Davide Proserpio

New working paper: "Do Modern Difference-in-Differences Estimators Change Published Findings in Business Research?"

Peter Hull's recent meta-analysis found that in economics, two-way fixed effects and heterogeneity-robust DiD estimators rarely disagree. Using Claude, I ran his protocol on seven business journals, screening 1,829 papers, and re-estimated every eligible design in Management Science and Marketing Science, the two with public replication archives. That gave 355 published staggered-adoption specifications from 40 papers, re-estimated with Callaway-Sant'Anna (CS) and Borusyak-Jaravel-Spiess (BJS). On the comparisons that pass Hull's leverage screen, from 28 (CS) and 29 (BJS) papers:

- Correlations between TWFE and modern estimates are 0.61 and 0.70, against 0.88 and 0.93 in economics.

- Estimates change sign in 25 and 18 percent of comparisons, and 10 and 7 percent show a statistically significant difference.

- Modern standard errors are 1.6 and 1.3 times those of TWFE.

- Disagreement concentrates in unbalanced panels and in designs without never-treated units.

Every table and figure rebuilds from the public repository; re-estimating a specification from scratch requires the journal's replication package.

Paper: lnkd.in/gXmtvfBC
Code, protocol, and estimates: lnkd.in/gZijx_dy

Can someone endorse me so I can submit to arXiv econ? Endorsement Code: FJ3FZQ

Inspired by @instrumenthull, I am doing the same replication exercise for business research using TWFE (mostly Management Science and Marketing Science papers so far)

@SpaceX @elonmusk

Did the SpaceX IPO move the LA housing market?

In June, SpaceX priced the largest IPO on record. Its Hawthorne campus employs roughly 7,600 workers, a few miles from some of the most expensive beach towns in Southern California. Within days, local brokers reported calls from SpaceX employees about $5M+ homes. So naturally, Marco Giacoletti and I decided to see whether there was any truth to it.

In a new R.E.A.L. post, we ask whether the data back up these anecdotes. We combined Census data on where SpaceX Hawthorne workers live with ZIP-code-level housing data from Redfin and Zillow. We did find suggestive evidence that SpaceX did increase home sales:

1. Home sales in the six expensive ZIP codes where SpaceX workers concentrate rose 24% year over year in June–August while only 2% in 51 comparable expensive ZIP codes.

2. The 22-point gap is the largest in any three-month window since 2017.

3. Prices have barely moved so far, consistent with evidence from past California IPOs.

Full post: realab.blog/p/did-the-spac…