DACA and Human Capital Investment Decisions

Economists love to repeat (over and over) that incentives matter. This leads to one way of examining differences in educational achievement by investigating differences in the return to education among different people. The logic goes as follows: If people like me don’t tend to land high-paying or satisfying jobs after completing some level of education, then why would I expend the resources to attain that education level? In short, for some people, the costs of attaining higher levels of education may exceed the benefits. This is a really simple model of decision making, so it is worth asking: Is this really how people make decisions in the real world?

In a new NBER Working Paper, Elira Kuka, Na’ama Shenhav, and Kevin Shih, examine this question by making use of the Deferred Action for Childhood Arrivals (DACA) program. You may already know about DACA (because it is in the news quite a bit lately), but it basically provided temporary work authorization and diminished the risk of deportation for undocumented youth living in the United States. Therefore, a couple things could be occurring that increase the return to education for DACA-eligible kids. First, by authorizing work DACA could have led to an increase in expected wages. Second, by reducing the risk of deportation DACA could increase expected lifetime earnings.

So, what happened? The authors estimate the following impacts of DACA: (i) High school graduation rates increased by 15%, (ii) teenage births declined by 45%, (iii) college attendance increased 25% among women, and finally (iv) the same individuals who acquire more schooling also work more which indicates an increase in overall economic productivity.

Here is a figure from the paper, but if this interests you, check out the entire paper:

A couple my own comments about this paper:

First, and perhaps most obviously, this paper has very pertinent insights for a deeply contested issue in United States politics at the moment. This paper suggests that the DACA program has positive benefits in promoting higher acquisition of education – in the form of increased high school graduation rates and college attendance. Additionally, DACA also seemed to reduce risky behavior among undocumented youth – in the form of decreased teenage pregnancy rates. Finally, all of these benefits side step potential costs since DACA-eligible kids who go to school more are also working more. In the face of frustrating discussions around politics in the United States right now, this is a really pertinent study.

Second, the authors highlight innovative methodology when using a difference-in-difference empirical strategy in the face of differential pre-treatment trends in outcomes. The authors show that the pre-treatment trend in their outcomes, differentiated by DACA-eligibility status, are not strictly speaking parallel. Therefore, before estimating their regressions, the authors perform a method of de-trending the trends for DACA-eligible and non-eligible individuals. (See more information about this here.) This is relatively novel methodology that, it seems to me, others (myself included) should apply to their own work.

Third, the authors use data from the American Community Survey (ACS) from IPUMS USA and showcase the possibility of performing valuable and interesting research using administrative data. (Full disclaimer, I currently work for the IPUMS Team.) Indeed this is a reminder to myself, as I work with IPUMS data almost every day and should think a bit harder about useful research topics using IPUMS data.

Finally, this paper is still a working paper, so the peer-review process may refine some of these core findings. For example, it could be that the process the authors use to identify who is DACA-eligible will be called into question. In particular, the authors could try to improve their DACA-eligibility variable by excluding individuals who are clearly not DACA eligible but may look like it in the ACS data (e.g. refugees, for instance). Additionally, the statistical methodology will likely be interrogated. In particular, is clustering standard errors at the state level valid, when this could be done at the PUMA level with IPUMS USA data?

Nevertheless, this is a valuable study and I hope more research in this topic is completed soon.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s