The following is a draft of advice for undergraduates working with me on writing senior theses. I’m intentionally leaving all the citations and academia flag-planting out of this; moreover, this is advice to my students, not anyone else’s, so don’t take this as gospel or hegemonic domineering.
What is a “thesis”?
You will spend this semester writing a “thesis”. By convention, we refer to the 35-50 pages you will produce with that word, but it’s worth unpacking why such a long paper is referred to by that name. After all, a “thesis” is just a claim that you want to put forward (and, ideally, test). You can state a thesis easily: for instance, “Hell, purgatory, and heaven seem to differ the same as despair, fear, and assurance of salvation” or “The economic effects of oil rents include the perpetuation of patriarchy.” So what about the other 34.9 pages you’ll write?
The answer is that we don’t care about them separate from your thesis. They are only the supporting stonework in the arch for which the thesis is the keystone. And like a keystone, your thesis will both be supported by and support the rest of the work. We refer to the whole work as a “thesis” in recognition, conscious or otherwise, of the importance of this central claim.
Theses–and here I’m referring to the claims, not the document you’ll produce–may be stated at any length. There is often a ‘gut’ version–“Voters blame politicians for random events”–and a more precise version, as in Achen and Bartels’s seminal paper on “blind retrospection”:
Switch analogies for a moment: regard the thesis as a Slinky that you can stretch or compress. Both “gut” and “precise” forms of the thesis express the same argument, but they have been reshaped to fit the space available (as well as the intellectual requirements of your task). But whatever shape or size the thesis appears to take at the moment, it must remain the same argument. A chief quality-control task for you as knowledge worker is ensuring that your expression of the thesis remains as analytically precise and consistent as possible, whether said in punchy Anglo-Saxon words or elaborated in scientific (or pseudo-scientific) Latinate verbiage.
Any number of claims could be made to serve as a thesis, but in this course we are interested in theses that make a claim that relates to a particular theory. In particular, to complete an empirical senior project, your thesis must be:
- falsifiable (there might exist some observation that could show that your claim is not true)
- testable (related, but different: the observations that could falsify your claim actually exist)
- important (we care about the implications for a thesis that is true or false)
- nonobvious (the answer is not trivially clear–that is, research is genuinely valuable)
- puzzle-driven (your thesis must answer your puzzle)
And it should probably also be
- general (a thesis that accounts for phenomena in a larger range of observations is, ceteris paribus, more valuable than one that accounts for phenomena in a smaller range)
- limited by evidence (but you shouldn’t go too far in proposing general theses if you can only test your claims on a severely restricted sample — e.g., don’t generalize about ‘human’ cognition if your test subjects are all Western, educated, individualistic, rich, and democratic)
- humble (you should know what other folks have said that is relevant to what you’re working on–don’t write as if you’re the first to claim something that everyone has heard of)
- confident (but on the other hand, don’t assume that because you’re young and obscure you have no thoughts worth communicating — quite the opposite, all of social science is understudied and you may well make a contribution)
Note that a thesis need not be particularly original–at least, not at the grand level. Indeed, most great puzzles are not original (“given the costs and risks, why is there war?” or “given the costs of casting ballots and irrelevance of a single vote, why does anyone vote?”), and that means that many of the theses that seek to respond to them are also not particularly original — in the abstract. In the specific, of course, I expect that your own thesis will be original in the specific, even if that means that it’s closely related to one or another existing set of claims that you uncover.
The crucial idea, though, is that your thesis represents a claim that you are making–and a wager that you assert is true. That claim, in turn, must be justified by some theory.
How does a thesis relate to theory?
Theses must derive from some theory. Theories offer general accounts of wide classes of behavior. At each stage, theories identify explanandums (things to be explained) and explanans (factors that explain them). For instance, rational-choice theory assumes that people make decisions on the basis of their rational matching of strategy, circumstance, and outcome, choosing strategies to yield the best-possible outcome under the circumstances. The explanandum, then, might be an agent’s choice, and the explanans would be the circumstances under which the choice is made (including game structure, knowledge, beliefs, best-response of other players, etc.) and the agent’s utility function. If we argue, for instance, that variations in private costs affect citizens’ propensity to turn out to cast ballots, then we would take as an explanans “casting a ballot” and we would find our explanans in factors such as the distance to the polling site, the resources the voter possesses (e.g., does she have a car or must she take public transportation?), the strength of private utility from voting (does he proudly display an “I voted” sticker or does he think voters are suckers?), attachment to a particular political party, and so on.
Theories therefore differ from ad hoc explanation. The best shorthand I know of for expressing when an explanation has become a theory is that it lacks proper names or specific nouns. Saying that “Dave buys ice cream because he likes it” is just a description; arguing that “People eat foods that are bad for them because they value short-term gains more than long-term costs” offers a more general argument, which rests on plenty of arguments about how people value near- and long-term gains, how people evaluate costs, and so on. The goal of this intellectual game is often described as coming up with “law-like generalizations” (think of the Tom Friedman “first law of petropolitics”, but done better).
Theory-making is always a creative process. Theories and their analytical elements do not correspond to directly observable features of the real world. They are always selective about what they explain. A theory explaining voter behavior in U.S. presidential elections, for instance, does not have to account for voter behavior in Toronto’s municipal elections. If observations truly fall outside the scope of a theory, then they are irrelevant to its usefulness or veracity. Similarly, to prevent problems that arise from infinite regress, some things simply have to be assumed to be true — that the “United States” exists, that “votes” have actually occurred in the past, that we do not live in a computer simulation that began five minutes ago, and so on into infinity (really).
If a puzzle falls within a theory’s ambit, then the theory should supply a clear thesis. In the vote-turnout example, for instance, a theory that sees citizens as social actors would lead to the thesis that social factors would prove important in turning people out to vote. For instance, if voting is socially desirable, perhaps threatening to publicize someone’s vote history will lead them to turn out from fear that their neighbors will think less of them if they don’t vote. Or perhaps vote turnout is determined by bigger institutional features, and a decline in unionization and other organizational factors explains variation in turnout over time. Closer to our knowledge base, Michael Ross argues that “petroleum promotes patriarchy” because of the ways in which oil rents undermine alternative pathways toward women’s empowerment, such as employment in light manufacturing, which is the first step toward joining unionization.
It should be clear, then, that a theory’s choice of explanatory factors will affect the sorts of explanations it can offer. For example, theories of war that begin from the presumption that leaders are irrelevant would offer different explanations about whether “Hitler causes the Second World War” than theories that believe that individual leaders can make a difference. Indeed, to someone who believed that leaders are irrelevant to world politics, the argument that “World War II would not have happened without Hitler” makes no sense, as the explanatory factors they can bring forward to address the question of “Why World War II?” can’t include individuals. Similarly, if a theory assumes that individuals make decisions rationally, it cannot include purely nonrational explanations for phenomena within its scope without conceding the field altogether.
What should your theory look like?
For this week’s assignment, you should work through a theory that addresses your puzzle and delivers a clear thesis that summarizes your argument. That theory need not be complete but it should offer as many answers to as many “why” questions as possible — and it should be as general and abstract as you feel you can handle. (You can always offer illuminating, real-world examples, but you should remain at an abstract level here.)
The good news is that, once you have done this, writing hypotheses is easy. You simply think about what must be true if your theory is right–and, ideally, what would not be true under competing explanations–and set that out. So, if we think that private costs matter to voters in choosing to turn out, an experimental manipulation that lowered those costs for some votes (perhaps subsidized Lyft rides to the polling place?) should yield higher turnout rates. Or if our argument is that a loss of community organization matters most, we should not expect to see turnout falling at similar rates in communities that have not experienced a downturn in institutions like churches and unions as in those that have.
The bad news is that writing a general explanation for complex social phenomena is hard.
The better news is that it’s February and you have months to fix what you come up with!