Economic research: causation

One great thing about studying economics is that career prospects are quite broad. Whether you are interested in business, insurance, data science or politics even, a degree in economics likely makes you a suitable candidate. If you'd rather not face the challenges of real-life and having a job straight after your Bachelors you can always extend your stay in the academic bubble and continue on a Masters course (obviously this isn't only an option for economists). In this case, you are likely to come across doing economic research, and if you are doing your Bachelors dissertation (like me) it is sooner rather than later.

Research is a pretty broad term and there are various types of research out there. You could be a scientist and trying to figure out how a specific drug works or how this star has come to be (as you can tell I know A LOT about astrophysics). In the social sciences - so anything from social anthropology to political science - you are likely interested in understandings society and human interaction. Now economics sits somewhere between the two. To some degree, economics would love to be considered a science such as physics or mathematics but this is complicated by its subjects. The economy is not some natural thing that is guided by the laws of nature (although we do try to sound sciences by calling them the laws of supply and demand, although I would consider them more guidelines than actual laws) which makes it so much more difficult to find a definite answer to an economics question. If I drop my pen gravity will ensure that it always falls according to some rule. But humans don't always behave according to specific rules. The rational agent in economics is an unrealistic chimera that serves as a simplification more than anything else. Nobody (I hope) in economics believes anymore that humans actually behave like rational agents. Yet, the rational agent can be useful in helping us model how we would behave if we were rational and then understand why that might not be the case. But I digress. What all this is trying to say that economics is a social science in terms of the things it studies. But in doing so, economists use methods that aim to make the conclusions as scientifically robust as we can.

One thing economists are particularly interested in is causation. This happened because of this: "the price increased because more people wanted to buy this good". In this sense economics is similar to the natural sciences: "the pen fell down because of gravity". Causation is very, very difficult to identify because in the social sciences there are so many things that might be the reason for something to happen. Here's a real-life example. A group of economists wanted to study the effectiveness of political protests. Demonstrations and protests are recognized as being an important aspect of democracy, but are protest actually the reason (the cause) for political change? Or are protest not the actual cause of political change, but instead people have changed their beliefs and preferences and protests are just a symptom of this change in opinions? The paper "Do Political Protests matter? - Evidence from the Tea Party movement" tries to answer this question. More simple: are political protests the reason for political change. So they try to measure whether bigger protests resulted in more political change. The problem is that they can't just estimate an equation for this, because maybe it's actually a change in opinions that causes BOTH bigger protests AND more political change. If this is true, it's just that bigger protests and more political change are related* but one isn't really the cause of the other and remembers we want to identify causation. So what to do about all this?
Luckily, these researchers came up with a cool way to identify causation. Instead of estimating the effect bigger protests had on political change, they estimated the effect of rainfall on political change. Huh? Why would they do that? The reasoning goes like this: We know (or more we make the reasonable assumption) that more rainfall doesn't have a direct effect on political change, because why would it. But we also assume that more rainfall means smaller protests, because we might be passionate about a cause but less motivated to actually join a protest if it rains like crazy. So we assume that more rainfall means smaller protests, assuming that everything else (in terms of peoples beliefs or other things that might have an impact on their voting behavior such as income) stays the same. Now we estimate the effect of rainfall on political change, also assuming everything else stays the same. Remember, if you think about it rainfall should not really have an effect on political change because there's not any reason why more rain would cause - or be the main reason for - political change. That is unless there is something going on in the background. The something that is going on in the background is the relationship between rainfall and the size of protests - more rain means smaller protests and vice versa no rainfall means bigger protests. Bringing it altogether if we find that an absence of rain (compared to when there is high rainfall) leads to more political change we can make the conclusion that the size of protests does cause policial change. And this is exactly what the researchers found. On rainy days there were smaller protests and in those areas, there was less political change (in the context of the tea party movement which their article was about). But when protests were on sunny days there were bigger protests, and suddenly they saw quite a bit of political change. Everything else is the same, people have the same political opinions, same income, etc, the only thing that was different was the weather. This allowed the researchers to conclude that political protests do matter.

This example illustrates how tricky it can be to identify causation but how smart economic techniques (and researchers!) can help overcome this burden and make robust conclusions about social issues. Of course, this isn't always the case and not all researchers truly identify causation. At the same time not all research is on causation, economists might also be interested in evaluating the effects of policies and programs or many other issues. But the issue of causation is a good example to show how economists research social issues using scientific methods.

*Footnote: There are many things that are correlated but where one thing doesn't cause the other. Correlated means for example that as one thing increases, the other does as well. Tyler Vigen has collected a number of these "spurious regressions" where things are correlated but obviously have nothing to do with each other like cheese consumption and the number of people dying from bedsheet entanglement.

No comments:

Post a Comment