Educational Assortative Mating and Income Inequality in Denmark

Of course, changing the distribution of types will not change the income gaps between any pair of types, but it may change the average gap by making some types more numerous and others less so.

Details are available online (http://​en.​iu.​dk/​education-in-denmark/​the-danish-education-system).

Just over 2% of men and women are missing information on education. These are immigrants who were educated in their home country before they came to Denmark and whose educational activities were not recorded in the registers. Statistics Denmark conducts regular surveys of this group to collect the missing information, but the data are not yet complete.

This growth of single-person households could be an indication of the reduced utility of forming partnerships and/or of sex differences in life expectancy and an aging population. But over this period, widows and widowers declined, respectively, from 20% to 10% and from 5% to 3% of the population, while unmarried women and men increased from 32% to 43% and from 56% to 62%, respectively, suggesting that the increase in single households is the result of changed mating behavior rather than population aging.

Obviously, employers may fail to report employer’s earnings, just as self-employed persons may report only part of their earnings. However, in Denmark, the tax authorities are extremely vigilant, and tax fraud is punishable by heavy fines and possible imprisonment.

The reason for the high share of Danish citizens with zero earnings is quite complex. Generous welfare benefits increase reservation wages and may reduce benefit recipients’ incentives to work, while the rules often prevent recipients taking a low-income job if they wish to maintain their entitlement to benefits. Furthermore, in Denmark there is not as much stigma attached to receiving income replacement benefits as there is in the United Kingdom or the United States.

For deflating, we use the official consumer price index published by Statistics Denmark (http://​www.​dst.​dk/​da/​Statistik/​emner/​prisindeks/​forbrugerprisind​eks.​aspx).

For a report in Danish on this, see Økonomi- og Erhvervsministeriet (2002).

To understand how we constructed these counterfactual distributions, consider the case in which we allowed only the distribution of couple households to change and kept single-person households fixed at their 1987 distribution. We first constructed the square 64-cell table of household types (where the 64,64 cell contains a structural zero representing the “Absent,” “Absent” category) in which the last row and last column contains the proportion in the single household types in 1987 and the remainder of the table contains the proportions of couple households in 2006. The entries of this table will not necessarily sum to 1, nor will their row or column totals match those of the 1987 or 2006 observed distributions of household types. So we used the Deming-Stephan (1940) algorithm (also called iterative proportional fitting) to adjust the entries in the table so that their marginal distributions matched that for 2006 (Deming and Stephan 1940). This was then our counterfactual table. An analogous method was used to form the counterfactual table in which only the pattern of educational homogamy changes. Further details of the method are available on request from the first author.

The lower age limit for this analysis is now 28 rather than 30, but analyses using 30 as the lower limit yield identical results.

As shown in Fig. 6, inequality tends to be quite high for couples in which one or both partners have missing information on education.

The detailed results are available from the authors on request.

Holding \( \overline x_j \) at its 1987 value and allowing both p _j and T _j to take their 2006 values yields a counterfactual Theil value of 0.1916, an increase of 20% compared with the observed increase of 21%.