RaboResearch - Economic Research

Mexico/Canada: are free-trade agreements with the EU beneficial?

Economic Report

  • In this Economic Report, we evaluate the impact of the FTA between Mexico and the EU on bilateral trade. This effect is subsequently used to proxy a potential substitution effect of Mexican exports to the EU in the potential event of a NAFTA breakdown
  • We have constructed a strongly balanced panel of quarterly data ranging from 1980 until 2017 covering 20 trading pairs of countries/regions (Mexico, the US, Japan, Canada and the EU), yielding 3,020 observations
  • We find a robust and statistically significant impact of the FTA between Mexico and the EU on exports, raising total bilateral trade between 4 and 5 percent. In terms of 2016 trade figures, this comes down to approximately USD 750mln and USD 1bn additional export due to the FTA
  • In light of these outcomes, the current negotiations between the EU and Mexico to renew and upgrade the bilateral trade agreement that came into force in 2000 could potentially lead to a small boost in trade between both partners. Moreover, an improved FTA between the EU and Mexico could mitigate some of the negative fallout of a NAFTA collapse, albeit to a small extent


On 23 January, the US, Canada and Mexico will continue their renegotiations on NAFTA in Montreal. Progress has bogged down since the fourth round of talk in Washington. Moreover, there are more and more indications that the US is planning to pull out of NAFTA (Reuters, 2017), which makes an assessment of potential economic fallout of such an event even more opportune. Erken et al. (2018) ran scenarios of a hard and soft NAFTA breakdown, and made an assessment of the impact of rising trade barriers on declining export volumes from Canada and Mexico to the US (and vice versa). One important element which we will reflect on in this Economic Report is the mitigating effect of free-trade agreements that Mexico has with the European Union. And since September 2016, Mexico and the EU have been negotiating a bilateral trade deal, which will upgrade the current one dating from October 2000.

Free-trade agreements can cushion the negative economic effects in Canada[1] and Mexico caused by a potential NAFTA breakdown, as lower trade with the US is compensated by increasing exports to the EU. However, it is hard to isolate the impact of free-trade agreements on trade volumes, as we do not have a so-called counterfactual situation, where we know what the export patterns of trading partners would be without these trade agreements in place. This complicates the calculation of clean-cut export substitution effects for Mexico and Canada. However, it is possible to use a workaround by calculating the impact of free-trade agreements on export volumes within a panel of countries over time, and at the same time control for many other important determinants for exports. The elasticities that we derive from this exercise at least provide some guidelines of the possible substitution effects that we can subsequently use as input for our broader scenario analysis on a NAFTA breakdown (see Erken et al., 2018).


The literature is broadly in consensus about the determinants that are relevant in identifying the impact of FTAs on trade (see, for instance, Baier and Bergstrand (2007) and Aitken (1973). Gravity models are often used as workhorse models to estimate the impact, including the distance between the countries, gross domestic product (growth) of both the exporting country as well as its trading partners, common language, common borders and transportation costs.

GDP of the importing country serves as a measure for import demand, as higher income is positively correlated with higher imports. GDP can also serve as a proxy for the potential market size of the exporting country, as smaller countries are relatively less attractive as an export destination. In addition, GDP of the exporting country is considered to reflect the potential capacity of export supply. Distance can serve as a proxy for transport costs and cultural proximity to trade, where a longer distance is negatively associated with trade. Furthermore, two countries that share common cultural values are found to trade more by means of common preferences and lower perceived trade costs (Felbermayr and Toubal, 2007). A similar mechanism holds for sharing a common border or language, as trade costs are lower and trade is perceived as being easier.

The literature shows mixed results as far as the magnitude and statistical significance of the impact of FTAs on trade is concerned. Back in the 60s, Tinbergen (1962) was one of the first to document the impact of FTAs on trade flows, but found economically insignificant results. Tinbergen’s study was later corroborated by e.g. Bergstrand (1985), who did not find an economically and statistically significant impact of FTAs on trade. Nevertheless, more recent research by Baier and Bergstrand (2007) shows that an FTA approximately doubles bilateral trade after 10 years of member countries. Kohl (2014), however, again finds mixed results. Although Economic Integration Agreements (EIA’s), such as FTAs, raise trade by 50%, more than half of the examined EIAs appear to have no impact on trade at all. In a similar fashion, Kohl et al. (2016) show that heterogeneity of trade agreements matter for international trade.

Model, data and methodology

In line with Baier and Bergstrand (2007) we analyze panel data. They argue that this approach is preferred over cross-sectional gravity estimation in gauging the impact of FTAs on trade, as unobserved heterogeneity causes an estimation bias. In order to calculate the impact of a free-trade agreement on exports, we will estimate a simple model that captures export dynamics of country i to trading partner country j over time t. This model has the following functional form:

where X represents exports (in USD), Y is gross domestic product (in USD), REER is the real effective exchange rate, D measures distance, NAFTA is a dummy variable measuring if NAFTA applies to two trading partners (value of 1) or not (value of 0) and FTA is a dummy variable which captures the impact of the trade agreement between Mexico and the EU. Lastly, t is a time-specific index, D indicates year-on-year mutations and log is the natural log. In addition,  and  respectively reflect the country and time fixed effects that are incorporated in our model specification. The REER measures the weighted average of bilateral real exchange rates with trading partners of country i:

where T measures the trade weight of country i with country j and N indicates the number of countries. The second part of the equation (after the first multiplication sign) captures nominal bilateral exchange rates E adjusted for domestic and foreign prices (P).

We use quarterly time series data over years 1980 to 2017 for the US, Canada, Mexico, Japan and the EU to estimate equation (1). The fact that we only use a limited amount of countries in our panel complicates the inclusion of too many ‘fixed’ variables with limited or no variance over time, such as common language and common borders. Instead, we prefer to include FTAs in our estimations and experiment with distance as our static time-invariant variable. Note that in the fixed effects specification, opposed to the random effects specification, distance cannot be incorporated as, by definition, time-invariant variables drop out of the equation.

Export data is taken from the IMF Direction of Trade Statistics (DOTS), GDP data originates from the OECD Quarterly National Accounts, data on real effective exchange rates is produced by the BIS and distance[2] data is taken from .

We estimate both a fixed effects as well as a random effects specification. The random effects specification is preferred, as the null hypothesis of the Hausman test cannot be rejected. In contrast to Baier and Bergstrand, who propose to use a first differenced panel, a random effects specification with differenced variables is preferred. Differencing ensures stationarity as the data follows a unit root process.

In the country fixed effects specification we control for country-specific characteristics that do not vary over time but might affect trade such as country-specific institutional characteristics. Thus, by including country fixed effects one controls for the unobserved heterogeneity in the cross-sectional dimension. In both the random and fixed effects specifications time dummies are incorporated to control for unobserved heterogeneity over time. In addition, we have lagged variables for GDP and the Real Effective Exchange Rate (REER) in order to prevent potential endogeneity, as for example lagged GDP does affect current exports, but current export does not affect lagged GDP.


The estimation results are presented in Table 1. Column (1) shows the full model estimate. Both GDP variables show a significant impact. These coefficients indicate that a 1% increase in GDP of the exporting country leads to an increase in exports of 0.96%. Similarly, an increase in GDP of the trading partner by 1% leads to 0.89% additional exports. The real effective exchange rate and the distance variable show the correct sign, but are statistically insignificant. The NAFTA dummy does not have a statistical significant effect, but the FTA does. The estimated coefficient of the FTA variable suggests that the FTA between Mexico and the EU has resulted in 4% higher exports.

Table 1: Model estimates
Table 1: Model estimatesNote: Standard errors in parenthesis. Significant at * 5%; ** 1%.
Source: Rabobank

In column (2), we again estimate the Random Effects model and remove our distance variable, as it might have distorting effects on our trade agreement variables (NAFTA and FTA). Moreover, the coefficient is not statistically significant. Besides again a robust and stable effect of the FTA dummy on exports, the NAFTA dummy is also statistically significant at the 95% confidence band. Finally, we estimate a fixed effects model as a robustness check for our model in column (2). All estimated coefficients remain stable and both GDP and FTA continue to show a statistically significant impact on exports. Our FTA coefficients imply that in terms of 2016 figures, Mexican export have benefited from the FTA with the EU by USD 750mln and USD 1bn.

NAFTA coefficients, although positive and significant in our first model specification, show mixed results. This result is exemplary of the tough spot NAFTA is in at the moment, with some economists defending the benefits of the free-trade agreement, whereas other dispute the desirability of keeping NAFTA under current condition.


In this study we estimate the impact of FTAs on bilateral trade between Mexico and the EU. This effect is subsequently used as a proxy for potential substitution effects of Mexican exports to the EU in the event that NAFTA negotiations fail. We have constructed an extensive dataset ranging from 1980 until 2017 including bilateral trade and country characteristics that is analyzed in a panel setting. In our study we document a statistically significant impact of the FTA between Mexico and the EU on exports, raising bilateral trade with 4%. This means that in terms of 2016 figures, Mexican exports to the EU are raised between USD 750m and USD 1bn. We find this result to be robust under all model specifications. This finding implies that there could be scope for substitution of Mexican exports to the EU. Nevertheless, this estimate still only remains a rough proxy while a backward looking estimate can only be a rough indication what might happen to future exports given the largely unpredictable nature of trade flows, especially given the substantial improvement that the current free trade agreement will probably undergo after negotiation will be concluded between Mexico and the EU.

Our methodology deviates from other studies, as we use a panel data setting rather than a cross-sectional gravity model, which usually do not take into account time-varying determinants of trade. By using a panel data setup, we follow up on the critique by Baier and Bergstrand (2007). The authors obtain a much larger estimate and document a 92% increase in trade due to FTA formation. This result is incomparable to our results as we only focus on bilateral trade between the EU and Mexico while they focus on all FTAs gauging an average treatment effect rather than a trading partner specific effect. Moreover, Kohl et al. (2016) showed that heterogeneity of trade agreements matter substantially, which implies that a simple comparison of a FTA-specific impact on trade with average treatment effects would be inappropriate.


[1] 21 September 2017, the Comprehensive Economic and Trade Agreement (CETA) between Canada and the EU came into force as well. However, the time span of the trade agreement is too short to assess the impact on the trade in this exercise.

[2] Distances are calculated following the great circle formula, which uses latitudes and longitudes of the most important city (in terms of population) or of its official capital.


Aitken, N.D. (1973). The effect of the EEC and EFTA on European trade: A temporal cross-section analysis. American Economic Review, 63(5), 881-892.

Bergstrand, J.H. (1985). The gravity equation in international trade: some microeconomic foundations and empirical evidence. Review of Economics and Statistics, 474-481.

Baier, S.L. and J.H. Bergstrand (2007). Do free trade agreements actually increase members' international trade? Journal of international Economics, 71(1), 72-95.

Felbermayr, G.J. and F. Toubal (2010). Cultural proximity and trade. European Economic Review, 54(2), 279-293.

Kohl, T. (2014). Do we really know that trade agreements increase trade? Review of World Economics, 150(3), 443-469.

Kohl, T., S. Brakman and H. Garretsen (2016). Do trade agreements stimulate international trade differently? Evidence from 296 trade agreements. The World Economy, 39(1), 97-131.

Tinbergen, J. (1962), Shaping the World Economy. New York: The Twentieth Century Fund.

Hugo Erken
RaboResearch Netherlands, Economics and Sustainability Rabobank KEO
+31 6 2223 1650
Marijn Heijmerikx
RaboResearch Global Economics & Markets Rabobank KEO

naar boven