Media reports on tariff policy changes and trade negotiations put the ‘effective’ tariff index under the same spotlight as the consumer cost-of-living index. The observed buyer expenditure shares adequately approximate price weights in cost-of-living index comparisons across time, but they are deficient for ‘effective’ tariff index comparisons of tariff heights across countries. Tariffs have large indirect effects on buyer shares due to trade diversion to third parties. Such effects ripple outward and rebound back within the world trade system. Thus, the ‘effective’ tariff index misinforms public opinion and policymakers. We (Anderson and Yotov 2025) offer a tariff index appropriate for international comparison purposes.

The True Cost of Protection (TCP) index

The True Cost of Protection (TCP) index uses weights that incorporate the effects of others’ as well as own tariffs on buyer trade shares. The difference in weights between the TCP and the import-weighted tariff indexes is illustrated by an example. US tariff increases on China’s exports to the US not only lower China’s share of US imports; they raise other countries’ US import shares. Third-party effects also matter: US tariff increases on Chinese exports divert some of China’s exports to Japan, raising Japan’s share weight on its tariff on China’s manufactures. Some crowded-out Japanese home-market sales may go to the US, raising the Japan-US trade share. Ripples continue ad infinitum.

The economic structure underlying the TCP is the constant trade elasticity ‘gravity’ representation of spatial equilibrium (Anderson and van Wincoop 2003, Arkolakis et al. 2012). Gravity is the current workhorse model of trade policy analysis because it fits the data extremely well, and its use in projections is remarkably more successful than most economic models. Efficient spatial arbitrage equilibrium determines the distribution of given supplies of goods to multiple destinations. Thus, buyers’ willingness-to-pay is equal to the product of the origin seller’s net payment times a markup factor comprised of ‘iceberg trade frictions’ times the bilateral tariff factors. The constant elasticity of substitution demand system is assumed to generate the buyers’ willingness-to-pay at all destinations. Tariff average indexes for each importing country are calculated as the uniform tariff on all goods from all origins that yields the same value of imports at world prices observed with the actual highly heterogeneous tariffs.

The TCP index is equal to the product of each country’s buyers’ incidence of (its own and the rest of the world’s) tariffs and its sellers’ incidence of sales costs (of its own and the rest of the world’s) tariffs. The indexes are interdependent across countries, since a small rise in one country’s tariff index raises its buyers’ incidence and also raises other countries’ seller incidences. Consequently, the other sellers’ tariff indexes rise.

Despite its underlying computational complexity, TCP tariff interpretation for public opinion is reasonably simple. Actual applied tariffs are hypothetically replaced by nationally uniform tariffs that maintain each country’s observed value of imports at world prices and given incomes. Tariff heterogeneity across countries and products is indexed in border wall height measures for each country.

Equal exchange of tariff changes measured by the TCP – tariff reciprocity – satisfies a notion of fairness. Local changes can be chosen that satisfy the trade-volume-reciprocity principle, the equal exchange of market access. The TCP by construction satisfies equivalence to the non-discrimination or most-favoured-nation principle. These principles in the US Trade Agreements Act of 1934 remain as relevant in the protectionist post-2016 era of bilateral threats and negotiations as they were in the protectionist 1930s. The WTO principle of reciprocal exchange of market access is locally equivalent to TCP tariff reciprocity because equal-percentage TCP tariff changes produce equal-percentage trade-volume changes.

Bagwell and Staiger (2002) show that reciprocity in trade agreements offsets the tendency for tariffs to be overused by governments because foreign sellers pay part of the tariff as its incidence falls partly on them. The TCP, as a guide to tariff reciprocity in trade agreements, has the same virtue because the decomposition of changes in TCP automatically measures the negative externality.

Empirical implementation

The TCP requires trade and production data, which must be matched with tariff data. For trade, production, and tariffs, we use three matching datasets of the US International Trade Commission. In addition, we use various datasets of proxies for bilateral frictions. We focus on 107 manufacturing sectors, since there are no tariffs for services, and exclude agriculture and mining.

We estimate gravity equations for each industry to construct bilateral trade cost vectors. Then, we construct the TCP by solving the gravity system with and without tariffs for each industry. To mitigate convergence issues (e.g. due to some very small countries), we select the 99 largest exporters, which account for more than 99% of world exports.

TCP tariffs and import-weighted tariffs can be very different

The number of TCP tariffs that we obtain is large (99×107 = 10,593), so we report summary statistics for the full set of TCPs.

TCP tariffs and import-weighted tariffs are far from perfectly correlated. The difference really matters. The average correlation across all sectors and countries is 0.65, but it varies by sector, between 0.10 (for ‘Distilling rectifying and blending of spirits’) and 0.86 (for ‘Starches and starch products’). The ranking of sectors by tariff height will thus differ considerably between indexes. Our index can thus better inform and facilitate trade relations management in the likely future of sequential bilateral negotiations.

TCP tariffs are predominantly smaller but can be larger than import-weighted tariffs. The predominance of smaller TCPs is due to the concavity of the constant elasticity of substitution price index function and Jensen’s inequality, which holds when sectoral trade balance prevails. Instead, the revealed comparative advantage (sectoral exports larger than sectoral imports) is negatively associated with the domestic expenditure share. The combination of revealed comparative advantage and its likely effect on reducing uniform measures of sectoral tariffs would reduce the TCP below the import-weighted average tariff, even in the linear case.

Figure 1 plots TCP tariffs against import-weighted average tariffs for US sectors. Interestingly, the US is one of the countries with TCP indexes that are mostly larger than the corresponding import-weighted tariffs. Consistent with our discussion above, this reflects the overall trade deficit.

Figure 1 True Cost of Protection (TCP) index vs import-weighted tariffs, US sectors

TCP tariffs vary significantly across countries

Figure 2 plots country aggregate TCP tariffs against import-weighted average tariffs. (For clarity, not all country ISO3 codes are listed on the x-axis of Figure 2.) The lowest uniform TCP tariffs (i.e. subsidies) are about -3% for Chile, Peru, and Romania, and the largest TCP tariffs are more than 10% for India, Kenya, and Sri Lanka. Notably, some of the largest tariffs are for Asian economies other than China.

Figure 2 True Cost of Protection (TCP) index vs import-weighted tariffs by country

The correlation between the TCP and the import-weighted country tariffs is 0.53. Most of the import-weighted tariffs are higher than the corresponding TCP tariffs. Consistent with Figure 1, the US is one of the countries whose TCP index is larger than the corresponding import-weighted average tariff. Interestingly, some of the other countries whose TCP tariffs are larger than their import-weighted tariffs are countries that are peripheral but not members of the EU, such as Norway, Switzerland, and Turkey.

The incidence of the TCP tariffs varies widely across countries and sectors

Finally, we calculate the variation of buyer and seller incidence of the TCP tariffs decomposition. The variation in the incidence of TCP tariffs on the consumers and the producers varies significantly by country and by sector. Producers in many rich countries (e.g. Germany, Italy, Japan, Sweden, US) will suffer higher trade costs. The US is an extreme example, with the highest share of industries where the producers will suffer higher trade costs. The large US trade deficit is the likely explanation. The opposite is true for China. Importantly, the buyers’ TCP incidence in the US is even larger; that is, US buyers pay most of the tariff.

Conclusion

TCP tariff indexes dominate other tariff indexes for cross-country comparisons. The difference matters substantially. Trade negotiations or disputes will be better guided by the use of TCPs.

References

Anderson, J E, and E van Wincoop (2003), “Gravity with gravitas: A solution to the border puzzle”, American Economic Review 93(1): 170–92.

Anderson, J E, and Y V Yotov (2025), “Tariff reciprocity”, NBER Working Paper 34052.

Arkolakis, C, A Costinot, and A Rodriguez-Clare (2012), “New trade models, same old gains?”, American Economic Review 102(1): 94–130.

Bagwell, K, and R W Staiger (2002), The economics of the world trading system, Cambridge, The MIT Press.