The impact of real exchange rate on international trade: Evidence from panel structural VAR model

Abstract

This paper investigates the importance of real exchange rates on export volumes by estimating a panel SVAR model using quarterly unbalanced panel data from 21 emerging markets over the 2005:Q1-2018:Q4 period. Although the results suggest no conclusive evidence that real exchange rate shocks do affect the export volumes in our sample of emerging markets, the responses of export volume to real exchange rate shocks are heterogeneous across countries in which commodity exporter countries, on average, have a lower response of exports to the real exchange rate movements. Furthermore, we find that while the magnitude of response of the export volumes to exchange rate shocks is positively related to exchange rate volatility, the higher export market penetration ratio helps insulate the economy from real exchange rate shocks. Overall, our results carry broad policy implications indicating that policymakers need to pay attention to the exchange rate volatility of their countries and expand their export competition in the world trade.

Keywords:

• Exchange rate volatility
• international trade
• real exchange rate
• panel SVAR

JEL Classifications:

• C22
• C23
• F14
• F31

Disclosure statement

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

Notes

1 See Gangnes and Assche (2018) for a review of the literature regarding the impact of integration to GVCs on the link between exchange rates and trade.

2 For a review of studies regarding the exchange rate volatility and trade flows, see Bahmani-Oskooee and Hegerty (2007).

3 According to the World Bank (World Bank 2016), approximately two-thirds of emerging and developing countries are commodity-exporters.

4 It is assumed that the structural shocks are orthogonal with respect to each other for each type. Put differently, the idiosyncratic shocks are mutually orthogonal to one another, as are the various common shocks to one another.

5 The i and t subscripts on the time and cross-section dimensions take into account the unbalanced panel data.

6 As suggested by Pedroni (2013), the imposed restrictions on the panel are consistent with the similar restrictions made upon the individual data if they were viewed as individual time series rather than as members of a panel.

7 The reason is that when we obtain estimates for the composite shocks ${\epsilon }_{it}$, and common shocks ${\overline{ϵ}}_t$, ${\widetilde{ϵ}}_{it}$ and ${\mathrm{\Lambda }}_i$ can be recovered using the orthogonality properties of the shocks.

8 As suggested by Pedroni (2013), we normalize the variances of both ${\epsilon }_{it}$ and ${\overline{ϵ}}_t$ to be unity, $E[{\epsilon }_{it,m}{\overline{ϵ}}_{t,m}]/E[{\overline{ϵ}}_{t,m}^2]$ is equal to simple correlation between ${\epsilon }_{it}$ and ${\overline{ϵ}}_t$.

9 Following Pedroni (2013), we re-scale the size of the idiosyncratic shocks in order to interpret all impulse responses as unit shocks.

10 For robustness, we generate an alternative real exchange rate volatility variable using ARCH/GARCH methodology following Zhang, Chang, and Gauger (2006), Baum and Çağlayan (2006), and Hooy and Baharumshah (2015). Table A2 in the Appendix presents the results from the robustness tests that support the baseline results.

11 In a county-level study for Brazil, Paiva (2003) reports that the significance of the exchange-rate movements on export volumes increases once the exchange rate volatility is controlled.