Exporters’ productivity and margins of trade deflection: theory and micro-evidence

Abstract

This paper examines how firm heterogeneity plays a role in trade deflection effects when exporters exit trading partners’ markets with trade-restrictive measures and deflect exports toward third markets with less stringent ones. We develop a Cournot-type three-country theoretical framework that highlights the role of firm productivity in trade deflection effects of trade-restrictive measures and empirically examine research hypotheses using a firm–product–destination level panel data of Korean exporting firms during the 1996–2010 period. We find that highly productive firms facing higher tariffs are more likely to deflect export to new third-country markets with lower tariffs as alternatives. However, once they enter the third destination, the positive effect of tariffs on the deflection of trade volume is less prominent for highly productive firms due to a lower trade destruction effect for them. Our results imply that the magnitude of trade deflection at both the intensive and extensive margins can be heterogeneous across firm productivity and multi-destination status.

KEYWORDS:

• Trade policy
• trade deflection
• exporter productivity

JEL Classification::

• F13
• F14

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Trade diversion effect, a term more frequently used in the literature and by the press, arises when an importing country reduces (or switches) imports from non-member exporting countries of a regional trade agreement to member exporting countries. In contrast, trade deflection effect occurs when an exporter reduces (or switches) exports from trading partners to a third country. The terms are similar in that trade is diverted due to a relatively higher trade policy implemented in the importing country compared to a third region.

2 Also refer to the following papers for trade diversion effects: Nicita (2019) for the trade confrontation between the U.S. and China, Ramasamy (1995) for ASEAN, Russ and Swenson (2019) for the Korea–U.S. FTA, Reynolds (2009) for U.S. trade liberalization, Kahouli and Maktouf (2015) for the Mediterranean area, and Saggi, Stoyanov, and Yildiz (2018) for FTAs during the 1980–2011 period.

3 We consider these conditions only for simplicity and thus obtain explicit solutions, implying that they are not necessary conditions to obtain the four main trade effects. The general conclusions will not change without these conditions, referring to Bown and Crowley’s (2007) implicit solutions.

4 Proposition 2.1 provides the explicit solutions for the general functional form of Bown and Crowley (2007). In our empirical tests, we focus on Proposition 2.1 (i) and (ii) rather than (iii) and (iv) due to data limitation. We leave empirical analyses for Proposition (iii) and (iv) for future studies.

5 Considering that each country sells domestically and country m produces, we can additionally obtain $\frac{\mathrm{\partial }{q}_k^k}{\mathrm{\partial }{\tau }_k^m}>0$, $\frac{\mathrm{\partial }{q}_j^j}{\mathrm{\partial }{\tau }_k^m}<0$, $\frac{\mathrm{\partial }{q}_m^m}{\mathrm{\partial }{\tau }_k^m}>0$, $\frac{\mathrm{\partial }{q}_m^k}{\mathrm{\partial }{\tau }_k^m}<0$, and $\frac{\mathrm{\partial }{q}_m^j}{\mathrm{\partial }{\tau }_k^m}<0$. These results imply that firm k reallocates its production from exports to country m to domestic sales and exports to country j, and firm j reallocates its production from exports to country k and domestic sales to exports to country m for an increase in country m’s tariff on country k. For firm m, $\frac{\mathrm{\partial }{q}_m^m}{\mathrm{\partial }{\tau }_j^m}>0$ shows that a tariff by country m against country k causes an increase in firm m’s domestic sales, which is the effects of protectionism. With $\frac{\mathrm{\partial }{q}_m^k}{\mathrm{\partial }{\tau }_k^m}<0$ and $\frac{\mathrm{\partial }{q}_m^j}{\mathrm{\partial }{\tau }_k^m}<0$, all these results imply that firm m reallocates its productions from exports to two partner countries to domestic sales for an increase in country m’s tariff on country k.

6 $\frac{({\tau }_k^m-1)}{(3{\theta }_k+{\tau }_k^m)}<1$as ${\theta }_k>0$. Therefore, $\frac{({\tau }_k^m-1){\theta }_k}{2(3{\theta }_k+{\tau }_k^m)(3{\theta }_k+1)}<\frac{{\theta }_k}{2(3{\theta }_k+1)}$.

7 Based on the purpose of the analysis of the trade deflection effect, this paper does not test Proposition 2.1(iii) or 1(iv). See Note 3 for more details.

8 We use median productivity as a measure of threshold level.

9 Fontagné and Orefice (2018) pointed out that highly productive firms will exit the market imposing the measure, but those complying with the imposed trade barrier may benefit from the reduced competition.

10 Using a restricted sample does not raise serious sample selection bias issues as the main topic of this paper is examining firm response given the condition of the relative size of tariffs between the imposing country and a third destination, particularly when tariffs imposed by the imposing country m on the exporting country k is greater than those imposed by the importing destination country j. However, for a robustness check, an empirical result using an unrestricted sample is provided in Table A6, which confirms the baseline result in Table 3.

11 Other conventional gravity variables, such as dummy variables for a common official language, a common border, and colonial ties, are excluded because most observations are zeros for Korea.

12 The NICE information service provides information on financial statements and the stock market of Korean firms listed on KOSPI, KOSDAQ, and externally audited markets.

13 As the export value in our empirical analysis is the predicted value from export credit insurance, one should be cautious when interpreting the coefficient of the effect of tariffs on trade deflection as the exact size of the impact of the tariffs, although one still can determine the direction and relative size of the impact within the country–HS2–year triplet.

14 As the number of instrumented variables and independent variables is identical, Hansen statistics from the test for overidentifying restrictions are not reported.