Roopnarine, Karen Anne
(2018)
Three fourths a penny for your thoughts?: gender pay differentials in Trinidad and Tobago: an empirical analysis.
PhD thesis, University of Nottingham.
Abstract
The Caribbean is an understudied region in terms of gender wage gaps and this research adds new insights into the sparse economics literature on this topic for the region, and in particular, for the two-island state of Trinidad and Tobago. Economic inequality between men and women is a pertinent problem deserving of in-depth study because it has far-reaching inter-generational consequences. Furthermore, gender inequalities in the labour market are considered as indicators that considerably restrain economic growth. Trinidad and Tobago’s economy has undergone tremendous strides in terms of economic growth over the past 20 years, and this study provides a deeper understanding of how the gender pay gap evolved over that time period. The present analysis of the gender wage gap has allowed us to ascertain if working women in Trinidad and Tobago were able to benefit from the country’s improved economic prosperity.
The present study employs 2012 Continuous Sample Survey of the Population (CSSP) data for Trinidad and Tobago to investigate the causes of gender income differentials. The CSSP is used to generate labour force statistics for Trinidad and Tobago, and provides a wide range of information, including data on wages, gender, employment, unemployment, hours of work, industry, occupation, and level of education. The CSSP has two main advantages that make it a good source of data for analysing labour market issues in Trinidad and Tobago. Firstly, it is a nationally representative population survey, and secondly, it is the most detailed population survey for the country.
The Blinder-Oaxaca and Neumark methods of decomposition were used to portion the wage gap into “explained” and “unexplained” components. The findings suggest that the differential is not well explained by differences in the levels of human capital (“explained” component) and indeed gender bias in favour of male workers seems to be prevalent (“unexplained” component). The raw wage gap in 2012 measured 11.4 per cent, and in the absence of gender discrimination women’s wages could increase by as much as 26 per cent.
In addition to decomposing the gender wage gap at the mean level of wages, the research also investigated the causes of gender income differentials along the entire distribution of wages. Two recent quantile decomposition techniques – developed by the Machado and Mata (2005)/Melly (2006), and Firpo, Fortin and Lemieux (2009) were used to portion the gap into “explained” and “unexplained” components. Similar to the findings from the Blinder-Oaxaca methodology, the results for this portion of the research suggest that the differential in wages is not well explained by differences in the levels of human capital and substantial gender bias in favour of male workers.
Quantile decompositions allow us to ascertain if there is a “glass ceiling” or a “sticky floor” in the labour market. Glass ceilings are said to exist when there is a larger unexplained gender wage gap at the top of the wage distribution, whereas sticky floors exist when there is a larger unexplained wage gap at the lower end of the wage distribution. The results suggest that female workers in Trinidad and Tobago face sticky floors rather than a glass ceiling.
Lastly, the well-known Heckman two-step procedure (sometimes referred to as the limited information maximum likelihood (LIML) estimator) was employed to test for the presence of sample selection. The test for selectivity was carried out for both men and women separately. The results indicated no evidence of sample selection in any of the model specifications, including Mincerian-type wage regressions with additional controls for occupation, industry, and sector of employment (public vs. private). However, the sample selection model did not consider any exclusion restrictions due to data limitations, and consequently the model proved to be weakly identified. The chapter concluded that the “uncorrected” OLS subsample is the more appropriate model to be used for analysis given that these estimates are more robust compared to a sample selection model without exclusion restrictions.
Actions (Archive Staff Only)
 |
Edit View |
Tools
Tools
