### Discussion Paper

## Abstract

The current economic crisis has witnessed a strong deceleration in the growth of international trade. This has been even greater in the cases of the European Union and the eurozone, where the rates of export growth have even reached negative figures. In this paper, the authors examine to which extent exchange rate volatility might account for the drop in the rate of growth of exports in the eurozone since the start of the crisis. To that end, the authors estimate export functions, augmented to include several measures of exchange rate volatility, for the four largest economies of the eurozone, i.e., France, Germany, Italy and Spain, for the period 1994:1–2014:4. In the empirical application, they make use of two alternative measures for exchange rate volatility, i.e., (i) the standard deviation and (ii) the conditional variance from the GARCH methodology, of the change in the logarithm of the exchange rate, for both nominal and real exchange rates, and in the latter case computed using as deflators both export prices and unit labour costs. The empirical results show no clear-cut evidence on the impact of exchange rate volatility on the exports of the countries analysed, suggesting that financial markets were developed enough so that exchange rate volatility does not hinder the evolution of exports.

## Comments and Questions

see attached file

First of all, many thanks for your comments. Notice that:

1. We present a suite of state-of-the-art unit root tests to support our results. Further unit root tests, including the KPSS test suggested by the referee as well as the DF-GLS test (see, Elliot et al, 1996) are made available ...[more]

... upon request but do not change the nature of the results reported here.

2. The GARCH model is used to generate the volatility series, which is then used in the main regressions. Moreover, the GARCH models are included as a robustness check on the use of the standard deviation as a measure of volatility and in themselves do not form the main point of interest. Indeed, the results in these tables are not GARCH model results and perhaps there is some confusion regarding this.

3. Finally, recall that in footnote 2 of the paper we write: “Some formal tests have been also performed, revealing there is no residual autocorrelation or heteroscedasticity; again, the results are not shown for space reasons, but they available from the authors upon request.”

The paper deals with an interesting and topical issue. The question has been the subject of numerous studies over the past decades for different countries. The paper is professionally written and the empirical results, economic conclusions and policy implications seem plausible. In what follows I suggest to the authors some ...[more]

... possible changes to improve the paper:

1. The paper reads basically as an empirical exercise as the theoretical framework for the relationship between exchange volatility and exports is not well established. As it is presented it is just a statistical nexus between the two variables in an otherwise very standard model and simple model. That may create some specification and omitted variable problems that could be addressed by the authors augmenting the model with other variables for the sake of robustness of the results.

2. The authors should clarify the value added brought up by their contribution to the literature. More specifically, they could try to place their paper within this burgeoning literature and signal the differences between their paper and other extant papers.

3. As the authors claim about the possibility of a non-linear (and possibly asymmetric) relationship, (they point to the existence of overshooting in the exchange rate) they could explore the existence of threshold and non-linear cointegration. My advice would be to test for threshold cointegration combining the methodology suggested by Gonzalo and Pitarakis (2002) to specify the number and location of possible thresholds and the threshold unit root test developed by Seo (2008) and Hansen and Seo (2002).

4. Another suggestion could be the use of the ARDL approach. This methodology developed by Pesaran et al (2001) has different advantages, as the endogeneity problems and inability to test hypotheses on the estimated coefficients in the long run associated with the Engle-Granger method are avoided; the long and short-run parameters of the model in question are estimated simultaneously, and which is more important in the present case, the ARDL approach to testing for the existence of a long-run relationship between the variables in levels is applicable irrespective of whether the underlying regressors are purely I(0) or I(1). Finally, the small sample properties of the bounds testing approach are far superior to that of multivariate cointegration.

References

Gonzalo, J. and Pitarakis, J. (2002) Estimation and Model Selection Based Inference in Single and Multiple Threshold Models. Journal of Econometrics 110, 319-352.

Pesaran, M. Hashem, Yongcheol Shin, and Richard J. Smith. (2001): "Bounds testing approaches to the analysis of level relationships." Journal of applied econometrics 16 (3), 289-326.

Shin, Y., Yu, B., Greenwood-Nimmo, M.J. (2013). Modelling asymmetric cointegration and dynamic multipliers in a nonlinear ARDL framework. In: Horrace, William, C., Sickles, Robin, C. (Eds.), Festschrift in Honor of Peter Schmidt. Springer Science & Business Media, New York.

Many thanks for your interesting comments. We comment now on each point:

1. In the paper, we do the same procedure than in most of the available empirical literature (we stress this in the introduction to the new version). Specifically, we start from the imperfect substitutes model, which is the ...[more]

... workhorse model of most empirical studies on international trade from a macroeconomic viewpoint, extended to incorporate the role of volatility. Of course, we could always add more variables to the model, and no doubt that some justification might be found for any of them. But, which variables should we include? And, which would be the criterion to include those variables? Our purpose is simply to assess whether the inclusion of volatility (i.e., our variable of interest) changes the results from an otherwise very standard empirical model.

2. We have incorporated a new sentence in the introduction on the intended contribution of the paper: providing some more recent evidence (including the crisis period) for the case of the eurozone, and using a number of alternative proxies (up to six for each country analysed) for exchange rate volatility.

3. Notice that we use dummy variables to capture the effects of the financial crisis and the start of EMU, but don’t argue for a nonlinear model as such. This would be an extension, but it is not clear (theoretically) what the threshold would be, and essentially would be a different paper.

4. Again, this could be an interesting future extension, but is beyond the scope of the current paper. Since our econometric procedure is correct, as you implicitly recognise, the ARDL estimation would not be strictly necessary, and should be taken as an extension to the current analysis. Notice also that we present up to 24 estimated equations for each of the four countries, which gives a total of 96 equations, so re-estimating everything by ARDL would increase enormously the length of the paper. As before, this would mean doing another paper.

The paper empirically analyzes a very interesting and relevant topic based on the relationship between exchange rate volatility and international trade. Authors use aggregate exports equations depending on the foreign real income, external competitiveness and exchange rate volatility and focus on the four largest economies of the Eurozone (Germany, France, ...[more]

... Italy and Spain).

Henceforth, the paper presents an adequate methodological approach and results show no clear-cut evidence on the influence of exchange rate volatility on the exports.

A strong point of the paper is the use of a wide set of different measures for exchange rate volatility: for nominal and real exchange rates, with alternative deflators in the case of real exchange rates; and applying both a simple measure, such as the standard deviation, and a more sophisticated one, such as the conditional variance from the GARCH methodology.

Many thanks for your generous comments. Indeed, we think that using several proxies for volatility is an “asset” of the paper. We want to stress this point in the revised version.