In many educational systems (Spain, United States, United Kingdom), students in private schools get best results on average (standardized test scores, college access, adult income) than students in public schools. This difference continues to exist even when we correct for the financial resources of both types of schools. All budgetary conditions being equal, private schools graduate students with better average results.
This observation has (at least) two possible very conflicting explanations. The first explanation is that private schools teach “better.” This may be due, perhaps, to the fact that private schools emphasize, on average, more discipline and traditional pedagogical methods or that they have less bureaucracy and are more flexible to respond to new circumstances such as COVID-19. The second explanation is that private schools deal with very different students. Sending a child to a private school is usually associated with a more advanced socioeconomic class and, probably, with a greater average concern of parents for the education of their children. In this explanation, the private school does not contribute anything special: it is simply lucky to have “better” students.
Which of the two explanations is correct (or the more important)? The answer is key, for example, to design an educational system. Although there are arguments to defend private or public schools that go further than the mere academic results that each one generates, if we come to the conclusion that private schools are better at educating on an equal footing. budgetary conditions that public schools, our view is likely to be different than if we conclude that public education offers better results.
How do we find the answer? Looking for the “causes” of the results instead of simply observing the differences between both types of schools. The problem is that we live in a world of high causal density. For any socioeconomic observation that you put on the table, I will come up with, without much effort, three or four explanations that “explain” this observation for very different “causes”. Are the richest countries democracies because democracy creates prosperity or because prosperity generates a popular demand for democracy? Do college students earn more on average than non-college students because college education has market value or because college students are smarter on average? Do severe criminal convictions reduce crime by incarcerating the most dangerous criminals or aggravate it by turning petty thieves into professional criminals? Does the welfare state help economic growth by better redistributing social risks or does it harm it by increasing taxes? Do emigrants create new jobs or do they only reduce the salary of national workers? You may have your answers to the above questions, but opinions are one thing and solid statistical evidence is quite another.
In natural science, figuring out why things can be done with experiments. What is the effect of a toxin on the growth of bacteria? In the laboratory, we have test tubes with the bacteria and toxin and test tubes with the bacteria, but without toxin, and we analyze the difference between one and the other. What is the effect of wear on a pipeline from constant use? In the laboratory we put a pipe, we subject it to a strong pressure of liquids and we measure when it breaks. Even in medicine we can see, as we have recently learned, the effect of an experimental vaccine among volunteers with a treatment group receiving the vaccine and a control group receiving a placebo. After six months, we see how many volunteers in the treatment group have suffered from COVID-19 and how many in the control group.
But in social life we are not usually able to do this (every now and then there is an exception, I’ll come back to that in a moment). You can’t take, let’s say, Portugal, divide it into North Portugal and South Portugal, put high taxes in the north and low taxes in the south (keeping all the rest of the economic policy constant) and see what happens. First, and most importantly, for a pure ethical question: we are playing with people, not with pipes. Second, because it is not feasible. What do we do? Block Portugal with the forces of the NATO to be sure that no one escapes while we dedicate ourselves to raising and lowering your taxes and waiting to measure the results?
For this reason, from the end of the 80s of the 20th century, economists began to think in detail how to find “causes” without being able to resort to experiments in a laboratory (in economics you can use a lab from time to time, but I skip that today). Much of the statistical methods of previous decades had served to measure correlations between variables (in my example, at the beginning of this post, the differences in results between private and public schools), but not to be able to determine “causality” (Are the “best” private schools or are the students in “best” private schools).
The three winners of the Bank of Sweden Award In memory of Alfred Nobel this year, David Card, Joshua D. Angrist, and Guido W. Imbens pioneered the search for empirical methods to jump from correlation to causation. Its influence on academic research has been so profound that many have spoken of the “Credibility revolution” to highlight the central role of causality in current economic analysis. In fact, it was a “sung” award: for a few years everyone who is dedicated to empirical work knew that Card-Angrist-Imbens was a winning horse and the only question was when he would receive the award.
What kinds of methods have they developed Card, Angrist and Imbens? It would take me a whole book to explain them all so I’m going to select a simple example. Let’s go back to the case of private schools. Imagine that a well-funded foundation decides to offer 250 scholarships to go to private schools. The scholarships are reserved for disadvantaged families whose children are currently enrolled in public schools. In order to allocate the scholarships as equally as possible, the foundation decides that the scholarships will be assigned with a lottery among all the families that apply as long as they meet the income requirements.
Let’s say 1,000 families request the scholarship. The lottery selects, in a totally random way, 250 of these families. The foundation has not designed a “near-natural experiment”, but in practice that is exactly what has happened. The 250 selected families are the treatment group and the 750 unselected families are the control group. As the selection has been random among all the families that have applied for the scholarship, we can assume, with a high degree of confidence, that the families in the treatment group have socio-economic conditions and are committed to the education of their children at a level mean that is equal to that of the families in the control group.
Furthermore, there are always refinements to correct for possible differences between the two groups if, for example, we are unlucky that the lottery selects many more immigrant children in the treatment group than in the control group. Now, all we have to do is wait a year, two or three and see the results of the students in the treatment group, who can now go to a private school, and the students in the control group, who are still in the public school.
One of the winners yesterday, Joshua Angrist, has focused much of his research on studying this type of lottery, for example, in Boston and its surroundings, with different types of schools (public but autonomous, private, etc.). The results, as is almost always the case in empirical work, are full of nuances. For example, the most disadvantaged students in urban areas improve a lot when they move from the traditional public system to a mixed system (public schools, but with autonomy in management and with teachers outside the collective agreement, called ‘charters’). Compared, in rural areas, the mixed system seems to be detrimental.
Of course, this “near-natural experiment” is not perfect. First, we are measuring the effect of moving 250 students from the public to the private system, not the effect of moving tens of thousands of students. For example, private schools may have the capacity to accept 250 more students and provide them with a quality education, but they do not have the infrastructure to digest tens of thousands of new students. The consequences of a policy at the national level can be very different from the consequences of a small policy change in a neighborhood of a big city.
Second, we are measuring the effect of moving to 250 students from the public system to the private of families who applied for the scholarship. Will the effect be the same in the families that did not request it? The same effect is larger (they are children who live in families that are not interested in education and, therefore, have more to gain by changing schools) or it is smaller (what matters is the interaction between motivation families and the school, not the school alone).
Third, the long-term effects are very difficult to measure. Do youLet’s wait 40 years before deciding whether or not to change the scholarship system? With the covid-19 vaccine we accepted (with good reason on the other hand) that after a few months the evidence was clear enough, the vaccine worked very well in the short term, but we still do not know its effectiveness in 10 years.
The reader may want to learn more about these new methodologies. The simplest book is precisely by Joshua Angrist with his co-author Jörn-Steffen Pischke: “Mastering ‘Metrics’: The Path from Cause to Effect“. A little more advanced, from the same two authors, is’Mostly Harmless Econometrics: An Empiricist’s Companion‘, although the years show a little. But the bible (and a much more advanced text) is by Guido Imbens and Donald Rubin: ‘Causal Inference for Statistics, Social, and Biomedical Sciences: An Introduction‘. As I joke to my students, it is one of those books that you have to put under your pillow when you sleep so that the most of it sticks to you. I have three copies: one in my office, one at home, and a third on Kindle.
I close with a little personal story. Guido Imbens, one of the three winners, is my “boss” in an economic magazine called ‘Econometric‘, in which I collaborate as associate editor. Working under his supervision is a pleasure: Imbens is an academic with a dedication of professional service that I can only aspire to in my best dreams. In my own editorial work I constantly try to imitate his seriousness and devotion to economics to the best of my ability. That is why this Monday, when I saw the award, my response was a smile of joy that I don’t usually have at six in the morning when the alarm goes off on Monday.
We would love to give thanks to the writer of this post for this remarkable web content
A Nobel Prize for the why of things