Yesterday, Austin Contrarian fielded three good criticisms of my scatterplot (reprinted above) comparing advanced degrees vs. median income. From the wonky blogfather himself:
My criticisms: (1) he should run his regression on metropolitan areas rather than municipalities because high-wage earners tend to cluster together in exclusive suburbs (this would require using a national panel since there aren’t enough Texas metropolitan areas); (2) he should use gross metropolitan product rather than household income; and (3) we should keep in mind that Austin has an unusually high proportion of masters and Ph.D. because of UT; it is an outlier.
The first point makes sense. We want the proverbial apples-to-apples comparison. Creating a certain type of housing mix is indeed a local policy choice that ensures high median household incomes. Given that the r-squared on the plot was only 0.48, the housing mix and zoning of the local community might be a strong indicator of median household income. In the future, I’ll make new plots that segment the underlying type of community. We better want to understand all of the major drivers. That said, advanced degrees are a major driver regardless of the underlying community, so I still advance that the plot allows us to challenge the idea that Austin policy is presently somehow exceptional at leveraging its human capital. Maybe we are good at attracting and retaining it, but we appear to be average at using it to create median household income. This does not indicate that Austin is “bad” at creating shared economic gains off its base of advanced degrees. It just isn’t a star.
The second point on gross metropolitan product versus median household income is also good, but Austin Contrarian and I might be interested in measuring different things. Looking at the gross product gives us a sense of the ability of the local economy to transform advanced degrees into economic growth. It could be highly unequal growth with the gains being enjoyed by a tiny few. As a result a look at total gross product or a per capita share can mask inequality. Since my focus was more on looking at shared gains, that’s why I selected median household income. Aaron M. Renn makes this point in a slightly different way in Austin Contrarian’s comments. Again, in the future set of plots, I’ll also look at gross product and per capita, but personally, I am most interested in shared prosperity as an evaluation of public value created by local policymakers.
The third point, however, I am a bit confused about. Granted, we have a high percentage of PhDs, but as the plot indicates, there are many communities with similar and higher shares of advanced degrees. Maybe the issue is that the regression should not be linear. Or perhaps, as M1EK explained in the comments, Austin has an unusual concentration of advanced liberal arts degrees (i.e. literature as opposed to materials engineering) that are not optimal for creating economic gains. Again, these are good points and I will try to engage them in future plots.