SPECIALISATION IN MANUFACTURING
INDUSTRY IN EUROPE(*)
Rafael Myro Sánchez M.ª Elisa Álvarez López
Universidad Complutense de Madrid Universidad de Valladolid
ABSTRACT
This paper explores the determinants of product specialisation in the manufacturing industries of European Union countries, on the basis of a Ricardian model for a continuum of goods. The work is based on new annual data corresponding to the period 1980-1997 for 10 European countries and 13 big industries defined by following the two-digit NACE-CLIO classification.
Although labour productivity clearly constitutes the most powerful impulse for enabling countries to increase their share in world manufacturing production, its role in the sectoral specialisation of manufacturing industry output is offset by other variables, mainly by differences in the degree to which sectors are open to foreign competition. On the other hand, additional variables that reflect the technological capital provision of different countriesappear to be crucial, showing that specialisation cannot be explained just from a Ricardian framework. The Heckscher-Ohlin-Vanek model also has to be taken into account. Contrary to our expectations, richer countries do not always show relatively higher labour productivity levels in high-tech sectors. What provides the base for their specialisation in these sectors is, above all, a high provision of technological capital.
(*) We want to express our gratitude to W. J. Baumol for his continuous support, valuable assistance and suggestions, to José María Labeaga, who helped us solve several econometric problems, and to Carmen Díaz Mora, who provided us with data relating to foreign trade in EU manufacturing industries and physical, human and technological capital. We are also indebted to Marc Siscart for his valuable comments concerning a second version of this paper.
1. INTRODUCTION.
In spite of the growth of international economic integration and globalisation, there are still important differences in inter-industry specialisation among developed countries and they do not show any clear tendency to diminish, even in the case of those involved in a deeper integration process, as is the case of the nations of the European Union (Dollar and Wolff, 1993; Sapir, 1996; Álvarez López y García Grande, 1998a).
Furthermore, on the basis of the most detailed sectoral classifications, a tendency towards a wider difference in production specialisation among developed countries has appeared in the last twenty years, in opposition to the general pattern of increased similarity that dominated the 1970’s. Thus, Wolff (1999) finds only 5 countries out of 14 whose production structures became increasingly differentiated from the pool as a whole from 1970 to 1979, whereas this happened to 10 of them between 1979 and 1993. In the same way, focusing their attention on Europe, Midelfart-Knarvik and others (2000) find that 8 out of 14 countries became increasingly differentiated from the average between 1980 and 1997, while there were only 4 in the 1970´s[1].
This development contradicts the theoretical perspective based on the Heckscher-Ohlin-Vanek framework (HOV), according to which inter-industry specialisation should become increasingly similar among countries withdeeper economic integration, as equalisation of factor prices takes place, and the sources of comparative advantage disappear (Wolff, 1999). The increase in intra-industry trade would just reflect and contribute to the similarity of industrial structures.
This process would be accompanied by a reduction in differences in labour productivity, either at a national or industrial level (Dollar and Wolff, 1993). However, once again the evidence goes against this prediction. There is no clear evidence of convergence in labour productivity among the OECD countries, as is well known from Dollar and Wolff's key contributions mentioned above[2]. For the EU countries, Álvarez López and López Morales (1999) did not find any convergence process at all in labour productivity between 1980 and 1995. Even among the five countries that founded the European Community in 1957, only in 3 industries out of 13 can convergence in labour productivity be found between 1985 and 1995.
These empirical facts, together with evidence that only a very general version of the HOV model might work reasonably well (Davis and Weinstein, 2001), encourage us to explore other perspectives, one of the possibilities being a return to a Ricardian type of world where sources of comparative advantage for a continuum of goods stem from differences in labour productivity based on differences in physical, human or technological capital. The role of increasing internal returns to scale, and of learning by doing, can also be emphasised, following Gomory and Baumol (2001) and Wolff (1999).
As less developed countries are characteristically less well endowed with physical, human and technological capital, they will enjoy comparative advantages in labour intensive and technologically standardised industries. We could, therefore, expect these countries to find it easier to achieve high labour productivity levels in these low-tech industries.
Consequently, this paper will explore the role of labour productivity in the production specialisation of European countries, using a Ricardian model. In fact, this approach is very close to that followed by Wolff (1999), but we want to explore the relationships betweenspecialisation and labour productivity not only within their time path, but also, and principally, at a cross-section level, an area where we encountered some problems in previous studies (Myro, 1992 and 1995).
2. THEORETICAL MODEL.
From a Ricardian perspective, a country will efficiently produce all those commodities for which domestic unit labour costs are less than, or equal to, foreign unit labour costs. A commodity z will be produced if:
a(z)w a*(z) w*[1]
Where a and a* denote the unit labour requirements at home and abroad, and w and w* the respective wages, measured in a common unit (Dornbusch, Fisher and Samuelson, 1977). With such a unit, the efficient production of a set of commodities by the home country is guaranteed; otherwise, movements in its rate of exchange would take place.
Given a ratio of wages, w/w*, the home country will produce all the commodities for which the ratio of unit labour requirements with respect to foreign countries is more than or equal to this wages ratio. That is:
a*(z)/a(z) w/w*
Defining y as labour productivity {y=1/a}, this rule may be expressed in the following different form:
y(z)/y*(z) w/w*
As we have to deal with huge branch aggregations, and countries tend to have some barriers to trade, as well as being of different sizes, we can assume that the home country always obtains some production in some sectors. Accordingly, the above constraint translates into the following rule with which we are already familiar: there will be a positive relationship between the share of the home country in the production of commodity z and its labour productivity with respect to foreign countries, which we can express in the following way:
Y(z)/Y*(z) = f {y(z)/y*(z)}[2]
Where Y denotes output, and the ratio between Y and Y* is an index of specialisation.
It might appear that the ratio of wages can be left out of this expression because it is the same for every commodity. However, things change when several countries are taken into account. As each of them can have a different wage, a fixed relation between two wages cannot be set. So the introduction of wages into the second side of the above expression is necessary. This means the use of relative unit cost, instead of relative productivity, as the fundamental variable, if we look at the original formulation [1]. In addition, the index of specialisation used in the first side of the expression has to be changed for a more conventional one, to account for the fact that countries vary in size.
Therefore, the new formulation has to be:
E(z) = f {a(z)w/a*(z)w*}[3]
Where f´<0 and E (z) is a particular conventional index of specialisation of the home country in industry z, i.e. defined as the share of the industrial branch in industrial output at home in relation to the same share for all foreign countries. Now a* refers to the average unit labour requirements in the whole geographical area under consideration, and w* to the average wage.
In the real world, wages are not the same for every commodity in one country because of market imperfections or the fact that different numbers of skilled workers are employed. Moreover, commodity trade is hampered because governments establish various barriers which permit the existence of differences in prices between countries, measured as a common unit of currency, and also produce wage differences.Furthermore, vertical differences in product quality increase differences in prices. Thus, real unit labour costs must be used for each branch (ratio of real wages to labour productivity).
On the other hand, it can be expected that greater or lesser opening up to foreign trade and foreign competition might affect the positive relationship between specialisation and unit labour cost. This leads us to a reformulation of the expression [3], to produce this possibility:
E(z) = - f {a(z)w/a*(z)w*; TO}[4]
TO being any measure involving opening up to trade, i.e. share of export plus import in production. The sign expectedfor this last variable is negative: less opening up to foreign trade for a given industry means more relative output for a specific relative unit labour cost.
3. ESTIMATION FOR TEN COUNTRIES IN THE EUROPEAN UNION.
The annual industrial branch database available corresponds to the period 1980-1997 and follows the two-digit NACE-CLIO classification, used by Eurostat, taking in 13 manufacturing sectors. In Álvarez López and García Grande's work (1998b) an account is given of the considerable effort involved in building up this homogeneous database.Nevertheless, the variable TO is only available for the period 1985-1996, so the estimations including it are restricted to this period.
We have estimated expressions [2], [3] and [4] by using as a dependent variable the index of industrial specialisation of each country with respect to European industry as a whole. That index is defined as the percentage of added value that the industrial branch i has in the manufacturing industry of the country h divided by the same percentage for all European manufacturing industry.
Independent variables are defined as follows:
-Labour productivity: the ratio of real added value per employee.
-Real unit labour cost: the ratio of real wages (deflatedby price indexes) to labour productivity.
-Opening up to foreign trade index: the ratio of imports plus exports to production.
They are all always measured in relative terms, that is to say, as the ratio of their value in each country to the EU as a whole, for each of the manufacturing sectors.
We have estimated the above three expressions for each country and the EU as a whole, and for all the countries taken together. Nevertheless, only the latter are included here, to avoid overloading the text with excess information that is not strictly necessary. Variables have been transformed into logarithms, so the estimated coefficients can be interpreted as elasticities.
In processing the estimations, we followed the panel data methodology of estimation. We first looked, therefore, for individual effects correlated with regressors through the 2 Hausman test. When this was the case, we kept the within group estimations, which are the only consistent ones. In the other cases, we improved the efficiency of the estimators through generalised least squared estimations, correcting the auto-correlation frequently found in the estimations. We always tested the exogeneity of independent variables, using the Hausman test. As this was confirmed in all cases, we did not use instrumental variables. Residue normality was always tested by means of the Jarque-Bera test.
Table 1 presents the results of the within group estimations for the whole panel data, that is, those taking in all the countries and industries under examination.
These estimations provide strong support for the hypothesis of labour productivity guidespecialisation and show that the model works well, although we had expectedthat the role of wages would be clearer. Nevertheless, the limited contribution provided by that variable in explaining specialisation patterns is probably due to the fact that it is closely related to labour productivity, as we have chosen the wage levels of specific industries instead of the overall average. Whatever the case, the estimations carried out for each country faithfully reproduce the more general estimation mentioned above.
These results also match well with those obtained by Wolff in 1999, as he uses first differences in natural logarithms. He also introduces productivity and wages separately, keeping them apart from the ratio of gross capital formation over the number of employees. His results are good, in particular as regards productivity, which is revealed to be the key variable. Wages are not significant for the more recent period, nor is the ratio of investment in capital.
Neither within group estimations, nor Wolff's first differences estimations bring the analysis of specialisation to completion, as it involves substantial complexities that are not captured by these approaches. Of course, labour productivity guides the time path ofspecialisation. However, what happens between manufacturing sectors? Do more advances in labour productivity in one sector than in others lead to higher specialisation in it? Can one country improve its specialisation in an industrial branch by raising relative labour productivity within it? Apparently so, according to the above estimation; but in reality this is not so clear, as can easily be seen by looking at the cross-sectional relationships between specialisation indexes and relative labour productivity ratios, explored formally through the so-called between groups estimation, in terms of an econometric panel data analysis.
In other words, if we use cross-sectional information, we can have a more accurate estimation, but in this case, relative labour productivity, while remaining very significant, is not a key variable for explaining specialisation indexes. In econometric terms, the Hausman test advises us to select Generalised Least Squared (GLS) estimators instead of the within group ones, because the former use all the available information, not only that referring to the time path. However, that way we get poorer results, in terms of explaining the dependent variable, as Table 2 shows. The GLS estimators take values near to the within group ones, but the adjusted R2s are almost zero. The estimations carried out for each of the ten countries confirm the results of the more general estimation above, although labour productivity has greater explanatory capacity in the cases of Germany, Denmark and Italy.
This seems to indicate that there are other variables besides productivity and wages that powerfully influence sectoral specialisation and we have to look for them.
Unfortunately, the opening up to foreign trade or foreign competition that was explicitly considered in our model can not be one of those variables because it was supposed only to capture for individual effects correlated with regressors, and the Hausman test told us there is none.
Nevertheless, is interesting to observe that when this variable is introduced, the explanatory power of the estimations improves greatly (Table 3), although the coefficients of the other variables become lower[3], and it appears individuals effects related to regressors. The degree of opening up to foreign trade carries the expected negative sign, indicating that it tends to produce a reduction in specialisation. In other words, the more one manufacturing sector is closed to foreign competition, the more production within it can be increased, along with its specialisation index. This offers evidence that the opening up to foreign competition tends to make closer the industrial structures of countries. Therefore, the absence of a clear convergence process in productive specialisation among developed countries in the last 25 years must be attributed to the influence of other variables we need to look for[4].
4. SECTORAL PATTERNS ON THE EVOLUTION OF LABOUR PRODUCTIVITY: SOME PARADOXES.
The difficulties we found in assigning a crucial role for labour productivity in sectoral specialisation arise from the fact that, contrary to our expectations, neither the less developed countries nor the richest ones achieve a better performance in labour productivity in specialised manufacturing sectors. To focus on this paradox, we begin by showing, in Chart 1, the specialisation indexes of each of the ten European countries considered, within the high-tech and medium-tech industrial manufacturing sectors[5].
In 1997, there were three countries that were clearly specialised in high-tech and medium-tech industries, Germany, France and the United Kingdom, with some differences between them: Germany produced more electric and electronic machinery products and fewer computers than the other two countries. At the same time, Germany was stronger than these in its medium-tech industries, particularly in non-electrical machinery and transport equipment.
On the other hand, the specialisation of three countries was centred in low-tech industries: Italy, Spain and Portugal. This is common knowledge. However, there were very significant differences between these three countries too, showing Spain and Italy in stronger positions than Portugal in medium-tech and high-tech industries. Within the low-tech industries, Italy and Portugal led in textiles and clothes, while Spain was the front runner in non-metallic minerals and products.
The Netherlands, Belgium and Denmark were situated in an intermediate position, between the two aforementioned groups of countries. The Netherlands, with a specialisation pattern very similar to that of the EU as a whole, was the leader in paper products and industrial chemicals, while Belgium led in basic metals and industrial chemicals and Denmark in other manufactured products (wood) and non-electrical machinery.
These patterns of product specialisation underwent few changes since 1980, as has been well documented by the authors mentioned at the beginning of the article. In the seventeen years between 1980 and 1997, the Netherlands fell behind from the position their high-tech industries formerly held, while the United Kingdom advanced rapidly in these sectors replacing medium-tech with high-tech manufacturing. Germany, like Belgium, reinforced its position in the medium-tech industries and lost industrial weight in the computer sector, whilst the United Kingdom gained ground in that sector. But, more significantly, so far as their manufacturing production structure is concerned, Italy, Spain and Portugal became more similar to the EU as a whole, developing their high-tech industries (and medium-tech in the case of Spain).