Archive for the ‘mere empirics’ Category
Over the weekend, I got into an exchange with UMD management student Robert Vesco over the computer science/sociology syllabus I posted last week. The issue, I think, is that he was surprised that the course narrowly focused on topic modelling – extracting meaning from text. Robert thought that maybe there should be a different focus. He proposed an alternative – teaching computer science via simulations. Two reactions:
First, topic modelling may seem esoteric to computer scientists but it lies at the heart of sociology. We have interviews, field notes, media – all kinds of text. And we can move beyond the current methods of having humans slowly code the data, which is often not reliable. Also, text is “real data.” You can easily link what you extract from a topic modelling exercise to traditional statistical analysis.
Second, simulations seem to have a historically limited role in sociology. I find this sad because my first publication was a simulation. I think the reason is that most sociologists work with simple linear models. If you examine nearly all quantitative work, you see that most statistical analyses use OLS and its relatives (logits, event history, Tobit. Heckman, etc). There’s always a linear model in there. Also, in the rare cases where sociologists use mathematical models for theory, they tend to use fairly simple models to express themselves.
Simulation is a form of numerical analysis – an estimate of the solutions of a system of equations that is obtained by random draws from the phase space. You would only need to do this if the models were too complicated to solve analytically, or the solution is too complex to describe in a simple fashion. In other words, if you have a lot of moving parts, it makes sense to do a simulation. Since sociological models tend to be very simple, there is little demand for simulations.
Robert asked about micro-macro transitions. This proves my point. A lot of micro-macro models in sociology tend to be fairly simple and stated verbally. For example, many versions of institutionalism predict diffusion driven by elites. Thus, downward causation is described by a simple model. More complex models are possible, but people seem not to care. Overall, simulation is cool, but it just isn’t in demand. Better to teach computer science with real data.
Loyal orgtheorista and sociologist Amy Binder has forwarded me this course syllabus for a course at UC San Diego. It is called Soc 211 Computational Methods in Social Science and was taught by Edward Hunter and Akos Rona-Tas. The authors are working on a textbook, the course was made open to a wide range of students, a and it was supported by the Dean at UCSD. I heard people had a nerdy good time. Click here to read the soc211_syllabus.
At Overcoming Bias, Robin Hanson observes that his fellow economists don’t always focus on the policies that have broad consensus, are easy to understand, and easy to implement. He uses the example of road pricing:
Heavy traffic is a problem every economist in the world knows how to solve: price road access, and charge high prices during rush hour. With technologies like E-ZPass and mobile apps, it’s easier than ever. That we don’t pick this low-hanging fruit is a pretty serious indictment of public policy. If we can’t address what is literally a principles-level textbook example of a negative spillover with a fairly easy fix, what hope do we have for effective public policy on other margins?
I agree. Think about status in economics – what sorts of work gets you the rewards? For a while, it was really, really hard math. Also, macro-economics, which is a notoriously hard field. Recently, insanely clever identification work. What do these have in common? They are hard. In contrast, how many Bates or Nobel prizes have been awarded for simple, high impact work, like road pricing? Nearly zero is my guess.
The same is true in sociology. Sociologists often imagine themselves coming up with marvelous approaches to solving deeply rooted social inequalities. For example, a few months ago, we discussed research on gender inequality and how it might be explained, partially, by the relative over- or under-confidence of men and women. In other words, it might be that women are overly cautious in terms of promotions.
One simple solution would be to require all eligible people to apply for promotions (e.g., require that all associate profs apply for full professorship after a few years). It is a simple rule and would almost certainly help. The response in the comments? The solution doesn’t remedy gender prejudice. Well, of course not, but that wasn’t the point. The point was to fix a specific issue – under representation of women in applicant pools. I have no idea how to eliminate the bias against women, but I can make sure they get promoted at work often – and it’s easy!
Bottom line: Social scientists have their priorities reversed. They get rewarded for trying to solve insanely hard problems, while leaving a lot of simple problems alone. That’s leaving cash on the table.
A few days ago, we discussed an empirical issue around Goffman’s On the Run ethnography. That work focuses on how police intervention cripples poor Black men. The issue is that other ethnography reports an under policing of poor Black neighborhoods. Earlier, I suggested a voter driven explanation – voters like to see young Black men arrested on drug charges and reward police for it.
Here, I’d like to raise a methodological issue. Goffman’s ethnography is not typical in the sense of studying a field site like a firm or a neighborhood. Rather, the ethnography is a study of a cohort of people. You follow them around. That is different than field site ethnography where you choose a location and focus on the action happening in a space. People come in and out. So it is not surprising that if you stand on a modal street corner in Philly, you won’t see many cops walk by. In contrast, if you follow people who are the target of police, then you will, not surprisingly, see a lot of police.
I’ve recently argued that sociology has an amazing opportunity. The emergence of data science means that you should have people who really understand research design and social behavior. It doesn’t mean that sociology will automatically reap the benefits. Rather, we’ll have to work at it. My suggestions:
- Sociology programs should now make basic programming a standard feature of the undergrad and graduate degree.
- We have to have an internal discussion about the strengths and weaknesses of Internet generated data, much in the same way that there is a literature on the pros and cons of surveys, experiments, and ethnography.
We should also reach out to our colleagues:
- Start cross-over workshops.
- Reach out to faculty who already work with behavioral data by offering to help with grants
- Personally, I’ve found it hard to work with CS graduate students. They have the normal level of grad student instability + six figure paychecks waiting for them outside of academia. But still, some are very curious, super smart, and willing to think about behavioral science.
The major barrier, in my view, is the differing publication style. CS happens very, very quickly – sometimes in a manner of weeks, while sociology is slow. We have to stop that.