Raj Chetty's search for economic opportunity

For anyone not aware, Raj Chetty is a professor of public economics at Harvard, and has been described, by Tyler Cowen, as "the single most influential economist in the world today". Whenever he publishes a new paper, the world listens. So also this time, when a two-paper series by Chetty and co-authors was published in Nature earlier this month (Paper I, Paper II). The papers revolve around economic inequality- and mobility and claim to have found that a major determinant of upwards economic mobility is the number of friends with higher socioeconomic status (SES) you are surrounded by during childhood. Economic connectedness (EC), as it is termed in the papers, is quantified based on Facebook friend-data, and is defined as high if people of low- and high SES are frequently friends on Facebook. EC for some regions/neighborhoods are higher than for others, and this coincides with those regions having higher upwards economic mobility. The authors go on to make the following statement:

If children with low-SES parents were to grow up in counties with economic connectedness comparable to that of the average child with high-SES parents, their incomes in adulthood would increase by 20% on average.

This phrasing sounds as if they are making a stronger claim, meaning they believe that high-EC areas cause children’s future income to rise. In sociological research, causality is not easily found. For example, the explanation for the correlation between EC and income mobility may be that people naturally sort into neghborhoods—making the mix of people inherently different in different areas. The authors explain further why they still reach their conclusion:

To test for sorting on such dimensions, many of which are unobservable, one would ideally randomly assign families to low-EC and high-EC areas—thereby ensuring that families in high-EC and low-EC areas are comparable—and examine whether their children’s outcomes differ in adulthood. We approximate this experiment using quasi-experimental estimates of the causal effect of growing up for an additional year in each county in the United States on household incomes in adulthood from Chetty and Hendren [link to study]

The suggestion of random assignment into neighborhoods would be an excellent, although hard-to-perform, experiment to examine if the correlation is in fact a result of a causal relationship. But they refer to the study by Chetty and Hendren from 2018, stating that this study shows that the relationship is causal. In this paper, they tracked families who moved from “worse” to “better” neighborhoods and what effect on the children’s future family income this had. Moving to a better neighborhood was associated with a higher future income, and importantly, this was still true using a between-sibling/within-family type of design. So, if one sibling had moved at e.g. age 15 and the younger sibling at age 10, the younger sibling (on average) earned more money as an adult. Chetty and Hendren also used a couple of more, so-called, quasi-experimental designs and reached the same results regardless of method used. Because of this previous research the authors concluded the following:

We conclude that the correlation between EC and mobility is not driven simply by differences in the types of families who live in high EC areas. Instead, growing up in an area with higher EC causes significantly higher rates of upward mobility.

What is the effect size?

There may still be alternative explanations as to why this relationship exists. But let’s say there is (1) a causal neighborhood effect, and (2) that EC is part of that causal effect. Then how big are the effects? In the quote at the top of the text, we learned that children would make 20% more in adulthood if they grew up in a high-EC area rather than a low-EC area. This certainly sounds like a large effect size but let’s standardise the measurements and make some comparisons. From the 2018 Chetty and Hendren paper they present the following figure:

As can be seen, children of parents at the 10th percentile are expected to earn an income at around the 35th percentile national income distribution. Compare this to children of parents at the 90th percentile - which are expected to earn an income around the 65th percentile. This 10th-to-90th percentile of parental income difference makes for a difference of 30 percentiles (65 minus 35) in expected future earnings. As a comparison, Chetty and Hendren provide an estimate that growing up in a 1 SD better neighborhood equals having 1/3 of a SD richer parents. So, growing up in the 90th percentile best neighborhood you are expected to gain around 10 percentiles on the income distribution. Lastly, growing up in the 90th percentile highest EC area you are expected, according to the authors, to climb just over 5 percentiles on the income ladder.

Now we can make our apples-to-apples comparison. The whole neighborhood effect is about 1/3 as big as having richer parents and the EC effect is about 1/2 of the neighborhood effect.

Do these results fit with the twin research?

As discussed previously (here and here) twin- and adoption research show that heritable factors (having richer parents) explain around 40-45% of the variation in income between individuals. Shared environmental effects, of which neighborhoods are a part of, explain around 5-15 % of the variation. In other words, these two separate lines of research do point in the same direction - and that direction would be that neighborhoods explain a small part of the variation in income.

Conclusions

After spending a fair bit of reading and thinking on the topic, my current conclusion is that Chetty, in his quest of explaining differences in economic opportunity may have shed a lot of light on the smallest piece of the puzzle. To be more precise, if shared environment explains 10 % of the variation in people’s income, neighborhoods may very well constitute the bulk of this shared environmental factor. Further on, economic connectedness may in turn constitute around half of this neighborhood effect.