Income inequality and social origins Updated

The relationship between income inequality and intergenerational mobility: What is known?

From a strictly statistical point of view, the concepts of income inequality and intergenerational mobility are independent. However, mounting evidence points toward a negative association between them. A graph that has become famous as the “Great Gatsby curve” lines up the intergenerational elasticity of income and the Gini coefficient of income inequality across countries [3]. The intergenerational elasticity of income, an inverse measure of intergenerational mobility, is a widely used measure of the association between parents’ outcomes and their offspring's outcomes. A value of 0.5, for example, means that a 10% change in parents’ income is associated with a 5% change (half of the percentage change in parents’ income) in their offspring's income in the same direction. A status quo society, in which income differences are fully transmitted from one generation to the next, would have an intergenerational elasticity value of 1. A perfectly mobile society in which family origins play no role in shaping one's fortunes would have a value of 0. The Great Gatsby curve shows that, across countries, there is a positive relationship between the level of income inequality within the parents’ generation (as measured by the Gini coefficient) and the degree of intergenerational transmission.

This relationship has received considerable political attention because it says that inequality across generations, which undermines equality of opportunity, is most persistent in countries with high levels of inequality. Cross-country comparisons could be biased, however, by country differences in factors affecting both inequality and intergenerational mobility, such as differences in institutional settings or cultural values, which would muddy interpretation of the inequality–mobility relationship. But recent evidence shows that cross-country variation is not the driver of the negative relationship between inequality and mobility. A study for the US finds that the probability that children whose fathers were in the poorest fourth of the income distribution will climb the income ladder (upward mobility) is inversely related to the Gini coefficient of inequality of parents’ income (Illustration) [1].

Figure 1 plots the Great Gatsby curve across US states. Data on intergenerational income mobility were computed by the Equality of Opportunity Project from millions of US tax records for 1996–2012 for sons born between the early 1980s and early 1990s [1]. The data, computed for commuter zones (geographic units of analysis that more closely reflect the local economy where people live and work than political boundaries do), were aggregated to the state level. Data on the Gini coefficient of household incomes at the state level are from the US Census Bureau.

The left panel of Figure 1, which plots the father–son intergenerational rank correlation of incomes (a measure of intergenerational immobility similar to intergenerational elasticity) against state income inequality measured in 1979 shows a positive relationship between income inequality and intergenerational immobility, in the spirit of the Great Gatsby curve. In 1979, the sons that would constitute the “destination generation” of the intergenerational transition measured by the Equality of Opportunity Project were not yet born, so the income inequality shown in the left panel refers to the generation of fathers or earlier. The right panel of Figure 1 relates intergenerational elasticity to the 2013 level of income inequality, offering evidence that states that had a high level of inequality in the past and that experienced a low level of intergenerational mobility also have high levels of inequality today. In this sense, the lack of intergenerational mobility acts as a conduit for income inequality from the past to the present.

Note that, in principle, mobility in any period may depend on past mobility: for example, because of high past mobility in one period, families or individuals may attain their “proper” position in society, resulting in less mobility in future periods, implying that looking in the longer term across multiple generations one should see more mobility than actually seen when comparing parents and children.

However, multi-generational mobility studies find the opposite, that is, there is less mobility over three or more generations than what would be implied by extrapolating from parent–child transitions [4]. One reason for this longer-term persistence is the existence of a dynasty-specific latent factor (e.g. unobserved ability) that is highly correlated intergenerationally, which in turn determines outcomes such as income or education that are typically considered by analysts [5].

From the evidence in the Great Gatsby curve in Figure 1, it seems fair to conclude that there is an empirical regularity associating income inequality and lack of mobility, with lack of mobility appearing to determine the persistence of inequality over time. While this evidence shows that high income inequality is a concern because it is matched by low intergenerational mobility, the analysis reveals nothing about any underlying causal mechanisms.

Arguments based on a belief that society should provide equality of opportunity for its members justify interventions aimed at “leveling the playing field” in high-inequality countries, such as education investments in children from poor families. But these policies can be expensive, and if they result in tax increases, especially for high-income families, they could, in theory, discourage their economic activities and compromise economic growth.

It is difficult to say, however, whether policies promoting equal opportunity would actually result in diminished economic efficiency. The answer would depend on whether a country is already using all of its resources in an efficient way. Where that is not the case, governments that promote intergenerational mobility might be able to kill two birds with one stone. This would be the case, for example, if intergenerational transmission reflects talent misallocation because the children of well-off families are hired into good jobs that they would not have obtained except for their family connections. There is evidence for Canada, Denmark, and Sweden, for example, that children tend to be hired by the same firm for which their parents work. Part of this intergenerational transmission of employment may be efficient (because parents know the productivity of their children and the needs of their employer and can favor good matches). But the intergenerational transmission may also reflect forms of nepotism if it results in people who are not the best candidate getting the job solely because of their connections, harming firm productivity and efficiency. Another argument against the existence of trade-offs between leveling the playing field and reducing economic activity relates to the children of poor families, for whom the knowledge that inequality is associated with lack of mobility may be a strong disincentive to exerting any effort in life. Recent evidence for Italy shows that indeed mobility is lower in areas that are performing poorly from the viewpoint of conventional economic indicators such as GDP growth, which suggests the equity-efficiency trade-off is not binding in that case [6].

Is it all in the family?

Intergenerational associations provide an important but still partial perspective on the dependence of individual success on social origins. In principle, there might be other factors operating outside the family and independently of it that influence an individual's ability to generate income. One important example is school quality, which influences human capital formation. To some extent, the quality of the school a child attends will reflect the parents’ income and choices. But there may also be dimensions of school quality that are independent of parental choice, such as teachers’ effort or disruptive classmates. These may affect income capacity in the long term independently of family background. Similarly, the quality of a family's neighbors can influence young individuals beyond the effects of family. These “youth community effects” may influence individual behavior and human capital acquisition early in the life cycle and exert long-term effects on incomes, thus contributing to income inequality. Similar to the family, youth communities to some extent represent circumstances that an individual does not choose but that contribute to the generation of inequalities that block the emergence of equal opportunity. Because these factors are not controlled by the individual, there are good arguments for government interventions to alter them.

Economists and social scientists summarize the joint effect of family and community on income inequality through a measure known as the “sibling correlation of incomes.” The sibling correlation is the share of income inequality that is due to all factors that siblings share, which include both the family and youth communities. A sibling correlation value of 0.5, for example, means that half the observed income inequality is associated with factors that are shared within families and youth communities.

Figure 2 provides some evidence of patterns in the sibling correlation for brothers in incomes across countries. The left panel associates the sibling correlation with the Gini coefficient and traces what is in effect a Great Gatsby curve of the sibling correlation. Countries cluster into essentially two groups. One group consists of Scandinavian countries, with a low degree of income association between brothers, and Canada and France, with a moderate degree of sibling correlation (around 0.3). All these countries also have moderate levels of income inequality, with a Gini coefficient below 0.35. Countries in the second group (Germany, Spain, Italy, China, and the US) have a high share of total inequality that is accounted for by factors shared by brothers (more than 0.4), along with moderate to high levels of total inequality (Gini coefficient). This panel shows that the association between total inequality and the sibling correlation is driven mostly by differences in (high and low) sibling correlation, suggesting that the evidence of a Great Gatsby curve is less clear-cut for the sibling correlation than for intergenerational elasticity.

The right panel of Figure 2 shows the association between two measures of social origins: intergenerational elasticity and sibling correlation. Any major discrepancy between the two variables would suggest that youth communities play an extra role in shaping individual incomes beyond family effects. Not surprisingly, the panel shows that countries in which intergenerational transmission is strongest are also countries in which the share of inequality that can be ascribed to sibling similarities is the largest. The pattern in the graph suggests that the factors that operate outside the family move in the same direction as the factors that make sons similar to their fathers. There are significant cross-country differences in the extent to which income inequality is due to social origins, and the graph suggests that community effects add little to what is inherited from the family.

One reason behind the sizeable sibling correlations found in the literature may be that siblings tend to influence each other through what economists call peer effects, that is, the effect of one's behavior on the behavior of others belonging to a given social network. There is extensive evidence of peer effects among schoolmates, for example. Researchers have considered the possibility that these effects carry through to the family social network, although of course contextual influences coming from family are more pervasive than those that come from school, complicating the analysis. Indeed, results from these studies suggest that sibling spillovers can be detected on various outcomes. This is relevant because it highlights the possibility that policy effects are amplified within the family as they may be transmitted from treated family members to non-treated ones [7].

There is an ongoing debate on whether the effects of youth communities are irrelevant to individual incomes in the long term. Some studies have compared the sibling correlation of incomes with an analogous correlation estimated across youth neighbors [8]. These studies find that the correlation between neighbors is smaller than the correlation between brothers but that it is still substantial, accounting for between one-half and one-third of the sibling correlation. One issue with these findings is that neighbors might be similar to one another not because of the effects of youth communities but because they come from similar families. In other words, the neighbor correlation might simply reflect the family effect.

An analysis of Danish siblings and youth peers shows that indeed the sorting of families into communities induces an overstatement of community effects, whose actual weight within the sibling correlation is less than one-tenth. Also, community effects on income inequality tend to be stronger among young individuals, typically at the time of labor market entry, but then fade away relatively quickly. Most of the sibling similarity in long-term incomes seems to come from the family environment rather than from the community that surrounds the family [9]. Another study exploits data from the Toronto Public Housing Program, which randomly allocated families to communities, thereby eliminating any effects that might stem from selection into neighborhoods [10]. The analysis shows that while neighbor correlations across all of Toronto account for one-fifth of the sibling correlation, the income correlation for randomized neighbors is zero, supporting the idea that community effects are not additional to family effects in shaping income inequality.

Other studies have looked at the effects of neighbors on income levels rather than on inequality. Evidence reported by the Equality of Opportunity Project points to the role of early exposure as the key mediating factor of the neighbor effects [11]. The study finds that more than half of the intergenerational elasticity of incomes is indeed accounted for by neighbor effects. Application of the same research design to Australian data finds similar results [12]. Other research looks at educational quality effects and controls for selection into classes of varying quality within a school. These studies find that the quality of education has effects on earnings [13].

Is it nature or nurture?

Besides the distinction between the effects of family and community, studies looking at the distinction between nature and nurture as the mechanisms of intergenerational transmission offer another perspective on the relationship between income inequality and social origins. Is the inheritance of income potential from parents determined before birth because of the transmission of genetic endowments (for example, of IQ), or is it the outcome of exposure to environmental influences occurring after birth—or both? The answer to this question is relevant because it could identify one of the channels of intergenerational mobility. Caution is required in drawing implications about the effectiveness or appropriateness of anti-poverty policies from studies of nature versus nurture. Even if poverty is “natural,” in that it is “genetically” inherited, that does not mean that income support programs would be ineffective as poverty alleviation tools.

In the absence of direct information on genes, answering this question is difficult. Researchers have contrasted income correlations across groups of people whose degree of genetic similarity is known (on average), such as identical twins, who share the same genetic make-up, and fraternal twins, who share only half of their genes. One challenge in this type of comparison is that differences in correlations for the two groups may reflect differences in the environment between identical and fraternal twins, not just genetic differences. Using information on whether siblings were reared together or apart can reduce the problem. A study that compares income correlations across Swedish siblings and twins exploiting all these sources of variation finds that the lower bound estimate for the share of income inequality that can be ascribed to genetic similarities is 20%. Contrasted with an overall sibling correlation for twins of about 33%, this finding points toward a substantial role for nature rather than nurture in shaping income inequality [14].

Another strand of twins studies has explored the causal process of intergenerational transmission. The main idea behind these studies is that intergenerational persistence due to genetic factors is not causal. Thus, if all persistence is genetic, then randomly increasing the income of parents will not increase the income of the next generation because income increases do not affect parental genes. Intergenerational persistence due to parents investing in their children's education is, however, an example of a causal mechanism. A random increase in parental income could increase the resources devoted to investments in children's education, creating greater income potential for the children. The research strategy of these twins studies is to compare intergenerational transmission among parents who are identical twins because any differences that are found cannot be due to differences in genes and therefore would support a causal interpretation of intergenerational transmission.

One limitation of this line of research is that it typically focuses on educational attainment rather than income as the outcome that is passed across generations. Findings from this literature, especially from studies using register data that refer to the entire population of twins, as is typical in Scandinavian countries, point toward a non-negligible causal role of parental education on children's education, accounting for between one-quarter and one-half of overall intergenerational transmission.

Another way that researchers have addressed the nature versus nurture distinction has been to examine outcomes for adoptees. Because adopted children do not share any genes with their adoptive parents, any parent–child similarity in outcomes has to be attributed to nurture. An important caveat in this strand of research is that it is based on the assumption that adopting parents are not different in ways that make them more likely than average parents in the population to transmit their income or education. Findings point to a large and significant association between parents’ education and children's education for adopted children, which is suggestive of causal effects [15].