The announcement earlier this year that the US and the EU will launch talks on a Transatlantic Trade and Investment Partnership has once more drawn attention to the continuing surge in regional free-trade agreements. Given the lack of progress in the Doha round of multilateral trade talks, most countries have turned their attention to such regional trade-liberalisation initiatives.
The hope behind this development is of course that free-trade agreements will be economically beneficial for the participating nations or trade blocs. Existing economic research has shown that this is indeed the case, at least for developed economies. For example, in an influential paper Daniel Trefler (2004) showed that the Canada-US Free Trade Agreement of January 1989 led to dramatic increases in Canadian manufacturing.
While knowing with hindsight that a particular free-trade agreement has been beneficial for a country is useful, we would ideally be able to make predictions about future agreements as well, such as the proposed US-EU free trade area. Traditionally, so-called computable general equilibrium trade models have been used to make such forecasts. While their use for this purpose has unquestionably been beneficial, not everybody is happy. For example, Kehoe (2005) and Balistreri et al. (2011) criticise the forecasting performance of such models and point out that they neglect the effects of trade on aggregate productivity and trade growth along the so-called extensive margin (i.e. trade increases due to more firms trading or due to firms trading more products).
The empirical observation that aggregate productivity and extensive margin effects are important in explaining the trade and productivity effects of trade liberalisation has also been one of the key motivations behind the recent development of heterogeneous firm models in international trade. These models generate such effects by combining within-industry productivity heterogeneity of firms with export market entry costs (see Melitz and Redding (2013) for an in-depth exposition of these models). Since the seminal contribution of Melitz (2003), a large number of heterogeneous firm models have been developed which are realistic enough to be amenable to quantitative analysis and forecasting. However, a thorough evaluation of the quantitative performance of these models with regards to the effects of trade liberalisation is still at an early stage. In a recent working paper (Breinlich and Cuñat (2013), we use the Canada-US Free Trade Agreement to carry out such a test.
Can heterogeneous firm models replicate increases in trade and productivity?
A quick look at the data suggests that the Canada-US Free Trade Agreement was associated with substantial trade and productivity gains in Canada. Average goods trade flows (Canadian exports plus imports to and from the US) increased by 118% over the period 1988 to 1996, while the increase in labour productivity in Canadian manufacturing was 30%. This compares to growth rates of only 44% (trade) and 17% (productivity) for the pre-liberalisation period 1980-88. As discussed, papers such as Trefler (2004) have also provided more rigorous econometric evidence that these increases can indeed be linked to the tariff cuts implemented under the trade agreement. Our goal is to see to what extent different heterogeneous firm models can quantitatively replicate these increases in trade and productivity, and would thus be potentially useful for forecasting purposes.
We start with a 'minimalistic' baseline model, in which Canadian tariff reductions reduce the profits of Canadian firms in their domestic market due to increased foreign competition. This crowds out the lowest-productivity firms, thus raising average industry productivity. At the same time, lower export US tariffs allow more Canadian firms to enter the export market and existing exporters can export more to the US. The same is true for US exporters, increasing overall bilateral trade flows.
While this baseline model gets the basic direction of effects right, it turns out that the predicted increase in trade for a given change in tariffs is way too large relative to the predicted increase in productivity. This is true even if we choose the parameters of the model to match the observed changes. For example, if we choose parameters to match trade flows, the model substantially underpredicts the productivity growth we observe in the data. This forecast bias is substantial. If we match trade flow increases, we obtain productivity increases of less than 2% (compared to 30% in the data). When we use parameter estimates obtained from the pre-agreement period (1980-1988) – as we would of course have to do for forecasting purposes – the results are even worse.
Where does the productivity increase come from?
This mismatch raises two possibilities. First, we might simply be too demanding of a rather simple model which abstracts from a number of factors which increase productivity and trade flows in the real world and have nothing to do with trade liberalisation (for example, ongoing technical change or business-cycle movements). To address this possibility, we 'clean' our data by removing all influences which are not related to tariff cuts, using techniques similar to Trefler (2004). This does help somewhat, but the basic finding that the model does not generate enough productivity gains remains.
A second possibility is that we might need to integrate additional mechanisms for productivity gains into the model. We experiment with a number of extensions, such as allowing the trade agreement to lead to the entry of new firms or to affect the overall wage level in Canada and the US. We also introduce intermediate inputs in our model which raise predicted productivity by allowing firms to use cheaper inputs from abroad. Finally, we generate within-firm productivity increases in response to trade, in addition to the productivity gains arising from the crowding-out of less productive firms described above. (The specific mechanism we use is a multiproduct firm set-up as in Bernard et al. (2011); but we conjecture that other mechanisms such as technology upgrading would yield similar results.)
It turns out that only the last of these modifications helps us significantly in improving the model’s performance. If we allow individual firms to become more productive, we are able to closely match the observed trade and productivity gains in Canada. Moreover, we also get good forecasts if we use parameter estimates obtained from the pre-agreement period. Again, this is important because in practice we will have to work with data prior to the particular free-trade agreement whose effects we want to forecast.
Overall, our results suggest that if we want to use the current generation of heterogeneous firm models for the purpose of forecasting the effects of trade agreements, we need to allow for sources of within-firm productivity increases. Otherwise, our models will tend to dramatically underestimate the effect of free trade on productivity and might consequently lead us to abandon promising free-trade initiatives.
- At least for the Canada-US Free Trade Agreement, there is independent evidence which suggests that within-firm productivity effects were actually very important (see Lileeva and Trefler 2011). Of course, this does not mean that within-firm productivity increases will be important in all contexts, and more standard heterogeneous firm models might do better in such cases.
- There is also evidence that within-firm effects were important in other liberalisation episodes (see, for example, Bustos 2011 for evidence of the impact of MERCOSUR on Argentinian firms).
In any case, our results are only a first step in the systematic evaluation of the quantitative predictions of heterogeneous firm models. Given that these models are consistent with a large number of documented facts about trade liberalisations, they certainly hold great promise.
Much more work is needed to find out which of the many mechanisms highlighted in the literature are important for accurate forecasting in practice.
Balistreri E, R Hillberry and T Rutherford (2011), “Structural estimation and solution of international trade models with heterogeneous firms”, Journal of International Economics, 83, 95-108.
Bernard A, S J Redding and P K Schott (2011), “Multiproduct Firms and Trade Liberalization”, Quarterly Journal of Economics, 126(3), 1271-1318.
Breinlich, H and A Cuñat (2013), “Tariffs, Trade and Productivity: A Quantitative Evaluation of Heterogeneous Firm Models”, CEPR DP 9579 and CES-ifo WP 4354, published in the Economic Journal, September 2016.
Bustos, P (2011), "Trade Liberalization, Exports, and Technology Upgrading: Evidence on the Impact of MERCOSUR on Argentinian Firms", The American Economic Review, 101(1), 304-340.
Kehoe, T J (2005), “An Evaluation of the Performance of Applied General Equilibrium Models of the Impact of NAFTA”, in T J Kehoe, T N Srinivasan, and J Whalley (eds.): Frontiers in Applied General Equilibrium Modeling: Essays in Honor of Herbert Scarf, Cambridge University Press.
Lileeva, A and D Trefler (2010), “Improved Access to Foreign Markets Raises Plant-Level Productivity … for Some Plants”, Quarterly Journal of Economics, Vol. 125(3), pp. 1051-1099.
Melitz, M J (2003), “The Impact of Trade on Intra-industry Reallocations and Aggregate Industry Productivity”, Econometrica, Vol. 71, November, pp. 1695-1725.
Melitz, M and S Redding (2013), "Gains from trade: Firms and productivity", VoxEU.org, 30 May.
Trefler, D (2004), “The Long and Short of the Canada-US Free Trade Agreement”, The American Economic Review, Vol. 94, September, pp. 870-895.