Gender Inequality and Labour Market

Abstract

Gender bias or preferential treatment towards the male child is deeply ingrained in socio-cultural milieu that has exacerbated the discrimination of other forms, namely unequal access to education, health, rights and freedom and eventually leading to labour market bias towards females. The discrimination is experienced not only at the household level but also outside household chores. The phenomenon of increasing gender bias leading to inequality in the labour market with widening differences between male and female is clearly discernible. Women are generally engaged in low-productive jobs in the informal sectors with low wages and earnings. Though, their presence in high-productive and modern-sector jobs has improved yet they form a minuscule proportion. This has widened income inequality in labour market with divergence in educational level, status of work (regular or casual) and work experience. This has been clearly noted even after positive policy initiatives and their improved participation in higher professional education and skill training. From the policy perspective, it is necessary to make secondary education universal and free so that they can move into a higher ladder in education pyramid. Investment in education and appropriate training is indispensable in order to widen their human capital and endowment base.

Earlier version of the paper was published in Economic and Political Weekly 52(8), February 2017 (Balwant Singh Mehta as principal author).

Appendices

Annexure 3.1: Technical Notes

Income Inequality Measures: The trends in inequality are examined using the Gini coefficient and three Generalised Entropy measures––the Mean Log Deviation (MLD), the Theil index and half the squared coefficient of variation.

The Gini coefficient can be computed as follows:

$$ {\text{Gini}} = \frac{{ - \left( {n - 1} \right)}}{1} + \frac{2}{{n^{2} \mu_{x} }}\sum\limits_{i = 1}^{n} {i*x_{i} } $$

Inequality trends according to the Generalised Entropy measures depend on the measures used because of the different weighting given to different parts of the income distribution. The formula for computing is:

$$ {\text{GE}}\left( \alpha \right) = \frac{1}{{\alpha \left( {1 - \alpha } \right)}}\frac{1}{n}\sum\limits_{i = 1}^{n} {\left[ {1 - \left( {\frac{{x_{i} }}{{\mu_{i} }}} \right)\alpha } \right]} $$

The parameter alpha (α) represents weight given to income differences at different points of the income distribution of workers. The GE(0), the mean log deviation, and it gives more weight on income differences at the lower end of the distribution, and is more sensitive to changes at that distribution. The GE(2), half of the square of the coefficient of variation, and it gives more weight on income differences at the upper end of the distribution. The GE(1), Theil index, gives equal weights on income differences across the entire distribution and exhibits constant responsiveness across all ranges of income.

Decomposition: Fields (2003) has proposed an alternative approach that considers simultaneously the impact of several given characteristics on incomes, and allows us to distinguish the contributions of each characteristic. The approach is useful as it helps us to factor in the contribution of different explanatory variables including variables with non-linear effects and categorical variables entered as a string of dummy variables.

As some of the differences in income between the different employment statuses can be attributed to workers’ educational attainment and to the occupation or industry, this approach allows us to simultaneously account for these differences. We adopt the method developed by Fields (2003), which decomposes the contribution of various explanatory variables to the level and change in inequality within a standard semi-logarithmic wage (or earning) regression model. The first step in the regression-based decomposition methodology is the estimation of a semi-logarithmic Mincerian (standard or augmented) wage/earning function:

$$ \ln Y_{it} \, = \,a_{t} z_{it} $$

where ln Y_it is the log-variance of earnings;

at = [αt β1t β2t … βJt 1] and

Z_it= [1 xi1t xi2t … xiJt ε_it] are vectors of coefficients and explanatory variables, respectively.

A general approach to analyse household earning inequality would be to regress the log income on the characteristics of the household head such as gender, age, socio-religious category, education, industry. (Katz and Murphy 1992; Gottschalk and Joyce 1995; Fields, 2003). However, we have modified this standard approach in two ways. One, as our interest is to understand the factors that contribute to inequality at the earning level; we have included the characteristics of the wage workers in the regression. Two, several other factors such as days of work and employment status are also included in the regression to understand the impact of changing work pattern on inequality.

In the second step, the estimated standard semi-log regression is decomposed to compute the relative factor inequality weights (i.e. the percentage of inequality that is accounted for by the jth factor), which is as follows,

$S_{j} \left( {\ln Y} \right) = {\text{GE}}\left( \alpha \right) = \frac{{\text{cov} \left[ {ajZj,\ln Y} \right]}}{{\sigma^{2} \left( {\ln Y} \right)}} = \frac{{a_{j} *\sigma \left( {z_{j} } \right)*{\text{cor}}[z_{j} Ln,Y]}}{{\sigma^{2} \left( {\ln Y} \right)}} ,$

where S_j (ln Y) denotes the share of the log-variance of income that is attributable to the j’th explanatory factor; cov [.] denotes the covariance, cor (.) the correlation coefficient and σ(.) the standard deviation. The above decomposition, in other words, computes how much income inequality is accounted for by each explanatory factor, which is the ‘levels question’. We have excluded the residual and made the total of subcategories of explanatory variables 100 and then calculate the contribution of each factors and later combined each attribute and plot graph to show the difference over the period.

Annexure Table 3.2

See Tables 3.9, 3.10 and 3.11.

Table 3.9 Occupational distribution of regular workers by socio-religious groups in India (15–59 age groups), 2011–12

Table 3.10 Salary of regular worker by occupational and by socio-religious-group by gender

Table 3.11 Share of workers by their occupations by gender