Literature Review: Are Tariff Shocks Deflationary?

We review “What Is a Tariff Shock? Insights from 150 Years of Tariff Policy” by SF Fed Barnichon & Singh (2025). The paper is here.

What the paper does

This paper explores a surprisingly underexplored empirical question: what are the aggregate macroeconomic consequences of tariff shocks? The methodological contribution rests on two parallel identification strategies. First, the authors construct a narrative series of tariff shocks, identifying moments when tariff legislation was driven by long-run political or ideological considerations rather than by the current state of the business cycle. Second, they argue that even the “non-narrative” tariff changes can be treated as quasi-random with respect to the cycle because historically the two major U.S. parties held diametrically opposed views regarding tariffs. Using both the recursive OLS approach and narrative IV, they estimate VARs including tariff changes, inflation, and unemployment, concluding that tariff increases systematically reduce inflation and increase unemployment, both in the U.S. and—using analogous historical episodes—in France and the U.K.

Main results

A tariff hike is followed by lower inflation and higher unemployment, both in the same year and in the year after (Figure 1 shows the main results of the paper). These results hold in the pre–World War II period, in the postwar era where tariff variation is more muted, and when using only the narratively identified shocks. The authors interpret this pattern as evidence that tariff shocks behave more like aggregate demand disturbances than cost-push shocks: a tariff increase depresses activity and reduces inflation instead of raising it. They suggest uncertainty and wealth channels—visible in their extended VAR including stock prices and volatility—as potential mechanisms consistent with the negative inflation response.

Figure 1. Main result of Barnichon and Singh (2025)

A critique of the paper

The empirical exercise is ambitious, but the identification strategy raises some concerns that might limit the strength of the causal interpretation.

The central assumption behind the OLS estimates is that tariff changes are quasi-random with respect to the business cycle because U.S. political parties historically took opposite stances on tariff policy. However, this assumption rests on correlations between unemployment and electoral outcomes that are statistically weak in their regressions, but weak correlation does not imply orthogonality. Elections reflect many factors other than the state of the cycle, and tariff policy is often intertwined with broader agendas (industrial policy, agricultural interests, geopolitical considerations). These determinants may correlate with unobserved economic forces that also drive inflation and unemployment, thus invalidating the claim that policy changes are exogenous. The fact that political turnover is not systematically predicted by unemployment does not imply that tariff policy is conditionally independent of macroeconomic conditions.

Moreover, even if the direction of tariff changes is not systematically correlated with unemployment at the moment of enactment, the magnitude and anticipation of tariff legislation may still be correlated with expected economic conditions. Tariffs are slow-moving policies involving lobbying, negotiation, and long periods of debate. The VAR treats tariff innovations as contemporaneously exogenous, but many tariff changes were anticipated for months or years. Anticipation undermines the recursive identification strategy and contaminates the OLS impulse responses, especially given the annual frequency of the data.

The narrative IV is intended to address these deficiencies, but it has its own limitations. The very small number of narratively identified shocks, together with heterogeneous historical contexts, makes the exclusion restriction highly implausible. Tariff changes motivated by “long-run considerations” cannot be assumed to be uncorrelated with long-run economic conditions or expectations. For instance, the 1890 McKinley Tariff, while politically driven, was enacted in an environment of structural transformation and industrialization that likely had independent effects on inflation dynamics. Similarly, the postwar GATT rounds reflect a deliberate strategy of institutional and geopolitical reconstruction, not a context devoid of macroeconomic relevance. Treating such reforms as shock-like IV events risks attributing to tariffs what may instead be the macroeconomic consequences of broader trade liberalization, geopolitical stabilization, or technological change.

The negative inflation response, which is the signature empirical finding of the paper, also raises conceptual issues. Cost-push channels from tariffs are a robust prediction of most structural models, and the authors’ interpretation relies heavily on secondary variables (stock prices and volatility) to justify the opposite sign. However, both stock prices and volatility may themselves respond endogenously to expectations of slower growth rather than to tariffs per se. The VAR ordering assumes that tariff shocks precede and cause financial market movements, but given the anticipation problem mentioned above, investors may react before tariffs are enacted. The identification scheme thus cannot definitively establish that tariffs cause uncertainty rather than merely coincide with periods of uncertainty for unrelated reasons.

A further identification concern stems from the fact that the VAR does not control for monetary policy, despite the central role interest rates play in shaping inflation and real activity. Because the stance of monetary policy systematically responds to both inflationary pressures and cyclical conditions, omitting interest rates risks conflating the direct effects of tariff changes with the endogenous reaction of the central bank. In practice, a tariff-induced rise in prices or slack would trigger a monetary policy response that subsequently affects inflation and unemployment, thereby contaminating the estimated impulse responses. Without conditioning on the path of interest rates, the VAR cannot separate the causal impact of tariff shocks from the policy reaction function, and the negative inflation response documented in the paper may in fact reflect monetary tightening rather than an intrinsic demand-like effect of tariffs.

Finally, the international evidence, while descriptively consistent, faces even more severe identification challenges. Tariff histories in France and the U.K. are deeply entangled with wars, colonial adjustments, reconstruction, and shifting trade regimes. In these contexts, it is difficult to argue convincingly that tariff movements can be isolated from other macroeconomic or political shocks.

Conclusion

This paper opens an important and overdue empirical conversation about the macroeconomic effects of tariffs. Its historical perspective and use of narrative methods offer valuable descriptive insights. However, the empirical identification is considerably weaker than the authors claim, and the interpretation of tariff shocks as aggregate demand disturbances is far from settled. The main result—tariff hikes reduce inflation—is intriguing but cannot be taken at face value as a causal effect. At a minimum, the evidence is compatible with the idea that tariff changes tend to occur in periods of heightened uncertainty or structural transition, which themselves drive both inflation and unemployment.

Contact

Want something more tailored?

We provide tailored consulting on ad-hoc projects.