Source
Juhász, LIT-1489, read in full in NOTE-1122 from the working draft of 17 December 2015 (CIRJE copy): pp. 20–40 and Tables 1–5. The version of record (American Economic Review 108(11), 2018) and the online appendices were not read.
What was actually shown
- Short run. Across 88 departments observed in 1803 and 1812, a larger rise in effective distance to London went with a larger rise in spindles per thousand inhabitants (coefficient 33.11, s.e. 9.78; Table 1). The interquartile difference in the shock predicts a gain about equal to mean 1812 capacity. No pre-trend once market potential is controlled, and no effect on wool spinning, leather tanning, capital-labour ratios or machine vintage (Table 2).
- Long run. Instrumenting 1812 capacity with the shock, one more spindle in 1812 meant about 2–3 more in 1840 and 5–6 more in 1887 (Table 3); 1840 firm productivity rose 0.2–0.5 standard deviations per standard deviation of 1812 capacity (Table 4); industrial value added per capita was higher in 1860 only (Table 5): “The point estimate decreases in size over time and from 1896 onwards, the coefficient is longer distinguishable statistically from zero” (sic, p. 38).
What could have come out otherwise: the blockade’s northern industry could have been Heckscher’s “hothouse” growth and collapsed when British yarn returned. It did not; the southern departments shrank instead.
What this does not say
- That a temporary tariff can be self-limiting. No tariff was chosen: the protection was a side effect of a war blockade, which Napoleon did not intend for cotton. And it was not followed by free trade: “tariff and non-tariff trade barriers were put in place between the two countries” (p. 33). The within-France results compare more and less protected regions under continued national protection.
- That protection pays. The author says the evidence “does not imply a case for infant industry protection in similar settings” (p. 8), and the aggregate gain faded within about forty years.
- That the instrument is strong. “The Kleibergen-Papp F-statistic for the first stage is generally below 10” (p. 35).
- That the mechanism is learning by doing. That is suggested by firm entry, productivity and historical sources, not separated from pecuniary agglomeration.
Connections
- CLAIM-445. The claim’s capability-building contrast case: protection that left a lasting capability. It supports that the contrast is real, and it says nothing about whether protection maintained itself. The petition by large spinners for “a complete ban on English cloth” (p. 13, fn. 7) shows the constituency forming; whether it shaped post-war tariffs is not studied.
- THEORY-885. A historical case of increasing returns locking in an early, accidental advantage.
Note of 2026-10-30: what holds it back
The long-run persistence rests on first stages the author concedes are weak (Kleibergen-Paap F 3–8), and the 2015 draft read is probably the version first submitted, so the AER tables may differ (NOTE-1122). The record has no stronger first stage, no weak-instrument-robust test and no reading of the version of record. If the title is narrowed, it should keep the weak-instrument hedge it now leaves out.