Principal investigator: Bruno Benevit
Original title: Educational Assortative Mating and Household Income Inequality
Authors: Lasse Eika, Magne Mogstad and Basit Zafar
Location of the Intervention: Denmark, Germany, Norway, United Kingdom, United States
Sample Size: 2.846.609 individuals
Sector: Education
Primary Variable of Interest: Income inequality
Type of Intervention: Marriage selection based on education
Methodology: OLS, DiNardo Decomposition
Summary
The persistence of inequalities among families is one of the most discussed topics in societies today. The selection of partners with similar social characteristics can be highlighted as a mechanism for perpetuating this phenomenon, also encompassing the level of education of individuals. This article sought to identify how income inequality among families is associated with marital selection based on the educational level of partners in five countries between 1962 and 2013. Through a descriptive analysis and the use of a semiparametric decomposition method, the evidence found demonstrated that individuals in all countries select partners with similar levels of education, and that this behavior explains a considerable part of the income inequality among families.
- Policy Problem
The wealth of households is closely related to the socioeconomic conditions of family members. Thus, income inequality among families can be explained by the characteristics of the couple that make up the head of the household and their spouse. To the extent that partners select individuals with similar socioeconomic conditions, this behavior can explain the persistence of income inequalities among families (Eika, Mogstad and Zafar, 2019).
In this context, selecting partners with the same level of education (educational assortative matingThis behavior can represent one of the factors inducing the maintenance and expansion of income inequality among families (household units). However, for the correct identification of the effect of this behavior on income distribution over time, it is necessary to consider aspects such as the expansion of access to higher education for both genders (Liu and Lu, 2006), and the returns from education in each period (Autor, Katz and Kearney, 2008).
- Policy Implementation Context
Between 1962 and 2013, the Gini coefficient for household income among married couples increased from 33,9 to 43,2 in the United States. Simultaneously, the proportion of couples in which both spouses have a college education rose by 22 percentage points during the same period. Similar trends were observed in Germany, Denmark, Norway, and the United Kingdom.
Specifically, the proportion of husbands with a university degree was 13% in 1962 and 37% in 2013, while only 7% of women held a university degree in 1962, approaching that of their husbands in 2013, reaching 40%. An increase in the premium for higher education was also observed over this period. Additionally, the ratio of male to female income also fell from 6,4 in 1962 to approximately 2,1 in 2013.
- Evaluation Details
The analysis of the United States was based on two data sources. The first analysis for the United States covered the period from 1962 to 2013, using the publicly available Monthly Population Survey (CPS). The CPS provides nationwide and representative data for each year, allowing for the identification of households and family members. The variables considered included individual demographic information (such as gender, date of birth, and marital status) and socioeconomic data (including educational level and income).
The second data source involved using data from the decennial United States Censuses. The main advantage of census data lies in its temporal scope, allowing the study of marriage selection by education since 1940. Due to the impossibility of identifying spouses, the 1950 Census was not considered. The analysis of family income inequality for the United States considered only CPS data.
The analysis related to other countries considered various data sources with a structure similar to that of the CPS, containing demographic and socioeconomic data representative of individuals and households, and covering data up to 2013. For Denmark and Norway, administrative registration data were used, with data available from 1980 and 1967, respectively. The data source for Germany is the German Socioeconomic Panel (SOEP), providing a panel since 1984. For the United Kingdom, data from the UK Labour Force Survey (LFS) starting in 1979 were used.
For the main analysis, the sample was restricted to couples where at least one partner was between 26 and 60 years of age. Individuals were separated into four mutually exclusive groups according to their highest level of education: (i) incomplete high school, (ii) complete high school, (iii) incomplete higher education, and (iv) complete higher education.
- Method
The process of matrimonial selection by education is defined as the tendency of people with similar characteristics to marry more than would be expected in a random scenario. This behavior can be identified and quantified by comparing the contingency table for the educational levels of spouses with a contingency table generated by random pairings between husbands and wives. In this way, parameters are calculated that allow for the evaluation of the probability of union between partners of the same educational level, indicating whether the frequency is higher or lower than randomly expected. Positive or negative values of this parameter indicate marriages that are more or less frequent than predicted. The joint distribution of spouses' education is described by these parameters and by the marginal distributions of education. To estimate these parameters, CPS data from 1962 to 2013 were used.
To verify the robustness of the parameters, they were re-estimated considering age adjustments, creating three age range groups, and incorporating unmarried cohabiting couples into the sample, using data from Norway. Additionally, the parameters were also checked considering race, a longer panel from decennial census data, data from other countries, and the field of study of higher education. Finally, the Altham index (1970) was estimated, providing a new measure for assessing marital selection by education based on the odds ratio of matched education groups (for each of the four education levels).
The authors also performed regressions for the wife's years of education (outcome variable) in relation to the husband's years of education (explanatory variable), and also regressions inverting the relationship between these variables. This exercise was performed for each year in the period from 1962 to 2013. Finally, the behavior of the coefficients associated with the explanatory variables was verified for each year of the two regressions (in relation to the wife's and husband's education).
To quantify the contribution to family income inequality of changes in returns to education, educational composition, and marital selection by education, the decomposition method proposed by DiNardo et al. (1996) was adopted. This approach provides income distributions under counterfactual scenarios, in which the distribution of one factor is fixed at a given base year, while the other factors vary over time. For the validation of the results of the decomposition method, different specifications, subsamples, and countries were considered.
- Main results
The results found for the education-based selection parameters in the United States showed the existence of positive selection at all educational levels. However, temporal trends varied according to educational level, indicating a decrease in the parameter among the most educated over the years. In 1962, individuals in the United States with a college education were almost five times more likely to marry someone with the same educational level compared to random selection. This relationship among couples with higher education decreased to approximately three and two times in the 1980s and 2013, respectively. Conversely, the education-based selection parameter increased from 1,6 in 1962 to 7,2 in 2013 when considering those with incomplete secondary education. When considering the average of all parameters across all educational levels during the period between 1962 and 2013, the authors identified a gradual increase in education-based selection.
Regarding the results of the robustness analyses for the education-based marriage selection parameter, all approaches indicated the consistency of the behavior and temporal trends identified for different countries, samples, specifications, and periods. Notably, the analyses showed a more pronounced drop in the parameter among couples with higher education in the United Kingdom and Norway, and also indicated that the process of marriage selection by education is even more pronounced when considering the type of undergraduate degree: law graduates in Norway were 73 times more likely to marry each other in 1967, declining to 26 times in 2013.
The results of the analysis on the determinants of income inequality among families indicated that marital selection based on education had an impact on income inequality when compared to the counterfactual, but with limited relevance. Increases in returns to education emerged as the main source of increased inequality. Furthermore, changes in educational composition offset some of the increase in inequality, especially at the lower end of the household income distribution. Thus, the estimated counterfactual indicated that inequality would have fallen continuously if returns to education at the 1962 level had been maintained over the years. Again, the results were corroborated by robustness analyses.
- Lessons in Public Policy
This study sought to address the relationship between income inequality among families and marital selection based on education. First, the degree of marital selection based on education was examined over time and across different countries, estimating parameters that would explain this relationship. Second, this parameter was used to quantify the explanatory role of various factors in household income inequality through a decomposition method.
The evidence presented in this article demonstrated that over the years, marital selection based on education increased for couples with lower levels of education and decreased for couples with higher levels of education. The results also indicated that educational selection contributed a significant portion to household income inequality in each country. However, the main factor explaining inequality among families was the increased return to education observed during the period. This evidence provides crucial information for the formulation of public policies aimed at reducing income inequality among families.
References
ALTHAM, PME The Measurement of Association of Rows and Columns for an r × s Contingency Table. Journal of the Royal Statistical Society: Series B (Methodological), v. 32, no. 1, p. 63–73, Jan. 1970.
AUTHOR, DH; KATZ, L. F.; KEARNEY, MS Trends in US Wage Inequality: Revising the Revisionists. Review of Economics and Statistics, v. 90, n. 2, p. 300–323, May 2008.
DELAVANDE, A.; ZAFAR, B. University Choice: The Role of Expected Earnings, Nonpecuniary Outcomes, and Financial Constraints. Journal of Political Economy, v. 127, no. 5, p. 2343–2393, Oct. 2019.
DINARDO, J.; FORTIN, N.M.; LEMIEUX, T. Labor Market Institutions and the Distribution of Wages, 1973-1992: A Semiparametric Approach. Econometrics, v. 64, no. 5, p. 1001, Sept. 1996.
EIKA, L.; MOGSTAD, M.; ZAFAR, B. Educational Assortative Mating and Household Income Inequality. Journal of Political Economy, v. 127, n. 6, p. 2795–2835, Dec. 2019.
LIU, H.; LU, J. Measuring the degree of assortative mating. Economics Letters, v. 92, no. 3, p. 317–322, Sept. 2006.
