VoxEU Column International trade

Bilateral and aggregate trade balances: Finding the right focus

Bilateral trade balances have come under scrutiny recently, with some policymakers concerned that their large and rising size may reflect asymmetric obstacles to trade. This column argues that macroeconomic factors – rather than bilateral tariffs – have been the key drivers of the evolution of bilateral trade balances. While tariffs have played a modest role in the evolution of bilateral balances, declines in tariffs have lifted productivity by allowing a greater international division of labour, including through participation in global value chains. A sharp increase in tariffs would therefore create significant spillovers, leaving the global economy worse off. 

The presence of large and rising bilateral trade balances has raised concerns that asymmetric obstacles to trade may distort the international trade system. Yet, rather than looking at bilateral balances, economists typically focus on understanding aggregate trade balances (that is, the sum of a country’s bilateral trade balances with all other countries) and their macroeconomic determinants (e.g. Obstfeld 2012; for an exception, see Davis and Weinstein 2002). This raises the question of whether bilateral trade balances are the right focus of policy discussions.  

In a recent study (IMF 2019), we delve into this politically important issue, focusing on three questions. First, we ask what drives bilateral trade balances. Second, what is the relationship between aggregate trade balances (and their drivers) and bilateral trade balances? Third, when bilateral tariffs are raised in an effort to target a specific bilateral trade balance, what are the consequences for the countries involved – and what spillovers arise for others when accounting for the presence of global value chains? 

The drivers of bilateral trade balances

A first look at bilateral trade balances suggests a number of driving forces. Figure 1 shows the aggregate trade balance and the three largest bilateral trade balances for a few large countries in 1995 and 2015. In general, countries have both positive and negative bilateral trade balances, reflecting the international organisation of production – lower trade costs allow countries to specialise in what they do best and then trade with each other. There are cases, however, where most bilateral trade balances are one-sided, either on the negative side (for instance, in the US) or on the positive side (for instance, in Germany). These are usually accompanied by large aggregate trade balances that tend to be shaped by macroeconomic factors such as aggregate demand and supply. 

Figure 1 Bilateral trade balances by major partners (percent of global GDP)

Sources: OECD Trade in Value Added database, and IMF staff calculations.
Note: Data labels use International Organization for Standardization (ISO) country codes. Top three partners shown per year.

To quantify the relative importance of the drivers of bilateral trade balances, we resort to the workhorse gravity model for bilateral exports. This model – which has solid microeconomic foundations – distinguishes three groups of drivers of bilateral trade flows, in line with our observations on Figure 1: macroeconomic factors, trading costs, and the sectoral composition of demand and supply. Macroeconomic factors are represented, essentially, by the level of aggregate demand and supply of each of the trading partners. Trading costs include bilateral tariffs and proxies for other trade costs (e.g., distance), as well as the average import and export costs for each country (i.e. the “multilateral resistance terms”). Finally, the sectoral dimension allows us to capture the potential effects of structural characteristics and transformations in shaping the sectoral composition of countries’ demand and supply, and hence, the international organisation of production. 

The model can be applied to both bilateral exports and imports and thus allows us to derive results for bilateral trade balances (see Davis and Weinstein 2002 for a similar approach). It is estimated over a large sample consisting of 63 countries, 34 sectors, and the period 1995-2015.

We find that changes in bilateral balances over the past two decades have been, to a significant extent, driven by macroeconomic factors – that is, the relative evolution of aggregate demand and supply in both trading partners. For example, macroeconomic factors explain 20% of the change in the US-Germany trade balance but over 95% of the change in the US-China trade balance (Figure 2). In the latter case, this reflected a larger increase in aggregate supply than aggregate demand in China and, to a lesser extent, a larger increase in aggregate demand than supply in the US. 

Figure 2 Drivers of changes in selected bilateral trade balances, 1995-20151 (billions of US dollars)

Sources: OECD Trade in Value Added database, and IMF staff calculations.
Note: Data labels use International Organization for Standardization (ISO) country codes. 1 Average value 2010-2015 minus average value 1995-1999. 2 This includes tariffs and free or preferential trade agreements. 3 This residual is the sum of the model residuals plus the approximate error.

The direct impact of changes in bilateral tariffs played a small role, reflecting the fact that tariffs were already at relatively low levels in 1995 in many countries and subsequent reciprocal tariff reductions had offsetting effects on bilateral trade balances.1

The link between bilateral and aggregate trade balances

Under mild assumptions, the gravity equation can be transformed to show more formally that the bilateral trade balance between two countries is a function of the two trading partners’ aggregate trade balances (which capture the imbalance between demand and supply in each country), and a bilateral trade intensity factor that summarises how a specific trade relationship is affected by the bilateral and multilateral trading costs and other determinants of comparative advantage and sectoral specialisation. The logic underlying the gravity model also suggests that to better evaluate the development of the bilateral trade balance over time, it should be weighted by the product of the two countries’ outputs (relative to world output) to reflect the fact that as the two countries grow, all else being equal, their bilateral trade balance would tend to increase in absolute value.

Applying this to the evolution of the US-China bilateral trade balance points to two interesting facts. First, the US-China bilateral trade balance widened from the mid-1990s until the mid-2000s but then subsequently has been shrinking (Figure 3). Second, the evolution of the appropriately scaled bilateral trade balance mirrors the pre-Global Crisis widening of external imbalances of the US and China and their post-crisis reduction. Holding the bilateral trade intensity constant, most of the evolution of the US-China trade balance can be explained by the evolution of the aggregate trade-balance-to-GDP ratios of the two countries (“predicted” line). 

Figure 3 US-China bilateral and aggregate trade balances (percent)

Sources: OECD Trade in Value Added database, and IMF staff calculations. 
Note: TB = trade balance. Data labels use International Organization for Standardization (ISO) country codes. 1 Predicted as the difference between aggregate trade balance-to-GDP ratios of both countries, holding bilateral trade intensity constant.

From a policy perspective, it is therefore important to understand what drives the macroeconomic imbalance between supply and demand in each country – also reflected in their aggregate trade balance. Drivers include fundamental factors, such as demographics and the level of economic and institutional development, but also macroeconomic policies, in particular fiscal policy and credit cycles, and – in some cases – exchange rate policies and domestic supply-side policies (for example, widespread subsidies to production costs) (e.g. IMF 2018). For instance, procyclical fiscal stimulus has been pointed out as a source of excessive US external deficit while China is being discussed as one example where widespread supply-side policies, such as production subsidies and regulatory policies, may be at play, artificially raising aggregate supply. 

A closer look at tariffs and their spillovers

While the direct impact of tariffs on the evolution of bilateral trade balances has been small relative to macroeconomic factors, this doesn’t mean that tariffs do not matter. Large and sustained changes in tariffs and other trade costs can shape the international organisation of production into global value chains, with important consequences for aggregate productivity and income growth. Indeed, since the mid-1990s, significant declines in trade costs (tariffs, transportation, and communications costs) have gone together with an increase in global value chain integration (Figure 4).

Figure 4 Tariffs and global value chain participation (value-added weighted average over countries and sectors, percent)

Sources: OECD Trade in Value Added database, World Bank, World Integrated Trade Solution (WITS), and IMF staff calculations. 
Note: GVC = global value chain. GVC participation is the backward and forward participation in GVCs as a percent of total exports. Tariffs are for agriculture, mining, and manufacturing sectors. 2012-15 extrapolated based on TIVA in 2018.

The integrated nature of the current global trade system suggests that a sharp increase in tariffs could have significant spillovers as it propagates across borders through upstream and downstream GVC links. To gauge this possibility, our work combines an empirical approach with the use of structural quantitative models. Using a large panel data set of 35 countries and 13 manufacturing sectors, we calculate upstream and downstream tariffs using the cumulative tariff formula derived by Rouzet and Miroudot (2013) and find that tariffs imposed further upstream or downstream in the value chain generally hurt domestic output, employment, and productivity. Thus, tariffs hurt economic activity not only for the countries directly imposing and facing them but also for other countries up and down the value chain. Simulations (based on partial equilibrium country-sector estimates) illustrate that the output cost of a generalised 1 percentage point increase in manufacturing tariffs would be larger today than it would have been in the mid-1990s, particularly so for countries highly integrated in manufacturing supply chains (e.g. Korea and Germany) (Figure 5).

Figure 5 Illustration of the effect of a 1% generalised tariff increase on real value added (percent of GDP)

Sources: IMF staff estimates.
Note: Data labels use International Organization for Standardization (ISO) country codes. Effects are partial equilibrium estimates based on a country-sector level analysis. The figure shows the change in the simulated tariff spillovers between 1995 and 2011, the last year for which such an exercise is possible given data constraints. 2011 is a good approximation of current global value chain links because most of the growth in global value chain integration took place before 2011.

However, when tariff increases are not deployed across the board but are targeted to specific partners, trade diversion may benefit countries that are not directly targeted by the tariffs, as it increases demand for their goods. Trade diversion effects suggest that attempts to target one bilateral trade balance through bilateral tariffs or other distortions will likely be met with offsetting changes in the trade balances with other partners – a point we take up next. 

Bilateral tariff wars

Three different general equilibrium models are used to simulate a hypothetical bilateral tariff increase by 25% on all goods traded between the US and China (Caceres et al. 2019, Hunt et al. forthcoming). The simulations suggest three main takeaways. 

  • First, they show that the US and China would be the countries facing the largest output losses (Figure 6). 
  • Second, while their bilateral balance declines somewhat, the change in their aggregate trade balances is negligible, as each country’s demand is diverted to other trading partners. Thus, bilateral tariffs are not only costly for the countries involved but they are also likely to be ineffective in addressing their external imbalances. 
  • Finally, while countries such as Mexico, Canada and to a lesser extent east Asia benefit somewhat from trade diversion, the global economy is left worse off. These losses are likely underestimated as they do not factor in the toll that trade tensions take on confidence and the adjustment costs associated with the domestic and international sectoral reallocations of global value chains.

Figure 6 Macro effects from a 25% increase in tariffs affecting all US-China trade, real GDP (percent change from baseline)

Source: IMF calculations.
Note: Effects are simulated from three general equilibrium models: CFRT = Caleindo and others (2017) model, GIMF = Global Integrated Monetart and Fiscal model, GTAP = Global Trade Analysis Project. LR = long run, NAFTA = North American Free Trade Agreement. In the figure, NAFTA is NAFTA countries excluding US and Asia is Asian countries excluding China. 

Policy implications

Two main policy conclusions emerge from our findings. 

  • First, macroeconomic factors are crucial drivers of both bilateral and aggregate trade balances. Unless macroeconomic conditions change, using bilateral tariffs to target a specific bilateral trade balance is likely to lead to compensating adjustments in other bilateral trade balances due to trade diversion, leaving the country’s aggregate trade balance broadly unchanged. The discussion of external imbalances should therefore focus on avoiding macroeconomic policies that generate distortions in the aggregate trade balance, such as procyclical fiscal policies or heavy subsidisation of export sectors.  
  • Second, multilateral reductions of tariffs and other non-tariff barriers would benefit trade and, over the longer term, improve economic outcomes. Policymakers should continue to promote trade openness by undoing recently enacted tariffs and enhancing efforts to reduce existing barriers to trade. At the same time, our finding that tariff changes lead to potentially large reallocations of production across sectors and countries also provides a cautionary tale on the possible side effects of both tariff hikes and trade liberalisations – trade policies can impose costly adjustments for some workers and communities that policymakers are called to address.

Authors’ note. The views expressed herein are those of the authors and should not be attributed to the IMF, its Executive Board, or its management.


Caceres, C, D A Cerdeiro, and R Mano (2019), “Trade Wars and Trade Deals: Estimated Effects using a Multi-Sector Model”, IMF Working Paper No 19/143. 

Davis, D R, and D E Weinstein (2002), “The Mystery of the Excess Trade (Balances)”, American Economic Review, 92 (2),170-174.

Hunt, B, D Mursula, R Portillo, and M Santoro (forthcoming), “Trade Tensions: Inspecting the Effects Through Two Lenses” IMF Working Paper.

International Monetary Fund (2018), External Sector Report.

International Monetary Fund (2019), “The Drivers of Bilateral Trade and the Spillovers from Tariffs”, in World Economic Outlook, Chapter 4, April. 

Obstfeld, M (2012), “Does the Current Account Still Matter?”, American Economic Review, 102 (3), 1-23.

Rouzet, D, and S Miroudot (2013), “The Cumulative Impact of Trade Barriers Along The Value Chain: An Empirical Assessment Using the OECD Inter-Country Input-Output Model”, Paper prepared for the 16th Annual Conference on Global Economic Analysis, Shanghai, China.


[1] Over the longer term, large and persistent changes in tariffs can also influence the international division of labour and thereby, potentially, macroeconomic factors—an issue we return to below.

2,100 Reads