## 6. The Effect of Tariffs using Industry-level Data ## Source details **Canonical URL:** [6. The Effect of Tariffs using Industry-level Data](https://www.imf.org/-/media/files/publications/wp/2019/wp1909.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2019/wp1909.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2019/wp1909.pdf.json) --- ### Introduction and scope - Objective: estimate macroeconomic and distributional effects of changes in tariff rates using panel local-projection impulse responses. - Sample coverage: an unbalanced panel of 151 countries (34 advanced and 117 developing) over 1963 through 2014. - Focus variables: GDP, labor productivity (GDP/employment), unemployment rate, Gini coefficient (inequality), real effective exchange rate, and trade balance (percent of GDP). - Time horizon for results presentation: truncated at five years after tariff shock. ### Empirical methodology - Primary estimation method: local projection method (Jordà, 2005). Baseline specification (equation (1)): - y_{i,t+k} - y_{i,t-1} = α_i + γ_t + β ΔT_{i,t} + ν X_{i,t} + ε_{i,t} - Key regressor: ΔT_{i,t} = change in tariff rate. - Controls X include two lags of changes in the dependent variable, the tariff, log output, log real exchange rate, and trade balance (percent of GDP). - Fixed effects: country fixed effects {α_i} and time fixed effects {γ_t}. - Instrumental variable (IV) approach: - Instrument I_{i,t} = Σ_{j=1..5, j≠i} ΔT_{j,t} · w_{i,j,t} - w_{i,j,t} = (Export_{i,j,t} + Import_{i,j,t}) / (Export_{i,t} + Import_{i,t}) - Instrument based on weighted-average changes in tariffs of top 5 trading partners; first-stage reported as “strong”. - Industry-level specification (equation (1’)): - y_{j,i,t+k} - y_{j,i,t-1} = α_{ij} + γ_{it} + ρ_{jt} + β^I ΔT^I_{j,i,t} + β^O ΔT^O_{j,i,t} + ν X_{j,i,t} + ε_{j,i,t} - Distinguishes input tariffs (T^I) and output tariffs (T^O). - Input tariffs computed as weighted average of output tariffs in supplier sectors using OECD input-output tables (Amiti and Konings (2007); Topalova and Khandelwal (2011)). - Industry panel: unbalanced panel with 16 manufacturing sectors for 39 countries over 1991-2014 (UNIDO INDSTAT2 2016, ISIC Rev. 3). ### Key aggregate findings - Definition of a one-standard-deviation tariff shock: a 3.6 percentage point increase in the tariff rate. - Output and productivity: - A one standard deviation (3.6 percentage points) tariff increase leads to a decrease in output of about .4% five years later. - Labor productivity cumulatively declines by about .9% after five years. - Per Table 6, the response to a 1 percentage point increase in tariffs: output declines by 0.119 percent at t = 5 (table reports responses at annual horizons t = 0..5). - Unemployment and inequality: - Tariff increases lead to a small (statistically marginal) increase in unemployment in the medium term. - Tariff increases raise inequality (Gini); effect becomes statistically significant two years after the tariff change. - Transmission channels: - Real exchange rate: tariffs tend to appreciate the real exchange rate; effect statistically significant in the short term. - Trade balance: net effects on trade balance are small and statistically insignificant. - Consumption and inflation: - Consumption: tariff increases lead to a statistically significant decline in consumption of around 0.4 percent after four years. - Inflation: higher tariffs lead to an increase in inflation after two years; inflation effect stronger during expansions than in recessions. ### Heterogeneity and nonlinearity - Asymmetry of increases vs decreases: - Sample composition: ~40% of observations are tariff rises (mean 1.7ppt, standard deviation 3.3), ~53% are tariff falls (mean -1.8ppt, standard deviation 3.4). - Tariff increases hurt output and productivity more than tariff decreases help them (asymmetric effects). - Advanced economies vs EMDEs: - Tariff increases have more adverse effects in advanced economies. - Example: for advanced economies output declines by about 1% after four years, compared to the .4% decline in the baseline. - Business cycle dependence: - Tariff increases during expansions generate larger medium-term losses in output and productivity than during recessions. - Tariff increases in expansions can raise inflation and induce contractionary monetary responses, magnifying negative output effects. ### Industry-level (sectoral) findings - Input vs output tariffs (one-standard-deviation equivalents): - Input tariff increase equivalent: about 0.4 ppt. - Sectoral output: decline of about 6.4% five years after the input tariff hike (statistically significant). - Sectoral productivity: decline of about 3.9% five years after the input tariff hike (statistically significant). - Output tariff increase equivalent: about 2.0 ppt. - Sectoral output: increases by 3.1% in five years (statistically positive). - Sectoral productivity: positive but not statistically significant. - Interpretation: the negative macroeconomic effects of tariff increases largely stem from increases in input tariffs that raise production costs and reduce sectoral output and productivity. ### Robustness checks and endogeneity concerns - Transformations of tariff regressor: percent change of tariffs and lagged tariffs produce results consistent with baseline. - Alternative estimation strategies: - VAR-style with Cholesky ordering. - Controlling for contemporaneous changes in trade balance and real exchange rate. - Controlling for expected future GDP growth (IMF WEO forecasts). - IV approach using top-5 trading-partner weighted tariff changes: IV estimates are noisier but produce even larger output declines within five years. - Sample perturbations tested (results persist across): - Dropping series with gaps (<20 consecutive years), dropping high inflation episodes (>100%), dropping small population countries (<1 million), dropping outliers (bottom/top 1% residuals), restricting sample to years after 1979, dropping high tariff episodes (>66 percent), dropping Americas, dropping Asian and Sub-Saharan African economies. - Conclusion from robustness: baseline results are reasonably robust to transformations, estimation techniques, IV strategy, and sample variations. ### Policy implications and substantive conclusions - Tariff increases have adverse domestic macroeconomic and distributional consequences: - lower output and productivity, - higher unemployment and inequality, - lower consumption in the medium term. - Tariff increases do not reliably improve the trade balance; appreciation of the real exchange rate offsets import reduction effects. - Negative effects are larger when tariffs are raised during expansions, in advanced economies, and when tariffs rise rather than fall. - Industry evidence highlights the importance of input tariffs: protecting inputs can substantially reduce sectoral output and productivity, amplifying aggregate costs. - Overall message: empirical evidence from this wide panel supports the conventional policy case for liberal trade and cautions against using tariffs as macroeconomic policy, particularly in advanced economies during cyclical upturns. ### Appendix I — Three Episodes of Tariff Hikes - Overview: three cases where tariffs rose in the sample—Denmark, Colombia, and India—considered typical; primary causes related to the balance of payments. - Denmark 1971: - Event: ten percent import surcharge on two-thirds of imports introduced Oct 19, 1971. - Purpose: strengthen the currency and improve the balance of payments before joining the European Economic Community; explicitly limited in duration. - Reported tariff series in the data: - 1971 7.3% - 1972 12.0% - 1973 12.8% - 1974 11.0% - Colombia 1964: - Trigger: after a central bank float that led to an almost 90% depreciation, imports were largely prohibited for 90 days and tariffs were raised. - Reported tariff series in the data: - 1961 31.5% - 1962 49.8% - 1963 47.3% - 1964 70.4% - 1965 55.3% - 1966 36.1% - India 1972: - Context: fourth five-year plan (1969-73) aimed for self-reliance; tariff spike also related to the Indo-Pakistani war of December 1971. - Reported tariff series in the data: - 1969 59.4% - 1970 58.7% - 1971 63.8% - 1972 91.4% - 1973 76.2% - 1974 64.9% ### Additional aggregate and methodological notes (excerpts) - Welfare and wages: - Canadian and Mexican welfare decline by about 2 percent; U.S. real wage declines by 0.3 and Canada and Mexico by 1.7. - Sample composition and averages: - Around 28 percent of sample observations consist of tariff changes for advanced economies (AEs) and 65 percent consist of tariff changes for other countries. - Average tariff change statistics: - Advanced economies: mean of -.3 percentage points and standard deviation of 2.2. - Other countries: mean of -.3 and standard deviation of 4.1. - Positive-to-negative tariff change ratios: AEs 65 percent; non-AEs 75 percent. - Identification and sensitivity: - Smooth transition autoregressive approach used; results robust to different θ and alternative indicator constructions. - Correlation between changes in tariffs and the smooth transition function F(z_it) is -0.001. - Kleibergen‒Paap rk Wald F statistic for each horizon of the IRF is always higher than the associated Stock-Yogo critical values. - Results robust to inclusion of contemporaneous controls, alternative lag structures, alternate distribution tail checks, and many sample exclusions. - Input vs output tariffs: input tariffs have a more detrimental output effect than output tariffs; similar pattern for productivity. *Source: wp1909 - 6. The Effect of Tariffs using Industry-level Data (https://www.imf.org/-/media/files/publications/wp/2019/wp1909.pdf).* ### References .........................................................................................................29 ### References .........................................................................................................29 ### List of Tables - 1. Data Sources for Country-level Analysis ..........................................................................31 - 2. List of Countries in Country-level Analysis ......................................................................31 - 3. List of Countries in Industry-level Analysis ......................................................................32 - 4. List of Industries ................................................................................................................32 - 5. List of Advanced Economies in Country-level Analysis ...................................................33 - 6. The Aggregate and Distributional Effects of Tariffs .........................................................34 ### List of Figures - 1. The Effect of Tariffs ..........................................................................................................35 - 2. The Effect of Tariffs – Tariff Increases vs. Decreases; Advanced Economies vs. Emerging Markets & Developing Economies ........................................................................................36 - 3. The Effect of Tariffs – Expansions vs. Recessions............................................................37 - 4. Robustness for Output........................................................................................................38 - 5. Robustness for Productivity ...............................................................................................39 *Source: wp1909 - References .........................................................................................................29* ### 6. The Effect of Tariffs using Industry-level Data .................................................................40 ### 6. The Effect of Tariffs using Industry-level Data ### Introduction and scope - Objective: estimate macroeconomic and distributional effects of changes in tariff rates using panel local-projection impulse responses. - Sample coverage: an unbalanced panel of 151 countries (34 advanced and 117 developing) over 1963 through 2014. - Focus variables: GDP, labor productivity (GDP/employment), unemployment rate, Gini coefficient (inequality), real effective exchange rate, and trade balance (percent of GDP). - Time horizon for results presentation: truncated at five years after tariff shock. ### Empirical methodology - Primary estimation method: local projection method (Jordà, 2005). Baseline specification (equation (1)): - y_{i,t+k} - y_{i,t-1} = α_i + γ_t + β ΔT_{i,t} + ν X_{i,t} + ε_{i,t} - Key regressor: ΔT_{i,t} = change in tariff rate. - Controls X include two lags of changes in the dependent variable, the tariff, log output, log real exchange rate, and trade balance (percent of GDP). - Fixed effects: country fixed effects {α_i} and time fixed effects {γ_t}. - Instrumental variable (IV) approach: - Instrument I_{i,t} = Σ_{j=1..5, j≠i} ΔT_{j,t} · w_{i,j,t} - w_{i,j,t} = (Export_{i,j,t} + Import_{i,j,t}) / (Export_{i,t} + Import_{i,t}) - Instrument based on weighted-average changes in tariffs of top 5 trading partners; first-stage reported as “strong”. - Industry-level specification (equation (1’)): - y_{j,i,t+k} - y_{j,i,t-1} = α_{ij} + γ_{it} + ρ_{jt} + β^I ΔT^I_{j,i,t} + β^O ΔT^O_{j,i,t} + ν X_{j,i,t} + ε_{j,i,t} - Distinguishes input tariffs (T^I) and output tariffs (T^O). - Input tariffs computed as weighted average of output tariffs in supplier sectors using OECD input-output tables (Amiti and Konings (2007); Topalova and Khandelwal (2011)). - Industry panel: unbalanced panel with 16 manufacturing sectors for 39 countries over 1991-2014 (UNIDO INDSTAT2 2016, ISIC Rev. 3). ### Key aggregate findings - Definition of a one-standard-deviation tariff shock: a 3.6 percentage point increase in the tariff rate. - Output and productivity: - A one standard deviation (3.6 percentage points) tariff increase leads to a decrease in output of about .4% five years later. - Labor productivity cumulatively declines by about .9% after five years. - Per Table 6, the response to a 1 percentage point increase in tariffs: output declines by 0.119 percent at t = 5 (table reports responses at annual horizons t = 0..5). - Unemployment and inequality: - Tariff increases lead to a small (statistically marginal) increase in unemployment in the medium term. - Tariff increases raise inequality (Gini); effect becomes statistically significant two years after the tariff change. - Transmission channels: - Real exchange rate: tariffs tend to appreciate the real exchange rate; effect statistically significant in the short term. - Trade balance: net effects on trade balance are small and statistically insignificant. - Consumption and inflation: - Consumption: tariff increases lead to a statistically significant decline in consumption of around 0.4 percent after four years. - Inflation: higher tariffs lead to an increase in inflation after two years; inflation effect stronger during expansions than in recessions. ### Heterogeneity and nonlinearity - Asymmetry of increases vs decreases: - Sample composition: ~40% of observations are tariff rises (mean 1.7ppt, standard deviation 3.3), ~53% are tariff falls (mean -1.8ppt, standard deviation 3.4). - Tariff increases hurt output and productivity more than tariff decreases help them (asymmetric effects). - Advanced economies vs EMDEs: - Tariff increases have more adverse effects in advanced economies. - Example: for advanced economies output declines by about 1% after four years, compared to the .4% decline in the baseline. - Business cycle dependence: - Tariff increases during expansions generate larger medium-term losses in output and productivity than during recessions. - Tariff increases in expansions can raise inflation and induce contractionary monetary responses, magnifying negative output effects. ### Industry-level (sectoral) findings - Input vs output tariffs (one-standard-deviation equivalents): - Input tariff increase equivalent: about 0.4 ppt. - Sectoral output: decline of about 6.4% five years after the input tariff hike (statistically significant). - Sectoral productivity: decline of about 3.9% five years after the input tariff hike (statistically significant). - Output tariff increase equivalent: about 2.0 ppt. - Sectoral output: increases by 3.1% in five years (statistically positive). - Sectoral productivity: positive but not statistically significant. - Interpretation: the negative macroeconomic effects of tariff increases largely stem from increases in input tariffs that raise production costs and reduce sectoral output and productivity. ### Robustness checks and endogeneity concerns - Transformations of tariff regressor: percent change of tariffs and lagged tariffs produce results consistent with baseline. - Alternative estimation strategies: - VAR-style with Cholesky ordering. - Controlling for contemporaneous changes in trade balance and real exchange rate. - Controlling for expected future GDP growth (IMF WEO forecasts). - IV approach using top-5 trading-partner weighted tariff changes: IV estimates are noisier but produce even larger output declines within five years. - Sample perturbations tested (results persist across): - Dropping series with gaps (<20 consecutive years), dropping high inflation episodes (>100%), dropping small population countries (<1 million), dropping outliers (bottom/top 1% residuals), restricting sample to years after 1979, dropping high tariff episodes (>66 percent), dropping Americas, dropping Asian and Sub-Saharan African economies. - Conclusion from robustness: baseline results are reasonably robust to transformations, estimation techniques, IV strategy, and sample variations. ### Policy implications and substantive conclusions - Tariff increases have adverse domestic macroeconomic and distributional consequences: lower output and productivity, higher unemployment and inequality, and lower consumption in the medium term. - Tariff increases do not reliably improve the trade balance; appreciation of the real exchange rate offsets import reduction effects. - Negative effects are larger when tariffs are raised during expansions, in advanced economies, and when tariffs rise rather than fall. - Industry evidence highlights the importance of input tariffs: protecting inputs can substantially reduce sectoral output and productivity, amplifying aggregate costs. - Overall message: empirical evidence from this wide panel supports the conventional policy case for liberal trade and cautions against using tariffs as macroeconomic policy, particularly in advanced economies during cyclical upturns. *Source: wp1909 - 6. The Effect of Tariffs using Industry-level Data (https://www.imf.org/-/media/files/publications/wp/2019/wp1909.pdf).* ### Appendix I—Three Episodes of Tariff Hikes ### Appendix I—Three Episodes of Tariff Hikes ### Overview - The appendix describes three cases where tariffs rose in the sample: Denmark, Colombia, and India. - The authors do not claim these episodes represent all tariff variation but consider them typical. - In all three cases, domestic macroeconomic considerations do not appear overwhelming; primary causes are related to the balance of payments. ### Denmark 1971 - Context: Denmark’s newly elected Social Democratic government imposed a ten percent import surcharge on two-thirds of the country’s imports on its first day in office, Oct 19, 1971. - Purpose: Introduced to strengthen the currency and improve the balance of payments before joining the European Economic Community; explicitly limited in duration. - Trigger: Partial response to the “Nixon Shock” of August 1971, which imposed a ten percent surcharge on all dutiable American imports, intended to force a real American depreciation. - Reported tariff series in the data: - 1971 7.3% - 1972 12.0% - 1973 12.8% - 1974 11.0% ### Colombia 1964 - Context: Some protectionist surges in developing countries were triggered by balance of payments issues rather than only import-substitution strategies. - Trigger: In late 1964, the central bank of Colombia floated an over-valued exchange rate, which then depreciated excessively, by almost 90%. Colombia prohibited the free import of almost all foreign products for a 90-day period, pursued a standby package with the IMF, and raised tariffs. - Reported tariff series in the data: - 1961 31.5% - 1962 49.8% - 1963 47.3% - 1964 70.4% - 1965 55.3% - 1966 36.1% ### India 1972 - Context: Indian development was guided by five-year plans; the fourth plan (covering 1969-73) aimed for self-reliance through import-substitution. - Trigger: As part of the plan, tariffs were imposed on all goods other than grain and a few smaller exceptions; the more immediate reason for the protectionist spike was the Indo-Pakistani war of December 1971. - Reported tariff series in the data: - 1969 59.4% - 1970 58.7% - 1971 63.8% - 1972 91.4% - 1973 76.2% - 1974 64.9% *Source: wp1909 - Appendix I—Three Episodes of Tariff Hikes* ### 0.2 percent, and Canadian and Mexican welfare by about 2 percent, with a real wage decline of 0.3 in the US ### wp1909 - 0.2 percent, and Canadian and Mexican welfare by about 2 percent, with a real wage decline of 0.3 in the US ### Key findings - Welfare losses: Canadian and Mexican welfare decline by about 2 percent; U.S. real wage declines by 0.3 and Canada and Mexico by 1.7. - Aggregate sample composition: Around 28 percent of sample observations consist of tariff changes for advanced economies (AEs) and 65 percent consist of tariff changes for other countries. - Average tariff change statistics: - Advanced economies: mean of -.3 percentage points and standard deviation of 2.2. - Other countries: mean of -.3 and standard deviation of 4.1. - Positive-to-negative tariff change ratios: AEs 65 percent; non-AEs 75 percent. ### Methodology and identification - Smooth transition autoregressive approach: Equivalent to Granger and Terävistra (1993); results robust to different values of θ and to substituting F(z_it) with a dummy taking value 1 for F(z_it) greater than 0.5. - Correlation with macro state: No statistically significant correlation between changes in tariffs and the measure of state of economy used; correlation between changes in tariffs and the smooth transition function F(z_it) is -0.001. - Instrument relevance: Kleibergen‒Paap rk Wald F statistic for each horizon of the IRF is always higher than the associated Stock-Yogo critical values. - Treatment of contemporaneous channels: To address potential endogeneity from changes in trade partners’ tariffs affecting domestic output via real exchange rate, estimations control for contemporaneous changes in other control variables, including real exchange rates; results robust and similar to those in Figure 5. - Lag specification: Equation modified to allow all explanatory variables (including changes in the tariff) to enter with a lag; robustness noted. ### Robustness checks and sensitivity - Extreme trade-balance observations: Baseline results robust when including observations with trade balance between +/-50 percent of GDP. - Alternate distribution tail checks: Similar results obtained when top and bottom 5th percentiles of the same distribution are considered. - Alternative orderings: Results are robust to alternative orderings. - Contemporaneous controls by dependent variable: - For regressions on output, productivity, unemployment and inequality: control for contemporaneous changes in trade balance and real exchange rate. - For regression on trade balance (real exchange rate): control only for simultaneous changes in the real exchange rate (trade balance). - Instrumental concerns: Modified specifications controlling for contemporaneous changes in other variables produce results very similar to Figure 5. - Equivalence across specifications: Analogous results for unemployment, inequality, real exchange rate and trade balance reported in Appendix II and Appendix III; baseline results described as basically insensitive. - Input vs output tariffs: Input tariffs have a more detrimental output effect than output tariffs; consistent with prior empirical work (Ahn et al. 2016; Amiti and Konings 2007). Similar results also obtained for productivity. ### Additional notes - Frequency and magnitude: While tariff changes have been less frequent in rich countries, the average magnitude of the changes is similar across AE and non-AE samples. - Appendix references: Appendix I contains examples of major tariff increases in the sample; Figures in Appendix II provide equivalent results for unemployment, inequality, real exchange rate and trade balance; Figures AIII, 1-4 of Appendix III report analogous results for other variables of interest. *Source: wp1909 - 0.2 percent, and Canadian and Mexican welfare by about 2 percent, with a real wage decline of 0.3 in the US* --- _Source: https://www.imf.org/-/media/files/publications/wp/2019/wp1909.pdf_