EUROPEAN UNION WORLD

European Central Bank Monetary Policy

Within- Sectoral Reallocation

We use sectoral data to test the hypothesis that the end of competitive devaluations has induced a restructuring process in the EA firms. We begin by describing the empirical approach and the data and then move on to the results. Finally, we perform a series of extensions and robustness checks.

 The Empirical Approach and the Data

We test the effects of the euro on within-sectoral restructuring using sectoral data from different countries. Ideally, one would like to use direct measures of reallocation, such as job creation and destruction, entry, exit, and so forth. Unfortunately, such measures can only be constructed from firm- level data and so are not available for a cross- section of countries. Accordingly, we use an outcome variable that should be closely related to reallocation (i.e., productivity growth). In fact, if reallocation and restructuring bring about productivity increases, then the country- sectors that restructured more should have recorded a higher growth rate of productivity. We measure productivity as real value added per hour worked. We also consider growth in employment (more precisely, the number of hours worked) growth: in fact, productivity increases might have been due simply to a reduction in the employment level, connected with the exit of the less- productive plants and workers, the reorganization of production, and off shoring.

One important feature of this approach is the inclusion of both country and sector dummies. Country dummies ensure that the results are not driven by specific country characteristics that might potentially be related to the devaluation measure: rather, we use within- country differences in sectoral growth rates to identify the parameters of interest. The same applies to sectors: we do not compare different growth rates of productivity across sectors, as these might be dictated by sectoral characteristics potentially related to the variables we use to classify them. As such, this approach is robust to the main criticisms of the cross- country regressions with aggregate data, such as omitted- variable bias and reverse causality. Although the inclusion of country and sector dummies controls for the most likely omitted- variable problems, one could still argue that we might just be capturing an underlying process that would have occurred even without the euro. For example, the intensifying competition from emerging countries might have forced restructuring regardless. Such a process might have been more pronounced precisely in those countries and sectors that relied more on competitive devaluations, potentially more vulnerable to such competition. This is indeed a very serious concern. To address it, we take the three countries that did not adopt the euro as a control group and compute the effect of the interaction for the EA in deviation from non- EA countries.

The idea is that the latter countries did not give up the possibility of devaluing but are similar to the EA countries from an economic point of view, because as members of the EU, they are subject to identical foreign trade rules, with the exception of the exchange rate. Differences in the degree of restructuring according to the interaction term can therefore be attributed to the euro. As discussed, this control group is probably the best available, although it can be criticized both for its small size and its not necessarily random selection. To make sure that our results are not totally dependent on the control group, we also estimate equation on EA members only-that is, considering the absolute effect rather than the deviation from the control group. In this case, we are not controlling for potential confounding factors. However, we still control for fixed country and sectoral attributes so that these estimates allow us to assess the extent to which our results depend on the control group.

In terms of the country- level indicator, we want to capture the reliance on competitive devaluations. From the theoretical standpoint, it is unclear whether real or nominal devaluation is the relevant variable. Consider a country that kept a fi xed nominal exchange rate with the DM but gained competitiveness by curbing price rises. For it, the euro should not represent much of a change, as the exchange rate was already stable, and using real devaluation might overstate its reliance on devaluations. On the other side, consider a country with relatively rapid price inflation that used devaluations to limit the effects on competitiveness. For such a country, appreciation was already under way before the euro, and using the nominal exchange rate would overstate the reliance on devaluations. These examples suggest that the ideal indicator should consider real devaluations that were due to changes in the nominal exchange rate. To capture this, in our basic specification, we introduce both the nominal exchange rate and the degree of relative producer price inflation in order to allow for potentially different dynamics of the two components of the real exchange rate. We test whether the coefficients of the two variables are opposite in sign and equal in absolute value, in which case the real exchange rate can be used directly.

For the sectoral indicators, we assume that price competition is more relevant in activities with a low human capital content (i.e., in which low- skilled workers are prevalent). The products of low- skill activities are likely to compete more in price than in quality relative to high- skill products. For a sector with low human capital content, the end of devaluations should have represented a stronger incentive to restructure; other things being equal, these sectors should have recorded higher productivity increases. Our main indicator is thus the skill content at the sectoral level. Following Rajan and Zingales (1998), in order to avoid endogeneity problems, we use the U.S. measure on the assumption that skill content is largely a technological characteristic, so the measure computed for the United States also applies to other countries.

This assumption is particularly suitable for the EA countries, whose level of development is comparable to the United States. In accordance with our interpretation, we use sectoral low- skill intensity-that is, (1- skill intensity). This makes it easier to read the regression results. We also experiment with other measures of sectoral dependence on devaluation. Following the same reasoning as before, high- R&D activities should also compete less on price and more on quality and technological content, reducing the price sensitivity of demand and hence the effects of exchange rate movements. Low-R&D activities should be characterized by greater price elasticity of demand, intensifying the response to terms of trade movements. We also use ICT intensity on the assumption that this is related to technological content. As before, we define sectors in terms of low- R&D and ICT intensity: (1- R&D content) and (1- ICT intensity), again computed for U.S. sectors.

Underlying our approach is the idea that in low human capital activities, the end to competitive devaluations has deprived EA countries of an instrument for meeting the competition from low- wage emerging economies. An alternative way to rank sectors, then, is to look directly at the importance of those economies in world trade. We take the most important of them, China, and compute its share of world exports in 1998. In this case, we are testing whether restructuring has been more intensive in countries that had relied on devaluations more heavily and in sectors where China’s export share was larger.

The bottom part of table 3.1 reports the correlation coefficients between the sectoral indicators. As expected, the correlation between the first three indicators is high, ranging from 0.6 to 0.8. That between China’s world market share and the others is negative. That is, the Chinese share is inversely related to the human capital content of production, but correlation is low in absolute terms: – 0.3 with ICT and skill intensity and – 0.1 with R&D intensity, suggesting that to see China simply as a low human capital good exporter might be to miss some important features of its economy. We also run the same regression for EA countries in the period before the introduction of the euro. The assumption is that at that time, the competitive pressures were mitigated by competitive devaluations. In this case, we expect no particular difference between the study and the control group. In the language of the policy evaluation literature, we make sure that we are not simply capturing preexisting trends and that the euro did indeed induce a structural break.