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

While the economics profession largely supports free trade, this consensus hasn't effectively persuaded the public and policymakers. The economic argument for free trade rests on theoretical models of comparative advantage and microeconomic evidence demonstrating the inefficiencies and welfare losses from protectionism. However, public discourse often focuses on macroeconomic indicators like GDP, and the existing literature lacks sufficient empirical macro analysis to address this gap. This paper investigates whether aggregate data can illuminate the costs of tariff increases. Existing literature offers limited conclusive macroeconomic evidence, with many studies finding small or insignificant effects. This study counters this by employing a more comprehensive dataset covering 150+ countries and over five decades to analyze the impact of import tariffs on GDP growth. The analysis aims to provide a broad macroeconomic perspective using traditional macro approaches and a large dataset, offering a crucial missing piece in the ongoing debate. The researchers anticipated that using this extended dataset might yield more precise estimates and better inferences regarding the macroeconomic impacts.

Early debates on import protection trace back to Britain's tariff imposition during the Great Depression, aiming to reduce unemployment. Some studies supported this by showing faster growth in protectionist countries during the 19th century. However, Eichengreen (1981) demonstrated that while tariffs might boost short-term output and employment, long-term production could decline. Subsequent research yielded mixed results, with many finding limited or no negative effects from tariffs. Ostry and Rose (1992) highlighted the theoretical ambiguity surrounding the output effects of tariffs, emphasizing the influence of factors like timing, duration, real wages, exchange rates, elasticities, and institutional factors. Their empirical work found no significant impact of tariff changes on exchange rates, trade balances, or output. The 1990s and 2000s saw a focus on trade's impact on growth, generally finding a positive relationship, although this was debated. Microeconomic studies using industry-level data provided further insights into the impact of trade policies, but their country and sector-specific focus limited generalizability. Recent studies, spurred by renewed interest in tariffs, mainly used disaggregated sectoral data for a few countries, largely the United States, demonstrating welfare losses but again lacking broader scope. This research addresses this gap by examining broad macroeconomic evidence on the output effect of tariffs.

The study utilizes two approaches to assess the effect of tariffs on output growth: (1) visual analysis of growth trends following tariff hikes, and (2) formal econometric analysis using vector autoregression (VAR). Data Collection: The researchers compiled tariff data from multiple sources, including the IMF Research Department, World Integrated Trade Solution (WITS), and World Development Indicators (WDI). The dataset encompasses 151 countries (34 advanced, 117 developing) from 1963 to 2014. Data cleaning involved extensive checks to ensure accuracy and identify spurious tariff changes using cross-referencing with policy reports. Descriptive statistics and visualizations illustrated tariff trends across different income groups. Stylized Facts: The researchers created graphs illustrating the evolution of output growth following substantial tariff increases. These were “residualized” growth rates, which account for the influence of country and time effects. The visualization showed the average residualized growth for countries experiencing tariff increases above one and three standard deviations. This provided an initial assessment of the relationship between tariff increases and output growth. VAR Analysis: The study employed a VAR model to analyze the dynamic response of output to tariff changes. The VAR included variables such as output growth, tariff changes, real effective exchange rate changes, and trade balance changes. A Cholesky decomposition was used to identify orthogonal shocks. The VAR results showed the dynamic response of output to a one standard deviation tariff increase over a five-year horizon. Robustness checks were conducted using alternative model specifications (e.g., local projections and instrumental variables), which verified that the results remained consistent.

The study's key findings strongly suggest a significant negative impact of tariffs on output growth. Both the stylized fact analysis and the VAR model consistently indicated a detrimental effect. Stylized Fact Analysis: The analysis of residualized output growth following tariff increases showed a persistent negative effect, with the magnitude of the decline increasing with the size of the tariff hike. For instance, after four years, tariff increases exceeding three standard deviations were associated with a 1.5% annual decline in output growth. VAR Analysis: The VAR model estimated that a one-standard deviation increase in tariffs (roughly 3.6 percentage points) led to a 0.4% decline in output five years later. While the initial impact wasn't statistically significant, the effect accumulated over time, becoming both economically and statistically significant after four years. The negative effect persisted beyond the five-year horizon. Robustness: The negative relationship between tariffs and output growth remained robust across multiple model specifications, including local projections and instrumental variables approaches. Further robustness checks involving variations in key regressors, samples, and the inclusion of control variables (e.g., crisis dummies, political regime, M2 growth, real exchange rate shocks) confirmed the findings. Channels: A companion paper explored potential channels through which tariffs affect output. These include reduced labor productivity, higher imported input costs, and real exchange rate appreciation. The negative effect on labor productivity was particularly substantial (a 0.9% decline after five years).

The findings strongly support the view that tariffs have substantial and persistent negative effects on output growth. This contradicts some previous research which found minimal or insignificant impacts. The use of a comprehensive dataset spanning a long period and many countries likely explains this difference. The results directly challenge the assertion that the macroeconomic effects of tariffs are minimal, providing robust evidence of the costs associated with protectionist policies. The substantial and persistent nature of the negative impacts suggests significant macroeconomic risks associated with protectionism. The relatively muted impact of the recent trade wars on global growth, as compared to the findings of this study, is plausibly explained by the countervailing effects of monetary policy easing by major economies. However, the study’s results clearly suggest that the benefits of such policies may be limited by the negative consequences of protectionism.

This research offers compelling evidence that increasing import tariffs leads to a significant and sustained reduction in output growth. The study's robustness across various econometric techniques and its large dataset provide a strong basis for policymakers to reconsider protectionist policies. Future research could explore the distributional consequences of tariffs and the interaction between tariffs and other macroeconomic policies.

While the study utilizes a large dataset and various robust econometric techniques, it is important to acknowledge some limitations. Firstly, the analysis focuses on tariffs as the primary trade barrier; the potential impact of non-tariff barriers remains a topic for further investigation. Secondly, while the study considers potential mediating factors, it may not capture all channels through which tariffs affect economic growth. Finally, the analysis is correlational, not causal, implying further research may be necessary to confirm that the observed relationship truly reflects a causal effect.