In a 2020 NBER Working Paper (alternative ungated version here), Janet Currie, Henrik Kleven, and Esmée Zwiers (all Princeton University) looked at how the methods employed in applied microeconomics research have changed over time. They mined text-based data from a sample of 10,324 NBER Working Papers over the period from 1980 to 2018, and 2,830 journal articles published in the top five economics journals from 2004 to 2019. They find a number of interesting trends, including:
...a virtually linear rise in the fraction of papers, in both the NBER and top-five series, which make explicit reference to identification. This fraction has risen from around 4 percent to 50 percent of papers...
...a somewhat slower rise in the use of experimental and quasi-experimental methods... Currently, over 40 percent of NBER papers and about 35 percent of top-five papers make reference to randomized controlled trials (RCTs), lab experiments, difference-in-differences, regression discontinuity, event studies, or bunching...
...a very similar pattern in references to administrative data. The NBER series starts increasing in the mid-1990s, rising to about 30 percent today. The top-five series shows a similar increase, but with a lag of about three years... The term Big Data suddenly skyrockets after 2012, with a more recent uptick in the top five...
The importance of figures relative to tables has increased substantially over time and in two phases. The first phase happened in the 1990s and likely reflects the diffusion of new software such as STATA that made it easier to create impactful figures. The second phase has happened in the last 10-12 years and is still ongoing. This is likely due to the increasing use of administrative datasets, which lend themselves to compelling graphical representation using raw data and non-parametric approaches...
...a sharp rise in the fraction of NBER working papers discussing randomized controlled trials since 2005, and especially since 2010...
Laboratory experiments have grown steadily in popularity since the late 1990s, which is connected to the rise of Behavioral Economics during this time period...
...authors have become increasingly concerned with whether their estimates are precisely estimated, and not merely with whether they are significantly different from zero in a statistical sense...
...a sharp rise in references to confidence intervals since the mid-1990s...
...after year 2000, there has been a massive increase in attention paid to clustering of standard errors...
Currie have clearly identified the most important changes in the types of research methods used in applied microeconomics over the last 30 or so years. One thing they haven't noted is the rise in the use of textual analysis, including sentiment analysis, as a trend within the use of big data. This is somewhat ironic, since that's what their paper uses!
It would be interesting to see a similar exercise conducted for applied macroeconomics.
[HT: Marginal Revolution, last year]
No comments:
Post a Comment