Research

The Effect of Temperature on Fertility in an Industrialising Economy

This paper studies the relationship between temperature and fertility rates. Exploiting exogenous temperature variation across 81 French départements from 1851 to 1911, I examine how changes in income, driven by temperature variations, can lead to either a dominant income effect or substitution effect in fertility decisions. I find that a one standard deviation increase in temperature increases fertility by 19-31\%, corresponding to 4.7-7.5 additional births per 1,000 people. The climatic impact, driven specifically by growing season temperatures, is most pronounced in regions most reliant on agriculture. Higher temperatures negatively affect wheat prices, suggesting increased agricultural productivity. This temperature-price relationship is consistent with the hypothesis that income gains dominate substitution effects on fertility choices. Using a mediation analysis I rule out internal migration, infant mortality, and education as alternative mechanisms.

Vulnerability to Climate Shocks in Early 20th Century Spain 

This paper estimates the causal effect of annual temperature and precipitation shocks on provincial GDP per capita growth in Spain between 1904 and 1934. Following Nath et al. (2024), we extract weather shocks using province-specific autoregressions that separate exogenous variation from serially correlated climate dynamics, applying this identification strategy to a historical sub-national setting for the first time. Using a panel of 48 provinces, we find that a 1°C temperature shock reduces GDP per capita growth by 3.3--3.4 percentage points, more than twice the estimates for modern economies, reflecting the absence of buffering inputs in Spain's organic farming economy. The effect operates primarily through agriculture, with rural GDP falling by 5.6--5.8 percentage points against a modest 0.5--1 percentage point decline in urban GDP. Precipitation shocks have no significant aggregate effect. Both findings conceal substantial geographic heterogeneity rooted in aridity. Temperature effects are concentrated in the driest provinces, where dryland farming systems were most exposed to heat stress. Strikingly, excess precipitation is damaging rather than beneficial in the most arid quartile, reducing growth by 13.2 percentage points per 100mm shock. This aridity result is robust to controlling for income, sectoral structure, and geographic position. We interpret this as reflecting aridity amplifying a broader fragility of dryland cereal systems to high precipitation. Crop-level evidence supports this interpretation: excess precipitation damages dryland cereals in arid provinces while legumes and perennials show positive or neutral responses. The null aggregate precipitation result reflects two mechanisms: the cancellation of opposing crop-level effects within provinces, and the concentration of net negative effects in the most arid provinces where that cancellation breaks down.

Temperature and Wage Dynamics in Early 20th Century Spain

This paper estimates the effect of temperature shocks on wages across nine occupations in early twentieth-century Spain using local projections and province-level panel data from 1915 to 1931. Temperature shocks are extracted as innovations from province-specific autoregressions following Nath, Ramey and Klenow (2024), isolating exogenous weather variation from serially correlated climate dynamics. The results show substantial and persistent wage losses across occupations, with agricultural workers bearing the largest effects. Peak agricultural wages fall by around 0.17 log points at the two-year horizon, and a broader set of urban construction and craft trades are also significantly affected. The contemporaneous effect is negligible across all occupations, with losses emerging from the first year onwards and showing no reversion over a five-year horizon. Three transmission mechanisms are examined and largely ruled out: aggregate demand contraction is inconsistent with simultaneous increases in both rural and urban price indices; the price channel alone cannot account for the losses, as nominal wage adjustment accounts for 65 to 94 percent of the real wage decline across affected occupation-horizon cells; and seasonal labour market integration is rejected by formal tests of equality between peak and off-peak wage responses. The evidence is most consistent with an agricultural supply shock transmitting through an income channel, in which harvest failures reduce rural purchasing power and contract demand for urban goods and services with a lag of one to two years.

Spanish Climate at the Provincial Level: 1500-2000

This paper constructs and characterises a provincial-level dataset of seasonal temperature and precipitation covering all 47 mainland Spanish provinces and the Balearic Islands over the period 1500--2000. Drawing on the gridded climate reconstructions  Luterbacher et al. (2004), Xoplakiet al. (2005), and Pauling et al. (2006),, it extracts area-weighted provincial means for eight seasonal climate variables across 501 years. Principal component analysis reveals a fundamental asymmetry in the spatial structure of Spanish climate. Temperature variation is overwhelmingly driven by a common national signal, with PC1 explaining between 88 and 95 per cent of provincial temperature variance across all four seasons, meaning that temperature shocks are felt almost uniformly across all 48 provinces simultaneously. Precipitation is substantially more spatially fragmented: PC1 explains only 45 per cent of summer precipitation variance, with a structured secondary mode capturing a further 20 per cent. For summer, this secondary mode places the north and northeast in opposition to the south and interior. Two historical episodes validate these structural findings. The Year Without a Summer of 1816 produced a near-uniform cooling signal across all 48 provinces, consistent with the high national coherence of temperature shocks. The late Maunder Minimum drought of 1700--1750 produced a spatially heterogeneous precipitation deficit whose distribution maps almost perfectly onto the summer precipitation PC2 loading structure, confirming that the decomposition captures persistent climatological structure rather than a statistical artefact. These findings have direct implications for the use of Spanish provincial panel data to estimate the economic effects of historical climate shocks: temperature-based estimates are conservative lower bounds on the true causal effect, while summer precipitation provides considerably more scope for cross-sectional identification.