Till Labor Cost Do Us Part. On the Long Run Convergence of EMU Countries

Pancotto, Francesca; Pericoli, Filippo Maria Working Paper Till Labor Cost Do Us Part. On the Long Run Convergence of EMU Countries Quaderni - Working Paper DSE, No. 759 Provided in Cooperation with: University of Bologna, Department of Economics Suggested Citation: Pancotto, Francesca; Pericoli, Filippo Maria (2011) : Till Labor Cost Do Us Part. On the Long Run Convergence of EMU Countries, Quaderni - Working Paper DSE, No. 759, Alma Mater Studiorum - Università di Bologna, Dipartimento di Scienze Economiche (DSE), Bologna, https://doi.org/10.6092/unibo/amsacta/4482 This Version is available at: https://hdl.handle.net/10419/159599 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc/3.0/ Till labor cost do us part. On the long run convergence of EMU countries Francesca Pancotto Filippo Pericoli Quaderni - Working Paper DSE N° 759 Till labor cost do us part. On the long run convergence of EMU countries Francesca Pancotto * , and Filippo Pericoli  June 7, 2012 Abstract A sustainable long-run pattern in the relative competitiveness of euro area countries is a key factor for the survivorship of the monetary union. We analyze the issue focussing on unit labor cost dynamics using cointegration analysis for the whole economy and for the manufacturing sector separately. Our ndings show that the introduction of the euro has increased, rather than decreased, the distance among member countries, as measured in the metric of unit labor costs. Dispersion of productivity rather than wage compensation suggests that persisting idiosyncratic dynamics are driven by real factors, i.e. diverging technological patterns rather than by monetary factors, expressed by wage compensation. JEL codes: E31, O47, C32. Keywords: Unit labor costs, Convergence, Competitiveness, Manufacturing Sector. June 7, 2012 * Corresponding author: [email protected], Department of Communication and Economics, University of Modena and Reggio Emilia, Viale Allegri, 9 , Reggio Emilia Bologna, Italy,  Department of Treasury of Italy's Ministry of Finance, Via XX Settembre 97, Rome, Italy and CEIS Economics Foundation of Tor Vergata University, Via Columbia 2, Rome, Italy 1 1 Introduction The sustainability of the monetary union is guaranteed in the long run if member countries converge in terms of relative competitiveness. Competitiveness indeed aects not only the rate of growth of a single member state but also the economic cohesion of the union, given the high level of interdependence associated to the single currency. Therefore it becomes essential to investigate over the existence of persistent divergences that might jeopardise its future. This fact has been largely recognized by the European Central Bank that has introduced a mechanism of systematic surveillance of member states competitiveness, with the aim of maintaining a common framework capable of identifying and correcting imbalances. Since 2007, the European Central Bank monitors the state of convergence of the member states by means of seven indicators of competitive gaps: current account decits, ULC, the stock of a country's net external debt as a ratio to GDP, the rate of ination, the current account decit as a ratio to GDP, the private and government debt ratios, the stock of credit to the private sector (ECB, 2005, 2007). Any divergence of these indicators from the union average, is a signal that should be taken into account when evaluating sustainability. Our choice is to analyze unit labor costs (ULC thereafter), that measure the average cost of labor per unit of output: it informs on the relative dynamics of wages and productivity in the countries of the union and on the relationship among them. It represents a direct link between productivity and the cost of labor used in generating output. As it is an important and relatively stable component of ination dynamics, with respect to more volatile determinants of ination such as raw materials and commodity prices, it gives a long run idea on how wages ination is steadily inuencing the general price dynamics. In the perspective of a monetary union, the relationship between labor costs among member countries takes an even more important role as it expresses the degree of homogeneity, integration (and/or complementarity) of the member states. Bertola (2008) shows some concern related to the appropriateness of ULC as an indicator of relative competitiveness of euro area members and wage dynamics, in particular in the comparison between tradable and non-tradable sector: his concerns are basically twofold. 2 First, the comparability of data among member countries is aected by a low degree of homogeneity of data collection mechanisms; secondly, the Balassa-Samuelson eect can bias the information contained in the available data. Notwithstanding these issues, we believe that an inspection of the behavior of ULC for the total economy and the manufacturing sector, could give important insights on the dynamics of competitiveness of the currency union members. In a recent paper Dullien and Fritsche (2008) analyze ULC trends in the euro area with the aim to evaluate the degree of convergence reached within the euro area in terms of both wage and productivity trends. They rst examine ULC developments before and after the introduction of the single currency and secondly compare the performance of euro area countries with other currency unions, namely the federal states in the United States of America and the Länder of the Federal Republic of Germany. They implement a cointegration approach on ULC growth rates and test convergence with respect to the union average. Their analysis nds cointegration and thus convergence of ULC but at the same time the comparison with the performance of the other currency unions is not in favor of euro area, where deviations from area-wide averages are much larger than in US regions as well as in German Länders. Moreover, it is of their concern, the presence of a tendency towards deviation in the last years of the sample, in particular for Germany. Another contribution in this direction is the one of Tatierska (2008), which disaggregates ULC in 4 sub-sectors and uses quarterly data from 1990 up to the second quarter of 2007. She assesses cointegration mainly by means of Engle and Granger (1987) methodology and a panel Pedroni test (Pedroni, 1999), using euro area countries and comparing them with newly entered countries, namely Slovakia, Czech Republic, Poland, Ungary. She explores convergence by means of cointegration methodology of dierent economic sectors, but focusing on the convergence of newly entered countries with respect to the rest of the members. She nds evidence in favor of cointegration for almost all countries. In this work, we extend these contribution in various directions. First, we investigate the existence of a long run relationship with the Johansen (1988) approach, which is more general than the Engle and Granger (1987) methodology implemented by Tatierska (2008) for various sectors. We verify cointegration with λ -max and Johansen trace tests between 3 national ULC of each country and the area average. We also test the hypothesis of weak exogeneity of euro area ULC (excluded the i-th country) in the model for ULC for the generic i-th country: the rejection of this hypothesis would imply that the extent of disequilibrium in the i-th country aects the adjustment toward the equilibrium of ULC in the remaining euro area countries. Third, we verify whether the cointegrating vector has an economically desirable content, i.e. it is of the type (1,-1) which would imply similar long-run growth of ULC growth rates and consequently a stable relative competitiveness relationship among the countries of the area. We also include tests of stability of parameters, as presented by Juselius (2006), based on the recursive likelihood on both the cointegration parameters and the feedback mechanisms.We also analyse the results of the estimates of the cointegrating vector, dierently from Dullien and Fritsche (2008) which analyze instead mainly the loading factors as the main drivers of the adjustment process. Because of the low power of bivariate cointegration tests, we also performed panel cointegration tests, rst Westerlund (2007) test in its normal version and in its modied version with bootstrapped critical values, in order to take into account cross section dependences. We also conduct the panel extension of the Johansen trace test proposed by Larsson et al. (2001) using both standard and bootstrapped critical values, simulated in order to correct the latter for cross-sectional dependence as well. We perform the aforementioned analysis on both the whole economy and the manufacturing sector separately, departing from Dullien and Fritsche (2008) which only analyse ULC convergence of the total economy. We believe that exploring convergence of ULC in manufacturing sector increases profoundly the comprehension of the convergence of the european productive system, which is what ultimately matters in the understanding of the future sustainability of the union. Moreover, we analyse a dierent time span. Indeed, we preferred to give higher weights to the years following the monetary union, once the minimum number of observations required to ensure an appropriate inference was guaranteed. We believe that a sample of this sort would help in understanding more clearly the evolution of ULC in the last years and ultimately explore convergence in the light of the more recent evolution of economies the euro area. 4 Finally, a work similar to our is a recent paper of Herwartz and Siedenburg (2011) which test convergence of ination dierentials in the monetary union using monthly relative normalized ULC indices for the manufacturing sector only from the IMF's international nancial statistics, for Austria, Belgium, Finland, France, Germany, Ireland, Italy, and Spain. Similarly to ours, they use manufacturing sector data, interpreting ULC ination dierentials in this sector as a direct indicator of the relative evolution of external competitiveness within the monetary union. They test convergence for the years from 1979 to 2010 and then separately in two subsamples, before and after the monetary union. The rest of the paper is organized as follows. In Section 2 we explore literature contributions related and relevant for our work. In Section 3 we describe the database used for the analysis with some preliminary statistical analysis and present the empirical methodology implemented in the following section. In Section 4 we report tests and estimates results. In Section 5 we draw some conclusions and policy implications. 2 Literature Review In a seminal paper, Baumol (1986) explains how convergence in industrialized economies is achieved when innovation and investment in one country generates spillover eects on near-by countries. Countries at a lower level of development absorb part of the eects of innovation and increase their productivity, fostering income growth and wage increases. Innovation and investment spillovers generate such eect if technology is identical or at least comparable in all the countries involved in the process. Indeed, countries with a lower technological advancement may not be completely capable to take advantage of these spillover eects and thus to catch up with the productivity advancements of the leader. The eects of this type of misalignment could be observed in the dynamics of labor costs, aected by productivity, by denition. If we hypothesize that tradable sector goods are more aected by innovation spillovers than non-tradable sector, we should observe a dierent behavior of the two labor costs when analyzed separately. Convergence in the tradable sector should consequently be more pronounced if the member countries are 5 moving towards a similar technological pattern. The existence of diverging technological patterns, could be explained by cumulative knowledge and increasing returns of scale driving innovation and technological change as in Arthur (1989). Indeed, countries characterized by a higher initial technological development, and/or knowledge advancement, would be already in a diverging path leading to a systematic better competitiveness performance, once the scope for beggar-thy-neighbor policies are removed, as it is the case for economies with a unitary monetary policy. Krugman (1991) points at pecuniary external economies as the source of possible divergence among regions in a core-periphery model characterized by increasing returns in the manufacturing sector. Convergence or divergence would be determined by the elasticity of the manufacturing labor force with respect to wage. If the share of manufacturing workers decreases as the relative wage in the central region increases, the dynamics would be convergent. Indeed, workers will migrate out of the region with the larger work force. On the opposite, if the share of manufacturing workers in central region increases with its relative wage, workers will migrate into the region that already has attracted more workers, thus increasing the extent of divergence. In this case wages would be steadily higher in the economy with a larger manufacturing market. In the peripherical regions, in order to guarantee employment, it would be required a negative wage dierential, that would be permanent. The aforementioned theoretical contribution explains how the the comparison between tradable and non-tradable ULC can play an important role in signaling an eventual divergence between EMU countries. Another channel of cross-country interaction might arise from the possibility that ULC increases in the non-tradable sector impact ULC in the tradable sector. Indeed, tradable goods are subject to higher degree of international competition and consequently adjust more strongly to shocks and uctuations from international markets. Non-tradable sectors instead, can benet from a more protected price dynamics and consequently have guaranteed a higher average level of wages. Salido et al. (2005) explore determinants and macroeconomic implications of persistent ination dierentials in Spain within EMU. They 6 show that aggregate demand for non-tradable goods and real-wage rigidities are crucial in explaining diverging price developments in Spain. This is due to the fact that ULC in non-tradable sector aect productions costs of tradable goods and reduce competitiveness in the tradable sector as well. Relatedly, Zemanek et al. (2010) investigate over the persistency of intra-EMU current account decits. In particular, they assess the impact of structural reforms in the public and the private sector onto current account balance. They nd that current account divergences may have been generated by inationary pressures originating in the non-tradable sector. Non-tradable goods are used as inputs for tradable goods, thus inuencing the price of tradable goods as well; moreover, wages in the manufacturing sector would imitate wage increases realized by workers employed in the service sector (where wages are more rigid). They call this mechanism reversed Balassa-Samuelson eect, "... where rising wages in the non-tradable sector trigger wage adjustment in the traded goods sector, which might reduce the current account balance (Zemanek et al., 2010)." In a dierent dimension, the comparison between tradable and non-tradable ULC, are relevant in the debate on the impact of wage developments in the public sector onto convergence. Public sector wages account on average for more than 10% of GDP and more than 20% of total compensation of employees. Clearly, public wage increases constitute a strong signal for private sector wage negotiations: the larger the public sector is, compared with the tradable sector, the stronger will be the signal. Hence, the larger the public sector, the more important, and the more challenging, will be its role in the overall evolution of cost competitiveness (Trichet, 2011). Empirical evidence supports the idea of the relevance of public sector wages in driving private wage-agreements in many euro area countries. Such spillovers seem to be particularly important in countries that have experienced high and volatile public wage growth. Other analyses conrm the public sector wages may be responsible for rapid increases in ULC and misaligned intra-euro area competitiveness (Perez and Sanchez, 2010; Lamo et al., 2008). 7 nonetheless on a fruitful pattern. With respect to this point, what we are interested in whether the countries are moving on a common path or trying eorts in that direction, independently from the positive or negative trend of such path. Clearly, the desired outcome of policy makers would be that countries converge towards a common and fruitful dynamicsSecondly, cointegration tests are sensitive to the particular sample considered. Indeed, we decided to employ yearly data for the timespan 1980-2011 for the total economy, 1979-2010 for the manufacturing sector. This is a rather homogeneous period, approximatively coincident with the `great moderation'(Stock and Watson, 2002). This is indeed a period characterized by a relatively stable macroeconomic environment, and at the same time it guarantees a minimal number of yearly observations for applying the Johansen's methodology and estimating the cointegrating vectors in a bidimensional system. 4 Results Cointegration tests The rst step of the analysis consists in determining the cointegrating rank of the bidimensional system constituted by ULC in the i-th country and ULC in the rest of the euro area. In our case, the cointegrating rank can be 0, 1 or 2. The only economically interesting case is that of a system with rank equal to 1, which means that even though both series are non-stationary, there exists a linear combination of the two variables which is stationary (or trend-stationary). In our case, the presence of a system of rank 1, would imply a long run and stable, equilibrium relation between ULC in the i-th country and the rest of the euro area. The results of the sequential testing procedure proposed by Johansen and Juselius (1990) of the trace and λ−max cointegration tests are reported in Table 1. For the total economy, the trace and the maximum eigenvalue statistics indicate that there exists cointegration in all the countries included in the sample, at 5% signicance level. For Austria, France, Germany, Greece, Ireland, Luxembourg, Netherlands, Portugal 14 and Spain we accept the hypothesis of cointegration even without a linear trend in the cointegrating space, while in the case of Belgium, Finland and Italy it is necessary to include a linear trend in the long run behavior of the system to achieve cointegration. As Table 2 shows, the results for the manufacturing sector are less favorable. We have excluded Luxembourg due to a high number of missing observations. With regards to the remaining countries, we accept the hypothesis of cointegration for Austria, Belgium, France, Germany, Greece, Ireland, Italy, Netherland and Spain, while in the case of Finland and Portugal we reject it. In detail, there is cointegration without a linear trend in the cases of Austria, Belgium, Greece, Ireland, Italy and Spain, while in the cases of France, Germany and Netherlands we included a linear trend in the cointegrating space to achieve cointegration. The existence of a statistical cointegrating relationship cannot be considered as a proof of sustainability because the shape of the cointegrating space may economically unsustainable in the long run. The presence of a trend in the cointegrating space is economically relevant in the analysis of the convergence because it implies that the trajectories are systematically diverging. In other words, the cointegration analysis may be regarded as a test of sustainability only under very special assumption on the values of the parameters of the cointegrating vector. Cointegrating vectors In this section we explore the parameters of the cointegrating vectors previously estimated. In order to facilitate the reading of our ndings, we rewrite Eq. (6) in scalar form. Let us dene ulci,t the logarithm of ULC in the i-th country at date t and by ulcemu−i,t the logarithm of ULC in the remaining countries of the euro area. Equation (6) can be rewritten as: ∆(ulci,t) =α11 (β11ulci,t−1+β12 ulcemu−i,t−1+β13 t+β14) +c11 ∆(ulci,t−1) + c12 ∆(ulcemu−i,t−1) + c13 +1,t (7) ∆(ulcemu−i,t) =α21 (β11ulci,t−1+β12 ulcemu−i,t−1+β13 t+β14) +c21 ∆(ulci,t−1) + c22 ∆(ulcemu−i,t−1) + c23 +2,t (8) 15 The rst element to consider is the evaluation of the sign and the value of the coecient β12 . A negative sign for this coecient suggests that in the long run there exists a positive loglinear function which links ULC in the i-th country with ULC in the euro area considered as a whole. Secondly, this relationship is stable and converging, the more this coecient is close to -1. On the contrary, the country would be on a systematically diverging path. The signicance of a deterministic trend in the cointegrating vector may cast some doubt over the economic sustainability of the observed dynamics. A stable relationship up to a deterministic trend in the cointegrating vector would imply a stable diverging dynamics in ULC dynamics between one country ULC and the rest of the area. Table 3 reports the cointegrating vectors obtained from the reduced rank estimate of the VECM models normalized on ULC in the i-th country. As regards the estimates conducted for the total economy, results show that in all cases the coecients have the right negative sign. Indeed, Germany and Austria are characterized by a stable tendency toward a relative decrease of ULC as their estimates for the coecient β12 are smaller than one. Luxembourg, Netherlands and Spain have a diverging pattern but less pronounced than that of France Greece, Ireland and Portugal, for which the coecient is way above 2. Finally the parameter estimates for Belgium, Finland and Italy are not directly comparable in terms of relative competitiveness due to the presence of a signicant parameter for the linear trend in the cointegrating space. This suggests that these three countries have the most diverging dynamics out of the countries of the area. The results are dierent in the case we consider relative ULC dynamics in the manufacturing sector alone. In this case we nd that Austria, Germany and Ireland are characterized by a cointegrating vector with the wrong `positive' sign: the higher ULC in euro area, the lower in these two countries. Germany and Austria exhibit a stable tendency toward increasing their relative competitiveness, which is positive for the future of their manufacturing sector, but relatively negative in terms of considering convergence with respect to the rest of the union. The positive sign of the coecient that we observe fore Ireland β12 is in line with the substantial divergence of the productive system of this country from the rest of the union, which is also well visible in the panel of Fig. 3 related to Ireland. 16 In the case of France and the Netherlands, the cointegrating vector shows a negative and lesser than one coecient, which means that they are in a converging and stable path with respect to the euro area, notwithstanding the presence of a signicant trend in the cointegrating space whose coecient is nonetheless trascurable in value. For what concerns Italy, ULC in manufacturing sector appears to converge more than ULC in total economy: the cointegrating vector shows the `right' negative sign even though the coecient is larger than one. This conrms what can be observed in Fig. 3 in the panel related to Italy: the ratio of italian ulc over euro area is approaching one for the manufacturing sector, while the same does not happen for total economy. This results may be explained by a catching up pattern that italian manufacturing system has been experiencing after the introduction of the euro thanks to the internationalization of italian productive system. A `correct negative' sign of the coecient β12 is present for Belgium and Spain as well, but for these two countries, the evolution of ULC for manufacturing sector and total economy are similar, as it can be seen again from Fig. 3 and can be conrmed from the similar results in terms of cointegrating vector estimates of the total economy (See in the top panel 3. Lastly, we denote the lack of a stable cointegrating relationship in the manufacturing sector ULC for Finland and Portugal: for these two countries it is not even possible to identify a stable relationship with the rest of the area - not even diverging. In the case of Luxembourg we cannot perform the estimation as data available are not sucient to perform a robust cointegration analysis. Weak exogeneity tests After having estimated the VECM models for the total economy and for the manufacturing sector, we proceeded to test some economically relevant hypothesis starting from the unrestricted version of the models. First we have conducted a test of weak exogeneity by verifying the likelihood of the assumption that, in the equation for ULC dynamics in the euro area, the loading factor of disequilibrium in the i-th country is equal to zero, i.e. we test that α21 = 0 , in Eq. (7). The results of the tests (table 4) imply that the 17 hypothesis of weak exogeneity is always rejected by the data for the total economy as well as for the manufacturing sector alone. This result may seem counterintuitive in a normal setting given that one generally expects that a small country such as Ireland or Belgium should not aect ULC dynamics of a big country such as Germany. However, our model is deliberately not structural as our goal consists in examining long run tendencies in ULC dynamics rather than understanding real data generating processes. This means that the rejection of the hypothesis of weak exogeneity should not be regarded as an evidence of the economic importance of a given country. Rather we believe that there may exist common factors which drive ULC dynamics in small as well as in big countries and that these factors render ULC dynamics in the remaining euro-area countries error-correcting with respect to disequilibrium of a given country. This result justies the adoption of the multivariate approach to cointegration by Johansen (1988), ruling out the possibility of conducting inference within the Engle and Granger's (1987) univariate framework. Relative convergence tests We have also tested the hypothesis that the β in Eq. (6) vector is of the type (1,-1), which means that the elasticity of ULC in the i-th country with respect to ULC in the area is unitary. From an economic point of view this means that the relative competitiveness of a given country with respect to the euro average is constant in the long run. Notwithstanding some limitations this test can be assimilated to a test of economic sustainability, for a given country, of the adhesion to the currency union. From the results of the test reported on table 5 it emerges that the hypothesis of relative convergence is always strongly rejected by the data. This means that even if we did nd a stable statistical relation between country i and euro area ULC, the shape of the cointegrating vector is such that euro area countries exhibit tendency to diverge in terms of relative competitiveness. These diverging dynamics may produce unsustainable eects on intra-area trade balances and resource allocations given that ULC represent the most important factor driving producer prices. 18 Parameters constancy We have assessed the stability of estimated parameters over the timespan considered, performing the test based on the recursive log likelihood function in the X-form, as described in Juselius (2006). With regards to the total economy (see Figure 5), the tests indicate that estimated parameters are substantially constant over time in ten out of the twelve countries considered, the only exceptions being Italy and Ireland. On the opposite, in the manufacturing sector, the tests reveal that for all nine cases (see Figure 6) where we have found cointegration, the parameters suer from non constancies and thus measure only average eects (Juselius, 2006). In detail, it emerges that the models for the manufacturing sector have been rather volatile at the beginning of the 1990s and at the beginning of the 2000s, after the introduction of euro. Panel cointegration tests In this section we verify the existence of a cointegration relationship among the countries of our sample by means of a panel cointegration test procedures: they allow indeed the analysis of both the time-series and the cross-sectional dimension of the dataset. First, we conducted the four standard residual-based panel cointegration tests proposed by Westerlund (2007) in the version with standard critical values and the one with bootstrapped critical values, which are implemented when cross-sectional dependency is suspected. The null hypothesis of residual based panel cointegration tests (Westerlund, 2007) is that long run residuals of the cointegrating regression are non stationary. Results are reported in Table 6: in the column named p-value we report the probability of accepting the null hypothesis of no cointegration with standard critical values (without correction for cross sectional dependencies). The column named robust p-value instead reports the probability of accepting the null of no cointegration, using bootstrapped critical values, which also in this case are calculated as they may take care of potential cross sectional dependencies. From the table we observe that the standard statistics accept the null of no cointegration six times out of eight for the total economy, and three times out 19 of eight for the manufacturing sector, at 5 per cent signicance. On the contrary, if we look at robust p-values we can never reject the null of no cointegration (all 16 cases): results of the Breusch-Pagan LM test(Breusch and Pagan, 1980) conrm the presence of cross sectional dependence among the countries in the panel and consequently ensures the appropriateness of the use of bootstrapped critical values as appropriate statistics to verify panel cointegration in this case. We also conducted the likelihood-based panel cointegration test proposed by Larsson et al. (2001), that is based on the average of the individuals trace statistics standardized with asymptotic moments. The null hypothesis of this test is constructed in the same way of Johansen (1988) test, and under the null the statistics is distributed as a normal standard. Results of this test are reported in Table 7. In order to calculate the Larsson statistics, we needed to compute the average of trace statisics of the individual countries, and use the asymptotic mean and variance as in Breitung (2005) that we report in the table. The resulting statisics is reported in the Table in the row where Larsson statistics is indicated. The value of the statistics in the case of r= 0 , corresponds to the case when the cointegration rank is zero. As the statistics in this case is 27.11, which is larger than the corresponding critical value at 95 per cent, we can reject the hypothesis of rank zero of the cointegration matrix. The procedure considers then the hypothesis of cointegration rank equal to one: given the results we can also reject the hypothesis that the cointegrating rank is 1 (value of the statistics 4.36). The same result holds for Model 2. Consequently, according to the result of this test, we cannot identify the presence of panel cointegration in the countries considered. In the case when cross sectional dependency is taken into account, boot strapped critical values are calculated and the interval is reported in the table 10 . There is 95 per cent of probability that the statistics lays in the interval reported in square brackets, which tells us that the hypothesis of cointegration is rejected also in the case when bootstrapped critical values are considered. On the whole, the tests reject the hypothesis of a common cointegrating rank among the 10 The procedure to obtain the bootstrapped interval has been obtained adapting the procedures written by Vinod and da Lacalle (2009) and using the R package `meboot' 20 countries of the euro are and indicate that the hypothesis of no panel cointegration can be accepted at high probability levels, when cross sectional dependency is considered, whose presence we have veried with the use of Breusch-Pagan LM test(Breusch and Pagan, 1980). Thus the panel cointegration analysis casts further doubts over the convergence and ultimately sustainability of diverging dynamics of national ULC within the euro area. 5 Concluding remarks We analyse ULC dynamics in the total economy and the manufacturing sector, trying to give an insight on the long-run sustainability of the monetary union. Our analysis shows that euro area countries are characterized by diverging tendencies in ULC dynamics which result in persistent dierences in competitiveness, which have increased rather than decreased with the introduction of a common currency. This nding is true for the economy as a whole but even more for the manufacturing sector. This result is in line with the ndings of Belke and Dreger (2011) which stress the increase in competitiveness imbalances in member countries since the introduction of the monetary union, emerging through strongly heterogeneous current account decits. They suggest that these imbalances originate from competitiveness idiosyncracies, where they measure competitiveness with ULC, as we do. They consequently suggest a policy of reduction of ULC in those countries with lower competitiveness. We suggest a dierent interpretation of ULC imbalances in the euro area. In order to do so, we explore the variability of the components of ULC- wage and productivity components - and explore its dynamics. We nd out that heterogeneity is indeed larger in productivity rather than in wage compensation, and much strongly for the manufacturing sector than for the total economy. Moreover the trend is increasing in particular in the recent years. We may interpret these results suggesting that the monetary union has generated real exchange rate appreciation that has reduced competitiveness of countries more specialized in low value added production. This eect has generated current account decits that 21 have created in the long run a spiral of competitiveness loss for these countries, also due to international competition much more erce for these types of products. Our results are in line also with the ndings of Fischer (2007) which nd ULC ination divergence in the manufacturing sector in the years following the monetary union and similarly identify a pattern of loss of competitiveness of some countries with respect to Germany in particular. Although the industrial sector is generally more prone to competition in prices, costs convergence almost never guarantees automatic convergence of productive dynamics, ultimately of technological trajectories(Dosi, 1988). In order to test this hypothesis, a more specic analysis of the situation of industrial competitiveness disaggregated at the sectoral level would be required, but this goes beyond the scope of the present analysis. We believe nonetheless that this contribution suggests a step up in the policy tasks of the union, plannig common industrial policies with a look at guaranteeing a sustainable competitiveness relationship among members. 22 References Arthur, W. B. (1989, March). Competing technologies, increasing returns, and lock-in by historical events. Economic Journal 99 (394), 11631. Baumol, W. J. (1986). Productivity growth, convergence, and welfare: What the long-run data show. The American Economic Review 76 (5), 10721085. Belke, A. and C. Dreger (2011). Current account imbalances in the euro area: Catching up or competitiveness? Technical report. Bertola, G. (2008, November). Labour markets in emu - what has changed and what needs to change. CEPR Discussion Papers 7049, C.E.P.R. Discussion Papers. Breitung, J. (2005). A parametric approach to the estimation of cointegration vectors in panel data. Econometric Reviews 24 (2), pp. 151173. Breusch, T. S. and A. R. Pagan (1980). The lagrange multiplier test and its applications to model specication in econometrics. Review of Economic Studies 47 (1), 23953. Dosi, G. (1988). Sources, procedures, and microeconomic eects of innovation. Journal of Economic Literature 26 (3), pp. 11201171. Dullien, S. and U. Fritsche (2008). Does the dispersion of unit labor cost dynamics in the emu imply long-run divergence? International Economics and Economic Policy 5 (3), 269295. ECB (2005). Ination dierentials in the euro area: Potential causes and policy implications. Technical Report , 1 30. ECB (2007). Speech by Lucas Papademos, vice president of the ECB. The ECB and its Watchers IX, Frankfurt am Main, 7 September 2007 23 . Engle, R. F. and C. W. J. Granger (1987, 2). Co-integration and error correction: Representation, estimation and testing. Econometrica 55 , 251276. 23 Austria 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 2.5 Belgium 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 2.5 Finland 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 2.5 France 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 2.5 Germany 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 2.5 Greece 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 2.5 Ireland 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 2.5 Italy 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 2.5 Luxembourg 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 2.5 Netherlands 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 2.5 Portugal 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 2.5 Spain 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 2.5 Figure 5: Recursive log likelihood function: total economy 30 Austria 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 Belgium 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 France 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 Germany 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 Greece 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 Ireland 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 Italy 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 Netherlands 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 Spain 1987 1993 1999 2005 2011 0.0 0.5 1.0 1.5 2.0 Figure 6: Recursive log likelihood function: manufacturing sector 31 Table 1: Johansen-Juselius Cointegration Test. ULC total economy, 1980-2011. Trace λ -max Model 1 Model 2 Model 1 Model 2 H0:r= 0 H0:r≤1H0:r= 0 H0:r≤1H0:r= 0 H0:r≤1H0:r= 0 H0:r≤1 H1:r= 1 H1:r= 2 H1:r= 1 H1:r= 2 H1:r= 1 H1:r= 2 H1:r= 1 H1:r= 2 Austria 39.20*** 0.65 51.37*** 11.98 38.55*** 0.65 39.39*** 11.98* Belgium 42.20*** 4.83** 46.25*** 6.23 37.37*** 4.83** 40.03*** 6.23 Finland 38.27*** 8.08*** 52.94*** 8.25 30.19*** 8.08*** 44.69*** 8.25 France 29.55*** 1.12 55.70*** 10.23 28.43*** 1.12 45.47*** 10.23 Germany 39.51*** 0.20 53.39*** 14.01** 39.30*** 0.20 39.38*** 14.01** Greece 33.08*** 2.74* 41.52*** 7.34 30.33*** 2.74* 34.18*** 7.34 Ireland 34.72*** 3.42* 51.41*** 10.39 31.30*** 3.42* 41.02*** 10.39 Italy 35.54*** 5.30** 46.28*** 7.64 30.24*** 5.30** 38.64*** 7.64 Luxembourg 36.08*** 0.30 45.87*** 4.72 35.78*** 0.30 41.15*** 4.72 Netherlands 37.21*** 2.05 50.31*** 7.30 35.16*** 2.05 43.02*** 7.30 Portugal 33.86*** 1.72 38.05*** 3.53 32.14*** 1.72 34.52*** 3.53 Spain 54.93*** 2.21 55.70*** 2.92 52.72*** 2.21 52.78*** 2.92 Notes:*, **, ***, denote statistical signicance at the 10%, 5%, 1% levels, respectively. 32 Table 2: Johansen-Juselius Cointegration Test. ULC manufacturing sector, 1979-2010. Trace λ -max Model 1 Model 2 Model 1 Model 2 H0:r= 0 H0:r≤1H0:r= 0 H0:r≤1H0:r= 0 H0:r≤1H0:r= 0 H0:r≤1 H1:r= 1 H1:r= 2 H1:r= 1 H1:r= 2 H1:r= 1 H1:r= 2 H1:r= 1 p-val H1:r= 2 Austria 22.49*** 2.61 32.67*** 12.79** 19.88*** 2.61 19.88** 12.79** Belgium 24.74*** 3.73* 25.62* 4.51 21.01*** 3.73* 21.11** 4.51 Finland 20.58** 4.06** 25.57** 7.06 16.52** 4.06** 18.51* 7.06 France 32.47*** 5.98** 37.21*** 10.69* 26.50*** 5.98** 26.53*** 10.69* Germany 26.61*** 5.06** 32.50*** 9.82 21.55*** 5.06** 22.68** 9.82 Greece 20.26*** 2.87* 26.75** 9.32 17.39** 2.87* 17.43* 9.32 Ireland 30.27*** 3.06* 37.32*** 10.03 27.21*** 3.06* 27.29*** 10.03 Italy 17.94** 1.28 23.26 6.60 16.66** 1.28 16.66 6.60 Luxembourg ° Netherlands 28.50*** 9.58*** 29.92** 9.60 18.93*** 9.58*** 20.32** 9.60 Portugal 19.57** 3.93** 23.01** 7.32 15.63** 3.93** 15.68 7.32 Spain 22.31*** 2.18 30.81** 9.43 20.13*** 2.18 21.39** 9.43 Notes:*, **, ***, denote statistical signicance at the 10%, 5%, 1% levels, respectively. ° We could not perform the test for Luxembourg because data availability was not sucient to perform the test. 33 Table 3: Cointegrating Vectors: total Economy and manufacturing Sector. Total Economy, 1980-2011 Ulc UlcEU ( β12 ) t-stat Constant ( β14 ) Trend ( β13 ) t-stat Austria 1 -0.60*** [-9.93] -1.07 - - Belgium 1 -4.42*** [-6.11] 8.15 0.02*** [-6.00] Finland 1 -6.47*** [-7.90] 12.63 0.04*** [4.63] France 1 -2.20*** [-8.36] 3.07 - - Germany 1 -0.29*** [-4.11] -1.91 - - Greece 1 -3.58*** [-12.06] 6.92 - - Ireland 1 -3.03*** [-10.44] 5.38 - - Italy 1 -16.99*** [-5.68] 38.13 0.10*** [3.18] Luxembourg 1 -1.94*** [-17.89] 2.49 - - Netherlands 1 -1.95*** [-13.86] 2.45 - - Portugal 1 -3.33*** [-13.99] 6.12 - - Spain 1 -2.01*** [-32.22] 2.68 - - manufacturing Sector, 1979-2010 Ulc UlcEU ( β12 ) t-stat Constant ( β14 ) Trend ( β13 ) t-stat Austria 1 0.56** [2.19] -4.28 - - Belgium 1 -2.40*** [7.65] 3.82 - - Finland - No coint. France 1 -0.54*** [-3.38] -1.51 0.01*** [3.16] Germany 1 0.41 [0.81] -3.73 0.00 [-1.26] Greece 1 -5.26*** [-6.14] 11.85 - - Ireland 1 2.77*** [8.36] -10.16 - - Italy 1 -5.16*** [-5.46] 11.55 - - Luxembourg - No data - - Netherlands 1 -0.31** [-3.03] -1.84 0.00 [-1.21] Portugal - No coint. Spain 1 -3.06*** [-8.38] 5.74 - - Note: *, **, ***, denote statistical signicance at the 10%, 5%, 1% levels, respectively. T-stats in brackets. 34 Table 4: Weak exogeneity test: total Economy and manufacturing Sector. Total Economy, 1980-2011 Manufacturing Sector, 1979-2010 χ2 1 p-value χ2 1 p-value Austria 34.17*** 0.00 16.07*** 0.00 Belgium 28.77*** 0.00 11.99*** 0.00 Finland 32.39*** 0.00 No cointegration France 17.46*** 0.00 21.40*** 0.00 Germany 25.34*** 0.00 12.81*** 0.00 Greece 27.58*** 0.00 15.50*** 0.00 Ireland 29.09*** 0.00 8.77*** 0.00 Italy 23.99*** 0.00 14.57*** 0.00 Luxembourg 13.08*** 0.00 No data available Netherlands 28.70*** 0.00 17.57*** 0.00 Portugal 28.93*** 0.00 No cointegration Spain 48.92*** 0.00 15.82*** 0.00 Note: *, **, ***, denote statistical signicance at the 10%,5%, 1% levels, respectively. T-stats in brackets. 35 Table 5: Relative convergence test: total economy and manufacturing Sector. Total economy, 1980-2011 Manuf. sector, 1979-2010 χ2 1 p-value χ2 1 p-value Austria 27.48*** 0.00 11.49*** 0.00 Belgium 9.02*** 0.00 10.36*** 0.00 Finland 26.44*** 0.00 No cointegration France 14.92*** 0.00 6.54** 0.01 Germany 39.00*** 0.00 3.54* 0.06 Greece 19.39*** 0.00 13.82*** 0.00 Ireland 27.56*** 0.00 21.02*** 0.00 Italy 19.70*** 0.00 14.46*** 0.00 Luxembourg 28.83*** 0.00 No data available Netherlands 24.16*** 0.00 5.82** 0.02 Portugal 20.41*** 0.00 No cointegration Spain 48.21*** 0.00 17.60*** 0.00 Note: *, **, ***, denote statistical signicance at the 10%,5%, 1% levels, respectively. T-stats in brackets. 36 Table 6: Residual-based panel cointegration test, Westerlund (2007) Total Economy, 1980-2011 No deterministic trend Deterministic trend Statistic P-value Robust P-value Statistic P-value Robust P-value Panel Gt-Statistic -1.89 0.33 0.62 -2.55 0.21 0.78 Panel Ga-Statistic -6.24 0.72 0.54 -9.35 0.91 0.66 Panel Pt-Statistic -6.44 0.08 0.46 -8.15 0.17 0.68 Panel Pa-Statistic -7.88 0.00 0.16 -12.09 0.04 0.25 Manufacturing Sector, 1979-2010 No deterministic trend Deterministic trend Statistic P-value Robust P-value Statistic P-value Robust P-value Panel Gt-Statistic -2.31 0.03 0.36 -2.87 0.02 0.72 Panel Ga-Statistic -7.10 0.51 0.52 -11.39 0.62 0.62 Panel Pt-Statistic -5.72 0.18 0.54 -9.32 0.00 0.50 Panel Pa-Statistic -6.36 0.06 0.37 -15.11 0.00 0.13 Note:The null hypothesis H0 is of no cointegration. Critical values robust to cross-sectional dependence are obtained by bootstrap, with 1000 replications. 37 Table 7: Likelihood-based panel cointegration test, Larsson et al. (2001) Total Economy, 1980-2011 Model 1 Model2 r=0 r=1 r=0 r=1 Average of trace statistics 37.84 2.72 49.07 7.88 Asymptotic mean 1 8.27 0.98 16.28 6.27 Asymptotic variance 14.28 1.91 25.50 10.45 Larsson et al. (2001) statistic 27.11 4.36 22.49 1.72 Bootstrapped interval (2) [14.46-22.56] [12.48-20.90] [12.76-22.29] [12.75-22.32] Manufacturing Sector, 1979-2010 Model 1 Model2 r=0 r=1 r=0 r=1 Average of trace statistics 24.16 4.03 29.51 8.83 Asymptotic mean 8.27 0.98 16.28 6.27 Asymptotic variance 14.28 1.91 25.50 10.45 Larsson et al. (2001) statistic 13.95 7.32 8.69 2.63 Bootsrapped interval [8.84-16.38] [8.85-15.32] [12.00-20.64] [12.68-20.78] (1) Asymptotic moments are from Breitung (2005) (2) 95% condence interval simulated with the maximum entropy bootstrap algorithm by Vinod and da Lacalle (2009) with 1000 replications. 38 