Some new research by Raj Chetty, Emmanuel Saez, Nathaniel Hendren, and Patrick Kline finds that the likelihood of poor children moving up the income ladder in early adulthood varies dramatically by metro area in the United States. In places like Salt Lake City, Utah or Bakersfield, California, a child born in the bottom quintile has a roughly 12 percent chance of moving up to the top quintile by adulthood. In places like Memphis, Tennessee or Atlanta, Georgia, the chance is 4 percent or less.
By now, you may have seen some of the extensive media coverage of this research – which coincided with a major recent speech by President Obama on economic inequality. What you may not realize is how unique the data used for this study are. They primarily come from federal tax return data. The children in the study were born in 1980 and 1981. They are identified as dependents on their parents’ tax returns from the late 1990s. The authors then track down the children’s own tax returns from 2010 and 2011 (when they are about 30 years old).
In a March working paper, the authors explored the relationship between state and local income tax deductions and intergenerational income mobility. These deductions include a variety of programs including state income taxes, non-federal earned income tax credits (EITC), and homeowner mortgage interest deductions. They create an aggregate, metropolitan-level measure of the average tax deductions for households at different income levels. The figure below represents the progressivity of local tax codes by metropolitan areas.
Parents are ranked based on their household income and wages in 1996 and children are ranked based on the same metrics in 2010/2011. The association between these two measures provides the basis for calculating an intergenerational income elasticity – the greater the association in an area, the lower the intergenerational income mobility. (Read the paper to get a sense for the full range of interesting sub-analyses that the authors conducted). The darker areas in the map below illustrate the wide variation in intergenerational mobility (the darker the area, the lower the mobility).
As the graphics would suggest, there is a strong positive correlation between higher income mobility and greater tax deductions, as well as between higher income mobility and more progressive tax deductions. Which tax policies drive this relationship? They examine three tax policies and find that they are all associated with higher mobility: mortgage deduction rates, state income taxes (both the levels and the progressivity of the tax code), and state EITC rates.
Whether these policies are causing the mobility is a bit harder to determine, since they are correlated with other features of states that probably also promote redistribution and support for lower-income households. They find that high mobility areas also have other features – like more spending on schools, higher income equality, and more social capital – so these features could also independently drive some of the association. However, as a sensitivity analysis they find that the relationship for EITC and mobility holds, even after these other features are controlled for in a regression model.
The main policy message is that we cut back progressive, redistributive state programs at our own peril. The short-term budgetary benefits of these cuts may be outweighed by lower income accumulation and upward mobility for the most disadvantaged households.
What excites me the most about this work is that it ushers in a bold new research enterprise on the local area determinants of income mobility. In the coming years, I believe that research with multiple sources of administrative data, permitting geographic linkages across places and years, will be critical to moving forward inequality research. There are many sources of data that will be required – tax data, mortality records, educational data, corrections data, health care records, and welfare records (to name a few).
The barriers to this kind of research are clear – these data are highly sensitive, (often) legally protected, and difficult to work with. The investment in building databases with these resources is massive, and probably out of reach for many researchers and their institutions right now. To move this research forward, the research community will need to accelerate the data dissemination process – re-packaging proprietary data for a broader set of consumers. Some of this is happening already. The authors have posted their study data (aggregated) on their web page, which is an important resource for other researchers.