Are tariffs bad for growth? Yes, say five decades of data from 150 countries

The paper investigates whether increases in import tariffs affect aggregate economic performance, focusing on GDP growth. Motivated by the gap between strong theoretical and micro-level arguments for free trade and the paucity of macro-level empirical evidence, the authors assemble a large cross-country, long-span dataset to assess the macroeconomic costs of protectionism. They aim to determine if raising tariffs boosts or reduces GDP and by how much. The study finds that tariff increases are associated with economically and statistically significant and persistent declines in output growth, suggesting macro-level costs from protectionism.

The paper reviews historical and contemporary research on the macroeconomic effects of tariffs. Early debates linked British interwar tariffs to unemployment reduction, and some 19th-century evidence suggested protectionist countries grew faster (Bairoch, 1972; O'Rourke, 2000). Eichengreen (1981) found tariffs could raise output in the short run but reduce it in the long run. Other studies found limited or no macro effects (Krugman, 1982; Reitz & Slopek, 2005). Ostry and Rose (1992) showed no theoretical presumption about tariff effects on output and empirically found no significant effects using VARs. Later work examined trade openness and growth (e.g., Sachs & Warner, 1995; Dollar, 1992; Feyrer, 2009), with debates on measurement (Rodriguez & Rodrik, 2001; Temple, 2000), and moved to micro evidence on trade policy impacts (Amiti & Konings, 2007; Topalova & Khandelwal, 2011). Recent tariff studies concentrate on sectoral data for a few countries, notably the U.S. (Amiti et al., 2019; Fajgelbaum et al., 2019), limiting generalizability. Hence, broad macro evidence on tariffs’ output effects remains limited, which this paper addresses.

Data: Annual macroeconomic panel of 151 countries (34 advanced, 117 developing) spanning 1963–2014. Tariffs are compiled from multiple sources (IMF Research Department reform dataset extended with WITS and WDI), aggregated from product-level to country-level using import-share weights (ad valorem). Extensive data checks validate large tariff changes with country and policy reports. Descriptive statistics and evolution of tariffs are reported, showing broad declines over time but substantial variation, with 40% tariff increases and 53% decreases in the sample. Other macro variables include real GDP (WEO/WDI), real effective exchange rate (IMF INS), trade balance (WEO/WDI), employment/unemployment, and controls. Empirical approach: Two complementary strategies. (1) Stylized facts: Compute residualized annual real GDP growth by regressing growth on country and time fixed effects; then average residualized growth around episodes of substantial tariff hikes (≥1 or ≥3 standard deviations; 1 sd ≈ 3.6 percentage points). (2) VAR analysis: Estimate a panel VAR to trace dynamic responses of output to tariff changes. Identification uses a Cholesky decomposition ordering: change in log output (growth), change in tariff, change in log REER, change in trade balance (% of GDP), extracting tariff shocks orthogonal to contemporaneous activity. Impulse responses are evaluated up to five years. Robustness: Results are robust to alternative orderings, local projections, and instrumental variables approaches, as well as to variations in sample, controls (e.g., crisis dummies, political regime, M2 growth), and contemporaneous REER shocks.

- Tariff increases have economically and statistically significant negative effects on output, which persist for years and grow with the size of the tariff increase. - Stylized facts: Residualized GDP growth is negative for at least four years after tariff hikes; after two years, residualized growth is about −0.4% for 1-sd hikes and −0.8% for 3-sd hikes. After four years, 3-sd tariff increases are associated with around −1.5% annual output growth. - VAR results: A one-standard-deviation tariff increase (≈3.6 percentage points) leads to a cumulative decline in output of about 0.4% five years later; effects become statistically significant within four years and continue to build. - Channels (from companion analysis): Tariffs reduce labor productivity by about 0.9% after five years and marginally increase unemployment; they raise imported input costs, appreciate the real exchange rate, and have a small, statistically insignificant net effect on the trade balance. - Data characteristics: Despite a general downward trend in tariffs, there is substantial variation; 40% of observations are tariff increases (mean +1.7 ppt, sd 3.3), 53% are decreases (mean −1.8 ppt, sd 3.4).

The findings directly address the core question: tariff increases reduce GDP growth in a persistent and economically meaningful way. The adverse growth effects are larger for bigger tariff hikes and remain evident several years after imposition. This macro-level evidence complements micro-level findings of efficiency and welfare losses and suggests that tariffs harm aggregate performance through multiple mechanisms: reduced allocative efficiency and labor productivity, higher production costs from pricier imported inputs, real exchange rate appreciation that undermines competitiveness, and intertemporal demand shifts. The results are robust across identification strategies and controls, reinforcing their relevance for policy debates. In the context of recent trade tensions, the slowdown in global activity is consistent with these mechanisms; the absence of crisis-level outcomes likely reflects offsetting accommodative macroeconomic policies rather than benign tariff effects.

Using five decades of data for 151 countries, the study provides broad macroeconomic evidence that tariff increases depress GDP growth in a persistent manner, with effects that scale with the size of the tariff change. The analysis, based on harmonized tariff measures and standard empirical macro methods, fills a gap in the literature by documenting aggregate costs of protectionism. Given that nontariff barriers likely impose even greater costs, the estimates here may represent a lower bound on the macroeconomic impact of protectionist policies. Future research could extend horizons beyond five years, incorporate comprehensive measures of nontariff barriers, explore heterogeneity across country characteristics and exchange rate regimes, and further unpack mechanisms using linked industry- and firm-level data.

- The tariff dataset, while extensively validated, is a collation from multiple sources and forms an unbalanced panel; measurement error and comparability issues cannot be fully ruled out. - The main macro analysis focuses on aggregate GDP growth and a five-year horizon; longer-term effects and distributional or sectoral heterogeneity are not fully explored here (examined in a companion paper). - Identification relies on a Cholesky decomposition in a VAR framework and complementary local projection/IV approaches; as with all non-structural macro methods, causal interpretation depends on these identification assumptions. - The study focuses on tariff measures; it does not comprehensively capture nontariff barriers, which may entail additional or larger effects.