Mobility: Money

[ Part of a sequence of posts on income mobility ]

The most "obvious" vector through which shared environment affects income is via money - if your parents have money, they can spend that to give you more opportunities. The only problem with this theory is the evidence, which is mixed at best in the developed world

Evidence Against:

• The Georgia land lottery winners won the equivalent of 5 years of unskilled labor (think ~$100k today). Researchers tracked 1758 of them and found 4 of the 36 children outcomes differed at the p=0.1 level - i.e. exactly what you would expect to happen by change Shocking behavior: Random wealth in antebellum Georgia and human capital across generations.
• The children of Swedish lottery winners had no improvements in drug consumption, scholastic performance, or skills Lindqvist. The study had 564 people who won at least $100k in prize money and 114 who won at least $1M.
• An adoption study found no association between parental income and eventual child income How large are the effects from changes in family environment? A study of Korean American adoptees.
• A study of millions of people's federal tax and welfare data matched fathers to have similar demographics and the same employers, but where one set was laid off and one wasn't in a given year. The authors employ a difference-in-difference model and find no "effect" on the children's income at age 25 Hilger. This is the sole study in these two lists that doesn't use an experimental or quasi-experimental design, but I felt compelled to include it since it does make use of an enormous sample.

Also, kind of tangential, but you have to wonder how important income inequality is at perpetuating itself given that pre-tax income inequality is just as bad in Western Europe as in the US Gini in the bottle despite a huge difference in welfare spending.

Evidence in Favor:

• Mothers' Pension recipients in the early 1900s got cash transfer that "generally represented 12-25% of family income, and typically lasted for three years." The children of these recipients earned ~14% more than the children of rejected applicants in early adulthood and saw other educational and health benefits Aizer.
• A quasi-experiment using Medicaid eligibility as a treatment finds no effect of it on adulthood income for males. For females, a year of additional Medicaid eligibility increased cumulative earnings by 1.3% by the age of 28 The Long-Term Impact of Medicaid Expansions on Eligible Children.
• Children raised in families that receive housing vouchers or public housing earned several percent more than other children depending on program/sex Moving to Opportunity. In Wikipedia Andersson.
• Using EITC variation as a a treatment, a study found that an additional $1000 per year from age 13 to 17 caused an increase in annual earnings by ~2.2% Bastian. Similar results were later found for younger ages Barr.
• An adoption study found r~0.1 association between father income and eventual child income Björklund.
• One study used a regression discontinuity design and found that removing a child from the Supplemental Security Income program had a negative effect on the child (insignificant) and their younger sibling (significant) Deshpande.
• A study followed Native Americans who started receiving a stipend from a Casino. They found a positive impact on whether the children were employed at age 25, though they didn't measure actual earnings Akee.

As you can see, the evidence is mixed, but, if we squint a little, we can see a trend: ignoring the two contradictory adoption studies, both the studies against involve a one-time wealth transfer, while all the studies in favor involve multiple (smaller) wealth transfers over several years.

Naively, there should be an elasticity between transfers in childhood and adulthood income - that is, if I give parents 1% more income over 18 years, their children will presumably make x% more per year.

I made a spreadsheet where I perform a fixed-effects meta estimate for this elasticity based on the four multi-year-transfer studies. The result? An estimated elasticity of 4.0 with SE=0.5.

This brings us to a weird issue: an elasticity of between 3.0 and 5.0 is several times higher than the theoretical maximum (1.0, where everyone's income is purely determined by their parents' incomes) and an order of magnitude higher than any reasonable estimate consistent with the twin studies discussed previously. This makes me very suspicious of these studies' results.

There is, however, a way to reconcile these contradictory pieces of evidence: one-time cash transfers are useless; income transfers are useless to most families; income transfers are very important to very poor families - to the extent that a 1% boost now will boost child incomes by more than 1% in adulthood.

I think this hypothesis and the null hypothesis (that all effects are ~0) are the only two reasonable ones to believe given the evidence.

Finally, I should mention that there exist studies that find cash transfers boost students' educational achievements. It's unclear, however, how much these boosts lead to higher earnings later in life. Moreover, there is a significant replication crisis in the nearby study of the impact of poverty on cognitive ability O’Donnell, which makes me weary of this entire field of research.