Tom Pepinsky

Author: tompepinsky

  • Selection Bias, Workload, and Teaching Evaluations

    This New York Times opinion piece proposes a reform to the tenure system in which there are three tracks: the “teaching track,” the “research track,” and the “teaching and research track.” Worth a read. But I’m not here to talk about that, I’m here to question one of the findings used to support this argument, that “students rarely favor teachers who grade leniently — and give higher ratings to teachers who assign heavier workloads.”

    This finding screams “selection bias.” Let me explain how.

    Let’s say there are two types of students: gluttons for punishment and easy riders. Gluttons for punishment prefer the challenge of a hard class, and in my experience, you can actually get them more excited for your class by telling them how hard it is and how much work you will assign. I was a glutton for punishment, and so were some of my friends—I remember my sophomore roommate and I comparing our battle scars from Engin Akarli‘s insane Middle Eastern History course. (“Some people say, no dates. I say, “know dates!”)

    Easy riders prefer an easier course, with lower workloads and easier grading. They like courses like my cousin Hal’s criminology courses.

    The problem is, so long as students have a choice in what classes they take, we should expect them to self-select into easy courses if they are easy riders, and into hard courses if they are gluttons for punishment. The results: hard courses full of students who like hard courses, and who evaluate them favorably, and easy courses full of students who don’t like hard courses, and who evaluate them accordingly. If it is the case that even gluttons for punishment take easy courses sometimes (ahem), but easy riders never take hard courses[1], then in aggregate, evaluations will suggest that students prefer hard courses because only the students who prefer hard courses will be evaluating them.

    What I want, of course, is for someone to exploit a natural experiment in which the same course is taught with different workloads to students who don’t have a choice to take the hard version or the easy version. But even that won’t be a relevant design if the rest of us have to anticipate the effects of workload on enrollment.[2]

    ***

    [1] Actually, what we really need to assume is simply that easy riders are less likely to take hard courses than gluttons for punishment are to take easy courses.

    [2] As Director of Undergraduate Studies in my department, I think about this.

  • Boycotting Elections as Procedural Politics

    an election results is always to some degree a signal about the electorate’s preferences about the issues being confronted at that time.

    That is John Patty discussing two theories of election boycotts: why it might be in the interest of an opposition party to ask its voters, instead of voting against the incumbent, to not show up at all. Go read that post, then come back here.

    OK, ready? Here is another perspective[1] on boycotts: they are procedural politics. By this I mean that a boycott might not be about any policies at all, but about the procedures through which societies use elections to make policies. Let’s assume, for now, that there are two parties competing over one policy: a reform that the incumbent supports and the opposition opposes. Let’s also assume that we know with certainty the distribution of popular support for each party: 60% support the incumbent (and, hence, the reform) and 40% support the opposition (and, hence, oppose the reform).

    We’re not done yet: we need an electoral rule—the procedure that we use to translate votes into seats in government[2]. We will say that the status quo electoral rule is majoritarian. That means that the party that gets the most votes gets the most seats in government[3].

    In this scenario, we know with certainty that any election will put the incumbent party in office, and that means that the reform will be implemented. Because (by assumption) we know the distribution of support for each party and each policy, boycotting an election cannot send any signal. It strikes me that this is the Thai case in a nutshell.

    So in this case, we have to add something else into the model to generate boycotts as a meaningful political strategy. What might that be? Let’s start by focusing on the minimalist conception of democracy as a system of government in which parties lose elections. You have to be willing to lose, and in this case, the opposition will lose. But you might imagine that the opposition thinks “wait a minute, what if we had a different electoral rule? Say, one in which our districts get more seats because they are urban and the urban areas are important? Or, two houses, one elected the regular way, another elected in a way that favors us? Or maybe there’s a council of notables, and we pick it.”

    Now normally, you can consider this to be a reform like any other reform. The incumbent will oppose it, and it will not pass under the status quo electoral rule. So what is your option if you are a motivated opposition? It must be to reject the procedures themselves. How do you do that? You reject the procedures: boycott the elections.

    This may seem unfair to our simple model, and it is. I have added a secret ingredient. And here it is: for this scenario to work, it must be the case that somebody (could be an external party such as a military, or it could be the voters themselves) has preferences over the procedures themselves. In other words, someone must believe that there is a turnout level below which elections do not serve the functions that they are meant to serve, or alternatively, a cost to rejecting the status quo electoral rule which is high enough to make them willing to trade representation under the status quo system for order[4].

    But that’s bad news for our simple model: once we start admitting multiple dimensions and politicking over the electoral rule then unless we start pinning things down with further assumptions the strategies spin out of control and the model blows up.

    That, in the Thai case, might be just what the opposition wants. Because if the model doesn’t pin things down with assumptions, then you could always have the military do it for you.

    [1] There are probably a dozen other possible models that one might write down in which boycotting sends some sort of signal about something: resolve, mobilizational capacity, something else altogether. All of these might be in play in the Thai case too, but I want to hold those aside for expository purposes.

    [2] We will work within a very simplified model of government. Voters are voting for a single legislative body, and the party with the majority of seats in that legislature elects a prime minister who always implements its preferred policies. That is, there are no politics within the legislature, and no delegation or monitoring problems from the perspective of either voters (vis-a-vis the legislature) or legislators (vis-a-vis the prime minister).

    [3] We will also abstract away from things like gerrymandering. We’ll say that it happens to be the case that voters are distributed across districts in a way that 40% of districts have an opposition majority, and 60% have an incumbent majority.

    [4] I think that this is what’s actually happening in Thailand. The opposition believes that there is a preference for order, and the boycotts are disorderly.