The Luxembourg Income Study provides an invaluable source of income-surveys that are made compable across countries and over time. Not all the measurements of income are directly comparable, however. In some datasets the income variables were measured gross of income taxes and social contributions, whereas in other datasets they were measured net of income taxes and social contributions. Researchers seeking to do comparative analyses using the LIS will have to account for this difference between net and gross datasets.

In a new publication, we present netting down procedures, which are statistical tools that help improve the comparability of net and gross datasets in LIS. The paper discusses the issues involved with comparing net and gross income data, as well as the assumptions that are required when applying a netting down procedure. Two netting down procedures are discussed, and their performance in reducing bias is evaluated. The paper was co-authored by Rense Nieuwenhuis (that’s me – Institute for Innovation and Governance Studies (IGS), Universiy of Twente), Teresa Munzi (Data Team Manager and Research Associate of LIS) and Janet Gornick (Director of LIS).

The results indicate that directly comparing data on net and gross earnings (as a specific source of income) introduces bias to the analysis. This was not a surprising finding, because it is well known that progressive tax systems result in net earnings to be lower and distributed more equal than gross earnings. Nevertheless, it underlines the importance of carefully comparing net and gross earnings. Applying the netting down procedures allows users to approximate net earnings based on gross earnings and variables on income tax and social contributions. The paper provides the program code for use with SPSS, Stata, R, and SAS. The results of evaluating these netting down procedures suggest that the application of netting down improves comparative analyses across net and gross datasets in the Luxembourg Income Study.

Our new publication is titled “Netting Down Gross Earnings Data in the LIS Database: An Evaluation of Two Procedures”. The paper was published in the LIS Technical Paper Series, and is available online: http://www.lisdatacenter.org/wps/techwps/6.pdf

The premise of my project was something that is commonly believed amongst sociologists – and something I took for granted for many years. I even remember when I first learned about it during a first-year course on inequality. Simply put: because spouses’ earnings are positively correlated, the earnings inequality between households is thought to be bigger than the earnings inequality between individuals. Intuitively, this makes sense: it is easy to imagine how the earnings gap between households with two top-earners and households with two low-income spouses is quite big. Recently, a new book by Esping-Andersen (The incomplete revolution) is getting quite some attention, and is reporting on how increased women’s earnings have contributed to income inequalities between households in several countries.

I thought it would be very interesting to delve into this issue more, and try to analyze differences between countries in the degree to which women’s earnings affects the inequality between households in these countries. Interestingly, though, once I started working seriously on the project, it quickly became clear that women’s earnings attenuate – rather than increase – inequalities between households. Virtually all authors analyzing the contribution of women’s earnings to household inequalities report this attenuating affect, and do so for various countries and for different points in time. So, the common conception that women’s earnings generally increase inequalities, turned out to be a misconception. No need to be concerned of this unintended consequence of women’s increased labour force participation.

So, I had to revise my plan. I turned my attention to a systematic review of the literature with a special focus on how it is possible that spouses’ earnings are positively correlated (which, indeed, they are), but that when women’s earnings are high the inequalities between households tend to be reduced. I won’t spoil the anticipation by disclosing my results here (but please, do contact me if you want to be informed when these results become available), but I figured it out. Also, I am still working on a systematic analysis comparing countries with very interesting preliminary results (again, no spoilers).

This concludes the most technical part of this series of blogs on my New York experiences. Tomorrow some notes on how PhD students are trained in New York.