Tariffs – taxes on imported goods – likely impose a heavier burden on lower-income households, as these households generally spend more on traded goods as a share of expenditure/income and because of the higher level of tariffs placed on some key consumer goods. This column estimates the tariff burden by income group and by family structure using a new dataset constructed by matching of granular data on trade and consumer spending. The findings suggest that tariffs function as a regressive tax that weighs most heavily on women and single parents.
By Jason Furman, Katheryn Russ and Jay Shambaugh*
The US collects more than $33 billion a year – or roughly 0.2% of GDP – in tariffs, which are taxes on US imports. A common argument in favour of trade protectionism supposes that increasing tariffs imposes a very small cost on many people to protect more concentrated populations within particular industries. Many overlook the fact that like any tax, the tariff burden does not fall uniformly across goods, but falls more heavily on particular goods and the populations that purchase them. Several existing analyses find that from a consumption perspective, low- and middle-income households benefit substantially more from trade than high-income households do, in large part because lower-income households spend more on tradable goods like food and apparel (Fajgelbaum and Khandelwal 2016). In addition, some research has attempted to document that tariffs themselves have a regressive incidence (Gresser 2002, Moran 2014). However, a comprehensive study of the incidence of tariffs themselves has been missing from the literature. We match import duties to standard consumer expenditure data to take a more detailed look and find evidence that low- and middle-income households do, indeed, spend a higher fraction of their income and non-housing expenditure on tariffs. The findings indicate that tariffs act as a regressive tax on American consumers and are distortionary in their variation across products.
The tariff burden as a regressive tax
To estimate the distributional effects of tariffs across households, we constructed a new dataset by matching tariffs collected by US Customs and Border Patrol as reported by the International Trade Administration’s (ITA) Trade Policy Information System (TPIS) database with goods categories in the 2014 Consumer Expenditure Survey (CEX) administered by the US Bureau of Labor Statistics. In total, the dataset we construct matches 381 consumption categories to their relevant tariff. By matching HS codes to CEX categories, we create an average effective tariff rate within each consumption category dividing tariffs collected by imports.1 The calculation will not include the impact of tariffs via intermediate inputs, only from imported consumer goods. The dataset with aggregated consumption categories is available publically for researchers here.2 Details on the construction of the dataset and its interpretation are provided in the Appendix.
For the US, we calculate conservative estimates, assuming that protection via tariffs does not induce domestic producers of similar goods to raise their prices at all. These estimates show that the poorest 10% to 20% of households in the income distribution pay about $95 a year due to tariffs, middle-income households pay roughly $190, and the richest 10% about $500. In the figure below, we also show a range of higher tariff burdens that reflect some impact on prices of domestic goods (see Appendix note 6 for details).
However, the burden is substantially higher for poor households than for the richest relative to their income. The following figures demonstrate the regressivity of tariffs by presenting the tariff burden across income deciles as a percentage of after-tax income and as a percentage of non-housing expenditures. The CEX suffers from well-known data quality issues with respect to the ratio of consumption to income for very low-income households. For this reason, the estimate of the tariff burden relative to after-tax income for the bottom decile should be interpreted with caution. Regardless, the broad pattern of regressivity is clear.
These conservative estimates of the direct tariff burden are large compared to other recent and hotly debated tax policies. Removing the tariff burden would have a considerably larger impact on poor households than the 2001 and 2003 tax cuts, which were estimated to reduce taxes for the bottom quintile of households by $28 to $87 dollars per year. If tariffs were raised by 10 percentage points across the board, the cost of households’ 2014 consumption bundles (all else equal, including exchange rates) would have risen by $301 and $307 for the lowest two income deciles, $611 and $716 for the fifth and sixth income deciles, and $1,462 for the households in the highest income decile, assuming the full tariffs are passed through to consumers and have no additional impact on domestic prices.
Tariffs have arbitrary and unintended consequences for consumers
Economists often think of tariffs in a stylised way, as though they are applied uniformly to broad swaths of the economy, or simply targeted to protect only the most deserving or vulnerable sectors. In fact, there is enormous variation in tariffs at the disaggregate level for reasons that sometimes are due to careful targets, but alternatively may favour some groups or products over others for reasons that are not always obvious. For example, for a family doing their back-to-school shopping, backpacks of man-made fibres carry tariffs of 17.6%; ballpoint pens 0.8 cent each plus an additional 5.4%; non-mechanical pencils and crayons about 4.3%; markers 4%; mechanical pencils 6.6%. The tariff code thus penalises students more for using ballpoint pens and mechanical pencils than markers. Meanwhile, there are no tariffs applied on imports of cross-country snow skis, sailboards, or archery equipment (USITC 2016: Chapters 42, 95).
Within consumption goods, categories like apparel and many types of home furnishings and other household goods, tariffs are sometimes higher on cheaper varieties, making tariffs on these goods more regressive than our benchmark estimates suggest. A look at disaggregated US import data reveals many goods categories with unit values (a proxy for prices) that have a strong negative correlation with statutory tariffs, implying lower-cost goods often face higher tariff rates. Goods are grouped at the HS 4-digit level and correlations are computed across HS 8-digit items. For the HS 4-digit categories classified as consumer goods by the United Nations that have a calculated correlation between tariffs and prices, the tariff scale decreases in prices for roughly half, and the list of categories with strong negative correlations is extremely wide-ranging across the activities of daily life. These goods include adult, child, and infant apparel; waterproof footwear; home furnishings and bedding; olive oil; processed tomatoes; wine; bicycles; camping gear; helmets and protective sporting headgear; small kitchen appliances and humidifiers; clocks, tableware and kitchenware; crafted wooden boxes including caskets.
Sometimes the variation in tariffs just appears to be an artefact, a legacy left from decades of many rounds of negotiations. Bags, suitcases, and other cases can have a completely different tariff ranging from zero to 20% depending on their purpose and materials (USITC 2016: Chapter 42). Tariffs on sports gloves, regardless of the materials used, range from zero to nearly 6% depending on the sport or even the position they are intended for – ice hockey gloves are duty free, but ski gloves have a tariff of up to nearly 6% and golf gloves have a tariff of 4.9%. Batting gloves have a tariff of 3%, while other types of baseball and softball gloves are duty free (USITC 2016: Chapters 42, 61, and 62). Tariffs on nappies made with paper pulp or cellulose are duty free due to the sectoral agreement on pulp and paper products in the Uruguay Round, while tariffs on nappies made with cotton cloth or other textile fibres can face tariffs ranging from 2.8 % to 16% (USITC 2016: Chapter 96). Anderson and Neary (1996 and 2007), Kee et al. (2009), and Kee et al. (2013), for example, discuss different ways to quantify the trade and welfare impact of these types of distortions and how they relate to overall trade restrictiveness.
Tariffs may have a disproportionate impact on single parents, and many goods consumed disproportionately by women face higher tariffs than goods disproportionately consumed by men
In addition, the average effective tariff on many categories of women’s apparel exceeds that for men’s apparel by a substantial margin. The effective tariff is 23% for women’s varieties versus 14% for men’s on suits; 21% for women’s versus 13% for men’s on sweaters, shirts, and tops; 21% for women’s versus 7% for men’s on active sportswear; 15% for women’s versus 10% for men’s pants and shorts; and 13% on women’s undergarments versus 7% on men’s underwear. And there is a ‘tampon tax’ in the tariff schedule: feminine hygiene products made of materials other than paper or cellulose (so cotton, synthetic, or artificial fibres or textiles) are subject to tariffs ranging from 3.6% to 16% (UISTC 2016: Chapter 96). The category which includes both feminine hygiene products and diapers, HS 9619, has an average statutory (MFN) tariff of 7.6% (WTO). Thus, the tariff burden among single parents may be even higher for single mothers than for single fathers.
Based on this initial analysis, it appears tariffs are imposed in a regressive manner – in part because expenditures on traded goods are a higher share of income and non-housing consumption among lower income households, but also due to explicit regressivity within categories. The analysis highlights an underexplored aspect of trade policy and its effects and leaves open a path for subsequent research. More research on this area would be welcome – and the new dataset created for this analysis hopefully will help further some of that research.
Authors’ note: We thank Lydia Cox, Conor Foley, Nataliya Langburd, Willie Powell, and Nirupama Rao for help in crafting this analysis and summary.
For this analysis, we construct a new data set (available here) that combines information from detailed tariff data from the Harmonized Tariff Schedule published by the US International Trade Commission (USITC) and consumption data from the CEX. The matching of the tariff lines to consumption categories allows us to calculate the tariff burden across groups. The dataset with aggregated consumption categories is available publically for researchers here, and the pre-publication tables containing the more detailed categories used in the crosswalk are available upon request from BLS. We encourage other researchers to use these data and or extend and improve the match process.
Tariff data is available at the Harmonized System (HS) code level, which is often more granular (e.g. Crustaceans, molluscs and other aquatic invertebrates, prepared or preserved) than the CEX category (e.g. Canned Fish and Seafood). For each CEX category, we calculated an average tariff rate, dividing total tariffs collected for goods in the expenditure category by total imports of those goods. This is an effective tariff rate. The CEX is a detailed survey of household expenditures with both quarterly and annual components. The consumption categories are in some cases broad, like “Other hardware” or “Other fresh fruit,” but generally are quite specific, for instance distinguishing between bath linens, bedroom linens, and kitchen or dining room linens.3
The effective tariff rate is generally lower than an average of statutory tariff rates or the true economic burden of tariffs because buyers tend to substitute away from goods when tariffs raise the price too much relative to other options. Using effective rates will systematically underestimate the full burden of tariffs on all households, since it adds only the cost of tariffs on goods Americans actually buy, not on goods they might have preferred to buy had there been no tariff. For example, a tariff rate of 1000% may collect no revenue as it is so high that it blocks all imports of a good. Effective tariff rates will not include any loss to consumers arising from the tariff-induced unavailability of this good or increased prices that result from reduced competitive pressures on rival producers.
Businesses selling goods produced outside the US have the option of passing on any tariffs they pay on goods they bring into the US to consumers by raising prices. Foreign firms may absorb part of the tariff burden by lowering their own markups at the factory gate to keep prices competitive after the tariff is applied, so the final tariff incidence may fall in part on both producers and buyers or consumers in the US. For simplicity, we assume that the cost of the tariff is borne entirely by consumers, which is consistent with US Congressional Budget Office (2011) and the US Congress’s Joint Committee on Taxation (2015) estimates of excise tax incidence. In practice, the possibility that some portion of the tariff is borne by the foreign firm is not relevant for our analysis of the distributional impact as it would simply raise or lower the total tariff burden, not change its distribution.5
Since consumers buy a mix of imported and domestic goods, we apply the tariff rate only to the share of goods that are imported, provided to us for key categories of goods by the Office of the United States Trade Representative (USTR). Because domestically-produced goods are substitutes for imported goods subject to tariffs, we also present broader analyses that assume some spillover of price impacts to domestic goods.6
Total tariff revenue collected by the US is more than $33billion, of which $25 billion is collected in categories the UN classifies as consumer goods. The total tariff burden we calculate based on household expenditures (which are largely classified as consumer goods) is over $27 billion. The difference from total tariff revenue reflects both the inability to match some goods, measurement error, and the fact that not all tariff revenue is directly collected on consumer goods. In particular, our analysis will not reflect the fact that tariffs on intermediate goods may raise the prices of downstream domestic consumer goods. In addition, as our tariff burden calculation applies the effective tariff rate to domestic expenditures, not import values, it will include retail markups or sales taxes which would increase the measured burden relative to tariff revenue. On net, though, the tariff burden we distribute across groups is quite close to total tariff revenue on consumer goods, suggesting it is a reasonable proxy for the direct burden.
*About the authors:
Jason Furman, Chairman, Council of Economic Advisers
Katheryn Russ, Associate Professor of Economics at University of California, Davis; Faculty Research Fellow, NBER
Jay Shambaugh, Member of the Council of Economic Advisers (on leave from George Washington University)
An, L and J Wang (2012), “Exchange Rate Pass-through: Evidence Based on Vector Autoregression with Sign Restriction.” Open Economies Review 23(2): 359-380 (April).
Anderson, J and P Neary (1996), “A New Approach to Evaluating Trade Policy.” Review of Economic Studies 63(1): 107-125 (January).
Anderson, J and P Neary (2007), “Welfare Versus Market Access: the Implications of Tariff Structure for Tariff Reform.” Journal of International Economics, vol. 71(2) (March), pp. 627–49.
Arkolakis, C, A Costinot, D Donaldson, and A Rodriguez-Clare (2015), “The Elusive Pro-Competitive Effects of Trade.” NBER Working Paper No. 21370.
Auer, R and R Schoenle (2016), “Market Structure and Exchange Rate Pass-through.” Journal of International Economics 98.
Congress of the United States Joint Committee on Taxation, Fairness and Tax Policy (2015), “Fairness and Tax Policy.”
Congressional Budget Office (2011), “The Distribution of Household Income and Federal Taxes.”
De Blas, B and K N Russ (2015), “Understanding Markups in the Open Economy.” American Economic Journal: Macroeconomics 7(2): 157-180.
Edmond, C, V Midrigan, and D Xu (2015), “Competition, Markups and the Gains from Trade.” American Economic Review, forthcoming.
Elmendorf, D W et al. (2008), “Distributional Effects of the 2001 and 2003 Tax Cuts: How Do Financing and Behavioral Responses Matter?” National Tax Journal: 365-380
Fajgelbaum, P D and A K Khandelwal (2016), “Measuring the Unequal Gains from Trade.” The Quarterly Journal of Economics, 131 (3): 1113-1180
Feenstra, R C and D Weinstein (2016), “Globalization, Competition, and U.S. Welfare.” Forthcoming, Journal of Political Economy.
Gale, W G and P R Orszag (2004), “Bush Administration Tax Policy: Distributional Effects.” Tax Notes 104.14: 1559-1566.
Gresser, E (2002), “Toughest on the Poor: America’s Flawed Tariff System.” Foreign Affairs 81: 9.
Hayakawa, Kazunobu, and I. T. O. Tadashi. 2015. Tariff Pass-through of the World-wide Trade: Empirical Evidence at Tariff-line Level. No. DP-2015-34.
Irwin, D A (2014), “Tariff Incidence: Evidence from U.S. Sugar Duties 1890-1930.” NBER Working Paper No. 20635 (October).
Jara, C C M and A M del Carmen Vera Ganoza (2014), “Estimation of the Pass-Through and Welfare Effects of the Tariff reduction for Yellow Corn in Peru Between 2000 and 2011.” Trade Policies, Household Welfare and Poverty Alleviation: 243.
Kee, L H, A Nicita, and M Olarreaga (2009), “Estimating Trade Restrictiveness Indices.” Economic Journal 119(1): 172-199.
Kee, L H, C Neagu, and A Nicita (2013), “Is Protectionism on the Rise? Assessing National Trade Policies during the Crisis of 2008.” Review of Economics and Statistics 95(1): 342-346.
Moran, T (2014), “Tariffs Hit Poor Americans Hardest.” Peterson Institute for International Economics Real Time Economics Watch. (31 July).
Rosenberg, J (2015), “The Distributional Burden of Federal Excise Taxes.” Urban Institute and Brookings Institution Tax Policy Center.
United States International Trade Commission (USITC) (2016), Harmonized Tariff Schedule.
World Trade Organization. Tariff Download Facility.
 This is lower than both the average statutory tariff and the actual burden from tariffs as consumers substitute away from high tariff goods and the process will not count the welfare losses of switching from preferred goods. For simplicity, we assume that tariffs are fully passed forward to consumers, this does not impact the regressivity calculations (see Appendix).
 We use the disaggregated categories, for which the expenditure breakdown by deciles of before-tax household income and demographic groups are available in the pre-publication tables upon request from BLS: http://www.bls.gov/cex/csxresearchtables.htm#allnew.
 Our match between the CEX categories and the tariff categories is necessarily imperfect, so we may miss some collected tariffs.
 See Irwin (2014) for a discussion of empirical findings on tariff incidence and analysis of a product market where the US is a large consumer in the global marketplace.
 While some papers in the trade literature point out that businesses need not pass on tariff cuts to buyers in the form of lower prices and can simply absorb them as higher markups, this argument holds mainly in the case where tariffs are lowered only for one trading partner, in a model with only two countries (Arkolakis et al. 2015, Edmond et al. 2015). De Blas and Russ (2015) show that when the tariffs are lowered toward more than one trading partner in a multi-country world, as in a case like TPP, the cuts are more likely to be passed on to the buyer due to the threat of competition from other suppliers. Similarly, the exchange-rate pass-through literature suggests that changes in relative prices are likely to be passed through in their entirety into prices charged to buyers when they occur across multiple U.S. trading partners (Auer and Schoenle 2016) and that more pass-through occurs when they are persistent (An and Wang 2012), which is certainly the case for tariffs.
 While tariffs only directly raise the prices of imported products, indirect effects arise from reducing competition for domestic products, potentially causing the prices of all goods in a category to increase. This suggests that our results likely represent a lower bound. The literature, such as Gresser (2002), Hayakawa and Ito (2015), and Jara and Ganoza (2014), estimate a pass-through rate in the 0.22-0.74 range. To account for these indirect price effects, we also estimate the cost of tariffs assuming that tariffs also raise the price of domestic goods by 25 and 50 percent of the tariff rate, with zero percent as the benchmark estimate referred to throughout the text. This benchmark is also consistent with estimates by Feenstra and Weinstein (2016), who link increased trade consumer and welfare gains to a drop in U.S. domestic markups between 1992 and 2005. A related study by the Peterson Institute assumes 100 percent spillover of tariff reductions into domestic prices (Moran 2014).