Weekly links September 11: the fragility of high growth firms, women’s work and families, how many papers use AI now to write, and more...

· The consequences of more women working: In the economist this week, a lengthy piece on how women’s work makes families more equal – with discussion of how legal changes, social norms, technology, and migration all have affected the ability of women to work and the implications of this on their role within the family. Kathleen gets a shout-out: “changing laws can help, as Saudi Arabia shows. But it is only a start. Social norms are often a deeper challenge, notes Kathleen Beegle of the World Bank. Things “like which household member is responsible for pounding the foo-foo in the morning and making sure that dinner is on the table at 5.30pm; it’s not something that is really in the traditional policy sphere for governments.” · High-growth happens only for short periods for most firms: On VoxEU, Giuseppe Berlingieri and co-authors report on their research on “innovation bursts” in firms. “policies often assume that large firms are large because they are good at innovating, so that supporting them means supporting innovation....The patterns we find sit uneasily with the conventional view of how firms grow. ...innovation bursts enable firms to rapidly expand their product portfolio, sales, and employment, even if they are initially small. Even among firms that initially produce a single product, some add tens of products in a single year – and of these firms, all product innovation comes from just 7% of them....Innovation bursts have a key implication for what separates large firms from small ones: if firms mainly expand through rare, outsized bursts, then differences in firm size may reflect the luck of drawing one of those bursts rather than persistent differences in the ability to innovate.”. This is consistent with the high growth firms report that Arti Grover and colleagues wrote at the World Bank, which found that high growth is very episodic “, the achievement of high growth in one period does not mean that firms are more likely to grow rapidly in subsequent periods—evidence shows that HGFs mostly turn out to be “one-hit wonders.” As many as 50 percent of firms that experienced a high-growth event in the previous three years are likely to exit the market altogether in the following three to six years, whereas fewer than 15 percent are likely to repeat a high-growth episode—illustrating the short-lived and episodic nature of firm growth.” · Mentoring opportunity: The Association for Mentoring and Inclusion in Economics (AMIE) is now accepting submissions from current job market candidates for having a one-on-one session (approximately 1 hour) with a senior economist (in academia/non-academic research/private sector etc.). Candidates will receive feedback on their job market application materials, including their cover letter, resume, and website. This also includes one practice mock interview. Apply here by October 5. · Are papers now written with AI? Paul Goldsmith-Pinkham has some interesting analysis of the rise of AI-written sections of papers on Arxiv and in the NBER series using pangram: “I spent roughly three thousand dollars in API credits on Pangram’s model to classify every NBER working paper and economics ArXiv paper from January 2022 to mid-August 2026...I chunk the text and focus on the introduction and a randomly selected section from the body of the paper. ...I find a couple pretty remarkable facts. First, almost none of the NBER and arXiv economics papers in 2022 are classified as AI. This has really shifted my priors on how well we can avoid false positives. Second, almost 50% of arXiv papers have at least one 300 word window classified as AI, and roughly 25% of NBER working papers. Finally, the acceleration really only started occurring in 2025. This acceleration coincides with the growth of agentic AI (Claude Code) rather than just web-based AI platforms like ChatGPT.” · Do I need covariates for my DiD? Scott Cunningham on four principles for deciding when and which covariates to include when doing Callaway and Sant’Anna DiD: “we don’t care about covariates that cause levels in the outcome to differ...we care about covariates that cause trends in the untreated potential outcome to be different, and we care then if they are imbalanced.”- although as I have argued before, I usually find it more plausible that counterfactual trends will be similar for more similar observations than for more different observations. · You will be in a RCT if you submit to these journals: The AEA announced that it is going to be starting a 3-year RCT randomizing whether manuscripts are sent to referees without the authors names or not (single-blind vs double-blind peer reviewing). You can predict whether you think this will make any difference here. The 1991 trial found acceptance rates were lower and referees were more critical when the reviewer did not know the author’s identity so I’m hoping my papers get assigned to the status quo control group. I’m also hoping this does not mean having to submit even more files – I hate it when journals make your submit a separate anonymous title page, a blinded version of the manuscript, an unblinded version, separate conflict of interest statements for every author, and a zillion other files. Just let me send the paper!