Pandemics and protectionism: evidence from the "Spanish" flu

The study asks whether and to what extent major health emergencies affect international macroeconomic policy, focusing on trade policy. Motivated by speculation that COVID-19 could accelerate isolationist trends, the authors turn to historical evidence from the 1918–1920 influenza pandemic to quantify its independent effect on protectionism in the 1920s, apart from the well-known consequences of World War I. Using cross-country panel data, they hypothesize that countries experiencing higher pandemic mortality adopted more restrictive trade policies after 1918. They preview evidence consistent with this hypothesis: countries hardest hit by excess mortality imposed higher tariffs in the 1920s even after controlling for war participation, with divergence from pre-pandemic parallel trends persisting until the Great Depression.

The paper situates its inquiry in literature identifying World War I as a watershed ending the late-19th-century era of globalization (e.g., O'Rourke and Williamson). It reviews mechanisms through which WWI fostered protectionism: destruction, fiscal strain, trade disruptions, postwar reconversion pressures, political changes (franchise extensions, labor parties), and nationalism. It also highlights institutional efforts and failures to maintain openness (League of Nations, Wilson’s Fourteen Points) and interwar policy dynamics, including later 1930s protectionism and the role of exchange-rate regimes (Eichengreen and Irwin). The authors note scant prior work on macroeconomic impacts of the 1918 flu, but draw on public health literature showing non-pharmaceutical interventions’ effects and on narratives linking pandemics to inward-looking attitudes and revenue needs. They provide primary-source evidence from U.S. Congressional debates (1919–1922) and Australian parliamentary debates connecting influenza experiences to arguments for protection, industrial self-sufficiency, and regulation. This background motivates testing whether pandemic severity independently contributed to tariff increases in the 1920s.

Data: The core sample comprises 12 European countries with annual tariff data (import-weighted average ad valorem tariffs, AVE) from Lampe and Sharp (2013), 1900–1939, and one-time country measures of 1918–1919 all-cause excess mortality from Ansart et al. (2009). War intensity is proxied by WWI battle deaths from the Correlates of War (Sarkees and Wayman, 2010). An extended sample uses Barro et al. (2020) mortality rates (flu-specific where available, otherwise excess mortality) and war death rates for 29 countries, expanding geographic coverage and including non-belligerents. The authors discuss AVE limitations (specific tariffs’ price-sensitivity, some interpolations, non-tariff measures) and argue that postwar European non-tariff restrictions were limited relative to tariffs.

Empirical strategy: A difference-in-differences (DiD) design estimates the average treatment effect of pandemic severity on tariffs. Treatment intensity is ln(excess deaths) at the country level; timing is post-1918. Baseline specification includes country and year fixed effects and clusters errors by country:

• tariff_it = β0 + β1 after1918_t × ln(excessdeaths_i) + β2 after1918_t + c_i + θ_t + ε_it. To account for war confounding, a parallel model replaces ln(excessdeaths) with ln(1 + battledeaths) and a combined model includes both interactions simultaneously. The post period covers 1919–1939, allowing effects to materialize while the pandemic was still ongoing and aligning with the war’s end.

Event-study tests: Flexible DiD models interact ln(excessdeaths) or ln(battledeaths) with year dummies to assess pre-trends and the evolution of effects, with 1918 as the base year. In the extended sample, analogous models use flu death rates (flurate) and battle death rates (battlerate), including a specification restricted to non-participants in WWI to better isolate pandemic effects. Robust standard errors are clustered at the country level throughout.

• Core DiD results (12 European countries, 1900–1939): A 1% increase in excess deaths is associated with a 0.017 percentage point increase in tariffs after 1918. Interpreted at observed dispersion, one standard deviation higher excess mortality implies 0.022 percentage points higher tariffs, around one third of a standard deviation of the tariff variable. WWI severity also matters: a 1% increase in battle deaths is associated with a 0.004 percentage point increase in tariffs; one standard deviation more battle deaths implies 0.026 percentage points higher tariffs. In the combined regression, both pandemic and war effects remain positive and statistically significant, with the pandemic’s coefficient stronger than that for battle deaths. Standard errors are clustered by country, and coefficients are significant at conventional levels (p < 0.01 for pandemic effect in combined model).
• Event-study evidence shows no significant pre-trends prior to 1919 for either ln(excessdeaths) or ln(battledeaths). Coefficients become significantly positive after 1918, with pandemic effects fading earlier; by the early 1930s, both effects dissipate, consistent with the Great Depression’s dominant influence on protectionism thereafter.
• Extended sample (29 countries): Using death rates, results mirror the core sample: positive post-1918 effects of both flu and battle mortality on tariffs. In a subsample of countries that did not participate in WWI, flu mortality still predicts higher tariffs post-1918, supporting an effect independent of the war.
• Historical primary-source evidence (U.S. Congress; Australian Parliament) documents contemporaneous links drawn between influenza experiences, industrial self-sufficiency (e.g., drugs, coal-tar products), regulation, and protectionist or isolationist sentiments, corroborating plausible mechanisms.

The findings demonstrate that the 1918–1919 influenza pandemic had an independent, statistically and economically meaningful effect on countries’ moves toward protectionism in the 1920s, beyond the well-documented impact of WWI. The DiD results with robust pre-trend tests strengthen a causal interpretation under standard assumptions. Mechanistically, the study highlights two likely channels: (1) fiscal needs to fund health and social responses in contexts of limited state capacity, making customs duties an attractive revenue source; and (2) inward-oriented attitudes and suspicion of foreigners following the pandemic, increasing political tolerance for tariffs. The war and pandemic were interrelated (e.g., disease spread by troop movements), but the extended-sample evidence from non-belligerents and combined models suggests the pandemic’s role was not merely a proxy for wartime damage. The temporal decay of effects by the early 1930s aligns with the onset of the Great Depression, when macro-financial constraints and exchange-rate regimes overshadowed earlier determinants. These historical insights bear on contemporary policy debates: without careful design, public health crises may catalyze longer-run isolationist trade policies, potentially offsetting domestic benefits of relaxed restrictions by harming international openness and economic recovery.

The paper contributes evidence that the Spanish flu, not only WWI, helped drive higher tariffs in the 1920s. A one standard deviation increase in pandemic excess mortality raised subsequent tariffs by roughly 0.022 percentage points, about one third of a tariff standard deviation, with effects emerging after 1918 and dissipating by the early 1930s. The results are robust across core and extended samples and persist when accounting for war mortality. The authors caution that pandemics can have enduring policy repercussions toward protectionism, urging today’s policymakers to consider the international macroeconomic consequences of health policy choices to avoid repeating interwar patterns. They call for future research to unpack mechanisms—especially the roles of public health policy, state capacity, and political attitudes—and to link pandemic responses more directly to subsequent trade policy formation.

• Measurement of protection: AVE tariffs (customs revenue over imports) can be biased by specific tariffs and price movements; some series involve interpolation (e.g., Germany 1914–1924) and source changes. Non-tariff barriers are imperfectly captured, though argued to be limited in the 1920s outside special cases.
• Identification challenges: Pandemic severity and war impacts are correlated (e.g., returning soldiers). Although combined models and non-belligerent subsamples mitigate confounding and event studies support parallel trends, residual endogeneity cannot be fully ruled out.
• Sample constraints: The core sample is small (12 European countries). The extended sample increases coverage but mixes flu-specific and excess-mortality measures and includes some extrapolated timings, introducing measurement error and heterogeneity.
• Mechanisms unobserved: Proposed channels (fiscal needs, attitudes toward foreigners, health system expansion) are difficult to observe directly in available data, limiting structural interpretation.
• External shocks: The Great Depression imposes time-varying confounds that overwhelm earlier effects by the 1930s, restricting the window for inference.