The following comes from David Card (1995) using NLSY for men. We consider the relationship of wages (log wages) to education and other controls. Of course, there is the confounding omitted variable “ability” of which little is known that’ll make our results biased. This is such an important empirical question it is part of the active research agenda till today.
Card makes the following argument, we can assume that some of the exogenous variation that education has on earnings can come from the difference in groups which leave near 4-year institutions and those who live far away from them. I have recreated the following from Card’s paper to make the case for the 4-year college distance treatment. Particularly due to it’s effect on poor and low educated households.
This is what what I want to note ever so slightly. Where colleges exist in cities it makes a difference in the lower quantiles of education fulfilment. Intuitively this makes a lot of sense. Costs to low income households are down if their children attend local universities. Running a regression with respect to the indicator variable for nearby 4-year colleges also shows a positive relation. Thus, we have a (positive) causal relationship between years of education and nearby colleges. This has one important implication in my mind, recall from my past post I mentioned the same technology hubs over and over. They came up in Enrico Moretti’s work. These are Austin, New York, Seattle, Bay Area, etc. All of them have large and highly reputable colleges. More than that they have several large and highly reputable colleges.
Thus, when we discuss the effects of human capital externalities we can probably see from the graph above a positive generational effect. If in San Francisco day labourers are attracted to the hustle and bustle of a burgeoning city their kids will have an easier time affording San Francisco University. This might be the machinery of innovative cities so many urban planners and developmental economists discuss. Don’t take my statement as definitive proof, just food for thought.
I have a STATA do. file here and Card’s data here (as a STATA .dta) for anyone interested in the IV regressions since I didn’t reproduce them in this post.
PS I remembered this paper only because I had a chance to read it as part of my advanced econometrics class. Just goes to show you it pays to read the suggested readings no matter how overwhelming the number is. Frankly, I sometimes felt like I spent more time reading them and not enough drilling some basic concepts into my head.
Enrico Moretti and other economists discussing spatial and urban economics have documented in the United States a divergence in earnings based on location. Hubs where there exists a great concentration of high skilled workers tend to be areas where all residents have high earnings relative to other regions.
This comes from the following two papers by Moretti: 2013 (with Thulin) and 2003. With this we know that high skilled labour in small cities transfers to these hubs. And married couples with high investments in human capital match up with each other and with high education hubs (Silicon Valley, Seattle, Los Angeles, Boston, etc).
It then seems important in the United States if this self-sorting has any consequences on the political objectives of cities and their citizens. The recommendations Moretti makes is increased investment in education. Competition for headquarters of a large internet firm or medical research company will result in outcomes that don’t benefit the residents.
Taking this one more step (maybe a step too far) we may find the voting results of residents different as the self-sorting takes place. If the country as a whole needs greater investment in education in “knowledge” sector the voting system may limit this. The electoral college will give the rural and low density populations a disproportionate voice. As these actors attempt to protect their fortunes, they’ll election politicians who’ll save their local industries. They’ll do so with protectionist policies, subsidies, tax breaks and so on. The point is the government’s budget is constrained, it can not give rural America tax breaks for their manufacturing jobs and institute a dramatic investment in high level education. But the voting system favors the local population in low density areas, so they’ll keep voting for immediate solutions that don’t harm their welfare in the near term.
That’s not to say that proportional representation is correct form of governance but rather that the prior uncertainty (belief) in a human capital investment by rural residents is very high. This is due to the natural uncertainty anyone can have in an investment and lack of smoothing by agents (contrary to theory).
I have discussed at length on this blog last year the ways in which students don’t anticipate debt burden in a manner similar to the consumption cycle in labour economics. I think an assumption can be made for the rural voters (or voters in general) when they vote in investments in education. However, we further assume that this isn’t shown by compensating for the investment with higher taxes or certain reforms but rather compensated by lack of belief in the investment itself.
How can we model such actors? We can give them imperfect memory or a bias to their immediate welfare changes. So, a off-shoring jobs to a third world country pains them much more than the change 10 years from said off-shore in earnings after re-training. Their posterior beliefs will change but that’ll take time and experience. Again, expectations in returns to education I have discussed previously are important research interest of mine because it is surprisingly important and could be informative to many issues.