What is fuelling the income disparities across households in countries around the world today? Factors such as globalisation, urbanisation and technological change are often cited as potential drivers, notes Ana Kujundzic. But what if our more intimate choices – such as the selection of long-term partners – are an important reason behind the levels of income inequality that we observe today?
High levels of income inequality – both within and between countries – have become a pressing concern all around the world. In the 2020 edition of their World Social Report, the UN reports that “the extraordinary economic growth and widespread improvements in well-being observed over the last several decades have failed to close the deep divides within and across countries”. It goes on to note how income inequality has increased in most developed countries and in some middle-income countries, including China and India, since 1990 – and that countries where inequality has grown are home to more than two thirds of the world population.
What is fuelling growing inequalities within countries and what can be done about it? Some of the drivers that are often discussed include globalisation, urbanisation, technological change, migration, and educational meritocracy. While each of these may promote economic growth and reduce poverty, they may also exacerbate income inequalities within countries if the benefits disproportionately favour the wealthy over the poor. Regarding educational meritocracy, for instance, Michael Sandel has argued that a system in which wealthy families can pay for superior education for their children serves to perpetuate existing inequalities by providing unfair advantages to those who already hold greater economic resources.
One could argue that each of these factors have tangible policy solutions if there is the political will. But what if our more intimate, personal choices, such as the selection of long-term partners, are an important factor behind the sizeable income disparities between households observed today? Is there a simple policy solution to address this?
Do people “sort” themselves into couples with similar characteristics?
The old saying that “opposites attract” has long been discredited. For decades, studies have shown that we tend to choose partners similar to ourselves. Perhaps this is why we read about celebrity “power couples” – and why Cosmopolitan last year even compiled a list of “lookalike” celebrity couples.
Physical appearance is just one factor in partner selection, of course, and researchers such as Hans-Peter Blossfeld have suggested that the level of education is another key criterion because it strongly correlates with employment opportunities and earnings potential.
This raises an important question: is choosing partners based on education level – educational homogamy – a key factor in household income differences? It seems logical to expect that higher-educated households would have a higher concentration of economic resources, while less-educated households would have fewer economic resources available to them.
In my research, I look at the empirical evidence for four middle-income countries. I find that educational homogamy decreased significantly between the 1970s and 2010s in Brazil and Mexico, and between the 1990s and 2010s in Indonesia and South Africa. The extent of this decline varies across countries, ranging from 14 per cent in Indonesia and South Africa to 23 per cent in Mexico and 35 per cent in Brazil.
One might expect that this reduction in couples “sorting” themselves by educational background would lead to a notable fall in household income inequality in these four countries. However, my findings reveal only a weak association between the two variables. If educational homogamy had not fallen as it had, inequality would only be 1-3% higher than observed. On the face of it, then, this seems to suggest that trends in educational homogamy are not a very big issue.
But digging deeper into each country’s specific trends – and the different group dynamics that are at play – uncovers a more concerning narrative. Highly educated individuals are sorting less among themselves but at the same time, the least educated are sorting among themselves more than before. In other words, the inequality-reducing impact of fewer highly-educated people partnering up among themselves is almost entirely offset by the inequality-increasing impact of higher levels of sorting among the least educated group.
Case Study: South Africa
This is perhaps best illustrated in the case of South Africa. Figure 1 compares levels of educational homogamy between 1996 and 2011 for (1) couples with less than primary school education, (2) couples with primary school education, (3) couples with secondary school education, and (4) couples with university degrees. The measure of educational homogamy shown is the likelihood ratio, defined as the likelihood of both partners having the same education level, relative to the baseline scenario of individuals sorting randomly with respect to education.
In 1996, for example, individuals with a university degree were about 8 times as likely to have a partner with a university degree, compared to the baseline scenario of random matching. We can see that the level of sorting declined for individuals with a university degree from 1996 to 2011, while it has increased for individuals with less than primary school education. The level of sorting for the other two groups slightly decreased, but not by much.
Figure 1: Trends in homogamy by educational group in South Africa
Although university degree holders are sorting less among themselves and thereby contributing to a decrease in household income inequality, those with less than primary school education are sorting more and thereby contributing to an increase in household income inequality. These group dynamics largely cancel each other out, resulting in a very small overall impact of educational sorting on household income inequality. Interestingly, Eika et al. (2019) reach the same conclusion for the United States and Norway between the 1960s and 2010s.
What should we make of all of this? It is certainly good news that concerns about educational homogamy having a potentially large effect on income disparities between households appear mostly unwarranted. However, the finding that the least educated are increasingly sorting amongst themselves – both in middle- and high-income countries – suggests that marginalisation processes are taking place at low levels of the educational distribution. The least educated among us are being left behind, facing limited labour market opportunities and diminished chances of achieving upward socioeconomic mobility by marrying more educated partners.
All articles posted on this blog give the views of the author(s). They do not represent the position of LSE Inequalities, nor of the London School of Economics and Political Science.
Image credits: Everett Collection via Shutterstock.
Interesting indeed, Ana. There’s a vivid discussion in the Netherlands going on as well, among others by Tim ‘S Jongers. He recently wrote a book “Poverty explained to people with money” (in Dutch only 🤷♀️). Thank you for sharing.
Ana, your research is really important.
I have two comments. The first relates to the interpretation of the nexus between assortative mating and inequality. My second comment concerns your empirical findings.
1. Comment on the interpretation:
While there seems to be a consensus in a certain strand of the literature on the primary role of assortative mating as a driver of income inequality, another strand of the literature offers a different perspective. According to this alternative view, a higher inclination of people in a given generation to “marry their like” compared to earlier generations’ inclination to do so is a sign of increasing overall inequality from one generation to the next (see Katrnak et al 2012). By overall inequality, it is meant all those dimensions of inequality that matter at couple formation (health, wealth, income, ability to generate income, opportunity, etc.).
This second strand of literature is interesting to those scholars who:
(i) are ready to move away from the old cliché of blaming assortative mating for the high levels of inequality in the generation of “birds that will hatch in the future from the eggs of the birds mating assortatively today”;
(ii) who find it limitedly exciting intellectually to explore whether a higher degree of sorting leads to higher income inequality among dual-headed households since they know that $D^2(w_h+w_w)= D^2(w_h)+ D^2 (w_w)+ D(w_h)D(w_w)r(w_h, w_w)$ is increasing in $r(w_h, w_w)$, the correlation between husbands’ and wives’ income (where $D(w_h)$ and $D(w_w)$ are the standard deviations of husbands’ and wives’ income, respectively), and who also know that income tends to increase in education;
(iii) and who are convinced that some policy-controlled phenomena, such as “globalization, urbanization, technological change, migration, and educational meritocracy,” etc., are primarily responsible for inequality.
2. Comment on your empirical findings:
You “find that educational homogamy decreased significantly between the 1970s and 2010s in Brazil and Mexico, and between the 1990s and 2010s in Indonesia and South Africa.
The extent of this decline varies across countries, ranging from 14 per cent in Indonesia and South Africa to 23 per cent in Mexico and 35 per cent in Brazil.”
Actually, the International Demographic Inequality Lab (https://idil.li/) measures the same changes differently by using the NM-method (see: https://en.wikipedia.org/wiki/NM-method; Naszodi and Mendonca, 2019,2021,2022,2023). For Brazil, they quantify a decline in educational assortative mating between 1970 and 2000 that was followed by an increase between 2000 and 2010.
Similar to Brazilians’ inclination for marrying someone from the same educational group, educational homophily in Mexico is also found to be non-monotonous: the overall change in the degree of sorting is quantified by IDIL.LI to have declined by 4 percentage points between 1970 and 2010, with a relatively sharp decline between 1990 and 2000 amounting to 3 percentage points. This 3 percentage points change can be roughly interpreted as follows: if the early GenXer young Mexican adults (whose marital/cohabitation status and education level were observed in 2000) had the same education levels as the late Boomer young Mexican adults (whose marital/cohabitation status and education level were observed in 1990), then the proportion of educationally homogamous couples would have been 3 percentage points lower among the early GenXers (relative to the late Boomers).
As for South Africa, similar to your result, IDIL.LI also finds a decrease in educational assortative mating (although the magnitude is different due to differences in the methods applied). The negative trend in assortative mating in South Africa is especially interesting in light of the fact that WID.WORLD finds the top 10 income share to have steadily increased in South Africa over the same decades by using macro statistics rather than micro data (as you do and IDIL.LI does).
For trends in Indonesia and an additional 77 countries, please see https://idil.li/assortative-mating/.