No 70. Male-female labour market participation and wage differentials in Greece

Transcription

1 CENTRE OF PLANNING AND ECONOMIC RESEARCH No 70 Male-feale labour arket participation and wage differentials in Greece C. N. Kanellopoulos K. G. Mavroaras February 1999 C. N. Kanellopoulos Senior Research Fellow Centre of Planning and Econoic Research K. G. Mavroaras Departent of Econoics, University of Newcastle 1

2 2 Male-feale labour arket participation and wage differentials in Greece

3 Copyright 2000 by the Centre of Planning and Econoic Research 22, Hippokratous Street, Athens, Greece Opinions or value judgeents expressed in this paper are those of the authors and do not necessarily represent those of the Centre of Planning and Econoic Research. 3

4 CENTRE OF PLANNING AND ECONOMIC RESEARCH The Centre of Planning and Econoic Research (KEPE) was established as a research unit, under the title "Centre of Econoic Research", in Its priary ais were the scientific study of the probles of the Greek econoy, encourageent of econoic research and cooperation with other scientific institutions. In 1964, the Centre acquired its present nae and organizational structure, with the following additional objectives: (a) The preparation of short, ediu and long-ter developent plans, including plans for regional and territorial developent and also public investent plans, in accordance with guidelines laid down by the Governent. (b) The analysis of current developents in the Greek econoy along with appropriate short-ter and ediuter forecasts; also, the forulation of proposals for appropriate stabilization and developent easures. (c) The further education of young econoists, particularly in the fields of planning and econoic developent. The Centre has been and is very active in all of the above fields, and carries out systeatic basic research in the probles of the Greek econoy, forulates draft developent plans, analyses and forecasts short-ter and ediu-ter developents, grants scholarships for post-graduate studies in econoics and planning and organizes lectures and seinars. In the context of these activities KEPE produces series of publications under the title of "Studies" and "Statistical Series" which are the result of research by its staff as well as "Reports" which in the ajority of cases are the outcoe of collective work by working parties set up for the elaboration of developent prograes. "Discussion Papers" by invited speakers or by KEPE staff are also published. The Centre is in continuous contact with siilar scientific institutions abroad and exchanges publications, views and inforation on current econoic topics and ethods of econoic research, thus further contributing to the advanceent of the science of econoics in the country. 4

5 DISCUSSION PAPER SERIES This series of Discussion Papers is designed to speed up the disseination of research work prepared by the staff of KEPE and by its external collaborators with a view to subsequent publication. Tiely coent and criticis for its iproveent is appreciated. 5

6 Abstract: This paper is an epirical study of the developent of labour arket participation and wage differentials between ales and feales in Greece between 1988 and There is little known about the position of woen in the Greek labour arket. This paper uses recent survey data generated by the National Statistical Service. The decision to engage in paid eployent and the resulting reuneration are studied using selectivity corrected earnings estiations. Oaxaca & Ranso decopositions and counterfactual analysis show that the adverse treatent of feale labour arket participation is the largest identifiable reason why the wage gap is there and why it increased between 1988 and

7 1. INTRODUCTION The Greek labour arket experienced drastic legislative, econoic and social developents during the last three decades. Legislative changes started taking place during the late 1970s as a result of the preparation for and the joining of the European Union, henceforth EU, in the early 1980s. Econoic changes have been taking place throughout the period as a result of the post-war transforation of the Greek econoy. Social changes of considerable significance, such as reductions in fertility and increases in feale education, have also ade a ark in the developent of the labour arket. Legislative, econoic and social changes have had a profound effect on the relative position of en and woen in the labour arket. On the legislative side, in the late 1970s there was a gradual abolition of differences in the binding iniu wage rates for en and woen fro collective bargaining. In the early 1980s new legislation put en and woen on an equal foothold with respect to faily, eployent, social affairs and ratified the EU sex equality directives. On the econoic side, feale labour arket activity rates experienced a drastic decline in the 1960s, which coincided with large eigration and urbanisation increases along with an exodus fro the agricultural sector. The rate of econoically active feales was stable in the 1970s and then started increasing considerably in the early 1980s. Note that increases in feale econoic activity rates in the 1980s took place during a period of rather poor perforance of the Greek econoy and an eployent shift away fro anufacturing and towards services. Feale econoic activity increased ainly in urban areas, aongst prie feale age groups (20-45) and in the paid eployent sector. To a large degree, the reasons for the increased feale econoic activity have been of a social nature. Eancipation iplied that feales not only chose to study ore often and to a higher level, but they also chose to try to use their studies in the labour arket. First arriage ean age increased and ean nuber of children decreased considerably, allowing feales to pursue their labour arket objectives. Feale labour arket econoic activity rate was 7

8 29 percent in 1981, increased to 41 percent in 1991 and reached 47 percent in However, this considerable increase in the supply of feale labour was not atched by corresponding deand changes, as the level of feale eployent saw only far saller increases. The last two decades experienced clear anifestations of feale excess labour supply in ters of both increased feale uneployent rates (fro 12.4 percent in 1988 to 14.9 percent in 1994), as well as persistently higher than average ale uneployent rates (feale rates were percent, while ale rates were 5-7 percent). It is iportant to note that, despite the recent rearkable increases in feale econoic activity, Greece still has one of the lowest feale econoic activity rates aong EU countries. By contrast to feale econoic activity rate changes, the ale-feale wage gap has seen few changes since the 1960s. During the two decades leading to 1982, the wage gap reained alost constant, with feale earnings soe 33 percent below ale earnings. A 40 percent one-off increase in the national iniu wage introduced in January 1982, reduced the wage gap to about 22 percent by Since 1985 the wage gap has been increasing slowly. Strangely, at least on first sight, post-1985 increases in the wage gap took place during the very period when extensive EU rules on sex equality were being incorporated in the Greek legal syste and ipleented in the Greek labour arket. The post-1985 slow increase of the wage gap, despite the newly ipleented EU legislation, has been viewed by policy circles as a anifestation of lack of effectiveness of the legislation. An interesting picture arises when one considers socioeconoic and legislative changes working in tande. Socioeconoic changes led to higher feale labour supply which was only partly absorbed by the arket, resulting in higher feale eployent and uneployent along with lower feale wages. One could try to explain the perseverance of the wage gap using the observation that new labour arket entrants typically coand lower wages, thus depressing the ean wage of the group they belong to. Legislative changes designed to proote the position and pay of feales in eployent clearly iproved the lot of eployed feales, but at the cost of aking 8

9 feale eployent ore costly than it was before. Clearly, other things equal, legislative changes led to ore feales seeking eployent, less feales being hired and higher feale wages. The net outcoe of socioeconoic and legislative changes working together is far fro clear on a priori grounds. This paper assues a beckerian huan capital theoretical fraework and estiates a two stage odel where the first stage estiates paid eployent participation and the second stage estiates the conditional paid eployent earnings. Estiations are carried out for two representative cohorts which are six years apart. The paper uses two faily expenditure surveys conducted by the Greek National Statistical Service in 1988 and The first part of the paper estiates earnings and participation rates and calculates ale-feale pay differentials 1. Differentials are decoposed in accordance with the technique developed by Neuark (1988) and Oaxaca & Ranso (1988) and (1994). The ale-feale wage gap is split into its explained (often referred to as characteristics or productivity) and unexplained (often referred to as arket or discriination) coponents. Differentials derived fro the 1988 and 1994 saples are exained to identify the individual constituents of the observed wage gap, with specific reference to direct and indirect reunerative effects of participation on individuals. The second part of the paper exaines the developent of participation and wages between 1988 and The use of the appropriate counterfactual analysis akes possible the exaination of the developent of the productivity and arket changes after controlling for possible changes in the saple coposition between the two points in tie under consideration. Moreover, the use of the Oaxaca & Ranso decopositions enables the analysis to control for possible changes in acroeconoic conditions facing the two cohorts. This last point is particularly iportant because of the ajor changes that the Greek econoy in general and labour arket in particular have been undergoing 1 Previous relevant studies on the ale-feale pay differentials in Greece are those of Kanellopoulos (1982) using data pertaining to 1964 and Psacharopoulos (1983) using 1977 data. Both estiate traditional incerian type earnings functions. They eploy the Oaxaca (1973) technique to decopose the gross gender pay gap into the part due to differences in productive characteristics and the rest which is identified as the upper liit of discriination. While the percentage of the gross gender gap explained by differences in productive characteristics depends upon the broad or narrow definition of such variables, both studies 9

10 since the early 1980s due to the new constraints and possibilities opened up by EU ebership and due to the policies followed by the post-1981 governent. This paper shows that socioeconoic and legislative changes resulted in increased feale econoic activity, which has been translated into excess feale labour supply. Despite extensive legal protection, the wage gap reains unchanged and feales have to pay a reunerative participation penalty in order to be in eployent. This participation penalty is shown to have a large discriinatory eleent in it. Once the participation penalty has been accounted for, the reaining wage gap can be largely explained by observed ale-feale productivity differences. The result in this paper show that, alost exclusively, discriination against feales confines itself to the participation process. This is a striking result because it questions the current anti-discriination rules, which concentrate on what happens within eployent, leaving participation issues relatively undisturbed. By contrast, the ain conclusion of the paper is that policies designed at cobating the ale feale wage gap in the Greek labour arket should address above all participation issues. docuented a large feale earnings net disadvantage. 10

11 II. THE MODEL A. Discriination and paid eployent participation The paper uses a beckerian earnings function as its starting point. ( Ws) ln ~ = X β s s =, f, p (1) s This paper applies the Oaxaca & Ranso (1988) and (1994) pooled odel decoposition technique which is based on estiating (1) for three different saples: ales, feales and pooled. The pooled odel has been developed as a consequence of appreciating how useful but how liiting at the sae tie the epirical Oaxaca/Blinder (1973) decoposition odels were. All discriination studies ake an iplicit assuption as to what earnings would be in the absence of discriination. This is called the non-discriinatory arket structure. It is iportant that the assued non-discriinatory wage structure is as realistic as possible. Note the role played by pooled estiation in this context. Pooling ales and feales together and estiating their earnings as if gender did not exist, provides estiates of the actual reunerative value attached by the (assued nondiscriinatory) arket to observed characteristics. It should be noted that using the pooled odel assues a non discriinatory wage structure, which on average coincides with the actual wage structure and a non-discriinatory labour arket, which pays the sae total wages as the actual arket. have to assue that in the absence of discriination, total wages paid in the arket would have to be other than the currently paid total wages. After all it is only sensible to expect that reducing discriination would not influence total deand for labour in the econoy in any wider way. Second, assuing a odel which explicitly estiates current total earnings in the econoy, allows counterfactual analysis to isolate the developent of acroeconoic changes in the Wage Gap using estiates fro equation (1). 11

12 The ale-feale wage gap, WG = ln( W ~ ) ln ( W ~ f ) three ters as follows, can be decoposed into ln( WG f 1) = X ( β β p ) + X f ( β p β f ) + (X X f ) + β (2) p Ters 1 and 2 of the right hand side of Equation (2) represent the unexplained part of the wage gap and ter 3 represents the explained part. The decision to take paid eployent or not is represented by Equation (3). Heckan (1979) showed that estiating Equation (3) is a necessary step for deriving unbiased earnings estiates fro the estiation of Equation (2). L = Z γ + us, s =, ƒ, p (3) s s s Applying Heckan s ethod of selectivity correction, Equation (2) is re-written as follows. ( W ) ln ~ = X β + σ λ s s =, ƒ, p (4) s s s s After the inclusion of selectivity ters in the earnings equation, the Wage Gap decoposition has to be re-calculated in order to take into consideration the ipact of the participation decision. Equation (2) is re-written as follows. ln( G + 1) = X ( β β ) + X ( β β ) + ( X X ) β + f p f p f f p λ ( σ σ ) + λ ( σ σ ) + σ ( λ λ ) p f p f p f (5) The econoic interpretation of the ters of Equation (5) is iportant. The first two ters jointly, represent the unexplained part of the wage gap that is not due to any participation effects (a conventional discriination effect). The third ter represents the explained part of the wage gap that is not due to any participation effects (a conventional characteristics affect). If the arket were to ignore participation differences, ters one to three would estiate the origin of the wage gap. 12

13 Ters 4 and 5 jointly, represent the part of the wage gap which can be attributed to the unexplained influence of participation differences (one could view these ters as an estiate of discriination). 2 Ter 6 represents the part of the wage gap which can be attributed to the explained influence of participation differences. B. Indirect Wage Gap effects of participation. Interpreting the last ter in Decoposition (5) as an explained (by participation propensities differences) part of the Wage Gap can be probleatic. The reason is that σp( λ λf) treating like a conventional characteristics estiate (that is, a part of the Wage Gap explained by observed data differences), is tantaount to treating the variable λ as observed individual data. Clearly, this is not the case, since λ is an estiate derived fro the first stage estiations. As an estiate, λ will be partly explained by the observed explanatory variables in the first stage and partly unexplained 3. It follows that the explanatory power of λ in the (second stage) earnings estiation will also be (indirectly) partly explained by the (first stage) explanatory variables and partly unexplained. In order to attribute effects correctly, what appears to be the fully explained effect of λ in the earnings decoposition σ λ λ p( f), has to be further decoposed into the part that can be explained by the first stage explanatory variables and the part that cannot. Only then will Decoposition (5) be able to provide a coplete overall picture of explained and unexplained wage gap differences. The necessary adjustent is relatively siple to perfor. Goulka and Stern (1990) 2 A detailed explanation and discussion of the decoposition of selection ters can be found in Mavroaras and Rudolph (1997). Note the iportance of the Neuark/Oaxaca & Ranso pooled coefficients in Decoposition 5. Unlike the ale or feale odels, here coefficient differences are ultiplied by the correct data, the difference between ale (feale) and pooled coefficients by the ale (feale) data, and the difference between the ale and feale data by the pooled coefficients. The fact that this is a uch ore realistic decoposition technique is revealed by the always higher explained estiates than either the ale or the feale odels can achieve. 3 The technique for estiating discriination effects in selectivity odels used in this paper is explained in Mavroaras (1999). This paper utilises the technique in order to provide an overall easure of discriination. The first stage reflects the selection / participation side and the second stage reflects the 13

14 have shown that differences involving non-linear ters of the σ λ λ p( f) type can be decoposed with the use of a siple counterfactual. Define counterfactual λ( Z f, γ ) = φ( Z γ ) / Φ( Z γ ), which represents the selection correction ter that f f would have been generated using feale data with ale coefficients, that is, a easure of the participation propensity that feales would have had, had they received the arket treatent of their ale counterparts. Siplify notation and rewrite the last ter of Equation 5 using counterfactual λ( Z f, γ ) as follows: σ { λ(z p p, γ = σ { λ(z, γ ) λ(z, γ )} f ) λ(z, γ f f )} + σ p { λ(z, γ f ) λ(z, γ )} f f (6) The intuition behind the way the RHS of Equation (6) enters the earnings estiations is iportant in the context of this paper. The first ter in curly brackets, { λ( Z, γ ) λ( Zf, γ )}, is an explained effect as it cobines the differences between ale and feale data ultiplied by ale coefficients. Crucially, note that the explanation provided in the earnings equation is based on observed characteristics differences in the participation process, the differences between Z and Zf. Hence, the ter in curly brackets estiates the indirect effect of the observed factors which deterine participation (the Zs) on earnings differentials. Call this the indirect explained effect of participation on the wage gap. σp{ λ(zf, γ ) λ(zf, γ f ) T he second ter in curly brackets, }, is an unexplained effect as it cobines the feale data set with the difference between the ale and feale coefficients. Since this part of the wage gap is clearly not explained by the participation estiation, it would be istaken to interpret it as a wage gap difference which is explained by the earnings estiation. Therefore, it would be wrong to include it in the explained proportion of the Wage Gap, as the interpretation of Decoposition (5) earnings side. 14

15 iplicitly suggested in the previous section. Intuitively seen, the two ters in the RHS of Equation 6 represent the indirect wage gap ipact of the explained and unexplained part of the participation process. Taking on board the changes introduced by Equation 6, Equation 5 is re-written in order to incorporate indirect explained and unexplained effects resulting fro the participation process. ln(g λ p f ( σ σ ( λ + 1) = X λ p f σ ) + λ ) + σ ( β f ( σ p p ( λ β ) + X ( β f p σ ) + f λ f ) f p β ) + (X f X ) β + f p (5.1) C. The developent of the Wage Gap and participation. Counterfactual analysis is used to study the change over tie in the explained and unexplained parts of the Wage Gap. Counterfactual analysis is useful in this context because it takes into account any observed changes in the coposition of the saples copared and changes in acroeconoic conditions 4, presented by the nondiscriinatory (pooled) arket structure estiates. The change in the Wage Gap between 1988 and 1994 can be written as follows. ΔWG 9/8 = WG 9 WG 8 (7) Define WG 8 9 as the wage gap that would have existed if the 1980s cohort were faced with the 1990s arket conditions. 4 For exaples of this technique see Wellington (1993), Dolton and Mavroaras (1994), Mavroaras and Rudolph (1997). This technique copares essentially different cohorts and their estiation results, accounting for the observable changes in the overall coposition of cohorts over tie. It is not the sae as panel analysis where individual observable changes can be accounted for and tie invariant unobservable individual differences are differenced out of the estiations. Although the use of one or the other estiation is ore than often data driven (for exaple, this paper could not use panel technique) it is useful to bear in ind the differences in the results produced by the. 15

16 W λ = X ( σ 9 8 ( β σ 9 p9 β ) + λ p9 f8 ) + X ( σ p9 f8 ( β σ f9 p9 β f9 ) + ( λ ) + (X 8 λ f8 8 ) σ X p9 f8 ) β p9 + (8) Cobine Equations (6), (7) and (8) to decopose the over tie change in the Wage Gap as follows. ΔWG X β β λ λ σ σ f 8 9 p9 9 f 9 p9 p9 ( β (X (X ( σ ( σ ( λ ( λ 9 /8 f p9 9 f9 = X β f 9 X X σ σ λ λ f 9 8 ( β ) X 8 f 8 p9 ) + β ) β ) λ ) λ f9 f 9 9 f 8 f 8 ) σ ) σ p8 ( β f 9 8 (X ( σ ( σ p8 p8 β p8 (X p8 ( λ ( λ f8 ) X β f 8 p9 X X σ σ f 8 λ λ 8 ) + f 9 f 8 p8 ) + f8 f 8 ) + ) + ) ( β ) + ) + p9 β p8 ) + (9) Decoposition (9) splits the change in wage gap into 16 separate ters. Ters 1 to 4 estiate the degree to which the observed changes in the wage gap have been the result of changes in earnings discriination. Note that these changes are net of any observed changes in the non-discriinatory (pooled) earnings estiates. The su of ters 1 to 4 reveals the net changes in the wage gap which were the result of changes in discriination. Ters 4 and 5 estiate the changes in the wage gap due to changes in ale and feale observed earnings characteristics respectively. The su of ters 4 and 5 reveals the wage gap ipact of observed productivity changes. Ters 7 and 8 estiate the changes in the wage gap due to changes in the non-discriinatory arket structure, by ultiplying the changes in the pooled coefficients by the observed ale feale data differences. The su of ters 7 and 8 represents the wage gap ipact of estiated changes in the non-discriinatory environent. The reaining 8 ters estiate the wage gap ipact of changes in the participation process. Epirically, ters 9 to 16 prove to be very influential. Ters 9 to 12 estiate 16

17 the direct wage gap ipact of the unexplained part of participation. Ters 13 to 16 follow the logic of Equation 6 by decoposing changes in the explained direct participation effect, into changes in indirect explained effects (ters 13 and 14) and changes in indirect unexplained effects (ters 15 and 16). The indirect explained estiates are based on changes in participation data. The indirect unexplained estiates are based on changes in participation coefficients. III. The Data The paper uses data drawn fro the two recent Faily Expenditure Surveys (FES) in Greece conducted by the National Statistical Service in 1988 and 1994 throughout the entire country 5. These are rando surveys of private households in Greece and include full tie ale and feale workers for who inforation on earnings and socioeconoic characteristics is available. Since the decision exained in this paper is that of eployent, persons below 17 and above 60 are excluded fro the saple used in this study. Full tie students and pensioners are also excluded. In order to ake the two data sets coparable for counterfactual analysis, only variables which have the sae definition in both surveys were used. Although this restricted the odels that could be estiated, experientation with the (richer) 1994 data set reveals that decopositions are not affected in any ajor way by the exclusion of the necessary variables to ake the two sets coparable. The log of weekly wages in 1994 prices has been used as the earnings variable for those in paid eployent. Apart fro the usual huan capital variables like education, age and other, the data sets contain a nuber of interesting variables regarding participation decisions. Household size and coposition are well described in both surveys. Type of accoodation and hoe ownership are also available in both surveys. These are extreely valuable identifying variables for the first 5 The 1988 survey was carried out between Noveber 1987 and October 1988, and the 1994 survey was carried out between Noveber 1993 and October For a full description of the survey s design, 17

18 stage estiation of eployent outcoes, especially where these outcoes ay be affecting wages in a gender-related anner. An overview of the data is provided in Table 1 and descriptive statistics for both surveys, along with variable definitions are in the Appendix. Table 1 Paid Eployent in 1988 and Cohort 1994 Cohort Paid Eployent Saple Coplete Saple Paid Eployent Saple Coplete Saple Cases Wage Cases Cases Wage Cases Males Feales Total Wage Gap Note: Log weekly wages in 1994 prices. Table 1 confirs a general trend towards paid eployent in the Greek labour arket. It is ainly feale eployees who take up the new opportunities. Although feale participation has increased, it still reains alost half of the ale participation. The successive austerity and stabilization prograes, cobined with recession between the years 1988 and 1994 have taken their toll regarding pay; ean 1994 wages are below the 1988 level. Interestingly, feales who ade considerable gains in eployent volue, lost soe 8 percent of their ean wage. By contrast, ales who gained very little in eployent volue, lost less than four percent of their ean wage. Gross wage and participation rates convey the essage of feales underbidding and displacing ales fro paid eployent, at the cost of lower wages. The next section exaines these facts in a ultivariate context. IV. Estiations and results coverage and response rate, see National Statistical Service of Greece (1992). 18

19 A. Introducing participation estiates In the discussion to follow participation estiates play a crucial role. This section explains the constituents of participation effects using the estiates of Equation (5.1) for both cohorts found in Tables A5 and A6. For each (sub)saple of each cohort a ean participation propensity, λ, and the earnings coefficient of the participation propensity, σ, is presented in Table 2 and it is necessary to exaine the in soe detail. Table 2 Participation propensities and their reuneration estiates Pooled Feale Male Counter -factual λ p σ p λ f σ f λ σ λ f 1988 cohort * * * cohort * * * Note: * denotes significance at the 1% level. Detailed results in Tables A5 and A6. By definition λ is a positive nuber for participants and tends to zero as the probability of participation tends to one. 6 Hence individuals with lower participation propensities have higher values of λ. Reebering that at this stage coparisons between the 1988 and 1994 cohorts cannot as yet be carried out, note that feales have larger values of λ in both cohorts. Counterfactual λ f (feale data with ale coefficients) in the last colun indicates that a large part of the difference in ale-feale participation propensities is due to differences in coefficients rather than data in the first stage estiations, especially 6 λ( x) = φ(x) / Φ(x) which is always positive. Note as x tends to infinity λ(x) tends to zero. 19

20 in the 1988 cohort. Coefficient σ is negative and significant at the one percent level throughout. The sign of σ suggests in an unabiguous fashion that the arket views λ as a negative attribute of eployees. The size of the negative σ represents the degree to which low participation propensity is penalised by the arket. Feales suffer a uch higher penalty in both cohorts. The overall picture that arises is that feales on average posses ore of the low participation attributes than ales (that is, λ f is larger than λ ), and that feales are also penalised ore per unit of each low participation attribute than their ale counterparts (that is, σ f is also larger than σ ). It is tie now to see how participation estiates fit into the wage gap picture. B. Wage Gap Decopositions Earnings Equation 4 is estiated for all three saples (ale, feale and pooled) and Equation 5.1 decopositions are presented in Table 3. Positive (negative) estiates represent factors which favour ales (feales). The explained and unexplained effects are presented separately for the first and second stage of the estiation in Rows 1 to 5, with the corresponding nubers of the Equation (5.1) ters they represent. Rows 6 and 7 provide the total explained/unexplained split of the wage gap. Table 3 Wage Gap Decopositions Estiation results Ters in Equation cohort 1994 cohort 20

21 First stage participation estiation Row 1: Participation indirect explained Row 2: Participation indirect unexplained Row 3: Participation direct unexplained 4, Second stage earnings estiation Row 4: Unexplained earnings 1, Row 5: Explained earnings Row 6: Total Explained (direct and indirect) 1,2, Row 7: Total Unexplained (direct and indirect) 3,4,5, Row 8: Total wage gap 1 to Note: Detailed estiation results can be found in Appendix Tables A5 and A6. This Table reports cobined nepotis and discriination coefficients for both stages as the unexplained part of the estiation. Table 3 contains several interesting results. Conventional second stage productivity estiates (Row 4) show that feales are lagging behind ales regarding observed huan capital. Conventional discriination estiates (Row 5) are negligible for 1988, showing a sall feale advantage in the 1994 cohort. Most of the wage gap is the result of differences in participation propensities (Rows 1 to 3). In both cohorts soe 70 percent of the decoposed effects are attributable to the indirect ipact of the participation process on earnings differences (through the selection variable). This is an iportant result which vindicates the use of the two stage process for 21

22 the analysis of participation and earnings differences. Only a sall part of the participation ipact on the wage gap is explained by individual characteristics, showing no observed differences in 1988 and only a ale advantage in The large unexplained part of the participation ipact on the wage gap strongly suggests that ales are favoured by the arket, especially in 1994 where ore than two thirds of the wage gap can be attributed to the estiates in Rows 2 and 3. The interpretation of the participation effects ust be done with care. If one accepts the arguentation introduced by Equation 6 (that is, that the only truly explained part of the participation effects is that which can be attributed to data differences in the participation estiations - presented in Row 1 of Table 3), then the participation process could be called highly discriinatory. However, one should be cautious about such an interpretation, as it assues that productivity differences have been explained adequately by the estiations. Although the data used in this paper includes several iportant productivity variables, the way legislative differences ay result in a arginal value product differential between ale and feale labour ay not be captured adequately by the data. Soe differences will be captured by existing variables. For exaple, the added cost to the eployer of staff pregnancies should be captured by age and sector duies, the added cost to the eployer of higher absences should be captured by the children and arriage duies. To the degree that the data used does not capture gender related differences in arginal value product, the ipact of such differences on the wage gap can be istaken as 22

23 discriination. At the sae tie, to the degree that eployers perceive that feales are in a harder position in ters of finding and retaining eployent, the estiates of this paper ay siply reflect rent seeking behaviour by eployers. The iportant point in the present context is that in Greece arginal value product differences generated ainly by EU legislation, coincided with rate of pay protection generated principally by national iniu wage rules. This paper reveals a coplex arket reaction in an environent with strongly increasing feale labour supply. Lower feale arginal value product depresses deand for feale labour, and better working conditions increase feale labour supply. The resulting excess supply depresses wages and increases feale eployent. Miniu wage rules kick in at soe point and stop wages fro dropping further. Feale eployent cannot iprove any ore and excess feale labour supply is aintained. Feales who wish to be eployed face tougher copetition and have to pay a preiu in ters of lower wage, soe of which ay well be the result of discriinating eployer behaviour. 7 C. Coparison of different decopositions The coparison of different decopositions based on the sae data in Table 4 can be inforative. Coparing the ale and feale decopositions with the pooled one reveals that the pooled explains ore of the wage gap. This coes as no surprise as the pooled odel uses ore inforation (the pooled estiation) and assues a ore realistic non- 7 A well infored eployer will spot such effects and act by offering lower feale wages. The observed arket response would be in this case that the botto of the wage distribution will be censored by iniu wage and all wages above that level will be depressed by tougher copetition. 23

24 discriinatory wage structure. One of the ain advantages of the pooled odel is that (unlike the ale and feale odels) it assues that in the absence of discriination the total wages paid in the econoy would be equal to the observed total wages paid. All that would happen in a non-discriinatory econoy would be that jobs would be re-shuffled in order to eliinate discriinatory pay differences, but the total capacity of the econoy assued by the non-discriinatory wage structure is identical to the actual capacity. Table 4 Coparison of several decopositions Model Male Feale Pooled (only direct) Pooled (direct and indirect) Effect/cohort st stage Unexplained st stage Explained nd stage Unexplained nd stage (Ex) (Un) (Ex) (Un) Explained % of Total Explained 59% 51% 42% 51% 78% 63% 28% 46% Total Wage Gap Investigation of the penultiate row in Table 4 shows how the pooled odel explains ore of the wage gap than either the ale or the feale odel. Notwithstanding the superiority of the pooled odel, it is worth noting how re-interpreting results through the inclusion of indirect effects in the analysis, decreases the overall explained part of the wage gap considerably. The proportion of explained to unexplained wage gap drops fro 78% (63%) to 28% (46%) for the 1988 (1994) cohort in Table 4. Clearly, restricting 24

25 indirect effects to be zero (that is, treating λ purely as data) seriously and istakenly under-estiates discriinatory effects. This is an iportant result the universality of which should be tested with further data sets. D. Counterfactual Estiates of the Developent of the Wage Gap Table 1 shows that the total wage gap between ales and feales changed between 1988 and 1994 by alost five log points in favour of ales. This section decoposes this change into its first and second stage explained and unexplained constituents. 8 Establishing the reason for wage gap changes can have iportant policy raifications. For exaple, if it is established that feales are losing out over tie because their observed huan capital is deteriorating in relation to that of ales, a sensible policy recoendation would be that education and training differentials should be looked at ore carefully. If, however, relative huan capital reains constant and the wage gap increases nonetheless, other factors should be exained, such as gender related eployer and eployee attitudes. Table 5 suarises the ters of Counterfactual 9. The essage is clear. Looking at the second stage estiates, one can see that, although feale productivity iproved between 1988 and 1994, ale productivity iproved faster. The net productivity effect was in favour of ales (the difference between Rows 5 and 6). A substantial reduction in the 8 It should be noted that while a large proportion of ales are self-eployed in the two cohorts, feales typically either engage in paid eployent or stay out of the labour arket altogether. This fact akes the deterioration of the relative feale pay even ore iportant because feales have fewer outside options regarding eployent than ales. An alternative way to look at the deterioration of feale earnings would be that, it is because ales have better outside options that they retain their reunerative advantage in paid eployent. It should also be noted that between 1988 and 1994 the Greek econoy perfored rather poorly. Any developent in the wage gap during the period could also be attributable to ales 25

26 wage gap is attributable to the unexplained part of the earnings estiates (Row 4). If anything, the arket sees to be reunerating the feales who are in eployent in 1994 better than it reunerated their 1988 counterparts. Hence, Table 5 suggests that, if participation effects were absent, relative feale earnings would have iproved between 1988 and and feales treated differently during difficult econoic ties. 26

27 Table 5 Counterfactual Wage Gap decopositions Total Wage Gap change Attributable to first stage selection estiates 1. Due to changes in unexplained participation Due to changes in unexplained indirect participation Due to changes in explained indirect participation Total WG change due to participation Attributable to second stage earnings estiates 4. Due to changes in unexplained earnings Due to changes in ale productivity Due to changes in feale productivity Due to changes in the non-discriinatory wage structure Total WG change due to earnings estiates Participation estiates convey a far less favourable picture regarding the relative feale position. Overall participation changes increased the wage gap considerably. Explained participation propensity changes favoured ales considerably, whilst unexplained changes were favoured the a lot less. The essage is that, whatever it is that akes feales observably worse participants than ales (one ust look at the participation 27

28 variables), they possessed a lot ore of it in 1994 and they have had to pay a heavy price as a result. Finally, changes in acroeconoic conditions captured by changes in the nondiscriinatory wage structure (Row 7) have worked in favour of feales between 1988 and This result accords with intuition as it suggests that the labour arket has oved towards the type of eployent and reuneration that suits feales. Indeed the data shows that a large ajority of the net eployent gains between 1988 and 1994 went to feales. Given the liited extent of feale eployent rate in Greece, this is a welcoe result as it shows that acroeconoic changes are working in the right direction regarding participation and the wage gap. 28

29 V. Conclusion This paper studied the link between labour arket participation and wages in paid eployent in Greece. It established the constituents of the ale-feale wage and participation gaps and found that the observed lower feale relative pay can be priarily attributed to the factors which deterine paid eployent participation. The participation process was found to be highly discriinatory in favour of ales, with feales paying a large preiu in ters of lower wages in order to be eployed. In the context of the fast changing socio-econoic and legislative environent of Greece in the 1980s and the 1990s, with very low but rapidly increasing feale labour supply, the paper established that the wage gap has been increasing priarily due to participation factors. Attepts to reduce the ale-feale wage gap in the 1980s through legislation on equal pay have been frustrated by considerable increases in feale participation as well as faily and aternity legislation which iproved the non-pecuniary position of eployed feales in the labour arket. The ain recoendation of this paper is that several costs which ake the eployent of feales ore expensive to the eployer, should be pooled in order to put ales and feales on an equal footing in the labour arket and reduce the incentive of eployers to eploy feales only when they are willing to accept lower wages. A prie target should be the costs of aternity leave, which can have a serious ipact in an econoy with predoinantly sall size eployers. As a quid pro quo, eployers should be subjected 29

30 to ore stringent regulations regarding feale hirings in order to proote feale eployent in a non-discriinatory fashion. 30

31 References Blinder A S (1973) Wage discriination: Reduced for and structural estiates, Journal of Huan Resources, 8: Becker G S (1971) The Econoics of Discriination, Revised Edition Chicago: University of Chicago Press. Blau F D and M A Ferber. (1987) Discriination: Epirical Evidence fro the United States Aerican Econoic Review, 77(2): Dolton P and G H Makepeace (1986) Saple Selection and Male-Feale Earnings Differentials in the Graduate Labour Market Oxford Econoic Papers, 38: Dolton P and G H Makepeace (1987) Interpreting Saple Selection Effects, Econoics Letters, 24: Dolton P and K G Mavroaras (1994) Intergenerational occupational choice coparisons: the case of teachers in the UK The Econoic Journal, 104: Dolton P, D O Neill and O Sweetan (1994) Gender Differences in the Changing Labour Market, ieo, University of Newcastle, pp Gerlach K (1987) A Note on Male-Feale Wage Differences in West Gerany, Journal of Huan Resources, 22: Goulka J and N Stern (1990) The Eployent of Married Woen in the United Kingdo Econoica, 57: Heckan J J (1979) Saple Selection Bias as a Specification Error, Econoetrica, 47: Hurd M (1980) A Copensation Measure of the Cost of Uneployent to the Uneployed, Quarterly Journal of Econoics, Septeber 1980: Jacobsen J P (1994) The Econoics of Gender, Cabridge: Blackwell Publishers. Kanellopoulos C N (1982) Male Feale Pay Differentials in Greece, Greek Econoic Review, 2: Licht G and V Steiner (1993) Male-Feale Wage Differentials, Labor Force Attachent, and Huan Capital Accuulation in Gerany ieo, ZEW Mannhei, pp Mavroaras K G and H Rudolph (1997) Wage Discriination in the Reeployent 31

32 Process Journal of Huan Resources, 32(4): Mavroaras K G (1999) Indirect re-eployent discriination Bulletin of Econoic Research, forthcoing. National Statistical Service of Greece (1992) 1988 Faily Expenditure Survey. Neuark D (1988) Eployers discriinatory behavior and the estiation of wage discriination Journal of Huan Resources, 23: Oaxaca R L (1973) Male-feale wage differentials in urban labor arkets International Econoic Review, 9: Oaxaca R L and M R Ranso (1988) Searching for the Effect of Unionis on the Wages of Union and Non-union Workers Journal of Labor Research, 9: Oaxaca R L and M R Ranso (1994) On discriination and the decoposition of wage differentials Journal of Econoetrics, 61: Psacharopoulos G (1983) Sex Discriination in the Greek Labor Market Modern Greek Studies, 2: Wellington A J (1993) Changes in the Male/Feale Wage Gap, Journal of Huan Resources, 28(2):

33 Appendix Description of Variables: CHBELOW6: Nuber of children in the household age 6 or less CH613: Nuber of children in the household age 6-13 HEAD: Household head. (1 if yes, 0 otherwise) MARRIED: (1 if arried, 0 otherwise) DIVORCED: (1 if divorced, 0 otherwise) RENTEDHO: Rented house (1 if rented, 0 otherwise) AGE24: Age between 17 and 24 AGE34: Age AGE44: Age SECHOME: Second hoe (1 if a second hoe at household s possession, 0 otherwise) HHSIZE: Nuber of total household ebers RETIRED: Nuber of retired household ebers. Education duies: SECEDUC: Copleted secondary education.(1 if yes, 0 otherwise) HIGHED: Higher education graduate. (1 if yes, 0 otherwise) Sectoral duies: TRANSP: Transportation and Counication. (1 if yes, 0 otherwise) MANUF: Manufacturing. (1 if yes, 0 otherwise) FINANCE: Banking and Finance. (1 if yes, 0 otherwise) UTIL: Public Utilities. (1 if yes, 0 otherwise) Urbanization variables: URB: Resident of area with ore than inhabitants SEMIURB: Resident of area. with inhabitants RURAL: Resident of area with less than 2000 inhabitants 33

34 Descriptive Statistics Table A1 Mean Values for Coplete Saple Variable Pooled Males Feales Pooled Males Feales CHBELOW CH URB SEMIURB RURAL HEAD SECEDUC HIGHED MARRIED DIVORCED RENTEDHO AGE AGE AGE SECHOME HHSIZE RETIRED PARTICIPATION RATE Cases Table A2 Mean Values for Paid Eployent Saple Variable Pooled Males Feales Pooled Males Feales SECEDUC HIGHED TRANSP MANUF FINANCE UTIL AGE AGE AGE MARRIED DIVORCED PUBLPRIV MANAGER λ Log WAGE SD of Cases WAGES

35 Probit Results Table A3 First-Stage Participation Probit Results 1988 Variable Coefficient Pooled 1988 Males 1988 Feales 1988 Standard Standard Error Coefficient Error Coefficient Standard Error CONSTANT CHBELOW CH URB SEMIURB RURAL HEAD SECEDUC HIGHED MARRIED DIVORCED RENTEDHO AGE AGE AGE SECHOME HHSIZE RETIRED Log-Likelihood Restricted log likelihood Chi-Squared Cases

36 Table A4 First-Stage Participation Probit Results 1994 Variable Pooled 1994 Males 1994 Feales 1994 Standard Standard Error Coefficient Error Coefficient Coefficie Standard Error CONSTANT CHBELOW CH URB SEMIURB RURAL HEAD SECEDUC HIGHED MARRIED DIVORCED RENTEDHO AGE AGE AGE SECHOME HHSIZE RETIRED Log-Likelihood Restricted log likelihood Chi-Squared Cases