VoxEU Column

Trade imbalances and preferential trade agreements: An empirical investigation

The creation of new preferential trade agreements remains a key driver in trade liberalisation, at least bilaterally. This column examines the creation of new agreements, highlighting the effects of pre-agreement bilateral imbalance and within-country income inequality on the likelihood of a preferential trade agreement being formed. The authors suggest that while the creation of preferential trade agreements continues to be an important avenue to liberalise trade, there are significant political constraints to be considered.

Even in the midst of a global pandemic, as the Word Trade Organization (WTO) faces unprecedented challenges (Matoo and Staiger 2020), countries around the globe continue to use preferential trade agreements (PTAs) as means to liberalise bilateral trade flows. In its latest report, the WTO (2021) highlights that 36 new preferential trade agreements were notified between 1 July 2020 and 1 January 2021. While many of the new notifications were the result of Brexit, the recent wave means that as of May 2021, a total of 349 preferential trade agreements were in force. The vast majority of these agreements are represented by free trade areas (88% of the total), whereas customs unions (CUs) and partial scope agreements (PSAs) represent respectively only 5% and 7% of the total, respectively (see Figure 1).

Figure 1 Types of preferential trade agreements in force as of May 2021



At the same time, as preferential trade agreements have boomed, trade imbalances have once again come to the forefront of the trade policy debate. In a famous tweet dating back to 2018, former US President Donald Trump claimed:

“When a country [USA] is losing many billions of dollars on trade with virtually every country it does business with, trade wars are good, and easy to win. Example, when we are down $100 billion with a certain country and they get cute, don’t trade any more – we win big. It’s easy!”

Aggregate (Dekle et al. 2007, Crinò and Epifani 2014) and bilateral trade imbalances have also gained increasing attention among academic economists (Feenstra et al. 1998, Davis and Weinstein 2002, Cunat and Zymek 2019, Eugster et al. 2019, Delpeuch et al. 2021). Still, surprisingly little is known about the role that they play in shaping the formation of preferential trade agreements.

In a recent working paper (Facchini et al. 2021), we tackle this question by developing a rich political economy model of preferential trade agreement formation and then assessing empirically its predictions using a comprehensive dataset of 187 countries. 

Theoretical framework

The theoretical model is based on a three-country, multiple good setting in which two prospective members interact strategically to determine the trade policy to be implemented towards each other and the rest of the world. The rest of the world is assumed – for simplicity – not to behave strategically, and instead to implement a most-favoured-nation (or MFN) policy. The underlying structure of the economy is based on the standard oligopolistic model of trade used in earlier analyses of regionalism (Krishna 1998), in which individuals derive income from labour supply and from the profits generated by oligopolistic firms, whose ownership is distributed unevenly among the population. Citizens choose the trade policy regime (a customs union, a free trade agreement, or the most-favoured-nation regime required by WTO membership), and elected representatives determine the actual tariffs to be implemented. The model focuses on three key drivers of trade policy choices: bilateral trade imbalances, the degree of geographic specialisation between prospective member countries, and the pervasiveness of income inequality within each member country.

Main predictions

Our analysis indicates that trade imbalances and income inequality play an important role in explaining the formation of preferential trade agreements – independently of their type. A critical feature of the model helps us understand the effect of bilateral imbalances. In the oligopolistic setting we consider, preferential market access obtained by a prospective member tends to increase that country’s welfare by increasing the profits earned by its citizens. At the same time, preferential access granted to the partner country tends to reduce it, as the losses in profits and tariff revenues tend to dominate the gains in consumer surplus. As trade becomes more imbalanced, the extent of market access exchanged by the prospective members becomes more unequal. As a result, the greater the imbalances, the less willing to join a preferential trade agreement is the country experiencing a deficit – making the agreement itself less likely to emerge in equilibrium. Turning to the role played by income inequality, greater inequality in the model implies that the distribution of profits becomes more concentrated. As a result, their level becomes less relevant for the median voter, and consequently, the formation of an agreement will be less likely. 

The pattern of geographic specialisation – i.e. the extent to which the sectoral production structures of the two prospective members overlap – plays instead an important role in shaping the choice of agreement type (i.e. whether a customs or an free trade agreement will emerge). To understand the intuition behind this result, note that when trade policy needs to be coordinated between member countries, like in the case of a customs union, the median voter in each prospective member has an incentive to delegate power to a representative who is more protectionist than herself. These incentives are not at work if the external trade policy is not coordinated – like in the most-favoured-nation regime or the case of a free trade agreement. Importantly, greater geographic specialisation tends to exacerbate the strategic delegation incentives in the customs union, leading to higher external tariffs, making this trade policy regime more distortionary and less desirable from the point of view of the median voter.


To assess the predictions of the model, we have assembled a comprehensive dataset, spanning the period 1960-2015 and covering 187 countries. Following the structure of the model, the decision to form a preferential trade agreement and the choice of its type is estimated as a two-stage process. In the first stage, countries decide whether to establish an agreement, and in the second whether it will take the form of a free trade agreement or a customs union.1  

The empirical results lend support to the predictions of the model. Greater bilateral trade imbalances and greater within-country income inequality are associated with a significant decline in the likelihood that two countries will form a preferential trade agreement. Conditional on two countries establishing a preferential trade agreement, we also find that greater geographic specialisation in their production patterns is associated with a greater likelihood of observing a free trade agreement emerge in equilibrium, rather than a customs union.

Figure 2 Determinants of the PTA formation and of the choice between FTA and CU



Notes: The figure presents estimates of the marginal effects of a one standard deviation in the corresponding variable on the baseline probability of forming a PTA (top panel) and a FTA (bottom panel) obtained from running a probit model with sample selection, on a sample of 187 countries spanning the years 1960-2015. The selection equation (top panel) accounts also for the standard determinants of the PTA formation decision (see Facchini, Silva and Willmann 2021 for more details). 

The effects we uncover are sizeable and are described in Figure 2. In the top panel, we report the estimated impact of trade imbalances and inequality on the average probability of preferential trade agreement formation. A one-standard-deviation increase in trade imbalances decreases the probability of forming a preferential trade agreement by 6.17% compared to the baseline. Analogously, a one-standard-deviation increase in inequality decreases the preferential trade agreement formation probability by 5.65% compared to the baseline.

The bottom panel of the figure reports the effect of geographic specialisation on the choice between a free trade agreement and a customs union, conditional on a preferential trade agreement being formed. The estimated effect implies that a one-standard-deviation increase in geographic specialisation increases the likelihood of observing a free trade agreement by 9.08% compared to the baseline.

The results we have uncovered are remarkably robust. They continue to hold when we use alternative datasets to document the existence of preferential trade agreements, when we focus on the years after 2000 (which have seen a very fast growth in the number of new agreements reported to the WTO), when we exclude the European Union (by far the most important customs union in the sample), and when we exclude other large agreements like NAFTA and ASEAN.


As we already noted, the WTO has come under increased scrutiny because of the failure of the Doha Round, and more generally, because of the difficulties it has faced in pursuing further multilateral trade liberalisation. In our paper, we highlight that while the creation of preferential trade agreements continues to be an important avenue to liberalise trade further, there are significant political constraints to be considered. In particular, if trade imbalances between prospective members are very pervasive, the creation of preferential trade agreements is unlikely to receive sufficient political support in the electorate. This is a potentially relevant lesson for countries like the UK, that, in the post-Brexit era, plan to use preferential trade agreements as a key policy tool. If bilateral trade is too unbalanced, that seems unlikely to succeed. 


Baier, S and J Bergstrand (2004), “Economic determinants of free trade agreements’, Journal of International Economics 64: 29-63.

Baldwin, R and D Jaimovich, (2012), “Are free trade agreements contagious?”, Journal of International Economics 88: 1-16.

Cunat, A and R Zymek (2019), “Bliateral trade imbalances”, CES-Ifo Working Paper 7823

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

Dekle, R, J Eaton and S Kortum (2007), “Unbalanced Trade”, American Economic Review, Papers and Proceedings 97: 351-355.

Delpeuch S, E Fize and P Marin (2021), “Trade Imbalances and the rise of protectionism”,, 12 February.  

Egger, P and M Larch (2008), “Interdependent preferential trade agreement memberships: an empirical analysis”, Journal of International Economics 76: 384–399.

Eugster, J, F Jaumotte, M MacDonald and R Piazza (2019), "Bilateral and aggregate trade balances: Finding the right focus”,, 10 September. 

Facchini, G, P Silva and G Willmann (2021), “The political economy of preferential trade agreements: An empirical investigation”, CEPR Discussion Paper 16098.

Feenstra, R C, W Hai, W T Woo and S Yao (1998), “The US-China Bilateral Trade Balance,” NBER Working Paper 5910.

Crinò, R and P Epifani (2014), “Trade imbalances, export structure and wage inequality”, Economic Journal 128: 507-539.

Krishna, P (1998), “Regionalism and multilateralism: A political economy approach”, Quarterly Journal of Economics 113: 227–250.

Matoo, A and R W Staiger (2020), “Trade wars: What do they mean? Why are they happening now? What are the costs?”, Economic Policy 35: 561-584.

Van de Ven, W and B Van Praag (1981), “The demand for deductibles in private health insurance: a probit model with sample selection”, Journal of Econometrics 17: 229-252.

WTO (2021), “Facts and Figures: Regional Trade Agreements”, 1 July 2020- 1 January 2021, mimeo.


[1] The econometric analysis is based on a probit model with sample selection (Van de Ven and Van Praag 1981). All the empirical models include the standard determinants of PTA formation typically accounted for in the literature (e.g. Baier and Bergstrand 2004, Egger and Larch 2008, Baldwin and Jaimovich 2012). See Facchini et al. (2021) for more details.

735 Reads