Gender Wage Discrimination and Poverty in the EU

Abstract

This paper analyzes the role of gender wage discrimination in household poverty rates in several European Union (EU) countries using the European Community Household Panel. In order to quantify the impact of discrimination on poverty, it proposes the construction of a counterfactual distribution of wages where discrimination against women has been removed. Using this new wage distribution, the study computes total household income and compares poverty rates in the absence of discrimination to those actually observed. The results show that, in general, discrimination against women plays a determinative role in the current levels of poverty in EU countries, although results by country show that this role differs in intensity and pattern. Further, the study finds that in EU countries the effect of discrimination on poverty risk dramatically increases for individuals in households that largely depend on working women' earnings.

Keywords:

• Income distribution
• labor participation
• gender
• inequality
• poverty
• wage discrimination

Acknowledgments

The authors would like to thank the Editor, the Associate Editor and the anonymous referees of the journal for their insightful comments, which have clearly improved the initial version of the manuscript. Financial support from the Ministerio de Ciencia e Innovación (SEJ2007-67911-C03-01/ECON) and the Xunta de Galicia (Programa de Estruturación de Unidades de Investigación en Humanidades e Ciencias Sociais 2006/33 and FEDER funds) is gratefully acknowledged.

Notes

¹ See Joseph G. Altonji and Rebbeca M. Blank (1999) for a good survey on this literature.

² This paper only focuses on gender discrimination, although workers can be affected by multiple sources of discrimination, as recently analyzed by Kanchana N. Ruwanpura (2008).

³ Except Sweden, all countries that were members of the EU-15 in 2001 are included in the Panel: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Portugal, Spain, the Netherlands, and the UK. Luxembourg has been excluded from this study because of its small size. The Netherlands has also been excluded because in the December 2003 release of the Panel, the share of people in that country with university degrees was small, and this problem was systematically repeated in all the information for the following waves in which education was updated.

⁴ For American women between 1970 and 1995, Francine D. Blau (1998) reports a substantial reduction in several gender gaps, after which, however, large gaps still remain (Blau and Kahn 2006).

⁵ Randy Albelda (1999) argues that in the US, this phenomenon is particularly large, in comparison to Europe, due to an inadequate social welfare system regarding families' needs.

⁶ Consider, for instance, the case where an unemployed woman receiving an unemployment social benefit accepts a job offer as a consequence of the higher market wage when discrimination is removed. In that case, t*_i would indicate the new number of hours worked without discrimination while Δt*_i would accommodate the reduction in social benefits.

⁷ Alternatively, if one regards g as a discrimination tax, it would be regressive (progressive) if I(x*)−I(x) < 0 (I(x*)−I(x) > 0).

⁸ Exp() is the expected value of the log-normal variable y_f conditioned on Z_fi in the OLS regression. Note that in the case of estimating wage equations using Quantile Regressions, expression (1) would become: where q stands for the corresponding quantile.

⁹ In our empirical analysis, we use OLS estimations and men's wage structure as a benchmark. Two different specifications of this regression model are displayed, depending on whether specific employment variables are included. In all cases, selection bias is corrected.

¹⁰ Luxembourg was excluded because of its small population size, while the Netherlands was excluded because its share of population with education level above the primary level in the ECHP is implausibly small – less than 2 percent of those aged 16–25 years. In this situation, it is impossible to differentiate the Dutch population by education, an extremely important variable in the analysis of wage gaps.

¹¹ Women's labor activity rates or labor force participation rates are commonly referred to as the ratio of women employed or looking for a job out of all women in the population aged between 16 and 65 years.

¹² Women's employment rate is then the ratio of employed women out of those who participate in the labor market (the employed and the unemployed).

¹³ Some countries seem to show a few odd figures if we compare them with those reported by EUROSTAT using the Labour Force Survey. This discrepancy is due, at least partially, to the fact that Table 1 focuses only on women wage earners between the ages of 25 and 55, a range that excludes both young and elderly employees and the self-employed.

¹⁴ Wage earners at both wage tails (first and last percentile) have been removed in an effort to reduce the undesirable effect of measurement error, a problem that often affects the information on income and wages in household surveys such as the ECHP.

¹⁵ Results of wage equations for women in the private sector are shown in Appendix Table 1. The rest of the regressions were omitted for the sake of simplicity. They confirm standard results in the literature. In general, higher education, especially holding a university degree, and long work experience increase the log-wage. However, results for Denmark and Belgium must be taken with caution given that they appear to have few variables that are statistically significant at 95 percent, even if all complete models are jointly significant. Probably the small sample sizes of these and other countries (such as Ireland and Finland) are responsible for this. We consider that the quality of these regressions is enough for comparability purposes, even if a detailed analysis with alternative data sources would be required for a deeper analysis of these particular countries. A complete report on all detailed results related to this Data Appendix is available upon request.

¹⁶ Note that if we restrict this comparison to wage earners in the private sector, these differences are even higher (more than 10 percentage points in Belgium, Ireland, Italy, Spain, and Portugal). Recall, however, that our regressions are controlling for selection bias in the working female population.

¹⁷ Variables are measured in terms of equivalent income. Equivalent household income is the current net monthly household income divided by the number of equivalent adults using a modified OECD scale (where 1.0 is the weight of the first adult, 0.5 is that of any other individual aged 14 or over living in the household, and 0.3 is that of children aged less than 14 years). Results in Table 4 (columns 3, 5, and 8) are obtained by calculating, for each household, either the total women's wages or total household gap relative to their particular household actual or counterfactual equivalent income. We then report the arithmetic mean of these relative contributions in the corresponding country.

¹⁸ Similar results are obtained if we decide to use other relative inequality indices, such as those from the family of generalized entropy.

¹⁹ This is precisely the case in the Spanish labor market, as reported by Coral del Río, Carlos Gradín, and Olga Cantó (2008).

²⁰ The poverty gap is defined as the distance between the poverty line and the income of each poor person below that threshold.

²¹ In any case, it is important to remember that Danish (and Austrian) child poverty levels are very low (9.7 and 8.7, respectively); thus, small absolute reductions cause high relative decreases.

²² Once again, some caution must be taken regarding the results for Denmark and Belgium, since few variables are statistically significant at 95 percent when including the employment variables in the wage equations. Unfortunately, the ECHP does not provide a large enough sample size for these countries in order to undertake a more detailed analysis.