Public–private wage differentials in Turkey: public policy or market dynamics?

Abstract

We evaluate public–private sector wage differentials by gender in Turkey between the years 2005 and 2013. Using micro data from Household Labor Force Surveys we find a positive premium for low wage earners and a penalty of working in the public sector at the higher end of the distribution. Although the penalty has not disappeared, the price effect has increased at both ends of the distribution. The increase at the lower tail is attributed to a higher price effect in the public sector, whereas at the higher tail it has been a result of a relatively uneven wage increase in the private sector along the distribution, rather than an explicit public wage policy.

Keywords:

• Public–private wage gap
• decomposition methods
• Turkey

JEL Classifications:

• J31
• J45
• C31

Acknowledgements

The authors gratefully acknowledge the support of Galatasaray University Research Fund [grant number 15.103.003].

Notes

1. Note that some studies use the term premium as equivalent to the total, unadjusted, gap (Borjas 2002).

2. Among recent comparative empirical studies, de Castro, Salto, and Steiner (2013), Christofides and Michael (2013), and Depalo, Giordano, and Papapetrou (2015) estimate and analyse the gap for respectively 26, 27 and 10 EU countries; Mizala, Romaguera, and Gallegos (2011) for 11 Latin American countries; and Lausev (2014) provides a literature survey comparing eastern European countries with developed economies. Notwithstanding discrepancies in sources, samples, definition of public sector employment and estimation techniques, the results from Christofides and Michael (2013), Depalo, Giordano, and Papapetrou (2015) and Mizala, Romaguera, and Gallegos (2011) suggest that countries with a stable or increasing premium across the wage distribution constitute a minority.

3. In most developing countries, childcare supply is greater in the public sector, either because it is not compulsory in the private sector or not respected. Similarly, maternal leave, even when legislation exists, is rarely fully applied in the private sector.

4. Both Tansel (2005) and San and Polat (2012) refer to the same data-set (1994), whose full name is the ‘Household Income and Consumption Expenditure Survey’ (HICES). The survey was not conducted between 1995 and 2001, and is called the ‘Household Budget Survey’ (HBS) from 2002 onwards.

5. Another option would be using panel data – a feature that is absent from Turkey’s HLFSs – and estimate fixed effects quantile regressions, which can still be a problem: ‘unlike in linear models where the unobserved individual fixed effects can be differenced out and are thus omitted from the estimation, the individual fixed effects in a fixed effects quantile regression model have to be estimated along with the coefficients. As a result, it is very difficult to implement when the number of parameters involved here is so large. Further, only when the endogeneity of sector choice results from time-invariant individual heterogeneity, can a fixed effects model (either linear or quantile regression) solve the endogeneity issue. When sector choice depends on time-variant unobservables, fixed effects models are not helpful, and instrumental variables are required to deal with the endogeneity’ (Cai and Liu 2011, 371–372).

6. Another solution to the selection issue are PSM techniques, which basically consist of finding individuals that match in both sectors according to control variables, who then constitute the non-biased sample used for directly measuring the premium along the new distribution. However, here there is a trade-off between having a large number of covariates, which is desirable for best controlling observables, and decreasing the number of individuals that fall within a common support region, especially in quantile regressions. The PSM then has two drawbacks: either the distributions obtained after the PSM will contain too few individuals (especially in the public sector), or, as a number of controls must be left out, the sample will actually continue bearing heterogeneities, or both. Unsurprisingly, the results we obtained from various PSM estimations leave us with an insufficient number of matching individuals (results available upon request).

7. For a comprehensive account of the various decomposition techniques see Fortin, Lemieux, and Firpo (2011).

8. In what follows we follow the notation used in Fortin, Lemieux, and Firpo (2011) and use the counterfactual package developed by Chernozhukov, Fernández-Val, and Melly (2013) for Stata.

9. This is also the approach adopted in the analysis below, we further exclude informal employees. Tansel and Kan (2012) find that there is a transition from informal to formal employment, but not the other way round, which suggests the existence of a segmented market.

10. As expected, the alternative public employment definition based on the classification of economic activities (data available at ILOSTAT, not reported here) does not always match the one adopted here, as the role of the public sector greatly varies across countries.

11. TurkStat established an Address Based Population Registration System (ABPRS) in 2007; from 2009 onwards HLFSs have been revised by the 2008 based population projection.

12. The classification of sectors of economic activity (NACE) has undergone a number of breaks throughout the period, such that the classifications were not compatible. Recently released micro data by TurkStat has been retrospectively revised to harmonize sectoral classification (NACE rev.2).

13. The aggregation tables of occupation and economic activity classifications are given in Appendices 1 and 2.

14. Although excluded from the empirical exercise, we give statistics for the informal sector wage earners for descriptive, comparative purposes.

15. Comparing Turkey and the United States, Koçer and Visser (2009) show that weak social policies or collective bargaining mechanisms are likely to be compensated by minimum wage legislation as in the case of Turkey.

16. See Duman (2014) for a recent comprehensive account of the weakness of collective bargaining institutions in Turkey.

17. Returns to tenure in the public sector are higher and follow an unconditional compensation scheme regardless of position switch.

18. On the prevalence of long working hours in Turkey see Toksöz (2008).

19. Differences in distributions (public versus private sectors and these versus the distributions of the public–private wage gap) explain why the deciles where the contribution of the price effect to total public–private wage gaps increase do not match exactly with the deciles at which the contribution of the price effect in the private sector goes below that of the public sector.

20. Comparing premia in France, Italy and Great Britain, Ghinetti and Lucifora (2013) highlight that in a context of centralized wage setting as observed in France, this is an expected outcome.

21. Unfortunately the HLFS questionnaires do not ask for union membership. Even then, it is doubtful whether these would control for what they are meant to. Typically, although there is some evidence that public employees are more unionized than their private counterparts, they do not have the right to strike; and ‘collective bargaining’ in the public sector is an institution that originates from a historically centralized public wage policy that is not noticeably correlated with the existence of unions, or rate of unionization.