We finished our “blog your job market paper” series for 2020 on Thursday. We hope you enjoyed reading all this interesting work as much as we did, and thank all those who submitted posts. I thought it would be useful to summarize some of the numbers and lessons for communicating your research.
The numbers
We received 48 submissions on time (and several after the deadline that were automatically rejected). These were balanced by gender (24 females, 24 males), and came from students at 33 different universities, covering research on 30 different countries. Many of those submitting liked to leave it to the last moment to submit, with 26 submissions coming in during the last day and a half before the deadline.
Out of these we accepted 21 submissions (44%) – of which 12 were from female students and 9 from male students, from 15 universities, and covering research on 16 countries (5 on India, and one each on other countries). The papers cover a broad range of topics, and use a variety of different identification approaches and measurement approaches. Papers use multiple approaches, but a simple classification is we have 8 RCTs, 3 Difference-in-Differences, 3 IV papers, 2 RDDs, 2 Fixed Effects, 1 Bunching estimator, and 2 papers that used model and descriptive-based inference.
We received many excellent submissions, and so made this a record number of posts to publish. For comparison, we published 17 out of 42 posts submitted in 2019, 20 out of 51 in 2018, and 12 out of 26 in 2017.
Reflections on writing compelling job market posts
Once we receive the submissions, I allocate them amongst the five core bloggers on the Development Impact team, who then make decisions on which posts they are most interested in running. We then provide comments to the authors to try to help them communicate their work more clearly. Looking through the comments sent, there are a set of issues that come up again and again that I thought might be of more general interest in communicating results in this format.
1. Make clear the sample, data, and identification method and assumptions:
a. Sample: Tell us how big the sample is, and what the sample looks like: For example, multiple posts do not make clear the sample size, or what some basic characteristics of the sample are. For example, we might get told it is an experiment on firms, but not what sectors the firms are in, or whether these are 1-person or 20-person firms on average, nor how many firms are in the study.
b. Data: Tell us the main data sources, and any issues with attrition: This helps in understanding what the key outcomes are, how they are measured, and over what time horizon.
c. Identification: make clear the method and assumptions: this is particularly an issue in non-experimental studies, and especially those using DiD and IV approaches, where the identification assumptions are not always made clear. If you are using IV, state explicitly the exclusion restriction, and make a case for it. If you are using DiD, explain why parallel trends seems reasonable, etc.
2. Communicate magnitudes, not just statistical significance: provide numbers, and help the reader understand how to interpret these.
3. Avoid the wall of text, and use a compelling figure or two if possible: some submissions just look a lot like the introduction to a paper, summarizing the paper over multiple paragraphs. But for a blog post, using bullet points, number lists, and especially one or two compelling figures to illustrate key findings is really helpful. Even if your paper presents all the results as regression output in table form, it is worth investing in ways to graphically display some key findings. This will be helpful for presentations as well as for the blog post.
4. Avoid over-hyperlinking the reader to distraction: several posts I read had an opening paragraph or two that aimed to set the paper in the literature, hyperlinking to a whole lot of other studies. While this is great in a research paper, in a blog post each hyperlink before the link to the job market paper is a potential trap that encourages readers to click away from reading about your research to read about someone else’s interesting research.
5. Make clear why you think we should care about this for development: What should a reader who cares about development takeaway from your post? Is it that this should just help us understand better the nature of some development outcome, or should it help better target policy, or does it show a way to better actively help poor people, or something else? My personal bugbear here is with some historical persistence/political economy-type papers, which e.g. show that something that happened 200 years ago shows up as correlated with some outcome today, but which don’t go further than that (of course tastes vary here) – it is worth making as clear as possible why you think this is an interesting and important question for development research and/or policy.
Here is the full set of posts:
1. Digitising microfinance loans to create female enterprise growth: Guest post by Emma Riley
10. Let the (P)rice Flow: Can Export Activity Benefit Domestic Consumers? Guest post by Utsav Manjeer
14. Surprise! When do they work best for auditors?: Guest post by Wendy N. Wong
21. Profit vs. Revenue tax: How to make corporations pay their fair share? Guest post by Thiago Scot
Join the Conversation