Labor Supply and Earnings Inequality: A Cross Country Study
Two facts lead us to think that this topic deserves to be explored. The first one motivates our decision to carry out a cross-country study and refers to the differences in average hours worked across countries. A large literature has proposed explanations for these differences, such as the tax system or the role of labor contracts and unions (See, for instance, Prescott (2004) and Alesina et al. (2006)). Whatever the reasons for these differences, countries seem to fall into one of two categories, "workaholic", such as the US and the UK, or "work-sharing", such as Germany, France and Italy (We take this terminology from Bell and Freeman (2001) and Hunt (1999) respectively). Average hours worked are higher in "workaholic" than in "work-sharing" countries and we are interested in examining to what extent these differences affect inequality. The second reason that motivates our analysis is the response of the labor supply to the changes in the dispersion of hourly wages. Some previous papers show that greater wage inequality is associated with higher average hours of work, meaning that the increase in wage inequality of the last decades is likely to have spurred an increase in hours worked (See, for instance, Bell and Freeman (2001) and Bowles and Park (2005)).
Our claim is that the impact of such a mechanism on overall earnings inequality strongly depends on the labor supply behavior of the type of country considered. More precisely, the labor supply of a "workaholic" country would seem to be characterized by a stronger substitution effect, meaning that hours worked tend to increase with the level of hourly wage. By contrast, the "work-sharing" group shows a labor supply where the income effect is dominant, this translates into a preference for leisure over hours worked for certain levels of hourly wages. In other words, we expect to observe a rise in inequality of hours worked to follow the increase in hourly wage dispersion, and that the impact of such a mechanism on overall inequality strongly depends on the evolution of the correlation of hours worked and hourly wage. In particular, the sign of the correlation between hours worked and hourly wage is likely to be positive for the first group of countries and rather negative for the second, therefore impacting overall inequality in opposite directions.
In order to capture the effects mentioned above, the core methodology of the paper consists of decomposing an index of earnings inequality, computed for each country and over different points in time. We use the Half Squared Coefficient of Variation (SCV), which belongs to the General Entropy family and respects all the suitable properties of an inequality index. Following Shorrocks (1982), it can be decomposed over additive sources into a scalar transformation of inequality in hourly wages, an equal transposition of inequality in hours worked and a third component that expresses the correlation of the two sources (In our case we would need to take logs to obtain additive sources. For detailed calculations of the decomposition please refer to Jenkins (1995) and Garcìa-Peñalosa and Orgiazzi (2013)). Those three terms give us, respectively, the contribution of inequality in hourly wages, of inequality in hours worked and of the correlation between hours and hourly wages to overall earnings dispersion. The reason why we are keen on using such an index is that it is fairly sensitive to extreme values and this could give us important information about the top and the bottom part of the distribution of earnings, but applying the SCV over the log of earnings may not respect the Pigou-Dalton principle of transfers. An additional decomposition à la Esteban and Duro (1998) will be introduce to check the direction of the results.
We use household surveys for Italy, Germany, France, the US and the UK, and different points
in time over four decades starting from the 1980's. In particular, we use the Survey of Household
Income and Wealth for Italy, the German Socio-Economic Panel for Germany, the Current
Population Survey for the US, and the Enquete Emploi for France. For the UK we combine
three different surveys to cover the decades we are interested in: the Family Expenditure Survey,
the British Household Panel Survey and Understanding Society. To homogenize the different
variables we will follow rigorous recommendations of the main institutions in this field (Such as the International Labour Organization (ILO) and the Institute for the Study of Labour (IZA).).
If we compare the US and Italy, for instance, we notice that the distributions of weekly hours worked for prime-aged workers differ substantially. The correlation between hourly wage and hours
worked also differs: it is equal to -0.32 and -0.13 in Italy for the years 1995 and 2004, respectively, while it amounts to 0.10 and 0.12 in the US for the same pair of years. That is, in the US,
medium-high hourly wage workers are likely to work more hours than poorly paid individuals
and this tend to magnify a given distribution of wage rates. By contrast, in Italy, working hours
decrease with the hourly wage level, therefore the labor supply response shrinks initial differences in wages.
The paper also considers differences across groups in hour responses, either by gender or by
education level. This allows us to understand whether a particular type of worker (eg. male
highly educated individuals) drives the labor supply behavior observed at the aggregate level in
a particular country.