When and how to include a prediction in your paper
As I posted earlier, the single part of my clients’ Introductions that most often requires significant work is the all-important hypothesis or prediction. But it’s crucial to understand which kinds of papers need to make a prediction and which ones do not. The rules aren’t written in stone, but here are some questions to ask yourself to determine whether you should include one or more predictions:
- Are you reviewing others’ work, laying out a guide or framework, or evaluating a program, or are you collecting new data to answer a research question? If you’re doing any of the first three, a prediction is probably not appropriate. If you’re collecting new data on outcomes of an intervention or relationships among variables, however, chances are good that your reviewers will expect you to predict what you’re going to find. If you’re not sure which of these activities describes your paper, a simple guiding question you can use is, “Do I have any reason to think that I’ll get a particular kind of result?” If the answer is yes, you can make a prediction. (If the answer is no, you either don’t need a prediction or you haven’t established a good rationale for doing the study you’re planning. Be sure you know which is true before you collect your data.)
- Does your paper fit into the Intro/Methods/Results/Conclusions format? If so, you probably need to make a prediction. If not, or if you’re not sure, ask yourself why you’re doing the study, and whether you have any good reason for thinking the results will turn out a certain way, based on what you already know from the literature.
- Are you testing a theory, attempting to replicate another study, or trying to expose a possible flaw in another study’s methods? If any of these are true, you will be expected to make a prediction.
The four major pitfalls in prediction are: failing to make one when it is called for; making one that is too vague or too obvious to be useful; making one that isn’t logical in light of the rest of your Introduction; and – the worst of all sins – making one after you’ve collected your data. Commit one of these errors, and score a major eye roll (at best) or an accusation of an ethical breach (at worst) from your reviewer.
After determining that you need to make at least one prediction, you’ll want to consider carefully how to craft its wording. Let’s walk through an example.
Cohen & Chaffee (2013) published a paper with the following stated purpose: to understand “how civic knowledge, civic attitudes, and civic behaviors [of adolescents] are associated with self-reported likelihood of future voting.” The statement of the study’s purpose clues us in that there will likely be at least three predictions made: one each about the effects of knowledge, attitudes, and behaviors. Actually, the authors used the existing literature to identify three types of knowledge, two attitude variables, and two behavioral variables that they had reason to believe were associated with intention to vote, and each variable was tested independently. Because the prediction was that each variable would be correlated with intention to vote, here’s how they stated their (single) hypothesis about the seven variables:
“We hypothesize that in this cross-sectional survey of urban youth, these civic constructs are each independently associated with future voting, adjusted for demographic and academic factors.”
Had I been a reviewer of this paper, I might have suggested that the authors be more explicit in stating the predicted direction of the associations (e.g., “… higher levels of [variable] are associated with [stronger/weaker] intentions to vote.” But the hypotheses are there, and the literature review that the authors have done addresses the reasons for including each variable in the hypothesis.
In a paper where more than one prediction is being made, it can be very helpful for the reader if you label or number each one (e.g., Hypothesis 1, Hypothesis 2, etc.). This way,when you are reporting and discussing your results, you can easily create a shorthand (H1, H2, etc.) to remind the reader of each prediction you made, and how each accords (or doesn’t!) with the results. In the case of Cohen & Chaffee’s paper, it is not necessary to label each hypothesis, because the prediction was in the same direction for each, and because the results overwhelmingly supported each. In this situation, it would have been unnecessarily repetitious (and tedious for readers) to plod through seven labeled hypotheses in turn. Instead, the authors presented a concise table showing clearly that all of the associations were in the same direction, and that all seven predictions were supported.
Cohen, A. K., & Chaffee, B. W. (2013). The relationship between adolescents’ civic knowledge, civic attitude, and civic behavior and their self-reported future likelihood of voting. Education, Citizenship and Social Justice, 8(1), 43–57. http://doi.org/10.1177/1746197912456339
 There were additional variables in this study, but their discussion is not essential to the point about generating and stating hypotheses.