer controversiële aspecten van marktgedrag betreft het al dan niet voorkomen evenals de gevolgen van agressief en vernietigend gedrag, waarbij bestaande sterkere bedrijven de vaak kleinere en nog makkere rivalen zouden vernietigen of verzwakken, een beetje zoals grote roofdieren doen, om sterk te blijven of te worden. Recent is er een hernieuwde belangstelling voor dit (vermeend?) gedrag, vooral wellicht tengevolge van de aanklacht en de juridische uitspraak tegen Microsoft. In vele landen is zulk een gedrag door dominerende spelers verboden, omdat het buiten de regels van normaie mededinging valt. De overheid is hierbij de scheidsrechter die via klachten van de makkere moet oordelen of er al dan niet sprake is van agressief gedrag. Dit is niet zo eenvoudig omdat het o.a. samenhangt met een visie over concurrentie. 91. VISIE OP CONCURRENTIE
Stel even dat we sommige sporten als een metafoor van concurrentie gebruiken, dan komt het erop aan om onsportief, "mil" of agressief gedrag buiten de arena te houden. De vraag is echter welke sport een goede vergelijking oplevert. Sommigen zullen concurrentie vergelijken met hardlopen, b.v. de 400 meter. Zeker nu we in een tijd komen van producten met kortere levenscycli. In zulk een wedloop wordt het zelfs maar eventjes verla* Departement Toegcpaste Economische Wetenschappen, K.U.Leuven, Leuven. Met dank aan Electrabel leerstoel Bedrijfseconomie en aan de IUAP P4128, "Strategische analyse van organisaties".
ten van de eigen baan al gezien als onsportief gedrag, dat indien het wordt opgemerkr tot uitsluiting zal leiden. Anderen zrillen creativiteit en onzekerlieid als belangrijke ingreditsnten van mededinging beschouwen. De vergelijking met voetbal lijkt dan meer aangewezen. In dit spel kunnen de spelers allerhande vuile en zelfs gevaarlijke trucjes gebruiken, maas dan wordt liet spel zeer ongenietbaar eri kan het uiteindelijk uiltaarden. Maar toch is voetbal geen zacht spei. Sci~oudertegen schouder is toegelaten en zolang de bal gespeeld wordt, mag relatief veel. "De bal spelen" is hierbij een goede regel die koelaat om het onderscheid te maken trissen hard en vrril spelen. Bij hardlopen en voetbal is het resultaat echter al bekend na één spei of ékn seizoen. In de economie is er onvermijdelijk een langere evolutie doorheen de tijd, waarbij sommige spelers in de markt nu eenmaal beter in staat zijn oin te overleven dan andere. Omdat ze b.v. beter inspelen op de wisselwerlcing tussen het bedrijf en de omgeving, b.v. via drastische innovatie of voortdurende verbeteringen. Bij de creatieve vernies~wingof het zelf organiseren van de economie zullen hierbij onvermijdelijk sommigen aan belang verliezen en anderen aan belang winnen. Misschien bracht een ingebouwde zwakheid een bedrijf dicht bij de afgrond, waarbij een klein "briesje" van de rivaal volstaat om het in de afgrond te doen tuimelen. Veel nieuwkomers weten vaak niet of ze we4 Bsek~aamzijn om te overleven, ze gaan het eens proberen. Zij die talenten en middelen kiinnen combineren tot toegevoegde waarde zullen overleven en groeien, maar daarvoor is meestal een voortdurende opbouw van kennis en kapitaal en een aangepast veranderen nodig. Na eer1 zekere tijd zullen zeer vele bedrijven het daarom vaak laten afweten, zonder clat dit iets te maken heeft met agressief gedrag van rivalen (in de zogeiiaamde "chalieout") (Klepper (1997) en Klepper en Simons (1997)). In de globale economie kunnen deze tendensen nog sterker worden, omdat vele markten "alles voor dc winnaar" karakteristieken gaan vertonen, zoals in software en trendy klcding. De klant kiest alleen het beste en de ietwat mindere lwaliteit belandt in niets betekenende niches of verdwijnt noodgedwongen. Naargelang concurrentie als een statisch, creatief of dynamisch geheel bekeken wordt, zal de "normale" markpaerkii~gdus iets anders gaan betekenen. Wat ongewenst gedrag is bij één spel, kan nodig zijn bij een ander om te overleven. Het opstellen van regels voor gedrag die ergens vermijden om het kind niet het badwater weg te gooien, is
dus niet zo eenvoudig. Zo zou koppelverkoop van verschillende software nodig kunnen zijn om de consument te overtuigen van de coinplementariteiten tussen de pakketten. En mogelijk kan alleen via zulk een bundelen een voldoend ecoi~omischrendement voor de innovaties en de gebruikte kennis gerealiseerd worden. Maar hierdoor kunnen spelers die maar één pakket produceren het nioeilijker krijgen. In de markt van Internet software was Netscape enkele jaren geleden zeer verbolgen over het bundelen van Microsoft (van Windows met o.a. Internet Explorer) (zie Daly (1994)). Maar bundelen is één van de routine tactieken geworden die wordt voorgeschreven aan bedrijven om te kunnen overleven in het "digitaal Darwinisme" (zie Schwarte (1999)). Hierbij komt dat er maar onvolkomen informatie bestaat over het eigenlijke gedrag van de spelers. In voetbal zal een gevallen speler misschien wat toneel gaan spelen om de scheidsrechter proberen wijs te maken dat de bal niet gespeeld werd. Hn de economie kan hehnoeilijk zijn om te wcten te konaen of lage prijzen een gevolg zijn van eerlijke csnci~rrentleof van een streven naar eliminatie. Rechtbanken en advocaten zullen het bewijsmateriaal beoordelen en toelichten. Er bestaat onvermijdelijk een ruimte voor het maken van fouten. Wet is mogelijk dat een "snscliuldig" bedrijf toch "schuldig" bevonden wordt (ten onrechte positief), of dat een "schuldige" toch vrijuit kan gaan (ten onrechte negatief). Een conservatieve regel zal proberen de waarschijnlijkheid van een ten onrechte positief oordeel klein te houden. Men zal niet "vlug" tot "schuldig" besluiten. De keerzijde is dat het waarschijnlijkcerwordt dat ten onrechte besloten wordt dat er geen vuiltje aan de lucht is. Eeii inspiratie voor deze houding is wellicht dat men zwaar tilt aan het feit dat ten onrechte oordelera dat er agressieve concurrentie is, succes era creativiteit zal bestraffen en ontnioedigen en de normale dynamiek of evolutie zal tegenhouden of teveel kunstmatig vervormen met onvoorzienbare gevolgen. Een mecr populistische aanpak zal vlugger besluiten dat het gedrag onaanvaardbaar is, en dus een kleinere Icans inhouden dat er ten onrechte "onschuldig" beslist wordt. Dit omdat men vuile concurrentie door de sterkere in het nadeel van de kleinere wil ontmoedigen of omdat men zwaar tilt aan welvaartsverliezen. Tegelijkertijd, echter. wordt de kans groter dat naen ten onrechte positief zal oordelen en dus dat inen de dynamiek van de concurrentie en innovatie ten onrechte aan banden zal leggen.
Soms is de maatschappij bereid om dit risico te lopen omdat agressief gedrag tendensen kan inhouden tot het monopoliseren van de markt. De essentie van een markteconomie is immers dat de keuze van de consument uiteindelijk bepaalt welke goederen blijven en verdwijnen. Het gedrag van de aanbieders kan als het ware, door het onverantwoord uitschakelen van concurrenten, een soort van "kunstmatige" keuze maken in de plaats van de consument. Mogelijk komt dit neer op het brutaal moeilijk maken van de keuze, alleen maar om de eigen positie te versterken, en met weinig of geen andere voordelen voor de economie. Maar hoe dan ook gebeurt de rechtspraak binnen een culturele omgeving die mogelijk mede een vertekening in één of andere richting zal inhouden, e11 die moeilijk volledig kan vermeden worden gegeven de onvolkomen informatie bij het observeren en beoordelen van het gedrag.
III. NIET-MTIOMEEL? Zowel in Europa als in de Verenigde Staten wordt het agressief gedrag van dominerende spelers kritisch bekeken. Voor Europa zijn weinig gegevens bekend maar aan de andere kant van de Atlantische Oceaan is de slinger duidelijk in de richting gegaan van: "Speel maar voort, er is niets (abnormaal) aan de hand". Dit tot grote ergernis wellicht van zij die dichter bij een populistische vertekening aanleunen. De geschiedenis van deze beweging is zeer leerrijk. Verkopen met verlies - prijs beneden gemiddelde variabele kosten - is een symptoom van mogelijk agressief prijsgedrag. Za zou het olieconcern van Rockefeller systematisch lage prijzen gehanteerd hebben om hier en daar concurrenten te elimineren, waarna ze de prijzen terug verhoogden. MC Gee (1958) herbekeek het bewijsmateriaal van het proces tegen het concern in 1911, en hij besloot dat het eigenlijk weinig bevat om te besluiten dat hier roofdiergedrag in het spel was. Anderen hebben later opgemerkt dat het concern misschien wat subtieler te werk ging en er in slaagde om b.v. rivalen hoge vervoerprijzen aan een kartel van spoorwegbedrijven te laten betalen, waarbij het concern zelf belangrijke verminderingen kon afdwingen (zie Granitz en H e i n (1996)). Belangrijker hierbij is echter zijn stelling dat zulk een agressief prijsgedrag vertonen, eigenlijk een beetje "dom" is. De implicatie zijnde
dat het niet zoveel kan voorkomen. En als het toch opduikt zal diegene die het toepast hier via de markt al een "prijs" voor betalen. Waarom zou een bedrijf met b.v. een 80% marktaandeel haar winstgevendheid vandaag opgeven om morgen een stuk van de resterende 20 % binnen te halen? De buit van morgen kan een hele tijd op zich laten wachten, en heeft dus een kleinere actuele waarde. Bovendien is ze onzeker. De verliezen vandaag zijn echter zeker, onmiddellijk en mogelijk groot (indien b.v. de bijkomende productie en verkoop maar mogelijk is mits toenemende bijkomende kosten). De aanvaller kan verkoop met verlies niet lang volhouden. De zogezegde slachtoffers weten dit ook en zullen dus niet zo vlug het terrein verlaten. Indien ze zelf doelmatige rivalen zijn, zal de kapitaalmarkt en kredietmarkt hun winstgevendheid op termijn inzien en ze blijven steunen. Zelfs indien ze het opgeven, heeft de aanvaller weinig bereikt. Van zodra hij terug de prijzen verhoogt, zet hij een magneet in werking die nieuwe concurrentie zal aantrekken. Vaak zullen de activa van de rivaal blijven bestaan. En dan kunnen die terug ingezet worden, misschien zelfs door het vroegere slachtoffer, of door een sterkere nieuwkomer die ze kocht van de benadeelde. Ondanks al de paniek rond Netscape (zie bv. Daly (1997)) verdween het bedrijf niet en werd het gekocht door American Online voor 10 miljard $, in het voordeel van aandeelhouders en consumenten. En het tijdschrift "Wired" kon in december 1998 niet minder dan 83 redenen opnoemen waarom de marktdominantie van Microsoft ten einde zou zijnL. Mogelijk zullen de klanten wel doorhebben dat de Lage prijzen tot doel hebben om een monopolie op te bouwen. Derhalve hebben ze er belang bij de slachtoffers toch te blijven steunen. Wanten kunnen ookvoorraden gaan opbouwen, waardoor het later moeilijker zal worden om, gegeven de kleinere vraag, de prijs te verhogen (Easterbrook (1981)). De benadeelde hoeft ook niet stil te zitten en kan mogelijk de aanvaller lik op stuk geven in een andere markt. Mogelijk hanteren de spelers een hardere vorm van wederkerigheid, waarbij ze starten met vijandigheid en pas normale concurrentie of samenwerking toepassen, wanneer ze elkaar de tanden hebben laten zien. Het is niet uitgesloten dat de spelers gewoon wat onwetend zijn over de voor- en nadelen en dus eigenlijk geen bewuste strategie tot uitdrijven volgen. Of dat ze juist slim genoeg zijn om te weten dat andere wegen beter zijn om hun markt te beschermen. Het opkopen van een kleinere rivaal b.v..
Ei1 tenslotte is het mogelijk niet geloofwaardig om via agressief gedrag rivalen uit de inarkt te houden of te drijven. Bekijk hiertoe een spel met een eindig aantal markten waarbij de spelers elkaar en de omgeving goed kennen, en ze achtereenvolgens een tactiek uittekenen voor elke markt! In de laatste inarkt waar ze elkaar tegenkomen, is het aangewezen om er het beste van te maken, want er komt toch nicts meer. In de voorlaatste markt zouden de spelers er even Bunnen aan denken oin agressief te spelen, om te suggereren dat ze dit later ook nog eens zouden kunnen doen. Z e zouden ze b.v. nieuwe rivalen Ituiiileri ontmoedigen om op de markt te komen. Maar rationele spelers weten dat het in de laatste markt beter is om eerder toegevend te zijn, bijgevolg heeft het ook geen zin om nu te suggereren dat later hard zal worden opgetreden. Men kan dit wel proberen maar de dreiging is niet gelaohaardig. Derhalve is het beter om ook in de oorl laatste markt norinaal te concurreren. En dus ook i11 de markt die komt voor de voorlaatste, enz. Zo ontrafelt het gehele spel tot normale conciirrentie in alle markten.
Men kan dus veawachten dat erkop kopen met verlies" met als doel de markt te monopoliseren, zelden zal voorkomen, gewoon omdat het een zwakke, misschien zelfs domine manier van werken is. Roofdieren moeten in de liatuur svedeven via het opeten van de andere dieren, maar zulk gedrag is geen metafoor voor een concurrentie die binnen normale wettelijke en culturele regels gespeeld wordt (wellicht is het wel van toepassing in de wereld van de "georganiseerde" misdaad). Consistent hiermee is het feit dat klachten over zulk agressief gedrag relatief onbelangrijk zijn in vergelijking met klachten over b.v. het weigeren van leveren (zie bv. PIilips (1995) en Bolton e.a. (1999)). Of 1n.a.w. misschien is de prijs niet liet beste wapen om de markt te beschermen of uit te breiden'. Bij de visie van concurrentie als een voortdurende evolutie, is het toch verleidelijk om de analogie met de biologie te maken. Mensen zijn natuurlijk anders dan dieren, maar de interacties tussen rivalen kunnen gelijkaardige kenmerken vertonen (zie Hofbauer en Siginund (1988)). Zo zal een conventioneel gevecht tussen "havi~"en " d u i l spelers, mogelijk grote kosten meebrengen voor de haviken die met elkaaa- vechten en verliezen. Duiven gaan uit de weg voor de havik en
krijgen dan weinig, maar als ze andere duiven ontmoeten, delen ze de buit. De evolutie zal leiden naar een stabiele verdeling van beide typcs, met weinig haviken en veel duiven. Hetaandeel van de haviken moet dalen naargelang de kost van liun verliezen na vechten toeneeint. Sterkere dieren vechten zelden tot de dood voor b.v. hun terrein of of gewenste "partner". Veelal worden allerlei dreigende sigi~ale~i kreten gehanteerd om de tegenstrever te imponeren. En als ze dan toch vechten, is het alsof ze in een soort van boksmatch zitten, waarin erge schade zeldzaam Ps. Wolven, b.v. zullen zich inhouden om de dodelijke beet aan de opponent toe te dienen, indieai die zijn nek uitsteekt in een gebaar van overgave. Een variante van bovenstaande havik duif metafoor laat toe om bepaalde vormen van conciarrentie te begrijpen, b.v. waarin bedrijven wedijveren voor de controle over een technologie (Bulow en Klemperer (8999)). Er zijn vele gegadigden maar weinig uitverkorenen, ei1 bij de selectie zullen velen de arena verlaten (in de zogenaamde "shakeout"). Het duurde tien jaar en er waren 300 tekstverwerkingsprograra?ma7svoor men bij "Word" en "WordPerfect" bleef. Het is gewoon beter voor de meeste bedrijven om vlug te stoppen en de concurrentie over te laten aan enkele. Maar zulk verlaten van de markt heeft niets te maken met agressief gedrag van de overbliijvers.
Maar hoe verklaren we dan de goed gedocumeriteerde en gedetailleerde gevallen waarin driiideldk sprake was van verkoop met verlies? Zo is er het voorbeeld van schepen die gebruikt werden door een karte4 van reders in de China-Engeland tlieehai~delrond de jaren 188385, vooral vanuit Hankow over de Yangtze riaar Shanghai en Londen. Een buitenstaande reder die vooral op Australie vaarde, werd niet toegelaten tot het kartel dat besloot: Y .ifany non-coilference sieamer shouldproceed to liankow to load independently the necessary number of conference steamers should be sent at the same time E'OManlco~,in order to underbid the fieight which the independent slzip owners might oDu; without any regard to whetlzer the fveight they should bid, would be rernunerative or nut" (Yamey (1972)).
Dit werd toegepast en sommige van de schepen vaarden leeg terug vanuit Hankow, omdat de buitenstaanders toch een lading hadden kunnen verkrijgen maar wel met veel lagere tarieven. Maar de structuur van het veivoersaanbod van schepen kan sterke prikkels inhouden tot prijsoorlogen (indien de vervoerder het schip moet vol krijgen) (Pirrong (1984)). Het zou dus kunnen dat het kartel zulke destructieve concurrentie gewoon wou vermijden. Meer empirisch onderzoek van Britse kartels op belangrijke routes naar ZuidAfrika, India en het verre oosten suggereert echter dat ze alleen optraden tegen jongere en makkere nieuwkomers die weinig of geen financieel vangnet hebben, ervaring en Itlantenbestand missen en die ze niet tegenkomen in andere van hun markten (Scott Morton (1997)). Er zijn ook nog andere goed gedocumenteerde gevallen van agressief gedrag in b.v. luchtvaart, lucifers, cement, teiecommunicatie, suiker en distributie door supermarkten (zie bv. Bolton (1999) e.a.). Er worden dus nu en dan toch wel klappen uitgedeeld, inaar als dit niet zo verstandig is, zoals hierboven uiteengezet, blijft de vraag waarom? VI. SLIMMER PRHJSGEDMG De oudere analyse houdt geen rekening met strategisch (tactisch) gedrag in markten met onvolkomen informatie. Dit is niet zo verwonderlijk omdat de nodige speltheoretische technieken voor een ontrafelen van de logica pas een goede dertig jaar bestaan en pas vanaf 1982 werden toegepast op verkopen met verlies (Milgrom en Roberts (1982)). Indien de spelers niet alles met zekerheid weten, is het moeilijker voor de markt om de agressieve strategie te ontzenuwen. Ook kunnen de gevestigde spelers gemakkelijker de toekomstverwachtingen van de nieuwkomers of andere rivalen vervormen, zodat hun financiers of zijzelf het totaal niet meer of niet meer zo goed zien zitten (zie Bolton e.a. (1999) voor referenties).
A. Roet in hel eten gooien De startende bedrijven zullen veelal een beroep doen op extern kapitaal en leningen. De transacties met deze geldverschaffers gebeuren met asymmetrische informatie en objectieven. Zo is de financier niet zeker dat het bedrijf alles zal doen om de lening terug te betalen. Naast flagrant misbruik - b.v. "weglopen" met geld of activa naar zonniger oorden - zou het bedrijf b.v. teveel risico kunnen neinen, of teveel of
te weinig hooi op de vork nemen, of zich al te veel vastpinnen op wat de pioniers en voortrekkers belangrijk vinden. Zulk al te sterk opportunistisch gedrag wordt in toom gehouden door o.a. het bestuur van het bedrijf wanneer het gaat om kapitaal4 of door het leningscontract, altijd binnen een bepaalde cultureie en maatschappelijke omgeving. Zo zal de financier een zekere waarborg eisen of zal hij aanvankelijk maar een beperkte steun geven. Bijkomende bedragen worden maar ter beschikking gesteld na een zekere tijd en wanneer de verdere levenskansen duidelijker zijn. In geval van wanbetaling bij een lening blijft de verkoop van activa mogelijk. Dit alles gegeven zijnde, kan een bestaande concurrent via zijn marktstrategie roet in het eten gooienvan de nieuwkomer. Zo zou de vroeger gevestigde concurrent via b.v. lage prijzen de mogelijkheden tot terugbetaling of de verwachte winstgevendheid zodanig kunnen aantasten dat het bedrijf in moeilgkheden komt met de geldverschaffers. Ruzie stoken tussen de nieuwkomers en hun financiers hoeft niet altijd te lukken. Mogelijk wordt gewoon het vertrouwen tussen de partijen moeilijker of zal de twijfel over de winstgevendheid toenemen. Dit kan zich vertalen in een nadelige opstelling van de financier. Nieuwelingen zullen vlugger dan nor-maalbeslissen om er mee te stoppen. Minder extreem, zal de nieuwkomer misschien vlugger afzien van verdere uitbreiding. E n als mogelijke nieuwkomers dit alles observeren, zullen ze twee keer nadenken voor ze op de markt komen en zullen ze b.v. niet zo vlug geneigd zijn de activa over te kopen van de weggepeste rivaal. En als ze ondanks deze slechte verwachtingen toch toetreden, zal de financier weer "moeilijker" doen, omdat die weet dat her moeilijk opboksen is tegen het agressief gedrag van de gevestigde rivaal. B. Schrik aanjagen Mogelijk zullen gevestigde bedrijven zich realiseren dat nieuwe concurrenten die in één markt succesvol zijn, nog nieuwe concurrentie in andere markten of in dezelfde markt zullen aantrekken. Hierbij zullen de rivalen mogelijk onvolkomen informatie bezitten over het S p e van rivaal waar ze tegenover staan (havik of duif b.v.) of over het werkelijke niveau van de kosten. In meerdere markten met volkomen informatie is het niet optimaal om agressief te zijn (zie hierboven). Maar met onvolkomen informatie is bet wel aangewezen om vooral in het begin van de levens-
cyclus van de industrie, nieuwkomers nu en dan eens wat harder aan te pakken. Experimenten suggereren dat mensen de neiging hebben om op die manier te spelen (Isaac en Smitli (1985)). Niet zozeer om bestaande rivalen te verdrijven, maar wel om een reputatie van een hardere concurrent op te bouwen, zodat toekomstige rivalen schrik Itrijgeii en twee of meer keren nadenken vooraleer ze toetreden. Het is natuurlijk zeer moeilijk, indien niet onrnogeiijk, orn zullie intimidatie te bewijzen. Maar vaak zal dil opbouwen vaxi dc repatatie andere meer gesofisticeerde agressieve tactieken versterken. M Smogelijke concurrenten schrik hebben bv., zullen er minder zijn en zullen er bijgevolg weinig gegadigden zijn oin de activa van bestaande falende rivalen over te nemen. De financiers zullen rekeiiiiig hoiaden met de minder rooskleurige mogelijkheden van liquidatie, en zullen een zekere compensatie vragen. Hierdool kunnen bestaande concurrenten het moeilijker krijgen. E r bestaan vele varianten op het geiien van zulke negatieve signalen aan concurrenten. Zo kan de gevestigde concurrent in het geheim zijn prijs verlagen in een testmarkt van de nieuwkomer. Dan heeft die het moeilijker om te weten of de vraag naar zijn product nu hoog of laag is. Of de dominerende rivaal kan opcnlijk de prijs verlagen om lage kosten te suggereren of oin de aanvankelijlie operationele resultaten van de nieuwkomer te vervormen. Deze laatste tactiek werd in het begin van de jaren zeventig toegepast door Maxwell House in het Oosten van de Verenigde Staten, om de testmarkt resultaten van filgers koffie in de war te sturen. Deze laatste besloot hierop om de expansie uit te stellen.
VHI. SLIMMER NIET-PRIJSGED-G De bedrijven beschikken niei alleen over de prijs als wapen orn he: anderen moeilijker te maken (zie Ordovcr en Saloner (1989)). Het kan voor een dominerende speler interessant zijn orra te trachten het kostenniveau van alle concurrerende aanbieders te verliogen. Zo zal een versneld innovatieritme vaak door bijna iedereen moeten gevolgd worden. De treuzelaars of de minder doelmatige spelers zullera hun competitiviteit zien afkalven. OS misschien zal de overheid ecn standaard van kwaliteit opleggen die gemakkelijker door grote dan door kleinere bedrijven Itan gerealiseerd worden (ten gevslge van schaaleffecten).
Het is echter verkeerd om te denken dat een goede tactiek of strategie. altijd deze is die de winsten van de rivalen doet dalen. Het doel is de eigeil winstgevendneid te verbeteren en de waarde van het bedrijf tc vergroten. En vaak kain dit zeer goed via een strategie waar de rivalen ook voordeel kunnen uithalen. Mogelijk kunnen nieuwkomers b.v. veel leren uit de innovatie van de dominerende speler en zo zelfs haasje over spelen, of kunnen ze mee profiteren vali tendensen tot prijsverhoging, in gang gezet door de grotere speler. Bovendien is het eigeirbelang vaak geen hiiiderpaal om een duurzame samenwerking te ontwikkelen (zie bv. De Bondt (1999)).
VBEI. BESLUIT
De letterlijke interpretatie van roofdiergedrag als l-iet opeten van de zwakl<ererivaal, alleeai om zelf sterker te worden (of te overleven), is niet echt nuttig om coaicurrentie in de markt beter te begrijpen. De markt werkt gcwoon niet op die manies. Wel kan aP eens een robbertje gevochten worden o i kunnen de bedrijven proberen roet in het eten te gooien of ruzie te stoken. En de makkere nieuwkoiners kunnen het dan wat vlugger moeilijk krijgen en er missclnien de brui aan geven of in ieder geval hun ambities terugdraaien. Het beoordelen van zulk gedrag hangt sterk af van de visie van concurrentie die impliciet of expliciet gehanteerd wordt. Wellicht is er niet zoiets als de juiste visie, maas in een goed werkende democratie met een evenwichtig recht kan dit op zichzelf geen groot probleem zijn. Indien een wet bestaat die verkopen met verlies of roofdiergedrag verbiedt, is het natuurlijk beter om ze goed toe te passen dan om dit niet te doen. En dus kan het waardevol zijn om het juridisch proces te inspireren cn te behoeden voor foutieve inzichtein. Wellicht kunnen de hierboven geschetste nieuwe inzichten over agressief gedrag hierbij nuttig zijn. We begrijpen nu inderdaad beter waarom slim en uitgekookt concurreren ook wel wat gerichte klappen kan uitdelen aan b.v. nieuwkomers. Maar volgt hieruit dat de maatschappij of eco~ioiniezal verbeteren door een meer populistische interpretatie van het hardere gedrag? Willicht niet. In concurrentie is nooit iedereen gelijk cn dit ongelijk zijn, is om allerlei redenen gezond. De bedrijven en de markt zijn geen weerloze prooien, maar kunnen mits anticiperen en creativiteit eventuele slagen vennijden of een tegenzet plaatsen. Vaak zal samenwerken hand in hand kunnen gaan lilet gezonde (zij het soms
harde) rivaliteit. En zelfs zonder agressieve tactieken zullen vele bedrijven na een zekere tijd verdwijnen. Eén van de robuuste tendensen in bedrijfstakken is dat waar veel toetreding is, er ook veel uittreding kan verwacht worden. Iets totaal anders is natuurlijk het crimineel vernietigen van een concurrent of medemens (b.v. door de "georganiseerde" misdaad). Dit kan en zal spijtig genoeg ook blijven voorkomen, vooral daar waar het staatsgezag geen doeltreffend tegengewicht is. Uiteindelijk heeft zulk een misdadig gedrag inhoudelijk veel te maken met de gebreken van de mens en van de cultuur waarbinnen hij leeft. Maar het heeft weinig te zien met de intrinsielte eigenschappen van de concurrentie in een markteconomie. NOTEN 1. De bedoeling is hier niet om het geval Microsoft ten gronde te bespreken (zie Van Cayseele (1999) voor een korte discussie van de economische pleidooien). 2. Dit is de "chain store paradox" van Selten (1 978). 3. Prijs wordt ook maar zelden gehanteerd als instrument om de eigen markt te beschermen (zie D e Bondt (1989)). 4. Naast de andere mechanismen om behoorlijk bestuur te helpen garanderen (zie Shleifer en Vishny (1997)).
REFERENTIES Bolton, P., J.E Brodley en M.H. Riordan, 1999, Predatory Pricing: Strategic Theory and Legal Policy, (Tilburg University). Bulow, J. en P. Klemperer, 1999, The Generalized War of Attrition, American Ecorlonzic Review 89, 175-189. Daly, J., 1997, The Robin I-Iood of the Rich, Wired 5, 109-111, 146-145. De Bondt, R., 1989, Het anticiperen van nieuwe concurrentie, Tijdschrift voor Ecorzomie erz Management 34, 4, 433-454. De Bondt, R., 1999, Cultuur en samenwerking, (Leuven, D.TE.W, K.U.Leuven), 19 p. Easterbrook, EH., 1981, Predatory Strategies and Counterstrategies, University oJClzicago Luw Review 48, 263-337. Granitz, E. en B. Hein, 1996, Monopolizatiotl by Raising Rivals' Costs: the Standard Oil Case, Joilrnal of Luw nnd Economics 39, 1-47. I-Iofbauer, J. en K. Sigmund, 1988, The Theoiy of Evolution of Dynamica1 Systems. Mathematical Aspects of Selection, (Cambridge, Cambridge University Press). Isaac, R.M., en VL. Smith, 1985, In Search of Predatory Pricing, J o ~ ~ r n ofPolitica1 nl Economny 93, 320-345. Klepper, S., 1997, Industry Life Cycles, Industrial and Corporate Cllange 6, 145-181. Klepper, S. en K.L. Simons, 1997, Technological Extinctions of Industrial Firms: an Inquiry into their Nature and Causes, Irzdiutrial und Colyorate Cllange 6, 379-460. Milgrom, P. en J. Roberts, 1982, Predation and Entry Deterrence, Joumal ofEcoizomic Theory 27, 280-312. Mc Gee, 1958, Predatory Price Cutting: the Standard Oil (N.J.) Case, Jouriznlof Law and Economics 1, 137-169. Pirrong, S.C., 1987, An Application of Core Theory to the Analysis of Ocean Shipping Markets, Journa1 of Law arzd Ecorzol7zics 35, 89-331.
Phlip" L., 1995, Competilioil Policy: a Game-Theorelic Pcrspectivc, (Cambridge, Cambridge Uiiivcrsily Press). Ordover, J.A. en G. Saloner, 1989, Predation, Monopolization, and Antitrust! iii SchrnaIcnsee, R. en R.D. Willig, Handbook of Ind~istrialOrganizatioil, (Amsterdam, NorthHolland), 537-596. Selten R., 1978, Tlie Cliain Store Paradox. Tl~eorynrzd Decision 9, 127-159. Schwartz, E.I., 1999, Digital Danvinism. 7 Breaktlirough Business Strategies for Surviving in thc Cutthroat Web Econoiiiy, (New York, Broadway Books). Scott Morton E, 1997, Entry and Predation: Britisli Sliipping CarteIs 1879-1929,Journalof Ecoi~onzics,&Inl~agenientclrlrl Stizrtegy, 6 , 679-721. Shleifer, A. en R.W. Visliny, 1997, A Survey of Corporate Goverilance, .loi~rilnlofFii?ance, 737-783. Vali Csyseelc, P, 1999, Microsoft: de verkeerde verdediging?, Tïrnds Review, 29-32. Yarney, BS., 1972, Predatoi-y Price Cutting: Notes and Commelits, JoirnznloJ'LnwnndEconolnics 15, 129-142.
Tijdschrift voor Economie en Mallagement Vol. XLV, 2, 2000
De ontwiUei(iaag van een generische impost-haactie voor Prologa: een koppeling tussen Prologa en Excel
VANDERBIST Dirk (K.U.Leuven, %ic.TEW (1999)) Deze verhandeling situeert zich binnen het studiedomein van dc beslissingstabellen. Een beslissingstabel is een tabel die een exchaustief geheel van elkaar uitsluitende voorwaardelijke uitspraken voorstelt binnen een vooraf bepaald probleemgebied. Door d e exhaustiviteit en exclusiviteit laten beslissingstabellen toe om waterdichte cn efficënte procedure o p te stellen. Wat toegepast kan worden bij het opstellen van wetten, iebleinenten e n voorschriften. Voor er iets over de inhoud van d e verhandeling kan verteld worden, moet deze eerst gesitueerd worden binnen d e Faculteit Economische en Toegepaste Econoniische Wetensclaappen aan d e R.U.Leuven. Reeds verschillende tientallen jaren wordt er binnen de Pnformation Systems Group onderzoekverricht in verband met beslissingstabellen. Het onderzoek naar beslissingstabellen heeft geleid tot de ontwikkeling van een soiiware tool Prologa (Procedurai Logic Analyzer) genaamd. Prologa groeit nog elk jaar onder meer door verhandelingen ais deze. Prologa is oildertussen uitgegroeid tot een volliiiassen tool. H e t is één van de weinige tools die het volledige ontwikkelingsiraject voor een beslissingstabel kan afhandelen. Het combineert hiertoe incrementele opbouw met rigoureuze validatie en verificatie tijdens de bouw. Bovendien laat het achteraf makkelijk aanpassingen toe zodat onderhoud mogelijk blijft. Prologa bevat reeds verschillende exportfaciliteiten maar d e importfaciliteiten zijn echter beperkt. Hierin probeert deverhandeling verandering te brengen. Het doel van d e verhandeling was dan ook: "Realiseer een import-f~znct-ievoor Proioga". H e t resultaat is een import-wizard geworden die geprogrammeerd is in Object Pascal, gebrtiik makend van Borland Delphi IV, waardoor een fraai grafisch programma voor een Windows-omgeving verkregen werd dat goed aansluit bij d e reeds grafische omgeving van Prologa. Er werd geopteerd voor een koppeling van Excel omdat d e vertrouwdl-ieid van de gemiddelde PG-gebruiker met hct Microsoft Office pakket groot is en zodoende een grote doelgroep bereikt kon worden. Verder is Excel in staat o p een makkelijke manier tabellen voor te stellen en kan het omgaan met grote verzamelingen gegevens. Deze kunnen dun in groep naar Prologa overgebracht wordcn omdat de import-functie een vorm van groepsgewijze (batch) verwerking ondersteunt. Ook gebeuren in Excel reeds primaire controles zodat d e gebruiker weet of d e opgestelde tabel aan de basiscisen voldoet om als beslissingstabel aanvaard te worden. Z o kan de gebruiker Excel als uitgangsbasis gebruiken bij het opstellen van nieuwe tabellen. D e import-ftinctie heeft de vorm aangenoinen van een import-wizard. Dit wil zeggen een programma dat over een hoge mate van "intel!igentier' beschikt zo-
dat de interactie met d e gebruiker optimaal verloopt. Met intelligentie wordt bedoeld dat het programma probeert te acherhaleii wat een gebruiker wil en zoveel mogelijk handelingen automatisch probeert uit te voeren zodat d e gebruiker zelf een minimum aan handelingen hoeft te stellen. De gebruiker zal dus stap voor stap door het invoer-proces geleid worden. Verder is de import-functie zodanig geprogrammeerd dat het een afzonderlijke component vormt die in andere programma's kan opgenomen worden. In dit geval wordt de component opgeroepen vanuit Prologa, werkt het de invoer zelfstandig af en g e e f het een resultaat-beslissi~lgstabeiterug door aan Prologa. Verder legt de import-functie beperkingen o p aan het soort beslissingstabellen die ingevoerd kunnen worden. D e beperkingen vloeien voort uit het feil dat Prologa o p zich beperkingen oplegt aan de beslissingstabellen waarmee deze werkt. Het heeft geen zin tabellen in te voeren die later niet voorgesteld kunnen worden binnen Prologa. Zonder hier verder in detail te willen treden kan worden gezegd dat d e import-functie geexpandeerde en samengetrokken multiplehit beslissingstabellen ondersteunt. Hierbij mogen acties enkel in limited-entry vorm staan en condities worden zowel in lirnited-entry als i11 extended-entry vorm geaccepteerd. Voor de bespreking hiervan wordt naar d e verhandeling zelf verwezen. Onderstaande figuur geeft schematisch de opbouw van d e import-functie. D e component bevat drie onderdelen.
FIGUUR 1 De relaiie tussen de yrogrammaonderdelez van de import-functie
Het eerste onderdeel is de unit Import die zorgt voor de comn~unicatiemet de buitenwereld. H e t regelt d e inieractie met de gebruiker en voorziet in de communicatie met Prologa en Excel. Het bevat onder andere de grafische gebruikersinterface. De commuiiicatie met Excel verloopt via een Dynamic Data Exchange koppeling, Texcel, die gehaald werd van de Deiphi Super Page en dus niet eigenhandig geschreven werd. De technische verfijnde onderdelen zijn de units Script-parser en Syntax-parcer. Zij vormen d e kern van het invoer-proces. Voor de iiivoer van een beslissingstabel werd gekozen o m gebruik te maken van een scripttaal. Eerst wordt een script opgesteld dat dc volledige definitie van de in te voeren beslissingstabel beschrijft. Daarna wordt de beslissiiigstabel geimporteerd op basis van die definitie. D e scripttaal heeft een syntaxis die volledig afgestemd is om beslissingstabellen te beschrijven en met andere woorden makkelijk verstaanbaar is. D e twee eerder genoemde unitsverzorgen de afhandeling. Het voordeel van het gebruik van deze scripts is de herbruiltbaarheid. D e import-functie zal de gebruikcr stapsgewijs helpen een script op te stellen dat hij later eventueel in Excel bij de tabel lían opsiaan wat resiiiteeri in een vorm van gegevenskoppeling. Vooor gevurderde gebruikers is er steeds de mogelijkheid scripts rechtstreeks o p te stellen. Het grootste deel van de verhandeling is verder gewijd aan de beschrijving va11 de programma-onderdelen die hef scripts controleren op syntaxis e n afhandelen ten behoeve van de import van d e beslissii~gstabel.Om d e genericiteit verder o p te drijven werd voor de scripttaal een parser ontwikkeld die de scripttaai controleert op zijn syntaxis door gebruik te maken van een meta-syntaxis. Hierin situeert zich dan ook liet zwaartepunt van de verhandeling hetgeen in de samenvatting niet in enkele woorden te beschrijven valt. Als concl~isievan d e verhaiideling kan gezeg/ worden dat d e import-functie behoorlijk werkt. E e n opmerking zou kunnen zqn dat aan de beslissingstabellen die ingevoerd kunnen worden sterke vormvereisten worden opgelegd. Deze zijn het resultaat van de afweging: weinig structurering versus hoge "intelligentie" van de import-functie. H o e minder de gegevens gestructureerd zijn hoe meer de tussenkomst van d e gebuiker nodig is om aan te geven waar wat zich bevindt en dus hoe lager de intelligeiitie. H o e meer structurering hoe meer import-functie eigen intelligentie kan aanwenden om dingen automatisch te vinden e n dus hoe lager de tussenkomsi van d c gebruiker. Om deze samenvatting af te sluiten wordt er nog verwezen naar een mogelijke aanvulling voor d e import-functie. De huidge import-functie werkt in één richting. Beslissingstabellen worden vanuit Excel naar Prologa overgebracht. D e uitbreiding zou twee-richtirigsve~-lieerrealiseren tussen Proioga en Excel waarbij d e beslissingstabellen in Excel een soort ingebedde logica zouden vormen. Het voigende voorbeeld maakt liet duidelijker. Er bestaan reedsverschillende programma's die beurskoersen cn beursverrichtingcn in Excel automatiseren, Stel nu dat men een beslissingstabel maakt waarin de voorwaarden en acties staan om ecn bepaalde transactie (verkoop, aankoop, ...) op een aandeel uit te voeren. D e beslissingstabel wordt via d e import-functie overgebracht naar Prologa. Ook de koersgegevens worden doorgegeven aan Prologa. Prologa evalueert met de líoersgegevens de beslissingstabel waaruit een te ondernemen actie volgt. Deze actie wordt aan Excel terug doorgegeven die de actie uitvoert. Deze verhandeling heeft d e basis gelegd die de volgende stap, een interactieve koppeling tussen Prologa e n Excel mogelijk inaakt.
Het HST-niet in Europa: impact op de economie, ruimten en geo-politieke verhoudingen: samenvatting
V M Z W G E N H O W N Gonda (IC.U.Leuven,Handelsingenieur (1999)) De vorming van een Europese eenheidsmarkt en de fundamentele veranderingen in Oost-Europa hebben geleid tot een toename van de internationale uitwisselingen in Europa en zo ook tot een stijgende behoefte aan internationale infrastructuren. Enerzijds om aan deze behoefte te voldoen, anderzijds - en vooral om de Europese economie te stimuleren, stelde de Europese Commissie in 1985 voor een HST-netwerk uit te bouwen tussen de verschillende lidstaten. De HogeSnelheidsTrein is een betrekkelijk recent fenomeen. Over deprecieze effecten van de ontwikkeling van een HST-netwerk, waarvan verschillende verbindingen nog onder constructie zijn, kunnen bijgevolg geen harde uitspraken worden gedaan. Op basis van de ervaringen in Frankrijk, waar de hogesnelheidstrein reeds 18jaar lang zijn diensteii aanbiedt, kunnen de effecten evenwe! ingeschat worden. een noodzakelijUit onderzoek komt naar voor dat transportinfrastruct~~ren ke, maar zeker geen voldoende voorwaarde vormen voor econoinische ontwikkeling. De bouw van hogesnelheidslijnen kan de lokale ontwikkeling bijgevolg maar aanzwengelen als deze gepaard gaat met een ruimer pakket maatregelcn dic een gunstig klimaat scheppen voor investeringen en uitbreidingen. Bovendien dient rekening te worden gehouden met de specificiteiten van de betrokken regio. Een goede regionale inbedding is onontbeerlijk. De economische effecten van de HST worden daarenboven niet evenredig over de Europese ruimte verdeeld. De komst van de HST bevordert voornamelijk de zuigkracht van economisch sterke gebieden, terwijl zwakkere of perifere gebieden daar veel minder voordeel uit halen. Deze ruimtelijke differentiatie kan enerzijds toegeschreven worden aan regio-specifieke factoren. Anderzijds versterkt de HST de ruimtelijke polarisatie omdat het bereilbaarlieidsvoordeel enkel geldt bij de stopplaatsen. Niet de fluxen, maar het tot stilstand komen van de fluxen opent kansen op ontwikkeling. Uit verschillende studies kan men afleiden dat het HST-netwerkvooral de bestaande metropolitane centra zal bevoordelen, en dan vooral deze in het econoinisch kerngebied. De doelmatigheid van een vervoerssysteem hangt bijgevolg niet alleen af van de aanleg van een moderne hogesnelheidsinfrastri~ctuur,maar ook van de aanwezigheid van secundaire verbindingen. Een verbetering van de verbindingen van en naar het nabij gelegen HST-station kan een bereiltbaarheidsvoordeel immers over een ruimer invloedsgebied verspreiden. Vanuit beleidsrniddens dient hier dan ook meer aandacht naar uit te gaan. Er bestaat e.chter geen wonderformule om, op basis van de HST Een duurzame stedelijlte economische groei tot stand te brengen. De integratie van het HSTsysteem in de stedelijke omgeving is maatwerk. Er dient rekening te worden gehouden met de specificiteiten van de betrokken stad en haar omgeving. Een uitgebreide analyse van de ontwikkelingen die gepaard gaan met de komst van de HST in Antwerpen-Centraal, toont aan hoe complex de integratie van de snelle trein is en hoeveel andere dan zuiver transport-elementen hierbij komen kijken. De aanpassingen in de stationsomgeving van Antwerpen-Centraal houden enerzijds verband met een reorganisatie en optimalisering van de bereikbaarheid en
ontsluiting van het station vanuit de ruimere Antwerpse omgeving. Anderzijds incorporeren zij aanknopingspunten voor de ontwikkeling van stedelijke activiteiten meegebracht door de vlotte treinverbinding met andere Noordwest-Europese centra. Niemand kan vandaag echter met enige vorm van zekerheid het effect van de verbeterde spoorontsluiting op het vlak van stedelijke ontwikkeling voorspellen. Deze onzekerheid is een bijkomende factor die de complexiteit van het project nog in belangrijke mate doet toenemen. Een goed doordachte en weloverwogen planning is bijgevolg aan de orde. Hiertoe bundelen de verschillende betrokken partijen hun krachten, hetgeen resulteerde in een overlegorgaan. Deze samenwerking vormt alvast een belangrijke troef voor een goede afwikkeling van het HST-project in Antwerpen. Een bijkomende positieve noot is het inschakelen van een objectief studiebureau. In een studie, uitgevoerd door "Mens en Ruimte" (1993) werd aangetoond dat de HST een bijdrage kan leveren tot een beter leefmilieu. De hogesnelheidstrein kan bovendien een nog grotere milieuwinst opleveren als hij een deelvan de vluchten naar nabijgelegen bestemmingen vervangt. Deze optie kan de laatste tijd op alsmaar meer sympathie rekenen. Heel wat luchthavens zien de HST niet meer als concurrent inaar als bondgenoot, m.n. omdat de HST luchthaveils met capaciteitsproblemen voor een deel uit de nood zal kunnen helpen. De snelle trein kan de passagiers die een trip maken die korter is dan 600 kilometer overnemen van de luchthaven en zo de catchment area van de Iiichthaven aanzienlijk vergroten. In deze context is het onbegrijpelijk dat de tweede HST-terminal van Brussel zich niet in de luchthaven, maar wel in Schaarbeek zal bevinden. De groei van Zaventem wordt belemmerd door capaciteitsproblemen, de milieubelasting en een gebrekkige ontsluiting. Een HST-terminal in de luchthaven zou voor elk van de knelpunten een mogelijke oplossing kunnen vormen. De keuze voor Schaarbeek blijkt dan ook enkel te verklaren vanuit het eigenbelang van de NMBS als grondeigenaar. De Vlaamse overheid zou zich geen NMBS-beslissing mogen laten opdringen die vanuit vervoerseconomisch oogpunt suboptimaal is. Inspraak van de Vlaamse regering in het beleid van de NMBS is daartoe onontbeerlijk.
Tijdschrift voor Economie en Management Vol. XLV, 2000,2
Handleiding Sociale Auditing Els Reynaeri (Acco (1998)) Om naar duurzaamheid te kunnen streven moet een onderneming naast het financiële inzicht ook een helder inzicht verwerven in haar sociale en maatschappelijke impact en ethisch gedrag. Social auditing is een holistisch en continu leerproces dat toelaat om deze niet-financiële aspecten van een organisatie op een systematische manier op te volgen en bij te sturen. Dit omvat een reflectie over de eigen identiteit, doelstellingen en waarden, het streven naar openheid en transparantie, maar ook zaken zoals het systematisch registreren, evalueren en verbeteren van de eigen prestaties. De kerndoelstelling van social auditing is de verbetering van de resultaten op sociaal en maatschappelijk vlak, door ze beter te begrijpen in relatie tot de gestelde doelstellingen en de visies van de stakeholders. De bedoeling van 'social auditing' is dus veel ruimer dan wat de term op het eerste gezicht laat vermoeden: 'sociaal' slaat ook op ethische, maatschappelijke en ecologische aspecten, terwijl 'auditing' slaat op een continu proces en niet enkel op het moment van doorlichting. De auteur stelt trouwens uitdrukkelijk dat de social audit in een perspectief van meerdere jaren moet worden bekeken, al was het maar omdat het zeer onwaarschijnlijk is dat een organisatie tijdens het eerste jaar alle vruchten van deze methode zou kunnen plukken. De bedrijfseconomische aspecten, ook al zijn zij vaak niet expliciet vermeld, blijven impliciet aanwezig op het achtergrond. Het audit-proces kan immers ook helpen om een duurzame balans te vinden tussen de sociale doelstellingen enerzijds, en de bedrijfseconomische resultaten anderzijds. Een social audit heeft trouwens een kost, dat niet kan gedragen worden als er geen voordelen aan verbonden zijn. Dit boek is bedoeld als een praktische handleiding voor bedrijfsleiders, auditors, en alle belangstellenden, en dat blijkt ook uit de toegankelijke taal en uit de zeer heldere structuur. Het proces van social auditing wordt immers stap voor stap toegelicht, en elk hoofdstuk komt met een van deze stappen overeen. De negen fases van het proces zijn: (1)de voorbereiding, (2) de definitie van sociale en maatschappelijke doelstellingen en actieplannen, (3) de dialoog met de stakeholders, (4) de identificatie van prestatie-indicatoren en vergelijkingspunten, (5) het sociaal-boekhoudsysteem, (6) de eindejaarsverrichtingen, (7) de externe verificatie, (8) de bekendmaking van het social-audit-rapport, en (9) de verbetering. Het concept 'stakeholders' loopt als een rode draad doorheen het verhaal. Social auditing vertrekt immers van de doelstellingen en waarden van de organisatie en die van haar stakeholders. Tijdens het proces krijgen zij een stem, waardoor er een grotere betrokkenheid en loyaliteit zowel binnen als buiten de onderneming kan ontstaan. Zij worden daarna betrokken tot de definitie van zowel
brede doelstellingeil als concrete indicatoren, en zij worden ook ingeschakeld bij de registratie van allc gegevens die in de audit zullen verwerkt worden. Tijdens de uiteindelijke externe verificatie wordt er op gewaakt dat het principe van meerstemmigheid tijdens het ganse proces geëerbiedigd is. Maar de problemen i.v.m. de identificatie van de stakeholders worden in dit boek weinig uitgewerkt. Het onderscheid tussen b.v. actieve en passieve stakeholders wordt hier niet gemaakt, met als gevolg dat diegenen die een keuze zouden willen maken tussen potentiële deelnemers aan de audit geen verantwoord hulpmiddel wordt aangercikt. Positief bij dit boek is dat het allemaal zeer concreet wordt uiteengezet, en dat de bestaande processen binnen de organisatie vaak als vertrekpunt dienen om social audit processen te ontwikkelen, zodat het boek inderdaad als een praktische handleiding kan gebruikt worden. Minder positief is dat de voorbeelden allemaal uit een beperkte reeks reeds uitgevoerde audits gehaald zijn, waardoor de lezer zou kunnen twijfelen aan de praktische bruikbaarheid van deze methode in zijn eigeil organisatie. Terwijl de auteur ons belooft dat het model in zeer uiteenlopende organisaties kan worden toegepast, door zowel kleine als grote organisaties, en dit zowel in de profit- als in de non-profitsector, halen kleine bedrijven uit deze laatste sector duidelijk de bovenhand in haar voorbeelden. Een verklaring hiervoor kan wellicht gevonden worden in het feit dat social auditing nog in zijn kinderschoenen staat, zodat er b.v. op internationaal vlak nog druk wordt gezocht naar algemene standaarden die de kwaliteit van het proces kunnen waarborgen. Daardoor wordt social auditing stilaan een beter omlijnd begrip, en verdient het ook een groeiende aandacht. Dit boek zal daartoe zeker bijdragen, want wie het gelezen heeft zal zonder twijfel zijn organisatie vanuit een nieuw perspectief bekijken. Stephan CLUDTS K.U.LEUVEN
Bansport Networks in Europe Concepts, Analysis and Policies K.Button, RNijkamp en R. Priemus (eds) (Edward Elgar (1998) xv-362 blz. Deze bundel vormt de neerslag van een colloquium dat in 1995 plaatsvond te Portugal. Verschillende aspecten komen aan bod, gaande van papers die eerder inzoomen op theoretische concepten als netwerken en netwerk synergieën tot case studies over bijvoorbeeld de Europese luchtvaartindustrie. Het is onmogelijk om elk van deze papers hier de revue te laten passeren. Gegeven de grote belangstelling die het mobiliteitsvraagstuk op dit ogenblik kent, Zal voornamelijk op dit gegeven worden ingezoomd. Zijn er oplossingen voor het mobiliteitsvraagstuk? Wie dagelijks naar Brussel of Antwerpen reist, zal zich wel eens wanhopig deze vraag hebben gesteld. De lezer die hoopt in deze bundel de oplossingen te vinden, zal van een kale reis terugkeren. Zo goed als alle auteurs wijzen op het complex karakter en geven aan dat simpele oplossingen niet echt voorhanden zijn. De goede oude tijd van het begin van de jaren '80 toen ons verkeersnetwerk zo goed als voltooid was en het aantal wagens op de weg relatief beperkt, met zo goed als geen files als resultaat, komt wellicht nooit meer terug. Meteen zijn de twee invalshoeken blootgelegd:
ofwcl probeert men de mobibliteit aan te pakken via de vraagzijde, ofwel via de aanbodzijde. Een derde piste situeert zich eerder op het beïnvloeden van de prijs van mobiliteit, waarmee de link met bijvoorbeeld rekeningrijden kan worden gelegd. In dit opzicht vormt d e paper van Stern, Bovy en Taclien een interessante bijdrage omdat zij o p een systematische wijze de verschillende instrumenten waarover de beleidsvoerder en de autogebruiker bescliiklíen, proberen te structureren. D e onderverdeling gaat van maatregelen die snel zijn te implementeren tot structurele ingrcpen die veel meer tijd in beslag nemen. Voor de autogebruiker is de eerste e n meest voor de hand liggende reactie, het aanpassen van het rij- eii reisgedrag (actief versus passief rijden, vervroegen/verlaten van het vertrekuur, het kiezen van een andere route o m tenslotte over te stappen op een andere modus zoals het spoor, de tram). Ingrijpender wordt het wanneer de eigenlijke werkverdeling wordt aangepakt (flexiwerk,...) om uiteindelijk de locatiekeuze van de school, het werk en in laatste instantie d e woonst te heroverwegen. Voor wat betreft d e overheid kan een gelijkaardige opsplitsing worden gemaakt. Relatief eenvoudige maatregelen zijn te relateren aan informatieverstrekking, een strengere controle of een andere prijszetting. Ingewikkelder wordt het wanneer men d e economische activiteit beter o p d e fileproblenlatiek (spreiding van d e vakanties, school, ...) wenst af te stemmen. Een laatste groep van maatregelen heeft te maken met de ruimtelijke ordening van het land. Structuurplannen alsmede de verdere uitbouw van het infrastructuurnetwerk kunnen in deze context worden geplaatst. Tal van maatregelen zijn dus beschikbaar om het mobiliteitsprobleem aan te pakken, maar het moet duidelijk zijn dat naast prijsregulering (rekeningrijden) ook maatregelen noodzakelijk zijn op het vlak van de ruimtelijke ordening en het sturen van de tijdsbesteding van de Belgen om in de toekomst onze maatschappij voldoende mobiel te houden. D e paper van Nijkamp, Rienstra en Vleugel leert ons dat er wellicht nog heel wat water naar de zee zal stromen vooraleer dit laatste kan worden bereikt. O p een boeiende wijze legt het drietal de inherente schizofrenie bloot waar veel economische agenten aan lijken te lijden wanneer het over mobiliteit gaat. O p basis van een sui-vey waar honderden Nederlandse transportexperten aan deelnamen, werden twee scenario's opgesteld: een optimale gewenste situatie enerzijds en een verwachte anderzijds. D e verschillen tussen beide scenario's zijn markant: enerzijds verwacht zo goed als nieniaild een verandering van de situatie in de nabije toekomst. Bestaande tendensen op het vlak van individualisering van de transportvraag e n privatiscring/deregulering van het transportsystecm zullen ook in de toekomst aanhouden. D e evoluties staan in schril contrast inet de optimale situatie die een antwoord Idealiter zou er een collectief zou kunnen bieden o p het niobiliteitsvraagst~~k. transportsysteem moeten worden uitgebouwd als oplossing voor het mobiliteitsprobleein. Hoe dit laatste val te combineren met een verdere privatisering van de transportvoorzieningen en individualisering van de transportvraag is onduidelijk. Samengevat: er is nog heel wat werk aan de winkel en belangrijke maatschappelijke keuzes zullen moeten worden gemaakt, wil men daadwerkelijk het mobiliteitsvraagstuk uit de wereld helpen.
1. Papers for publication should be cent in triplicate to: Tijdschrift voor Economie en Management Prof. Dr. i? Sercu p/a Mevr. A. Ronsinans Redactiesecretariaat Naamsestraat 6 9 - 3 0 0 0 Leuven, Belgi~im. Submission of a paper wil1 be held to imply that it contains original unpublished work and is not being submitted for publication elsewhere. 2. Manuscripts should be typed double-spaced on one side of the papev only, and record the author's name clearly.
3. Each paper should have an elaborate introduction and conclusion with summary. These have to contain the reasons and relevante of the research reported, as wel1 as its main findings and their policy relevaace.
4. Footnotes should be kept t o a minimum and numbered consecutively. They are p u i at the end of the text. 5. The conventions for references are those of the European Economic Review. In the text, references to publications should appear as follows: As argued hy Goldfeld and Quandt ( 1 9 7 3 ) ... or: "Decision tabies ... see Verhelst (1980). The author should make sure that there is a strict "one-to-one correspondence" between the names (years) in the text and those on the list. At the end of the manuscript (after any appendices), the complete references should be listed as: For monographs: Verhelst, M., 1 9 8 0 , De praktijk van beslissingstabellen (Kluwer, DeventerAntwerpen). For periodicals: Goldfeld, S. and Quandt, R.E., 1 9 7 3 , A Markov Model for Swiiching Regressions, Journai o f Econornetrics 1, 3 - 1 5 , For contributions to collective works: Taylor, B., 1 9 7 0 , Financing Tables and i h e Future, in Taylor, B. ed., Investment Analysis and Portfolio Management, (St. Martin's Press, New York), 3 7 8 - 3 8 6 . 6. Diagrams should be in a form suitable for immediate reproduction: 1 originai drawn in black ink on white paper and 2 photocopies. Care should be taken that lettering and symbols are of a comparable size. The drawings should not be inserted i n the text and should be marked on the back wit11 figure numbers, title of paper, and name of author.
7. Contributors are responsihle for the correction of galley proofs. Corrections other than printer's errors may be charged to the author. 3 copies are suppiied free; additional copies are avalaible at cost if they are ordered when the proof is returned. 8. Any manuscript which does not conform to the above instructions may be returned for the necessary revision before publication.
RICHTLIJNEN VOOR AUTEURS 1. Te publiceren artikels moeten in drie exemplaren gezonden worden aan: Tijdschrift voor Economie en Management Prof. Dr. P Sercu p/a Mevr. A. Ronsmans Redactiesecretariaat Naamsestraat 6 9 - 3 0 0 0 Leuven, Belgium. Elders gepubliceerde artikels worden niet aanvaard. 2. De teksten moeten getypt zijn op één kant van het papier, met dubbele interlinie met duidelijke vermelding van de auteursnaam. Artikels mogen in het Engels of het Nederlands gesteld zijn. Er kunnen nochtans slechts een beperkt aantal Engelstalige bijdragen per volume worden opgenomen. 3. Elk artikel dient een uitgewerkte inleiding en besluit met samenvatting te bevatten, zodat de essentie en relevantie van de probleemstelling alsook de eigen bijdragen van de auteurs duidelijk, overzichtelijk en summier aan de lezer kunnen overkomen.
4. Voetnoten moeten tot een minimum herleid worden en opeenvolgend genummerd. Ze worden achteraan de eigenlijke tekst samengebracht. 5. De conventies van referenties zijn dezelfde als deze van de "European Economic Review". In de tekst voorkomende referenties moeten als volgt vermeld worden: "Zo betogen Goldfeld e n Quandt ( 1 9 7 3 ) ... of "Deze beslissingstabellen zie Verhelst (1980). De lijst van referenties als volgt: Voor boeken: Verhelst, M., 1 9 8 0 , De praktijk van beslissingstabellen (Kluwer, DeventerAntwerpen). Voor periodieken: Goldfeld, S. and Quandt, R.E., 1973, A Markov Model for Switching Regressions, Journal o f Econometrics 1, 3 - 1 5 . Voor verzameld werk: Taylor, B., 1970, Financing Tables and the Future, in Taylor, B. ed., Investment Analysis and Portfolio Management, (St. Martin's Press, New York), 3 7 8 - 3 8 6 . 6. De in artikels voorkomende figuren moeten getekend zijn op een apart blad ( 1 origineel en 2 kopies) in zwarte inkt, op de achterzijde dient vermeld: naam auteur, titel artikel en figuur nummer. 7. De auteurs verbeteren de bandproeven. Extra-correcties op de drukproeven (d.w.z. verbeteringen die op afwijkingen van de ingezonden teksten neerkomen) brengen kosten met zich mee, ten laste van de auteurs.
8. De gebruikte spelling van de Nederlandstalige artikels is de voorkeurspelling 9. Elk artikel dat aan de hierboven beschreven instructies niet beantwoordt wordt voor nodige herwerking teruggezonden.
