Jeffrey Lin
@jeffrlin
Economist studying cities, regions, and growth. West Philly dad. Views are my own.
🛣️ 🌆 How do the quality of life costs of highways vary across cities? In a new review article (forthcoming at RSUE), Brinkman and I expand our Freeway Revolts methodology across US metropolitan areas.
In 1876, Bell & Gray collided because idea space was crowded. 150 years of spreading out have emptied it — and eroded the productivity of R&D. But spreading out isn't waste: it's how we get better solutions to more problems. The cost is in growth rates, not in what inventors produce.
Think of innovation as picking fruit: which tree, how high to climb. Prior work asks how expensive the equipment is (Jones 2009), how high you must climb as nearby fruit depletes (Kortum 1997), how fast new pickers arrive (Jones 1995). We add: how wide the forest is.
A calibrated decomposition attributes roughly 40% of the long-run US research productivity decline to spatial forces — alongside "fishing out" and the "burden of knowledge." Spatial geometry accounts for nearly half of what Bloom et al. (2020) measured.
The equilibrium geometry endogenizes parts of key primitives in canonical growth theory — spillover intensity, innovation step size, research productivity. Here, they're outcomes of where inventors stand relative to each other.
Spreading out unifies two patterns that previously seemed in tension: the secular rise in breakthrough inventions (Kelly et al. 2021) and patent values alongside the well-documented decline in research productivity (Bloom et al. 2020). Both are equilibrium outcomes of the same spatial geometry.
The model predicts that isolated inventors invest more and produce higher-quality inventions. The data confirm: inventor team size, organizational form, and KPSS (QJE 2017) patent values all rise robustly with spatial isolation.
We develop a spatial competition model to explain spreading out: inventors ("you", below) choose locations in idea space. As the frontier expands and entry costs rise, they spread out — capturing larger territories, raising R&D investment and invention quality even as knowledge spillovers attenuate.
We build the first systematic validation framework for patent text — interference cases, human annotation, patent classifications, spanning 1850–2023. GTE outperforms TF-IDF by 30% on detecting identical inventions and is uniquely robust to 19th-century language.
Measuring this correctly matters. Apply TF-IDF to patent text: similarity rises 1.5σ. Apply GTE neural embeddings to the same text: it falls 1.5σ. Same patents. Opposite conclusions about the central fact of the paper! (Our prior paper discusses this in more detail; see earlier thread.)
US inventions have grown increasingly dissimilar. Using validated neural embeddings on 11+ million patents (1836–2023), we find a 1.5 standard deviation decline in pairwise similarity since 1836 — corroborated independently by a 150-year decline in patent interference rates.
On February 14, 1876, Bell & Gray each filed for a telephone patent. Such collisions were once routine — up to 1 in 20 patents triggered an "patent interference." Over the next 150 years, that rate fell by more than 98%. In a new paper, we explain why — and what it might cost.
To measure neighborhoods where FAIR plans were likely offered, we digitized city directories for 26 cities, documenting private property insurer's postwar exit from central neighborhoods. Using these data, we measure neighborhoods experiencing insurer withdrawal symmetrically across states. 7/
Some contemporary observers blamed FAIR for the 1970s wave of urban arsons, although others have blamed disinvestment in fire protection. 5/
New working paper on 1960s insurance policy and its effects on housing (dis-)investment and neighborhoods, with Ingrid Ellen, Dan Hartley, and Wei You. jlin.org/papers/EHLY-...
Tune in tomorrow morning to see Vitaly present our paper at the I3 meetings. www.nber.org/conferences/...
Fascinating follow up to William Whyte’s Social Life of Small Urban Spaces by Salazar-Miranda et al. Using AI and historical and modern sidewalk video, compared with 1980, walking speeds up, lingering down, group encounters declined. www.nber.org/papers/w33185
And here’s a nontechnical companion piece with an analysis of burying the Delaware Expressway in Philadelphia: jlin.org/papers/BL-FD...
We had the privilege of hosting the 16th Conference on Urban and Regional Economics (CURE) last week here in Philadelphia. Great presentations by some of the field’s most promising young researchers. Thanks to all who participated and helped to organize! custom.cvent.com/9A0657ED9B01...
CFP: really nice small conference sponsored by CRED in Bern that I had the pleasure of participating in a couple years ago. Conference followed by an Alpine adventure! www.cred.unibe.ch/forschung/ev...
Found a free streaming TV channel that airs only Conan O’Brien clips 24/7! Currently airing only Jeff Goldblum interviews, one after another.
Q&A with my new colleague Allison Shertzer in the latest issue of Economic Insights. www.philadelphiafed.org/-/media/frbp...
Bryan Stuart has a review of the literature on gender gaps in the labor market in the latest issue of Economic Insights. www.philadelphiafed.org/-/media/frbp...
My colleague Jeff Brinkman on the increasing importance of amenities in regional dynamics. www.philadelphiafed.org/-/media/frbp...
Many of the best papers these days including exclamation points in their titles, you heard it here first