Presentation on theme: "Industry Wage Differentials, Unobserved Ability and Rent-Sharing Evidence from Matched Worker-Firm Data, 1995-2002 Robert Plasman (ULB, DULBEA) François."— Presentation transcript:
Industry Wage Differentials, Unobserved Ability and Rent-Sharing Evidence from Matched Worker-Firm Data, Robert Plasman (ULB, DULBEA) François Rycx (ULB, DULBEA, and IZA-Bonn) Ilan Tojerow (ULB, DULBEA, and IZA-Bonn)
2 1. Motivation Krueger and Summers (1988) have shown that substantial wage differentials exist in the US among workers with the same observed individual characteristics and working conditions, employed in different sectors. Since then, similar results have been obtained for many industrialised countries (Araï et al., 1996; Ferro-Luzzi, 1994; Hartog et al., 1997, 2000; Lucifora, 1993; Vaïniomaki and Laaksonen, 1995). Accordingly, the existence of sectoral effects on workers wages has become an accepted fact in the economic literature.
3 1. Motivation Nevertheless, the existence of inter-industry wage differentials remains a complex and unresolved puzzle. The role of unmeasured abilities in explaining inter-industry wage differentials is still unsettled (Abowd et al., 1999; Benito, 2000; Bjorklund et al., 2004; Carruth et al., 2004, Gibbons and Katz, 1992; Goux and Maurin, 1999; Martins, 2004). Studies hardly allow to discriminate among alternative models supporting the existence of an effect of the employers characteristics on wages, e.g. efficiency wages, rent-sharing (Benito, 2000; Krueger and Summers, 1988; Lindbeck and Snower, 1990; Thaler, 1989; Walsh, 1999).
4 1. Motivation What about Belgium ? The existence of inter-industry wage differentials in Belgium, for both male and female workers and across bargaining regimes, has been recently highlighted by Rycx (2002, 2003) and Rycx and Tojerow (2002). Using data for 1995, the latter show that their structure is comparable with that observed in the other industrialised countries and that they result in part from the characteristics of the employers in each sector (i.e., size of the establishment and level of wage bargaining). Moreover, findings support the hypothesis of a negative relation between the dispersion of inter-industry wage differentials and the degree of corporatism of the industrialised countries.
5 2. Aim of the Paper Research Questions: Are sectoral differences in pay a temporary phenomenon or do they persist over time ? Do they derive from sectoral differences in the unobserved quality of the labour force ? To what extent are they shaped by the sectors ability-to-pay, i.e. profits ? What is the contribution of rent-sharing to the observed industry wage differentials ? Data Sets: Structure of Earnings Survey (SES) and Structure of Business Survey (SBS) for the years 1995, 1999 and Representativity: all private sector establishments employing at least 10 workers.
6 3. Descriptive Statistics Number of observations in the SES: between circa 67,000 (in 1995) and 108,000 (in 1999). Average gross hourly wage (including bonuses): 15,5 EUR in Best paying sectors in 2002: Electricity, gas, steam and hot water supply (27 EUR/hour). Manufacture of coke, refined petroleum and nuclear fuel (23.1 EUR/hour) Insurance and pension funding (21.6 EUR/hour) Financial intermediation (21.4 EUR/hour) Post and telecommunications (21.2 EUR/hour) Worse paying sectors in 2002: Hotels and restaurants (9.5 EUR/hour). Land transport and transport via pipelines (11.1 EUR/hour). Manufacture of furniture (11.9 EUR/hour). Manufacture of wood and products of wood and cork (12 EUR/hour). Manufacture of textile / Retail trade (12.1 EUR/hour).
7 4. Magnitude, Dispersion and Stability ? Methodology: Krueger and Summers (1988), Zanchi (1998). Rests upon the estimation of the following wage equation: where, w i : gross hourly wage of worker i (excluding / including bonuses), X ji : vector of workers characteristics and working conditions (e.g., education, experience, tenure, occupation, sex, contract, hours), Y gi : industry dummies (Nace 2 and 3 digits), Z li : vector of employers characteristics (e.g., size of the establishment, region, level of wage bargaining), ε i : error term.
8 4. Magnitude, Dispersion and Stability ? Main Findings: In all periods (95, 99 & 02), substantial wage differentials are observed between workers employed in different sectors, even when controlling for working conditions, individual and firm characteristics. More than 70 per cent of these differentials are significantly different from zero (at the 0.05 level) Hierarchy of sectors in terms of wages is quite stable over time. (Correlation ranges between 0.70 and 0.90)
9 4. Magnitude, Dispersion and Stability ? Main Findings: Best paying sectors ( ceteris paribus ): –Electricity, gas, steam and hot water supply [+27; +31%]. –Manufacture of coke, refined petroleum and nuclear fuel [+20; +34%]. –Air transport sector [+12; +19%]. –Manufacture of chemicals and chemical products [+11; 12%]. –Financial intermediation [+6; 13%]. Worse paying sectors ( ceteris paribus ): –Hotels and restaurants [-11; -14%]. –Manufacture of wearing apparel, dressing and dying of fur [-11; -13%]. –Retail trade [-7; -12%]. –Manufacture of furniture [-8; -10%]. –Manufacture of textiles [-4; -8%].
10 4. Magnitude, Dispersion and Stability ? Main Findings: Dispersion of inter-industry wage differentials: Measured by the employment-weighted and adjusted standard deviation of the industry wage differentials (WASD). Belgium occupies a middle position among the industrialised countries with regard to the WASD. Depending on the specification adopted, the WASD decreased by between 10 and 20% between 1995 and Explanation : Increasing product market competition ?
11 5. Unobserved Ability Explanation ? Methodology: Martins (2004), quantile regressions. Intuition : Workers with better unobserved characteristics (e.g. ability, motivation, industry-specific skills) are likely to be found at the top of the conditional wage distribution. According to the unobserved quality explanation, workers with better unobserved characteristics are over-represented in high- wage sectors.
12 5. Unobserved Ability Explanation ? Methodology: If explanation valid, we expect : Industry wage differentials to be substantially larger at the top end of the conditional wage distribution. A bigger difference in industry wage premia across the wage distribution in high-wage sectors than in low-wage sectors. A strong positive correlation between industry wage differentials computed at the mean and at the 90th percentile of the wage distribution.
13 Main Findings: Industry wage premia are on average larger at the 90th percentile than at the 10th percentile. High-wage industries are on average characterised by bigger differences in wage premia between the top and the bottom percentiles. A positive relationship is recorded between mean industry wage premia and differences in wage premia across the wage distribution. These results are compatible with unobserved ability hypothesis. 5. Unobserved Ability Explanation ?
14 Main Findings: Yet, caution is required: Almost 40% of the industry wage differentials are not significantly different at the top and bottom of the conditional wage distribution. Top and bottom industry wage differentials are correlated with the same intensity to the mean industry wage premia. Correlation between mean and inter-percentile industry wage premia is lower than expected. Unobserved ability hypothesis may not be rejected. However, its explanatory power appears to be limited. Role of non-competitive forces should not be neglected. 5. Unobserved Ability Explanation ?
15 6. Sectoral differences in ability-to-pay ? Main Findings: Industry wage premia derive at least partly from sectoral differences in ability-to-pay, i.e. profits.
16 7. Contribution of Rent-sharing ? Objectives: Analyse whether firms do share rents with their workers and if the gains from profits vary along the wage distribution. Examine to what extent industry wage differentials are explained by a rent-sharing phenomenon. Investigate the significance, magnitude and dispersion of industry wage premia before and after controlling for rent-sharing.
17 7. Contribution of Rent-sharing ? Methodology: Re-estimate our benchmark wage equation including an additional explanatory variable, i.e. profits-per-capita at the firm level. Requires to: Correct for group effects in the residuals (Greenwald, 1983; Moulton, 1990). Control for the endogeneity of profits using instrumental variables.
18 7. Contribution of Rent-sharing ? Main Findings:
19 7. Contribution of Rent-sharing ? Main Findings: After controlling for rent-sharing: Substantial wage differentials are still observed between workers employed in different sectors. Hierarchy of sectors in terms of wages remains almost unchanged. However, Proportion significant industry wage differentials drops from around 75 per cent to 50 per cent. Size of (positive and negative) industry wage premia decreases sharply. Dispersion of inter-industry wage premia (WASD) diminishes by around 30%. Rent-sharing explains a large fraction of the observed industry wage differentials.
20 8. Conclusion Large and persistent wage differentials are recorded among workers with the same observed characteristics and working conditions, employed in different sectors. The unobserved ability hypothesis may not be rejected on the basis of Martins (2004) methodology. However, its contribution to the observed industry wage differentials appears to be limited. Workers earn ceteris paribus significantly higher wages when employed in more profitable firms. The instrumented wage-profit elasticity stands at This rent-sharing phenomenon explains a large fraction of the observed industry wage differentials. Indeed, the significance, magnitude and dispersion of industry wage differentials decreases sharply when controlling for profits.