Regional economic integration and foreign direct investment: the case of German investment in Europe.

Pain, Nigel ; Lansbury, Melanie

I. Introduction

The European Economic Community originally came into existence following the Treaty of Rome in 1957. Member states planned to harmonise their tariffs, pursue a common trade policy and liberalise intra-Community trade. All internal customs duties and quantitative restrictions on trade were successfully removed by 1968. However it subsequently became clear that this had not resulted in the full integration of product markets in Europe, with capital movements and trade continuing to be restricted by capital controls and non-tariff barriers. In the mid-1980s the European Commission identified around 300 areas in which legislative action could be taken to help the free flow of goods, services, capital and labour within the community. This led to the Single European Act in 1986 which aimed to complete the internal market by the end of 1992.

In this paper we concentrate on the relationship between foreign direct investment and the actions designed to ease non-tariff barriers in product markets. Measures introduced under the Internal Market programme include the removal of internal customs barriers and constraints on capital movements and moves to liberalise tendering procedures for public procurement contracts. Perhaps the most important changes have been the steps taken to harmonise technical standards and regulations, with the principle of mutual recognition of national standards allowing products that meet the legal standards of any individual member state to be sold throughout the European Union unless explicit minimum union-wide standards have been agreed. A more extensive description of all these changes is provided by Hoeller and Louppe (1994).

The economic impact of the Internal Market programme has been debated widely and the European Commission has recently completed the first detailed assessment of the measures taken so far (European Commission, 1996). This study contributes to the assessment by examining whether the Internal Market programme has affected the pattern of intra-European foreign direct investment. It has long been recognised that changes in regional economic institutions and trade policy can have a significant impact on both the level and location of overseas investment. The initial formation of the European Community prompted considerable empirical study into the question of whether investment was diverted into the region (Yannopoulos, 1990). More recently considerable attention has been paid to the impact of the North American Free Trade Agreement (NAFTA) on the location of production (Eden, 1994).

Our main focus is on the location and level of foreign direct investment by German firms within the European Union (EU). World-wide, German firms have the fourth largest stock of overseas assets of all investing countries. Within Europe they are the second most important investors after the United States. At the end of 1994 just under 1.2 million workers were employed in the foreign affiliates of German firms located in Europe. Of these some 180,000 were in the UK, representing around 1 1/2 per cent of private sector employment.

One striking feature of German FDI is the extent to which it became increasingly concentrated within Europe from the latter half of the 1980s. An important question of interest is whether this observed change in the pattern of direct investment is related to regional-specific factors or simply to differences in national characteristics and the extent to which Germany has become an unattractive business location (Jungnickel, 1995). To answer this we use a panel data set to analyse the determinants of foreign direct investment by German corporations between 1980 and 1992. The panel consists of six EU locations, plus Austria and the United States. The inclusion of the latter two locations allows us to examine whether investment has been diverted into the EU since the advent of the Internal Market programme. Within the EU we distinguish France, Italy, the UK, Belgium and Luxembourg, the Netherlands and the Iberian peninsula. We examine investments undertaken in seven broad sectors, chemicals, mechanical engineering, electrical engineering, transport equipment, 'other' manufacturing, distribution and financial services.

The remainder of this article is as follows. In the next section we review recent trends in intra-EU foreign direct investment and give an overview of the relevant literature on regional liberalisation and intra-regional investment. This suggests a number of specific hypotheses of interest that can be examined in any empirical analysis. Section III describes the basic model used and discusses ways of establishing the impact of the Internal Market measures. Section IV contains the main empirical results and provides a quantitative evaluation of the impact of the Internal Market programme.

II. Regional Integration and Intra-EU FDI

Foreign direct investment (FDI) has expanded rapidly throughout the world economy over the past two decades, particularly since the middle of the 1980s. The growth in FDI within Europe has been especially marked, with the proportion of the aggregate stock of world FDI located within EU member states estimated to have risen from 31 per cent in 1985 to 39 per cent by 1995 (UNCTAD, 1996). This appears to have come about due to rising levels of investment in the EU both by non-EU nationals and by EU firms themselves.

The flow of investment from EU firms into other EU member states and North America since 1982 is shown in Chart 1. Intra-EU investment has exceeded investment in North America since 1988, reaching a peak of 1 1/4 per cent of GDP in 1990. Over the five years to 1994, intra-EU FDI was equivalent to 4 1/2 per cent of gross domestic fixed capital formation in the EU on average. At face value these data appear to suggest that moves towards greater European integration have generated a structural change in the pattern of intra-EU investment (Hoeller and Louppe, 1994).

A simple descriptive analysis of the changing pattern of intra-EU flows of direct investment cannot provide any direct quantitative indication of the possible influence of the measures taken to ease product market barriers in the Internal Market programme. The observed pattern of investment by EU firms might in part simply be a temporary adjustment to the removal of national capital controls in many member states.(1) This provides one reason for focusing on a single country such as Germany, whose capital [TABULAR DATA FOR TABLE 1 OMITTED] market has been free of foreign exchange controls for some time.

A further handicap to the use of aggregate intra-EU investment in any empirical analysis is the well-known statistical difficulties that arise from inconsistencies in national definitions of foreign direct investment. Eurostat do attempt to construct harmonised data, but their series exclude reinvested earnings, an important component of total foreign investment, and are only available from 1984 onwards. If developments since the advent of the Internal Market programme are to be seen in an appropriate perspective it is necessary to be able to evaluate them in the context of an analysis over a longer sample period.

An alternative approach is to focus on large individual investors with detailed statistics on the location of foreign [TABULAR DATA FOR TABLE 2 OMITTED] investments. UK outward direct investment is considered in Pain (1997); here we utilise German data. The geographical pattern of the recorded stocks of German foreign direct investment is reported in Table 1. The figures provide a picture similar to those in Chart 1, with the proportion of investments held within the EU(2) rising steadily since 1986 and a corresponding decline in the share of investments held within North America. In contrast, the share of investment within the other Western Europe countries has remained relatively constant over time, whilst investment in Central and Eastern Europe has begun to expand since 1990. The long-term decline in the proportion of investments sited within other developing countries suggests that the relatively high level of outward investment from Germany is not simply a move towards low wage locations.(3)

The German investment data were amalgamated into seven separate sectors, five for manufacturing and two for services. The sectoral composition within the EU, the US and Austria is summarised in Table 2. Investments in energy, mining and holding companies have been excluded. The 'other manufacturing' group is given by total manufacturing investment less investments in the four separately identified industries. It is clear from the figures in Table 2 that investment in the two service sectors accounts for much of the growth in the overall proportion of investment within the EU. Within manufacturing, investment in the chemicals, electrical engineering and transport industries appears to have risen more rapidly in Europe than in the United States.

The Impact of Integration on FDI

The early studies of the likely impact of the Internal Market (IM) programme suggested that the measures should bring about a considerable degree of industrial restructuring. This was largely expected to come about through greater industrial specialisation, with firms able to produce in a single location, exploit any economies of scale arising from the existence of firm-specific fixed costs and serve the wider European market through trade (Emerson et al., 1988). Such a pattern is consistent with the findings of Brillhart and Torstensson (1996), who report a significant correlation between the geographical concentration of particular industries in Europe and a prior ranking of the importance of scale economies in those industries. Little mention was made of intra-EU foreign direct investment in the initial studies of the Internal Market. The implication of the specialisation argument is that intra-EU FDI might ultimately be lower than otherwise expected as a result of the Internal Market programme, although there could be an initial transition period with higher cross-border investment if the restructuring process required new investments to ensure that all stages of production were sited in the lowest cost locations. In North America there is some evidence that such a restructuring process has occurred following deeper regional integration, with a number of US multinationals closing subsidiaries within Canada and substituting exports for FDI as a result of the improvements in market access arising from the US-Canadian Free Trade Agreement in force since 1989 (Niosi, 1994).

The exploitation of comparative advantage was also expected to change the geographical distribution of industry within Europe. Labour-intensive assembly activities would be concentrated in sites on the periphery of Europe with relatively lower labour costs. Other, more capital-intensive, activities would be located closer to the industrial core of Western Europe. Relatively little was said about prospective developments within service sectors in the early evaluations of the IM programme, in spite of the measures taken to liberalise capital markets. Although the number of measures relating to the service sector are only a minority of the number of Directives issued as part of the IM programme, in many cases they represent the first significant attempt to reduce obstacles to the cross-border provision of non-tradeable services within the EU, and might therefore be expected to stimulate foreign investment.

In contrast the IM programme was widely expected to bring a higher level of inward investment in Europe from non-European firms by raising the relative locational advantages of the European market. The removal of internal barriers has taken place at a time of continuing uncertainties over the external commercial policy of the European Commission with increasing use being made of contingent protection (Barrell and Pain, 1997). An additional implication of this argument is that the Internal Market process may have served to divert investment into the EU at the expense of other locations (Baldwin et al., 1995). There is some support for these hypotheses in the econometric results of Aristotelous and Fountas (1996) and Pain (1997). The former find that the level of inward investment in Europe by US and Japanese firms has been significantly higher than might otherwise have been expected since 1987, whilst the latter finds some evidence that UK companies have raised investment in Europe since 1990 at the expense of investment in the United States.

A broad summary of the expectations at the start of the IM process was thus that the measures would raise inward FDI from outside the EU, but eventually lower intra-EU FDI, particularly in the manufacturing sector. Scope existed for greater investment in the service sectors, but few estimates were made of the likely effects. Dunning (1996) provides a more detailed discussion of these and other issues.

There are a number of reasons why this argument may understate the scope for intra-EU direct investment as product market barriers are removed. At a practical level, the IM legislative process has taken much longer than initially anticipated. By October of last year an average of 91 per cent of all IM directives were on member states' statute books (Commission, 1996, pg.8). Our sample period ends in 1992, by which time the Internal Market had initially been expected to be complete. The continuing existence of some impediments to trade at that time may make it difficult to get a full picture of the eventual impact of market integration.

Moreover it is clear that many national factors continue to impose costs on market access. Some of these are regulatory, arising from differences in environmental and health and safety provisions. Examples include the continuing use of national health insurance price controls within the pharmaceuticals industry and regulations on the movement of waste and standards of packaging. In other cases markets remain differentiated as a result of consumer preferences. The food industry is a widely cited example (Cantwell, 1992), with substantial national product differentiation in alcoholic drinks, sauces and biscuits. Another example is retail banking, where greater concentration has occurred within individual countries but not across countries, with consumers preferring to use banks located in the domestic market. In such cases, direct investments are often made either to enter local markets or to establish facilities for adapting products to local needs.

The impact of the IM measures on trade and investment could also be expected to vary across sectors. Some of the non-tariff barriers, notably customs controls, would previously have restrained trade linkages but not market entry by means of direct investment. Others, such as technical requirements and lack of competition in public procurement would have affected both exporters and foreign subsidiaries. Capital controls would have acted to constrain direct investment, but not trade. In other cases trade and direct investment need not be substitutes even in the absence of barriers to market entry. This is particularly true of many investments in the service sector, with investment in distribution facilities often designed to improve market access for traded goods. Cross-border investment in other industries such as corporate banking and advertising is often stimulated by the growing activities of home country clients in foreign markets.

Thirdly, a number of models arising out of the new literature on economic geography suggest that the initial stages of a decline in barriers to trade could simply generate increased agglomeration, rather than a broader geographical dispersion of industries in line with comparative advantage (Baldwin and Venables, 1995). Regional differentiation within such models is driven by the interaction between scale economies and transport costs. If trade costs are high, there is little separation between the location of production and consumption. If the costs to market entry via trade are minimal, production is extremely sensitive to differences in factor costs. However, agglomeration effects may arise under the combination of intermediate levels of trade costs and increasing returns to scale. Firms located in densely populated areas can economise on fixed costs by concentrating production in a single plant and on transport costs by locating close to a large market.(4)

These form of models support the original idea that the decline in barriers to trade will eventually lead to greater concentration of production within industries, but suggest that the initial stages of liberalisation could well see a one-off higher level of investment as firms relocate. For instance, trade liberalisation appears to have altered the economic geography of Mexico (Hanson, 1997), with Mexico City declining in importance relative to areas on the US-Mexico border as trade barriers between Mexico and the United States have been lowered.

However variants of these models which allow for the presence of firm-specific assets can generate different implications (Markusen and Venables, 1996). Such assets might include process innovations, marketing skills or managerial expertise. All serve to give economies of scale at the level of the firm rather than at plant level. Multinational firms thus have lower variable costs but higher fixed costs than national firms in different locations. If country characteristics differ then single-plant firms may remain more profitable than multinational firms, since the latter have to locate some capacity in a location with higher relative factor costs, as well as incur the fixed costs associated with the establishment of an additional production facility. If country characteristics are similar and transport costs (or more generally barriers to trade) are non-zero, then multinational firms may be better placed than single-plant firms if they have knowledge-based assets. These act as a joint input across plants, giving lower fixed costs per market. Thus changes in technology and production costs can help to support the existence of multinationals even at a time of reductions in barriers to trade.

III. The Basic Model

In practice there are a wide range of factors that may determine the pattern of specialisation and location over time. Multinational enterprises arise through a combination of industrial organisation motives that result in a number of activities being placed under common ownership and control, and comparative advantage reasons that cause these activities to be placed in separate countries (Krugman, 1995). We seek to capture these influences in the empirical work. The remainder of this section gives a brief overview of the factors we include. A more extensive discussion of our approach is provided in the accompanying article by Barrell and Pain.

Conventional supply-side models of the location of production determine direct investment using indicators of market size and relative production costs. In the empirical work we proxy market size by sector value added in the host location.(5) To capture the role of relative labour costs we use 'normalised' trend unit labour cost data for manufacturing produced by the IMF, converted into a common currency. This provides a measure of the real exchange rate. Unit costs are used so as to allow for differentials in productivity levels as well as wages and payroll taxes.

The decision to establish foreign operations is likely to also reflect factors internal to the firm, in particular the existence of firm-specific knowledge based assets. Such assets may be expected to stimulate foreign direct investment, both because they can provide a multinational firm with a cost advantage over two separate firms in different locations and because they affect the trade-off between direct investment overseas and the use of licensing agreements. We follow the approach of Pain (1997) and assume that the level of sector-specific assets is proportional to the cumulated stock of patents granted to German firms within that sector. Many German corporations, especially in industries such as automobiles and mechanical engineering, have been at the forefront of important innovations in business practice over our sample period (Patel and Pavitt, 1989).

The patents data are taken from US Department of Commerce statistics on the number of patents registered in the United States each year. A five-year cumulative measure is used as there is likely to be some time lag before the full commercial potential of most patents is realised.(6) Although it might be expected that innovatory activity is less important for service sector FDI, the available data in fact suggest that around three-fifths of total investment in distribution and one-quarter of investment in financial and other services is undertaken by manufacturing firms. We thus use a weighted average of manufacturing patents for the two service sectors.

Investment may depend upon the flexibility of the workforce as much as its basic cost. It is widely claimed that the greater flexibility of the UK labour market has been a primary factor behind the growth in inward investment over the past decade (Eltis and Higham, 1995). Flexibility is affected by a large number of institutional features in labour, product and housing markets, and it is difficult to capture these fully in any empirical study. Here we utilise data on the number of days lost through strikes in each of the host economies. We expect that more strike-prone locations will receive less inward investment.(7)

Following Barrell et al. (1996) we also seek to take account of exchange rate volatility. Such volatility can affect firms whose costs and revenues are denominated in different currencies. Whilst it is possible to insure against currency risk, this is not without cost. Thus exchange rate volatility may affect the foreign investment decision. To investigate this possibility we use a dummy set to unity for those host countries who have been members of the Exchange Rate Mechanism of the EMS during the sample period, plus Austria which has pursued a policy of closely shadowing the D-mark since 1981. The variable is zero for the UK and Spain prior to ERM entry in 1990, and zero for the US throughout the sample period.

A number of studies also suggest that the investment decisions of companies, both at home and abroad, are affected by domestic financial conditions, although there appears to be little agreement as to how these are best measured. Carlin (1996) provides an overview of the impact of corporate profitability on the pattern of domestic investment within Germany, while Dinenis and Funke (1994) obtain a significant effect from real equity prices on domestic fixed investment. We investigate the role of both these factors in this study. Fluctuations in domestic profitability and interest gearing have already been shown to have a significant effect on the level of foreign investment by US and UK firms (Barrell and Pain (1996) and Pain (1997)) and Heiduk and Hodges (1992) suggest that planned foreign investments are reduced before domestic ones at times of financial distress. Profitability is measured using the rate of return on capital in the German business sector, as reported by the OECD.

The basic form of the partial adjustment model we initially employ can be expressed as:

[Mathematical Expression Omitted] [1]

where [FDI.sub.ijt] denotes the stock of foreign direct investment

in sector j in country i at time t, OUTPUT is the measure of market size, RELCOST denotes unit labour costs in the host country relative to those in Germany, PATENTS is the cumulated measure of registered patents, STRIKES is the number of labour disputes in the host country, EXCH is the exchange rate dummy, REQP denotes German equity prices in real terms, RETURN is the rate of return on capital in the German business sector(8) and IM is the dummy variable(s) included to capture the direct impact of the Internal Market. We discuss this variable in more detail below. The sector and location specific fixed effects [[Alpha].sub.ij] allow for unobserved influences that remain constant over the whole of the sample period. All other influences will be contained in the disturbance term [v.sub.it]. The fixed effects may capture factors such as contiguous borders and language that are not reflected in the other variables.

Modelling the Internal Market

The most important factor in an exercise of this kind is to decide how to incorporate the effects of the Internal Market programme into the specification of the model. We argued above that it is to be expected that the impact of the Internal Market programme will have varied by sector. Thus it would not be appropriate to include either a single dummy variable set to unity from 1987 across all sectors, or to draw conclusions from a set of individual sector dummies, since these could capture other unobserved influences as well. Some indication of the relative sensitivity of particular sectors in the first place is required. Here we follow Pain (1997) and use an ordinal variable (ranging from 1-3) for the level of non-tariff barriers in particular sectors. Sectors where the programme is expected to have a high impact are given a ranking of 3. Sectors with moderate and little impact are given rankings of 2 and 1 respectively.

Two separate rankings are reported in Table 3. The first is constructed from European Commission estimates of the importance of non-tariff barriers at the industry level, summarised in Buigues et al.(1990). These estimates have already been successfully exploited in studies of trade patterns within Europe, see for instance Sapir (1996). Whilst directly comparable estimates are not available for the service sectors, it is possible to draw on the qualitative assessments in Sapir (1993). These indicate that the removal of non-tariff barriers could be expected to have a 'high' impact on banking, insurance and road transport, but a 'low' impact on hotels and catering.

An alternative strategy is to devise an indicator based on an approach that is readily comparable across both industrial and service sectors. Here we use data on the number of cross-border mergers and acquisitions within the European Community in the four year period, prior to the signing of the Single European Act in 1986 and in the four year period immediately after. Those sectors where the growth rate in the number of mergers was above the average growth rate of all sectors were given a ranking of 3. Sectors whose growth rates were at or just below the average were assigned a ranking of 2. The remaining sectors were given a ranking of 1.

One potential objection to the use of mergers and acquisitions data is that it comes close to using data on direct investment to devise an indicator with which to 'explain' direct investment. However mergers and acquisitions will not be reflected in the direct investment data unless they are financed in the home country of the acquiring firm. Moreover, at the econometric level, it should be emphasised that the mergers data relate to developments within the whole EU; there is no necessary reason why any indicator based on this data should be of use in explaining investment by German firms, even though they are important individual investors.

Table 3: Estimates of the Impact of the Internal Market By Sector

Sectors Internal Market Sensitivity

 Commission Mergers &
 Estimates Acquisition Data

Chemicals 2 2
Machinery 1 1
Electrical 3 3
Transport Equipment 2 3
Other Manufacturing 1 3
Distribution 2 2
Financial and Other Services 3 3

Source: Own calculations based on data from the Annual Report on
Competition Policy, various issues, Buigues et al. (1990) and Sapir
(1993).

The two Internal Market proxies are compared in Table 3. Both measures are broadly similar, with financial services and electrical equipment being shown as particularly sensitive sectors. Differences arise for (road) transport equipment and the 'other manufacturing' sector. In both sectors there was a marked increase in cross-border mergers and acquisitions after 1986, although the Commission estimates suggested that the removal of non-tariff barriers was expected to have little impact. Cantwell (1992) argues that considerable national differentiation remains in the food industry and also in products such as [TABULAR DATA FOR TABLE 4 OMITTED] professional and scientific instruments, both included in 'other manufacturing'.(9)

The econometric work in this paper relies primarily on the Internal Market indicator based on the pattern of mergers and acquisitions. Whilst this offers a consistent estimate for each sector, it should be recognised that the rankings are somewhat arbitrary, although this criticism could equally be made of the alternative measure as well. In practice the choice of measure appears to make relatively little difference to the econometric results.(10)

IV. Empirical Results

The data set has thirteen annual observations (19801992) for seven sectors in eight separate locations giving a total panel size of 728 observations. We use constructed data on the stock of direct investment at constant prices. This was obtained by converting the current price foreign direct investment stock into dollars using the end-year exchange rate and deflating the nominal dollar stocks by the GDP deflator of the host country. All the main explanatory variables are entered in logarithmic form, permitting direct estimates of their elasticities. We estimate a dynamic, partial adjustment panel model. The inclusion of a lagged dependent variable in [1] will induce small sample bias into panel estimates produced using OLS (Nickell, 1981), so that an instrumental variable estimator has to be employed. There are a number of potential instruments that can be used for the lagged dependent variable. In this paper we employ the rank order of the lagged dependent variable (Durbin, 1954). This latter instrument is clearly strongly correlated with the variable being instrumented, but has been 'cleaned' of the lagged disturbance term.(11)

The econometric results are summarised in Table 4. The first column (hereafter (4.1)) reports the parameter estimates for the basic panel model. All of the main explanatory variables are significant at conventional levels and there is no sign of any significant first-order serial correlation. Overall, the coefficients are in accordance with our priors, suggesting that foreign investment by German firms is driven by strategic factors and firm-specific competitive advantages as well as by a desire to relocate to lower cost sites. There appears to be an important role for the accumulated sectoral level of patents registered by German companies, with an implied long-run elasticity of 1.04 per cent. There are also well determined effects from labour costs and market size, with respective elasticities of 0.34 and 0.65 per cent, broadly in line with the results obtained by Barrell et al.(1996). There is also evidence that labour relations in the host country are of importance, with an implied long-run elasticity on the strike variable of -0.11 per cent, so that a rise in the number of strikes will reduce inward investment.

There are sizeable significant effects from both the growth in real equity prices and profitability, with a 1 per cent rise in profitability generating an expansion in foreign investment of 0.75 per cent. This suggests that financial factors have had an important influence on the timing and scale of direct investment by German companies, confirming the findings from the case studies cited by Heiduk and Hodges (1992).(12) These terms should be seen primarily as indicators of the extent to which changes in domestic financial conditions affect the timing and the size of the flow of direct investment. As neither can be expected to trend permanently over time they cannot be the primary factor behind the continuing upward trend in the stock of investment.

We also obtain a significant effect from our proxy variable to capture the impact of currency variability. The reported positive coefficient suggests that German corporations value exchange rate stability, and is consistent with the notion that Germany may be the final market for some goods produced elsewhere within Europe. An implication of this is that German firms may prefer to produce in countries who have made a commitment to link their nominal exchange rates to the D-mark.

The Internal Market variable has a significant positive coefficient, implying that German investment in the EU has, on average across sectors and countries, been higher than might otherwise have been expected since 1987. The continued presence of the lagged dependent variable means that the model has the implication that the impact of the Internal Market has accumulated over time. This equation has a marginally smaller standard error than an alternative one using the IM indicator based on the Commission rankings shown in Table 3. However the alternative measure also had a significant positive coefficient, suggesting that impact of the programme on within-sector developments cannot be estimated too precisely.

In equation (4.2) we investigate whether the Internal Market has had different effects on the industrial and services sectors by splitting the IM variable into two components, one for the industrial sectors, denoted IMIND, and one for the service sectors, denoted IMSER. Both are set to zero for Austria and the United States. Whilst investment appears to have been significantly higher in both sectors, the results provide support for the hypothesis that the Internal Market programme has had a significantly greater impact on the service sectors, with an attempt at imposing equal coefficients being rejected by the data [Chi-squared(1) = 6.36].

The tests reported so far have explored the possibility of variation between sectors across countries. It is also possible to allow for variation between countries across sectors. A test for variation within countries is reported in the third column of Table 4. Here we have included separate dummies for each EU country, set to unity from 1987 onwards for all sectors. The IM sector dummies are also included. This specification relaxes two implicit restrictions in the previous equations, the common size of effects across countries and the common direction of effects across industries. The impact of the Internal Market on each sector, within each country, is given by the coefficient on the country dummy plus the coefficient on the sector dummy multiplied by the ranking of that sector as shown in Table 3. Some care is required in interpreting the findings since the country dummies will also pick up any factors that are otherwise accounted for and have changed systematically within each host country since 1987.

The national dummies are jointly significant, suggesting that there is considerable variation across countries [Chi(6) = 28.38]. This was confirmed by the rejection of an attempt at imposing identical coefficients on the country dummies [Chi(5) = 26.10], implying that there have been significantly different patterns of direct investment in European countries after 1987. In most cases the dummies have a negative coefficient. However for all countries this is offset, in some sectors at least, by the positive coefficient on the IM variable. Even so, the implied Internal Market effects differ markedly between countries. For the UK, the combination of the positive coefficients on the country and sector dummies means that investment in all sectors has been higher since 1987. In contrast investment in Spain has only been higher in those sectors with a ranking of 2 or 3, whilst in Belgium and France investment has only been higher in the most sensitive sectors. We return to this issue in more detail below.

The Internal Market and Investment Outside the EU

In (4.4), we investigate the impact of the Internal Market on German direct investment in the non-EU locations. We initially introduce separate industrial and service sector dummies for Austria and the United States. These variables have the same sector pattern as the EU measures denoted IMIND and IMSER, and are set to zero for all EU countries. For both the US and Austria there is evidence that German investment has been lower than might have been expected in their manufacturing sectors since 1987. This is consistent with the hypothesis that the Internal Market programme has diverted investment into the EU. However direct investment in services within the EU does not appear to have been at the expense of investment within the non-EU locations. Neither of the IM dummies for Austria are significant, possibly because it became clear at an early stage that Austria would implement the Internal Market legislation in full, initially via membership of the European Economic Area established in 1992 and then as a result of entry into the EU itself (Baldwin et al., 1995).

In the final column of Table 4 we drop the two insignificant service sector dummies for the non-EU locations. This has relatively little effect on the coefficients on the remaining IM dummy variables and there continues to be [TABULAR DATA FOR TABLE 5 OMITTED] evidence that the level of German investment in the manufacturing sector of the United States has been significantly lower than might otherwise have been expected since 1987. It is worth noting that the introduction of country dummies has led to some changes in the other coefficients in the model as compared to the initial regression. In particular the size and significance of the exchange rate dummy has fallen, possibly suggesting that the higher level of German investment in the UK was previously being attributed to membership of the ERM between 1990-92, rather than to changes arising from the IM programme. The profitability measure has also become more significant, possibly because its coefficient was initially biased downwards as a result of the conjunction of the decline in profitability after 1990 and the continued high level of new investments since that time.

Evaluating the Impact of the Internal Market

T. he full matrix of coefficients implied by (4.5) for the impact of the Internal Market on investment by sector and country within Europe is reported in Table 5. This indicates that German investment appears to have risen much more rapidly since 1987 in the UK, the Netherlands and Italy than might have been expected. Investment in Spain has also risen on balance, although by a smaller amount. In contrast, investment in France and Belgium and Luxembourg has been lower than might have been expected in all sectors apart from financial services, although in some cases, the implied coefficients are insignificantly different from zero.

It is possible to use the estimated relationships with the sector and country dummies to calculate the direct effect of the Internal Market programme on the stock of German FDI in any particular sector within a EU member state.(13) Here we use equation (4.5), although there are obviously a number of alternative equations upon which the calculations might be based?I The methodology is based on that described in Pain (1997, Appendix 1), extended to allow for the country dummies as well.

The results for each EU member state and each individual sector are summarised in Table 6. These indicate that as of 1992, the Internal Market programme is estimated to have raised the stock of German FDI by some $13.7 billion, at constant 1990 prices, equivalent to 17 1/2 per cent of the aggregate stock level. Pain (1997) reports that the programme has raised UK direct investment in Europe by a similar absolute amount, although the proportionate effect is somewhat larger due to the lower value of UK investment in the EU.(15)

Table 6. The Impact of the Internal Market on the Location and
Composition of German FDI in the EU

 ($bn, 1990 prices)
 1992 Stock IM Effect

Country

Belgium & Luxembourg 21.9 2.1
United Kingdom 12.2 4.9
France 13.7 -0.5
Italy 9.0 2.5
Netherlands 9.5 3.0
Spain & Portugal 10.1 1.7

Sector

Chemicals 9.5 -0.1
Mechanical Engineering 2.4 -0.3
Electrical 4.8 1.0
Transport Equipment 4.0 0.4
Other Manufacturing 7.1 0.9
Distribution 17.3 2.9
Financial & Other Services 31.3 8.9

Total 76.4 13.7

The primary beneficiary of the higher level of outward investment by German firms appears to have been the UK, where investment is some $4.9 billion higher than otherwise, a little over one-third of the reported stock level. As might be expected, given the coefficients reported in Table 5, Italy, the Netherlands and Spain and Portugal have also gained additional investment. Belgium and Luxembourg are also estimated to have gained overall, although this is entirely due to the high level of investment in financial services in those countries. Investment in the other six sectors is estimated to have fallen. France is the only location where investment is estimated to be lower as a result of the IM programme. Although the estimated coefficients for France reported in Table 5 are similar to those for Belgium, France is less specialised in financial services, and the small gain made in this sector fails to outweigh losses in the industrial sectors, particularly chemicals and mechanical engineering.

The sector results indicate that the largest gains have arisen in distribution and financial services. Within manufacturing, the largest absolute gains are for electronics and 'other' manufacturing. In proportionate terms the Internal Market programme has had a particularly marked effect in the electronics sector, accounting for some 21 per cent of the outstanding stock of German direct investment in the EU. There is some evidence in the results obtained for chemicals and mechanical engineering that the removal of non-tariff barriers has led to greater concentration. Foreign investment in both sectors is estimated to have been lower than might otherwise have been expected.

The Impact of Labour Markets and Innovation on Inward Investment

It is also of interest to estimate the extent to which the growth in manufacturing inward investment from Germany across Europe can be directly attributed to developments in national labour markets and the research intensities of the investing sectors. We attempt to do this in Table 7 by using the estimated equation to calculate the effects of actual changes in relative labour costs, the level of strikes and patenting on the change in inward investment from Germany between 1986 and 1992. Any such calculations have to take account of the presence of the lagged dependent variable. For example our estimates of the extent to which movements in relative costs can account for the change in the stock of inward investment between 1986 and 1992 reflect the extent to which a distributed lag of relative costs from 1987 to 1992 differed from an equivalent distributed lag of relative costs from 1981 to 1986.

The results reveal some interesting differences between countries. The UK has clearly benefited more than any other country from labour market factors, with movements in costs relative to Germany estimated to have raised the stock of inward investment by 5.5 per cent, and the decline in the number of strikes estimated to have raised investment by 9.3 per cent. However similar outcomes were achieved by France, Austria and, to a lesser extent, Italy, even though labour market institutions in these countries differ markedly from those in the UK. In contrast, the cost position of Belgium and Netherlands also stimulated inward investment, but there was little improvement in labour relations. Movements in labour costs and labour relations both acted to restrain new investment in Spain over this period.

Table 7. The Contribution of European Labour Markets and German
Patents to the Growth of Manufacturing Inward FDI from Germany
between 1986-92

 (per cent change in stock of inward investment)
Country Relative Strikes German Patents
 Labour Costs

Belgium 3.67 -0.95 30.62
UK 5.52 9.34 28.03
France 5.90 4.84 30.20
Italy 1.58 6.17 31.73
Netherlands 5.54 -0.02 27.61
Spain -3.76 -2.14 33.09
Austria 5.96 4.80 34.10

A somewhat different picture emerges from the contribution of the growth in patenting. It is clear that for all countries this has had a much more substantial impact on the growth of the stock of inward investment than have labour market developments. This reflects the extent to which the accumulated number of patents registered by German firms has tended to rise over time, as well as the respective elasticities of the estimated model. The estimates indicate that the UK has fared relatively poorly in attracting investments from those sectors in which innovation within Germany has risen most rapidly. Overall, Austria appears to have gained the most, attracting research intensive investments and benefiting from improvements in labour market performance.

V. Conclusions

The primary objective of this study has been to quantify the impact of continuing moves towards European integration on the sectoral and geographical pattern of foreign direct investment by German corporations. We have used a panel data set to estimate a conventional econometric model for the stock of direct investment with allowance for sector-specific and country-specific effects, augmented by a set of constructed Internal Market indicator variables.

Our results suggest that the Internal Market programme has had a significant, positive impact on the aggregate level of intra-EU investment by German corporations in both industrial and service sectors as a whole, with some evidence of investment diversion from the US in the manufacturing sector. There is some evidence that investment in Europe has been reduced in the chemicals and mechanical engineering sectors, consistent with the hypothesis that the removal of internal barriers to trade could result in production becoming more concentrated. The largest single beneficiaries of the extra investment appear to be the UK, the Netherlands and Italy, with additional evidence of a modest increase in investment in the Southern periphery of the EU.

Overall, the combined results of this study and Pain (1997) suggest that the Internal Market programme raised the constant price stock of intra-EU investment from UK and German firms by some $27 billion as of 1992, equivalent to 0.5 per cent of EU GDP. The financial services sector accounts for half of this additional investment. These estimates should perhaps be regarded as providing a lower bound to the overall Internal Market impact, since the programme may also have affected the growth of output and research intensity. The conclusions of this paper and the accompanying paper by Agarwal in this Review also suggest that, for Germany at least, the Internal Market has been associated with investment creation in Europe, with little evidence that the higher level of foreign investment has been at the expense of domestic investment.

The results in this paper suggest that any overall evaluation of the 1992 programme should take into account developments in the service sectors as well as within the manufacturing ones. It would also be of interest to know more about the forms of activity undertaken by foreign subsidiaries within the EU. Our findings are consistent both with models of horizontal direct investment, such as Markusen and Venables (1996), in which firms produce similar products in different locations and models in which firms aim to widen their international division of labour. To address this question it would be of considerable interest to seek to augment the findings from sectoral econometric analyses with those from more detailed case studies of individual multinational firms.

Correspondence should be addressed to the authors at NIESR. Earlier versions of this paper were presented at the 1996 European Economic Association Conference in Istanbul and the winter seminar meeting of the International Economics Study Group. We are grateful to Ray Barrell, John Dunning, Peter Holmes, Helen Popper, Martin Weale and seminar participants for helpful comments and discussions and to the Economists' Advisory Group and the ESRC for financial support.

NOTES

(1) Arrowsmith et al. (1997) provide a detailed account of the evolution of EU capital markets and the pattern of capital flows since the start of the Internal Market programme.

(2) The EU figures are for the twelve member states as of 1994; they therefore include data for Greece, Spain and Portugal prior to their accession into the EU.

(3) It is possible that the rise in the EU share is exaggerated by a vintage effect. As the stock data are at book values, more recent investments can be expected to be closer to their present market values than older ones. To the extent that such investments tend to be within Europe, this will serve to raise the proportion of total investment held within Europe.

(4) Brulhart and Torstensson (1996) find that the industries with the greatest scope for scale economies tend to be concentrated within central EU countries and regions.

(5) The sector output data are for gross-value added by branch, obtained from Eurostat.

(6) Comparison of our results with those in Pain and Lansbury (1995) using a three-year cumulation of patents and in Barrell et al.,(1996) using a cumulative R&D based indicator suggest that the degree of cumulation makes little difference to the estimated long-run elasticities.

(7) However if there are other factors which attract companies to a particular location then it is possible that labour disruption could simply encourage capital-labour substitution.

(8) This variable is lagged one year to avoid possible simultaneity problems. If foreign investment was a perfect substitute for domestic investment, then a rise in the stock of FDI would be associated with a fall in the domestic capital stock and hence, for given income, with an apparent rise in the rate-of-return on domestic capital.

(9) There are other sectors, notably chemicals, with an intermediate ranking in spite of a high number of cross-border mergers since 1986. However these sectors also had a relatively high number of mergers prior to 1986 as well, implying that the recently observed restructuring is part of a longer term process of rationalisation (Thomsen and Woolcock, 1993).

(10) Further work might usefully seek to investigate whether alternative estimates could be obtained from a detailed analysis of industrial structure in the United States and Europe, possibly drawing on information on either scale economies or concentration within sectors prior to the start of the Internal Market programme.

(11) The rank order may be a weak instrument if there is substantial measurement error present in the instrumented variable and, hence, in the associated rank order. There are a number of alternative estimators available, but most would force us to estimate an equation for the investment flow rather than the investment stock as first differences of the data would have to be employed.

(12) It might be argued that the profitability measure is simply a general indicator for the rate of return throughout Europe, rather than a particular indicator of the financial health of German companies and that an explicit measure of the rate of return on foreign investments should be included. To investigate this (4.1) was augmented by an additional term for the rate of return in the business sector in the host location relative to that in Germany. This term was found to have a positive coefficient (of 0.10) but was insignificantly different from zero. The separate term on German profitability remained significant.

(13) The Internal Market may also have had an indirect effect to the extent that the programme has affected both the economic growth and cost competitiveness of the host location.

(14) Again it should be emphasised that such estimates need to be treated with a degree of caution because the country dummies may be picking up additional effects unrelated to the Internal Market.

(15) This estimate for Germany is much larger than that obtained in Pain and Lansbury (1995). This is because their estimates came from a model with common coefficients on the IMIND and IMSER dummy variables. In the alternative model used in this paper, such a restriction is rejected, with a larger effect being obtained for the service sector. Our earlier estimates formed part of the European Commission's initial assessment of impact of the Internal Market (European Commission, 1996, pg.86).

REFERENCES

Aristotelous, K. and S. Fountas (1996), 'An empirical analysis of inward foreign direct investment flows in the EU with emphasis on the market enlargement hypothesis, Journal of Common Market Studies, no. 34, pp. 571-583.

Arrowsmith, J., Barrell, R., Pain, N., Young, G. and Wlodek, K.

Baldwin, R., R. Forslid and J. Haaland (1995), 'Investment creation and investment diversion: simulation analysis of the single market programme', CEPR Discussion Paper no. 1308.

Baldwin, R.E. and A.J. Venables (1995), 'Regional economic integration', in G.M. Grossman and K. Rogoff (eds.), Handbook of International Economics Volume III, Elsevier, North Holland.

Barrell, R. and N. Pain (1996a), 'An econometric analysis of US foreign direct investment', Review of Economics and Statistics, vol. LXXVIII, pp. 200-207.

Barrell, R. and N. Pain (1997), 'Trade restraints and Japanese direct investment flows', European Economic Review, forthcoming.

Barrell, R., N. Pain and F. Hubert (1996), 'Regionalism, innovation and the location of German foreign direct investment', National Institute Discussion Paper no. 91.

Brulhart, M. and J. Torstensson (1996), 'Regional integration, scale economies and industry location in the European Union', CEPR Discussion Paper no. 1435.

Buigues, P., F. Ilzkovitz and J. Lebrun (1990), 'The impact of the internal market by industrial sector: the challenge for the member states', European Economy, Special Edition, pp. 1-114.

Cantwell, J. (1992), The effects of integration on the structure of multinational corporation activity in the EC' in M.W. Klein and P.J.J. Welfens (eds.), Multinationals In The New Europe and Global Trade, Springer-Verlag, Berlin.

Carlin, W. (1996), West German growth and institutions, 1945-90' in N.Crafts and G.Toniolo (eds.), Economic Growth In Europe Since 1945, Cambridge University Press.

Dinenis, E. and M. Funke (1994), 'Factor prices, employment and investment in UK and West German manufacturing', The Manchester School, vol. LXII, pp. 412-424.

Dunning, J.H. (1996), 'The European internal market program and inbound foreign direct investment', University of Reading Discussion Papers in International Investment and Business Studies no. 217, Journal of Common Market Studies forthcoming.

Durbin, J. (1954), 'Errors in variables', Review of the International Statistical Institute, vol. 22, pp. 23-32.

Eden, L. (1994), 'Who does what after NAFTA? Location strategies of US multinationals' in L.Eden (ed.), Multinationals In North America, The University of Calgary Press, Calgary.

Eltis, W. and D. Higham (1995), 'Closing the UK competitiveness gap', National Institute Economic Review, no. 154, pp. 71-84.

Emerson, M., M. Aujean, M. Catinat, P. Goybet and A. Jacquemin (1988), The Economics of 1992: The EC Commission's Assessment of the Effects Of Completing The Internal Market, Oxford University Press, Oxford.

European Commission,(1996), The Single Market And Tomorrow's Europe, Kogan Page, London.

Hanson, G.H. (1997), Increasing returns, trade and the regional structure of wages, Economic Journal, no. 107, pp. 113-133.

Heiduk, G. and U.W. Hodges (1992), 'German multinationals in Europe; patterns and perspectives' in M. W. Klein and P.J.J. Welfens (eds.), Multinationals in the New Europe and Global Trade, Springer-Verlag, Berlin.

Hoeller, P. and M-O Louppe (1994), 'The EC's internal market: implementation and economic effects', OECD Economic Studies, vol. 23, pp. 55-108.

Krugman, P.R. (1995), 'International trade theory and policy', in (eds.) G. Grossman and K. Rogoff, Handbook of International Economics Volume III, Elsevier, Amsterdam.

Markusen, J.R. and Venables, A.J. (1996), 'The increased importance of direct investment in North Atlantic economic relationships: a convergence hypothesis', in M.B. Canzoneri, W.J. Ethier and V. Grilli (eds.), The New Transatlantic Economy, Cambridge University Press.

Nickell, S. (1981), 'Biases in dynamic models with fixed effects', Econometrica, vol. 49, pp. 1399-1416.

Niosi, J. (1994), 'Foreign direct investment in Canada', in L. Eden (ed.), Multinationals In North America, The University of Calgary Press, Calgary.

Pain, N. (1997), 'Continental drift: European integration and the location of UK foreign direct investment', The Manchester School, forthcoming.

Pain, N. and M. Lansbury (1995), 'The impact of the internal market on the evolution of european direct investment'. NIESR report prepared for Economists Advisory Group.

Patel, P. and K. Pavitt (1989), 'A comparison of technological activities in West Germany and the United Kingdom', National Westminster Bank Quarterly Review, May, pp. 27-42.

Sapir, A. (1993), 'Sectoral dimension' in Market Services and European Integration, European Economy, vol. 3, pp. 23-40.

Sapir, A. (1996), 'The effects of Europe's internal market programme on production and trade: a first assessment', Weltwirtschaftliches Archiv, vol. 132, no. 3, pp. 457-475.

Thomsen, S. and S. Woolcock (1993), Direct Investment and European Integration: Competition Among Firms and Governments, Pinter Publishers, London.

UNCTAD (1996), World Investment Report 1996, United Nations, Geneva.

Yannopoulos, G.N. (1990), 'Foreign direct investment and European integration: the evidence from the formative years of the European Community', Journal of Common Market Studies, vol. XXVIII, pp. 235-259.