The domestic performance of UK multinational firms.

Girma, Sourafel

This paper analyses the domestic performance of UK multinational firms from two perspectives: (i) their productivity relative to foreign multinationals, domestic exporters and non-exporters, and (ii) their ability to benefit newly acquired affiliates. Nonparametric analysis shows that the productivity distribution of UK multinationals is dominated by their foreign counterparts (especially US firms), and quantile regressions reveal that the performance disadvantage of UK multinationals is more pronounced at the higher end of the productivity distribution. Using a difference-in- differences methodology, it is found that, unlike foreign acquisitions, take-overs of domestic non-exporting firms by UK multinationals do not appear to lead to any productivity improvements.

I. Introduction

Between 1996 and 2000, UK multinational corporations had spent 42.6 billion [pounds sterling] in foreign direct investment, and by the end of this period they owned a staggering 600 billion [pounds sterling] worth of assets abroad. (1) Yet, in contrast to the literature that documents the link between foreign multinationals in the UK and economic performance, (2) research on the domestic efficiency advantages or disadvantages of those multinationals has been very limited. Using data on quoted companies, Kumar (1984) finds that UK multinationals earn higher returns on assets and on sales than their purely domestic counterparts. Grant (1987) and Grant et al. (1988) report a positive association between overseas investment and firm-level financial performance in a sample of large UK manufacturing enterprises. Recently Criscuolo and Martin (2002) provide cross-sectional evidence that the average labour productivity of UK multinationals is higher than domestically oriented firms, and not any worse than non-US multinationals operating in the UK. But Han and Lee (1998) contend that multinationality in UK and other G-7 countries' firms does not generally translate into superior performance compared with otherwise similar enterprises.

This paper undertakes an empirical investigation into the performance of manufacturing affiliates owned by UK multinationals relative to foreign and purely domestically-owned firms, based on firm-level panel data over the period 1989-98. In so doing, it adds to the existing literature in three respects. First, instead of focusing on average productivity differentials, it evaluates the effects of multinationality on the entire productivity distribution. This is achieved through the use of two statistical approaches: the Kolmogorov-Smirnov nonparametric testing procedure and the semiparametric method of quantile regression. The Kolmogorov-Smirnov tests employ the concept of stochastic dominance to rank the cumulative distribution functions of total factor productivity, whilst quantile regressions trace the entire distribution of productivity, conditional on a set of regressors. This enables one to identify not just the average effects of multinationality, but also where in the productivity distribution these effects are more pronounced.

The second novelty introduced in this paper is the explicit comparison of the productivity dynamics of UK multinationals with domestic exporters. While there are a number of papers documenting that exporters are more productive than non-exporters, (3) none of them distinguishes between domestic multinationals and simple exporters. Some scholars have argued in the literature that profit stability (and hence efficiency) from foreign direct investment exceeds that from simple exporting since multinationals diversify in both product and factor markets, whereas exporters diversify in product markets alone (e.g., see Rugman, 1974). We subject this notion to empirical investigation, as it may have some implications for policies that seek to encourage domestic firms to optimally engage in foreign market activities.

The paper also explores the relative performance of domestic multinationals vis-a-vis their foreign counterparts from a different perspective; it asks whether UK and foreign multinationals are equally effective in transferring their firm-specific advantages (e.g. special know-how) to their newly acquired subsidiaries. Hard evidence on this issue is extremely limited, (4) and we contrast the productivity effects of foreign and UK multinationals' acquisitions of domestic firms that had no previous foreign exposure. This is the third contribution of the paper.

Our findings from the nonparametric test statistics indicate that the productivity distribution of foreign multinationals stochastically dominates the productivity distribution of UK multinationals, both in levels and growth rates. Our results also point out that the productivity distribution of domestic exporters is not below that of UK multinationals, whilst the latter stochastically dominate the productivity distribution of domestic non-exporters. Allowing for a host of firm-level and industry-wide variables, the results from the quantile regressions confirm the above findings. Whilst the growth rate of UK multinationals is comparable to (even more favourable than) their foreign counterparts at the lower productivity quantiles, foreign firms exhibit a faster growth rate at the higher end of the productivity distribution.

We also find evidence that foreign establishments tend to acquire the better domestic firms. Accounting for this potential endogeneity problem, we uncover the discernible productivity-enhancing impact of foreign acquisitions. In sharp contrast, UK multinationals appear to target firms with lower levels of efficiency, consistent with the notion of 'acquisitions as disciplinary transactions' (Shleifer and Vishny, 1988). However, we fail to identify any subsequent improvements to the productivity of the acquired firms.

The paper proceeds with Section 2, which briefly considers the theoretical literature concerning the domestic efficiency advantages/disadvantages of multinational firms. Section 3 discusses the data, and Section 4 presents the findings from the Kolmogorov-Smirnov tests. Section 5 reports the empirical results based on the quantile regressions. Section 6 considers the determinants of acquisitions by foreign and UK multinationals, and presents the evidence on the post-acquisition productivity trajectories of the acquired firms. Section 7 concludes.

2. Performance advantages/disadvantages of multinationals: theoretical considerations

Theories concerning foreign direct investment (FDI) emphasise that foreign firms must have inherent firm-specific advantages that allow them to overcome the higher costs of becoming a multinational (Hymer, 1976; Kindleberger, 1969). (5) Higher costs result because domestic firms are already familiar with suppliers, demand conditions, culture, language, the legal environment and product standards, while foreign firms are not. The firm-specific advantages of multinationals may be tangible--for instance an improved production process or a product innovation--or intangible. The latter include brand names, better management structures or the technical knowledge of employees. But firm-specific assets are frequently intangible, which provides one motivation for firms internalising transactions and becoming multinationals rather than licensing their assets to local firms. The intangible nature of the assets means that a market may not exist for them, increasing the difficulty of writing contracts and entering into licensing agreements. It is a combination of these factors that leads to FDI--the advantages of the firm, the need to internalise the transaction, and the advantages of the country in which the FDI is located (Dunning, 1977).

The assumption that multinational firms have firm-specific assets implies that they may also possess substantial competitive advantages over their purely domestic rivals. A number of arguments are advanced in the literature linking multinationality with performance advantages over otherwise comparable firms operating in the domestic market alone (see inter alia, Kogut, 1985; Benvignati, 1987; Grant, 1987; and Gomes and Ramaswamy, 1999). These include: (i) better access to technological knowledge and foreign product innovation; (ii) wider international networks and management structure to meet domestic competition; (iii) benefits from economies of scale and scope; (iv) dampening domestic business fluctuations by using foreign market outlets; (v) taking advantage of factor cost differentials across multiple locations; (vi) tapping into corporate specific strategies such as transfer pricing and tax accounting practices; and (vii) greater geographic dispersion, which facilitates the undertaking of domestic ventures that are high-risk but also highly profitable.

But the literature has also identified several reasons as to why the costs associated with expanding abroad might outweigh the potential benefits. Grant (1987) suggests that limits to the capacity of managers to cope successfully with greater complexity associated with multinationality may exert a negative impact on performance. Hitt et al. (1997) also argue that multinationality may generate a decline in performance due to excessive spreading of managerial capabilities and co-ordination problems. Greniger et al. (1989) conjecture that an inverted-U relationship exists between the degree of multinationality and performance, where beyond an optimal degree of internalisation, profit advantages start to be eroded. In the same vein, Gomes and Ramaswamy (1999) contend that, above a threshold level of multinationality, transactional and informational costs may exceed the benefits of diversification. However, Mathur et al. (2001) point out that due to potentially high initial costs of establishing abroad, there might be an initial performance decline followed by improvements, as average costs decline once economies of scale and scope are achieved. This line of argument predicts that the relationship between the degree of multinationality and performance is a U-shaped one.

3. Database construction

The primary source of information for this study is the OneSource database of private and public companies, which is derived from the accounts that companies are legally required to deposit at Companies House. All public limited companies, all companies with more than 50 employees, and the top companies based on turnover, net worth, total assets, or shareholders' funds (whichever is largest) up to a maximum of 110,000 companies are included in the database. Companies that are dissolved or in the process of liquidation are excluded. Thus the database is heavily biased towards the larger firms. This may or may not be a major problem depending on the purpose at hand. We believe that the fact that OneSource under-represents firms with fewer than 50 employees is not necessarily a problem for the purposes of this paper. Multinational firms generally employ more than 50 workers, and comparing their performance with similar (in terms of size) domestic firms actually makes sense.

The database provides information on employment, physical capital, output and cost of goods sold in a consistent way both across firms and across time. It is also one of the very few databases with firm-level export data. However, each edition of OneSource contains foreign-ownership indicators only for the latest year, (6) so that it is not possible to identify when a firm became a subsidiary of a foreign multinational. To track the dynamics of ownership, we matched the population of manufacturing firms in the database to a list of UK firms acquired by foreign multinationals. (7) OneSource does not provide any information on the multinational activity of UK-owned firms either. Fortunately, we were able to merge it with a newly created database of foreign multinational activity--the European Linkages and International Ownership Structure (ELIOS) database built at the University of Urbino. (8) The information obtained from ELIOS was also complemented by a list of UK firms that had made foreign acquisitions, compiled from various issues of Acquisitions Monthly.

We focus on manufacturing subsidiaries of domestic and foreign companies, and independent domestic producers that do not own any subsidiaries. (9) Firms with annual employment or output growth exceeding 100 per cent were omitted, given doubts about the reliability of these extreme data points. Our final sample contains information on an unbalanced panel of more than 11,000 firms over the period 1989-98. Table 1 gives the frequency distribution of these firms by type of ownership and year.

Whatever the object of the productivity analysis, it is very important to obtain consistent estimates of the parameters of the production function. Using log values, we write the production function as [y.sub.it] [equivalent to] f([l.sub.it],[m.sub.it],[k.sub.it],[r.sub.it],[TFP.sub.it]), where y is output (10) and TFP is a firm and time-varying productivity shock. There are four factors of production: labour (l), material or cost of goods sold (m), capital (k) which is measured by the book value of fixed assets, and intangible assets (r). The intangible assets variable in OneSource is an estimate of the firms' investment in R&D and marketing, and the value of patents, copyrights and goodwill. Braunerhjelm (1996) argues that intangible assets more closely correspond to the theoretical notion of 'firm specific assets'.

For estimation purposes we employ a translog flexible functional form approximation of the production function, and TFP is assumed to follow the following AR(1) process:

(1) [TFP.sub.it] = [rho][TFP.sub.it-1] + [delta][D.sub.t] + [f.sub.i] + [v.sub.it]

where D is a common year-specific shock, f is a time-invariant firm specific and v a random error term. We estimate the production function for each three-digit (11) SIC92 industry available in our sample. To reflect multinationals' use of different technologies, they are allowed to have distinct factor elasticity parameters. (12)

Table 2 provides summary statistics for some variables of interest. The raw data confirm previous findings in the literature that multinationals are generally bigger, more capital intensive and productive. Within domestically oriented firms, exporters exhibit higher productivity levels compared to their non-exporting counterparts. It can also be seen that there is a considerable variation in the data. It is especially worth noting that within-industry productivity variation exceeds between-industry heterogeneity, which is in line with findings reported elsewhere (e.g. Bartelsman and Doms, 2000).

4. Evidence from the nonparametric analysis

A common method of analysing productivity differentials across distinct groups of firms is to estimate linear regression models of productivity with appropriately defined dummies. This method yields ceteris paribus average productivity differentials. Here we take an alternative nonparametric approach, and analyse productivity differentials over the entire productivity distribution rather than just at the conditional mean. To this end, we employ the Kolmogorov-Smirnov one-sided and two-sided tests (e.g. Covonor, 1999) to rank productivity distribution using the concept of stochastic dominance. (13)

Let F denote the cumulative distribution function of UK multinational firms, and let G be the corresponding function of the comparison group, which consists of US-owned firms, non-US multinationals, domestic exporters or non-exporters. First-order stochastic dominance of F with respect to G is defined as: F(z)-G(z) [less than or equal to] 0 uniformly in z [member of] R, with strict equality for some. The two-sided Kolmogorov-Smirnov statistics test the hypothesis that both distributions are identical, and the null and alternative hypotheses can be expressed as:

(2) [H.sub.0]:F(z)-G(z) = 0 [for all]z [member of] R vs. [H.sub.1]:F(z)-G(z) [not equal to] 0 for some z [member of] R.

By contrast, the one-sided test of stochastic dominance can be formulated as:

(3) [H.sub.0]:F(z)-G(z) = 0 [for all]z [member of] R vs. [H.sub.1]:F(z)-G(z)>0 for some z [member of] R.

For the one-sided test, for example, the Kolmogorov-Smirnov statistic is given by

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where n and m are the sample sizes from the empirical distributions of F and G respectively, and N = n + m. Acceptance of the null hypothesis (3) implies that the distribution of F is to right of G. In this case, F is said to dominate G stochastically.

Because the limiting distribution of the Kolmogorov-Smirnov statistics is only known under independence of observations, it is more appropriate to apply the testing procedure on cross-section by cross-section basis, rather on the pooled panel where within-firm dependence is likely to exist.

For representative years, table 3 reports hypothesis test statistics for productivity level and growth differentials between UK multinationals and the various comparison groups. As far as the productivity level is concerned, the hypothesis of equality of distribution of foreign and UK multinationals cannot be accepted at conventional levels of significance. Furthermore, the productivity distribution of UK multinationals is stochastically dominated by the distributions of both US and non-US firms. Chart 1 presents graphical depictions of pairs of estimated distribution functions for 1998, to allow for easy visual comparisons of productivity distributions. It can be seen that the productivity difference in favour of foreign multinationals is more pronounced at the higher end of the distribution, especially for US firms.

[GRAPHIC OMITTED]

The equality of distribution test between the productivity growth of US and UK multinationals is also rejected in every year of the sample period, and the former stochastically dominates the latter. These differential productivity growth rates provide suggestive evidence that UK multinationals are not catching-up with US-owned firms. The hypothesis of equality of non-US and UK multinationals' productivity growth distributions was not rejected for some years, but we did not find an instance where the productivity growth distribution of UK multinationals was above that of their non-US counterparts.

As mentioned in the introduction, a number of empirical studies across a whole range of countries document the existence of a particularly robust correlation between exporting and firm-level productivity. (14) But the twist here is the explicit comparison of the performance of domestic exporters and domestic multinationals. The argument of Rugman (1974) suggests that multinationals are likely to enjoy greater competitive advantages than exporters since they diversify in both product and factor markets. This conjecture does not seem be borne out by the data, however. The Kolmogorov-Smirnov statistics indicate that the productivity distributions of UK multinationals is not above that of simple exporters, and in all years we fail to reject the hypothesis of equality of distributions. This suggests that, as far as domestic performance is concerned, the evidence in favour of becoming multinational rather than engaging in arms-length exporting is not overwhelming. Finally, we establish stochastic dominance of the productivity level and growth distributions of UK multinationals with respect to non-exporting domestic firms. (15) But, by the end of the sample period, the productivity growth distribution of UK multinationals was not above that of non-exporters (see also chart 2).

[GRAPHIC OMITTED]

Using a nonparametric productivity measure, Girma et al. (2003) also reach the conclusion that the differences in growth rates between different domestic firms (exporters and non-exporters) and UK multinationals are not statistically significant. They interpret this result as evidence against convergence in the productivity levels.

5. Quantile regressions of productivity growth differentials

The nonparametric approach presented in the previous section is a useful analytical tool in summarising and contrasting the productivity dynamics of different groups of firms. This section refines the analysis, and investigates differentials across a range of productivity distributions by controlling for a host of factors impacting on firms' productivity trajectories. The total factor productivity (TFP) growth equation

(4) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.]

forms the basis for our empirical work. Here i and t index firms and time periods respectively; [TFP.sub.it-1] denotes initial level of TFP, and X is a vector of variables hypothesised to impact on plant level TFP growth, namely plant age, changes in export intensity, four-digit industry growth and concentration (captured by the Herfindhal index). OWN denotes ownership dummies for the four comparison group of firms, and MD is the degree of multinationality of UK multinationals proxied by the number of countries they have affiliates in. The rationale of including a quadratic term in MD is to test the hypothesis that, beyond a certain threshold, the advantages of multinationality start to dissipate (e.g. Greniger et al., 1989). Finally D represents the full set of time dummies to account for common macro shocks affecting the productivity path of the firms, and [epsilon] is a random error term.

As mentioned in Section 3, the productivity variation in our sample is in line with recent literature on firm-level productivity dynamics which has established a large and persistent heterogeneity across firms, even within narrowly defined industries. (16) This has an implication for productivity growth empirics: standard techniques, such as OLS and GMM, which concentrate on the conditional mean function of the dependent variable, are unlikely to be adequate analytical tools. In the presence of heterogeneous productivity processes, it is more appropriate (and arguably more interesting) to examine the dynamics of productivity at different points of the distribution, rather than 'average' properties (i.e. conditional means).

Accordingly, this paper employs the quantile regression technique introduced by Koenker and Bassett (1978). Denoting the vector of regressors in equation (4) by Z, the quantile regression model can be written as

(5) [DELTA][TFP.sub.it] = [Z.sub.it][[beta].sub.[theta]] + [[epsilon].sub.[theta]it], [Quant.sub.[theta]]([DELTA][TFP.sub.it]]|[Z.sub.it]) = [Z.sub.it][[beta].sub.[theta]]

where [Quant.sub.[theta]]([DELTA][TFP.sub.it]]|[Z.sub.it]) denotes the conditional quantile of [DELTA]TFP. The distribution of the error term [[epsilon].sub.[theta]] is left unspecified, so the estimation method is essentially semiparametric, (17) and the [theta]th quantile regression, 0 < [theta] < 1, solves

(6) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.]

By increasing [theta] from 0 to 1, one can trace the entire distribution of firm-level productivity growth, conditional on the set of regressors. Thus quantile regressions allow us to focus attention on specific parts of the productivity growth distribution, and identify where in the distribution multinationality exerts the greatest impact. In this study we consider regression estimates at five different quantiles, namely, the 10th, 25th, 50th (median), 75th and 90th percentiles of the TFP distribution.

As reported in Table 4, the estimated coefficients of initial TFP are negative across all quantiles. This is consistent with the notion of [beta]-convergence, and the catch-up rate, which denotes the speed at which a firm approaches its steady level of TFP ranges from 6.9 per cent to 13.1 per cent. Conditional on initial TFP, older firms grow at a slower rate, but the magnitudes of the point estimates imply that the between-ages difference might not be practically important. The results also suggest that the share of exports in total shipments exerts a growth-enhancing influence at the higher end of the productivity growth distribution. As expected, firms in fast growing industries enjoy higher productivity growth rates, whereas the concentration variable attracts negative coefficients across all specifications. This indicates that the absence of strong competition in an industry has detrimental effects on the evolution of firm-level productivity.

Focusing on the role played by ownership, it is apparent that the average productivity growth differentials presented in the first column of table 4 hide some interesting variations across the productivity quantiles. (18) For example, compared with UK multinationals (the base group), the productivity growth of US and non-US owned firms is, on average higher by 1.1 and 1.5 percentage points respectively. However, at the lower end of the productivity distribution (i.e. at the 10th and 25th percentiles) foreign multinationals actually experience slower growth rates. By contrast, at the upper end of the distribution (i.e. at the 75th and 90th percentiles) the performance advantage in favour of foreign multinationals is quite remarkable: it is two to four times greater than the average productivity growth regression results would have suggested. For instance, at the 90th percentile, the productivity growth of US and non-US multinationals is higher by about 4 percentage points compared to otherwise similar UK multinationals.

In accordance with the nonparametric analysis, the semiparametric quantile regression estimates also show that the productivity growth of domestic multinationals is not any faster than their arms-length exporting counterparts: the coefficients on the domestic exporters dummy fall short of significance at all points of the productivity distribution, consistent with the nonparametric analysis. However, non-exporters exhibit lower levels of growth, and their performance disadvantages are more pronounced at the upper end of the productivity growth distribution. It seems that non-exporting domestic firms are not particularly well represented in the more dynamic end of the productivity spectrum.

There is some evidence that the productivity growth of UK multinationals is positively related to their degree of multinationality. However,

the absolute sizes of the coefficients suggest that the productivity effect of increasing the number of countries in which the firms operate is unlikely to be economically significant. But more interestingly, we did not find any non-linearity in the relationship between performance and the extent of multinationality: the squared multinationality term in equation (4) is statistically insignificant in all but one specification. (19) Thus the data do not lend support to the notion that the benefits of multinational diversification diminish beyond a certain level.

A legitimate concern when using quantile regressions based on panel data is that firms move across quantiles. For example a firm at the bottom 10 per cent of the productivity distribution in the previous period might find itself at the top half of the distribution in the current period. Consequently, it might be problematic to make predictions about individual firms. But we believe that this is not too problematic for the purpose of this paper, because the analysis of the relative performance of different groups of firms does not require that individual firms stay in the same quantile. Nonetheless, as a robustness check we also run quantile regressions based on within-firm time averages. This approach ensures that each firm is confined to a single quantile, but at the expense of losing the time series variation in the data. The empirical estimates are reported in table 5 and the results are qualitatively similar.

6. The productivity effects of acquisitions by UK and foreign multinationals

Are domestic and foreign multinationals equally effective in transferring their firm-specific assets to their newly acquired subsidiaries? In this section we document evidence on the link between acquisitions of domestic firms with no previous foreign exposure (i.e. non-exporters) and productivity growth. We confine our analysis to the acquisitions of domestic non-exporters, as we are interested in isolating the productivity effects of foreign exposure. Foreign multinationals and UK-owned companies that are deemed to undertake internationally mobile projects (20) in underdeveloped regions receive substantial subventions from the UK government under its Regional Selective Assistance scheme (RSA). For example, some half a billion pounds was paid in grants for internationally-owned companies alone between 1991 and 1995. (21) One of the principal reasons behind this type of discretionary grant is the expectation that this may help reduce the 'UK competitiveness gap' by exposing indigenous establishments to better technological knowledge and foreign product innovation. It is interesting to note that two-thirds of foreign-owned projects and three-quarters of UK-owned internationally mobile projects that received public grants in the above period involved expansions at existing sites, acquisitions and buy-outs rather than new start-ups or branch plants. The focus on the productivity growth impacts of acquisitions by multinationals is therefore practically relevant.

The sample of acquired firms was screened for data availability for at least three consecutive years surrounding the acquisition event, and in the final analysis we identified 882 acquisitions made by foreign and domestic multinationals over the period 1989-96. (22) The frequency distribution of these acquisitions is given in table 6. The control group consists of about 4500 non-exporting firms that did not experience any type of ownership change.

We start by making an exploratory investigation into the determinants of acquisitions in our sample of firms via a multinomial logit model with three mutually exclusive events: no change of ownership, acquisitions by a foreign multinational and take-over by a UK multinational. The purpose of this analysis is to assess whether productivity can be considered exogenous to the process of acquisitions.

The literature on corporate take-overs has identified a number of variables that are likely to impact on the probability of being acquired, and this guided our choice of covariates in the multinomial model. Palepu (1986) conjectures that size is an important determinant of acquisition, as it may be more difficult for larger firms to be acquired due to financial constraints. Older firms are also thought of as less likely take-over targets. The 'value maximisation hypothesis' contends that acquisition is a disciplinary device whereby less efficient or profitable firms are taken over by the most efficient institutions (Shleifer and Vishny, 1988). Firms with a high level of liquidity are also identified as being attractive take-overs targets (Palepu, 1986). However, a high level of liquidity could also signal a lack of investment opportunities, thereby making the firm less attractive to potential bidders. The hypothesis put forwards by Shleifer and Summers (1988) posits that take-overs are motivated by the opportunity they offer to renege on implicit labour contracts and reduce extra-marginal payments. This suggests that high wage firms are likely to be acquired. Among industry-level determinants, industry concentration has been hypothesised to present a greater deterrence to domestic acquisitions (Shapiro, 1983). Furthermore, previous foreign presence in a sector is thought of as capturing the attractiveness of the sector from the perspective of new foreign entrants (Head et al., 1995). Also sectors experiencing higher growth rates are supposed to be attractive to market-seeking entrants. Finally we add a UK Assisted Area dummy to account for the possibility that firms in underdeveloped regions are more likely to be acquired due to financial incentives available to multinationals.

Table 7 presents the marginal effects from the multinomial model. UK multinationals appear to target smaller and less efficient firms. They also tend to acquire firms with cash-flow problems, but neither wages nor the sectoral and regional variables are found to exert any discernible influence on their decision to acquire. By contrast, wages, sectoral growth rates and the assisted area dummy play a significant and positive role in the acquisition decision of foreign multinationals and, more interestingly, foreign multinationals seem to 'cherry pick' the most efficient domestic firms from the pool of potential targets. In sum, this exploratory analysis suggests that it is probably erroneous to assume that acquisitions are exogenous to the process governing the productivity dynamics of the firms.

In order to isolate the ceteris paribus impact of multinational acquisition activity on the productivity growth of domestic non-exporting firms, we adopt the difference-in-differences methodology (e.g. Meyer, 1994). This approach proceeds in two steps. Firstly, the difference between the average productivity growth rates before and after the change of ownership, say [[DELTA].sup.a]TFP, is calculated. However this difference cannot be exclusively attributed to acquisition since the post-acquisition period growth rate might be affected by factors that are contemporaneous with ownership change. To cater for this, the difference obtained at the first stage is further differenced with respect to the before and after difference for the control group of non-exporting firms. (23) The resulting difference-indifferences estimator, [delta] = [[DELTA].sup.a]TFP-[[DELTA].sup.c]TFP, therefore purges the effects of common shocks, and provides a more accurate description of the productivity impact of acquisition. This procedure is equivalent to running a regression of the following form:

(7) [DELTA][TFP.sub.it] = [delta][A.sub.it] + [u.sub.it]

where A is a vector of post-acquisition dummies whose coefficient [delta] should produce the average percentage point change in productivity growth that can be attributed to multinational acquisitions. To allow for the differential acquisition effect, we construct two separate dummies for foreign and domestic multinational acquisitions. We also control for other factors that may be correlated with changes in productivity by including age, initial TFP and sectoral growth in equation (7).

The regression results are presented in table 8. Controlling for initial TFP, age, industry characteristics and autonomous technical changes (via time dummies), and assuming exogenous acquisitions, we find a 4.4 percentage points increase in total factor productivity growth due to foreign take-overs. In sharp contrast, no efficiency enhancing impact of domestic multinational acquisitions is obtained.

If productivity plays some role in driving acquisitions as implied by the multinomial logit analysis, the post-acquisition dummy may be endogenous. To break the stochastic dependence between the acquisition variables and the disturbance term in the productivity growth equation, we instrument acquisition by the probability of take-over computed from the multinomial models. Vella and Verbeek (1999) have recently shown that this method of dealing with endogenous dummies generates estimators that are equivalent to Heckman's (1978) endogeneity bias corrected OLS estimator. In other words, the exogenous variation in the covariates included in the multinomial models reported in table 7 is used to deal with the endogeneity of acquisition. The instrumental variable estimates show that foreign acquisitions lead to a 5.4 percentage point growth in the post ownership change period, whereas domestic acquisitions appear to cause a reduction in performance.

The positive productivity effect following foreign acquisitions is consistent with the internalisation theory of FDI, which postulates that multinational firms transfer a range of intangible proprietary assets to their affiliates. But how can one explain the lack of success by UK multinationals to transfer such assets to their newly acquired subsidiaries? One reason behind the absence of an average productivity effect could be that UK multinationals tend to target less efficient firms having a low level of absorptive capacity (Cohen and Levinthal, 1989). Without the necessary absorptive capacity, acquired firms might not be able to fully benefit from their new association with multinationals, at least in the short run. One way to explore empirically the above conjecture is by interacting the post-acquisition dummy variables with the pre-acquisition TFP of the acquired firms. The results from this experiment are reported in the last two columns of table 8, and they are supportive of the notion that greater absorptive capacity (proxied here by pre-acquisition TFP) increases the degree of technology transfer. A calculation, based on the point estimates from the last column of table 8, indicates that firms need to have an initial TFP level above 0.40 to start having a positive effect from acquisitions by UK multinationals. The raw data show that at the time of take-over, fewer than 40 of the 326 firms acquired by UK multinationals had TFP levels above this threshold. This may explain the finding of insignificant or negative average acquisition effects. The absence of efficiency-enhancing effects of domestic acquisitions is also consistent with the widely held proposition in the literature that some (and perhaps most) deals are motivated less by considerations of shareholder value, than by managerial aspirations for empire-building and associated private gains (e.g. Jensen, 1986). (24)

7. Conclusion

Based on a firm-level panel over the period 1989-98, this paper provides detailed statistical analyses of two aspects of UK multinational firms' domestic performance: their relative productivity dynamics and their ability to benefit newly acquired affiliates in the form of higher productivity growth.

Using the concept of first order stochastic dominance to rank productivity distributions, we find that the productivity distribution of foreign multinationals dominates that of UK multinationals. This conclusion is corroborated by results from quantile regressions, which also reveal that the performance disadvantage of UK multinationals is more acutely felt at the higher end of the productivity distribution. This suggests that the technology frontier is typically dominated by foreign multinationals, and UK multinationals do not appear to catch up fast enough with their foreign counterparts. The paper also finds that take-overs of non-exporting domestic firms by UK multinationals are not accompanied by significant productivity improvements, whereas foreign acquisitions are generally beneficial to the productivity trajectory of the acquired firms. This would appear to vindicate the effort undertaken by policymakers in the UK to attract incoming FDI, at least as far as the direct efficiency gains from FDI are concerned. However, this conclusion should be somewhat tempered by the realisation that foreign acquisition activity tends to target the better domestic firms.

Table 1. Frequency of firms by type of ownership

 UK-owned Foreign

Year Non- Export- MNE USA Others Total
 exporters ers

1989 1722 780 204 208 468 3382
1990 2780 1058 248 226 493 4805
1991 3043 1267 263 236 514 5323
1992 3260 1417 269 244 529 5719
1993 3444 1560 288 259 547 6098
1994 3596 1686 305 267 567 6421
1995 3736 1831 334 277 582 6759
1996 3831 2012 355 302 592 7092
1997 4725 2939 393 310 486 8853
1998 4876 2009 396 290 464 8935

Table 2. Summary statistics

 Domestic non- Domestic
 exporters exporters

Variable Mean Std. dev. Mean Std. dev.

Labour A 4.189 0.786 4.251 0.688
productivity B 0.445 0.387
 C 0.729 0.621
Total factor A -0.061 0.637 0.039 0.543
productivity B 0.253 0.287
 C 0.614 .513
Log employment A 4.329 1.322 4.629 1.155
 B 0.616 0.544
 C 1.257 1.098
Log capital A 6.867 1.146 6.864 1.024
intensity B 0.651 0.654
 C 1.039 0.909

 UK MNE USA MNE

 Mean Std. dev. Mean Std. dev.

Labour A 4.255 0.709 4.464 0.692
productivity B 0.554 0.678
 C 0.536 0.539
Total factor A 0.071 0.561 0.158 0.554
productivity B 0.420 0.437
 C 0.482 0.492
Log employment A 4.926 1.299 5.361 1.415
 B 0.882 1.223
 C 1.104 1.179
Log capital A 6.885 1.133 7.270 0.996
intensity B 0.952 0.855
 C 0.913 0.780

 Other foreign
 MNE

 Mean Std. dev.

Labour A 4.453 0.687
productivity B 0.429
 C 0.601
Total factor A 0.182 0.580
productivity B 0.307
 C 0.532
Log employment A 5.079 1.217
 B .732
 C 1.111
Log capital A 7.171 1.129
intensity B .725
 C 0.959

Notes: A = Overall; B = Between industry; C = Within industry. TFP so
measured can be interpreted as efficiency relative to the mean. For
example US multinationals are on average 18.2 per cent more efficient
than the 'average' firm within their industry. The average TFP
advantage of foreign multinationals vis-a-vis UK multinationals is
statistically significant. The latter average TFP is in turn
significantly higher than the TFP of arms-length exporters and domestic
firms.

Table 3. Kolmogorov-Smirnov test statistics of productivity differences
for selected years

 US MNES vs UK MNEs Non-US MNES vs UK MNEs
 Null hypothesis Null hypothesis
 Equality of Difference Equality of Difference
 distribution in US MNEs' distribution in non-US
 favour MNEs' favour

Productivity
 level
1990 0.1252 0.025 0.087 0.015
 (0.045) (0.871) (0.077) (0.930)
1994 0.1249 0.032 0.098 0.009
 (0.026) (0.749) (0.047) (0.965)
1998 0.126 0.000 0.128 0.001
 (0.012) (1.000) (0.002) (1.000)
Productivity
 growth
1990 0.139 0.026 0.068 0.389
 (0.044) (0.879) (0.575) (0.068)
1994 0.186 0.031 0.183 0.027
 (0.000) (0.772) (0.000) (0.757)
1998 0.015 0.034 0.093 0.041
 (0.001) (0.695) (0.064) (0.503)

 Domestic exporters vs UK Domestic non-exporters vs
 MNEs UK MNEs
 Null hypothesis Null hypothesis
 Equality of Difference Equality of Difference in
 distribution in UK MNEs' distribution in US non-
 favour MNEs' favour

Productivity
 level
1990 0.077 0.077 0.106 0.005
 (0.197) (0.099) (0.014) (0.989)
1994 0.075 0.003 0.134 0.004
 (0.12) (0.006) (0.000) (0.993)
1998 0.042 -0.019 0.112 0.008
 (0.601) (0.013) (0.000) (0.962)
Productivity
 growth
1990 0.054 0.049 0.0502 0.047
 (0.792) (0.021) (0.078) (0.470)
1994 0.055 0.047 0.151 0.034
 (0.471) (0.052) (0.001) (0.503)
1998 0.043 0.027 0.059 0.014
 (0.605) (0.037) (0.224) (0.057)

Notes: P-values are given in parentheses. The null hypotheses in the
above table are rejected when the p-values are less than some
pre-specified critical values (e.g. at 5 per cent level of
significance, p values greater than 0.05 suggest that the corresponding
null cannot be rejected.)

Table 4. The domestic performance of UK multinationals: quantile
regression estimates

 At the mean 10%

Lagged TFP -0.153 -0.122
 (27.17) ** (32.06) **
Age -0.001 -0.001
 (14.53) ** (3.22) **
Export intensity 0.036 0.002
 (2.10) * (0.10)
Industry growth 0.039 0.036
 (4.58) ** (2.60) **
Concentration -0.137 -0.105 *
 (4.12) ** (1.99)
USA MNEs 0.011 ** -0.031
 (2.35) (2.35) *
Non-USA MNEs 0.015 ** -0.025
 (1.96) (2.16) *
Domestic exporters -0.010 -0.009
 (1.55) (0.94)
Domestic non-exporters -0.023 -0.023
 (3.36) ** (2.35) *
Degree of multinationality 0.002 0.003
 (2.64) ** (2.76) **
Square of degree of multinationality -0.0001 -0.0002
 (1.54) (1.92)
Constant 0.023 -0.236
 (3.07) ** (21.11) **
Observations 46131 46131

 Quantiles
 25% 50%

Lagged TFP -0.087 -0.069
 (50.69) ** (49.27) **
Age 0.002 -0.002
 (0.53) (8.40) **
Export intensity 0.012 0.012
 (1.09) (1.36)
Industry growth 0.027 0.031
 (4.15) ** (6.09) **
Concentration -0.081 -0.102
 (2.96) ** (4.61) **
USA MNEs 0.006 0.013
 (0.88) (2.37) *
Non-USA MNEs 0.003 0.012
 (0.47) (2.53) *
Domestic exporters 0.000 -0.001
 (0.08) (0.13)
Domestic non-exporters -0.004 -0.005
 (0.83) (-1.23)
Degree of multinationality 0.001 -0.000
 (2.14) * (0.31)
Square of degree of multinationality -0.0001 0.0001
 (0.76) (1.33)
Constant -0.121 -0.019
 (20.90) ** (4.03) **
Observations 46131 46131

 75% 90%

Lagged TFP -0.082 -0.131
 (35.00) ** (19.36) **
Age -0.001 -0.002
 (15.85) ** (14.36) **
Export intensity 0.047 0.066
 (3.68) ** (2.14) *
Industry growth 0.032 0.049
 (4.29) ** (2.94) **
Concentration -0.088 -0.117 *
 (2.65) ** -1.96
USA MNEs 0.028 0.040
 (2.24) * (2.38) *
Non-USA MNEs 0.024 0.038
 (3.49) ** (2.44) *
Domestic exporters -0.012 -0.019
 (1.92) (1.46)
Domestic non-exporters -0.047 -0.037
 (2.87) ** (2.08) *
Degree of multinationality -0.001 0.002
 (1.78) (1.08)
Square of degree of multinationality 0.0001 -0.0002
 (2.13) * (0.58)
Constant 0.111 0.298
 (16.60) ** (20.51) **
Observations 46131 46131

Note: UK multinationals form the base group in all regressions. The
full set of time dummies is included throughout. * significant at 5
per cent; ** significant at 1 per cent.

Table 5. Quantile regression estimates based on within-firm time
averages

 At the mean 10%

Initial TFP -0.133 -0.076
 (17.57) ** (26.19) **
Initial Age -0.001 0.001
 (8.50) ** (3.43) **
Average export intensity 0.038 -0.058
 (2.41) * (4.46) **
Average industry growth 0.033 0.237
 (0.99) (10.00) **
Average concentration -0.414 -0.613
 (2.54) * (5.04) **
USA MNEs 0.005 -0.031
 (0.37) (2.12) *
Non-USA MNES 0.004 0.020
 (0.34) (1.53)
Domestic Exporters -0.024 -0.004
 (1.68) (0.37)
Domestic Non-exporters -0.021 -0.028
 (2.88) ** (2.49) *
Degree of multinationality 0.002 0.003
 (1.00) (1.99) *
Square of Degree of multinationality -0.0001 -0.0003
 (0.69) (1.13)
Constant 0.040 -0.118
 (3.39) ** (10.23) **
Observations 8925 8925

 Quantiles
 25% 50%

Initial TFP -0.065 -0.066
 (55.55)** (53.66) **
Initial Age 0.001 -0.002
 (0.03) (7.13) **
Average export intensity -0.006 0.011
 (1.01) (1.99) *
Average industry growth 0.142 0.025
 (15.98) ** (3.04) **
Average concentration -0.440 -0.184
 (10.34) ** (4.64) **
USA MNEs 0.016 0.006
 (2.43) * (0.87)
Non-USA MNES 0.015 0.009
 (2.52) * (1.50)
Domestic Exporters -0.002 -0.007
 (0.39) (1.38)
Domestic Non-exporters -0.009 -0.07
 (1.91) (2.37) *
Degree of multinationality 0.001 0.001
 -1.75 (0.05)
Square of Degree of multinationality -0.0001 0.0001
 (0.91) (0.68)
Constant -0.056 0.009
 (10.94) ** (1.83)
Observations 8925 8925

 75% 90%

Initial TFP -0.085 -0.127
 (35.14) ** (15.57) **
Initial Age -0.001 -0.001
 (11.19) ** (7.99) **
Average export intensity 0.016 0.049
 (1.96) * (2.23) *
Average industry growth -0.056 -0.146
 (4.23) ** (3.48) **
Average concentration -0.042 -0.021
 (0.64) (0.09)
USA MNEs 0.016 0.014
 (1.98) * (2.57) **
Non-USA MNES 0.008 0.011
 (1.96) * (3.53) **
Domestic Exporters -0.013 -0.031
 (1.85) (1.72)
Domestic Non-exporters -0.04 -0.03
 (2.52) ** (2.18) *
Degree of multinationality -0.001 -0.003
 (0.62) (1.38)
Square of Degree of multinationality 0.0001 0.0002
 (1.11) (2.49) *
Constant 0.082 0.204
 (10.72) ** (10.38) **
Observations 8925 8925

Notes: * significant at 5 per cent; ** significant at 1 per cent.
Initial age and initial TFP correspond to values of the first year
the firm is observed in the sample.

Table 6. Frequency of acquisitions by year

 Acquiring multinational

Year Foreign Domestic

1989 64 52
1990 60 49
1991 74 44
1992 70 48
1993 51 30
1994 94 41
1995 81 32
1996 58 30
Total 552 326

Table 7. The probability of acquisition: marginal
effects from the Multinomial Logit Models

 Acquisition by

Regressor UK multinationals Foreign multinationals

Size -0.0022 (2.70) ** 0.009 (8.69) **
Age -0.0002 (3.34) * -0.0001 (2.12) *
Wage 0.0039 (0.38) * 0.0154 (3.17)
TFP -0.004 (2.16) * 0.136 (4.94) **
Liquidity -0.0126 (3.68) ** -0.008 (1.60)
Concentration -0.0204 (1.46) 0.013 (0.51)
FDI 0.001 (0.30) 0.041 (1.50)
Sectoral growth 0.002 (0.51) 0.006 (1.97) *
Assisted Area -0.0004 (0.18) 0.003 * (1.98)
Observations 21838
Log likelihood -2487.3
p-value for test
of IIA 0.71 0.89

Notes: 'No change of ownership during the year' is used as the base
category. All variables are lagged by one period. Liquidity is proxied
by total assets minus current liabilities. The absolute values of the
t-statistics are reported in parentheses. IIA stands for the assumption
of Independence of Irrelevant Alternatives. In our context, this
implies that the effects of the regressors on the probability of being
taken over by a domestic multinational is not affected by the inclusion
of foreign acquisitions as a possible outcome. We fail to reject the
validity of this assumption by conducting Hausman-type tests suggested
by Hausman and McFadden (1984).

Table 8. Post-acquisition productivity trajectories

 Baseline specification

 Exogenous Endogenous
 acquisitions acquisitions

Lagged TFP -0.174 -0.178
 (19.22) ** (19.13) **
Age -0.001 -0.001
 (10.71) ** (10.70) **
Industry growth 0.050 0.050
 (4.61) ** (4.60) **
Concentration -0.119 -0.120
 (2.67) ** (2.69) **
Foreign acquisition Effect 0.046 0.054
 (5.72) ** (4.34) **
Foreign acquisition * pre-acquisition
 TFP
Domestic acquisitions effect -0.007 -0.018
 (0.91) (2.16) *
Domestic acquisition * pre-acquisition
 TFP
Constant 0.054 0.055
 (3.03) ** (3.03) **
Sargan test of validity of instruments .318
 (p-value)
Observations 21838 21838

 Specification with pre-
 acquisition TFP
 interaction

 Exogenous Endogenous
 acquisitions acquisitions

Lagged TFP -0.179 -0.178
 (19.09) ** (18.95) **
Age -0.001 -0.001
 (10.66) ** (10.63) **
Industry growth 0.050 0.050
 (4.58) ** (4.58) **
Concentration -0.119 -0.120
 (2.67) ** (2.69) **
Foreign acquisition effect 0.006 0.007
 (0.73) (0.56)
Foreign acquisition * pre-acquisition 0.118 0.096
 TFP (5.72) ** (4.65) **
Domestic acquisitions effect -0.019 -0.032
 (2.37) * (3.42) **
Domestic acquisition * pre-acquisition 0.068 0.080
 TFP (2.87) ** (3.39) **
Constant 0.055 0.055
 (3.07) ** (3.06) **
Sargan test of validity of instruments .214
 (p-value)
Observations 21838 21838

Notes: Robust t-statistics in parentheses. ** significant at 5 per
cent; ** significant at 1 per cent. The full set of time dummies is
included throughout. Unlike the specification in the previous sections
export intensity is not included in the vector of regressors as we
focus on the acquisition of non-exporting domestic firms by
multinationals.

NOTES

(1) Source: Office for National Statistics, Foreign Direct Investment inquiries.

(2) A robust finding reported in this literature is the existence of significant productivity differentials in favour of foreign multinational firms operating in the UK. For recent micro evidence, see Girma et al. (2001) and Griffith and Simpson (2002). For evidence based on US data see Doms and Jensen (1998).

(3) Table 1 of Girma et al. (2002) summarises details of recent papers on nine countries.

(4) Conyon et al. (2002) and Harris and Robinson (2002) report evidence on the productivity implications of acquisitions. None of these studies was able to isolate acquisitions made by UK multinationals.

(5) The development of the concept of firm-specific or proprietary assets can be attributed to a number of authors. See also Dunning (1977), Buckley and Casson (1976) and Caves (1971).

(6) For this study we used the OneSource CD-ROM entitled 'UK companies, Vol. I', for October 2000.

(7) This information, which is in hard copy format, is obtained from the Office of National Statistics upon special request. The matching process required considerable effort, and I wish to thank Mehtap Hisarciklilar for helping me in this regard.

(8) I wish to express my gratitude to Davide Castellani and Antonello Zanfei for allowing me to use some information from this database.

(9) Parent companies were omitted if they have consolidated accounts, as this leads to double counting.

(10) To allow cross-time comparisons, we converted current values of output and inputs using highly disaggregated price indices obtained from the Office for National Statistics. Some extrapolation is done for missing years/sectors.

(11) Estimation of production functions is not performed at the more disaggregated four-digit level to maximise the number of observations available for estimation.

(12) TFP is therefore generated econometrically from a dynamic translog production function with four inputs. We also experimented with other productivity measurement approaches, namely labour productivity and Solow residuals, and generally find that they are highly correlated. Results based on these measures are available from the author upon request.

(13) Recently Delgano et al. (2002) examined productivity differentials between exporters and non-exporters in a sample of Spanish manufacturing firms, using the same approach outlined in this section.

(14) For example, see Bernard and Jensen (1997) for evidence on the US; for Columbia, Mexico and Morocco, Clerides et al. (1998); for Spain, Delgano et al. (2002); and for the UK Girma et al. (2002).

(15) One should keep in mind, however, Melitz's (2000) observation that, in an industry with differentiated products, comparing the productivity of firms based on a common price index might be misleading. For example Melitz shows that if the elasticity of substitution is higher in the international market, then measured productivity differences between exporting and non-exporting firms will be underestimated.

(16) See Bartelsman and Doms (2000) for a comprehensive review.

(17) See Buchinsky (1998) for an excellent overview of quantile models.

(18) Although it is not possible to test formally for the difference between the mean effects and the effects at each quantile, it is possible to test for the significance in the difference between pairs of quantiles. The test results vindicate the use of quantile regressions.

(19) At the 75th percentile the square of the multinationality variable is positive and significant, whilst the linear term is negative and weakly significant. If anything, this suggests a U-shaped, rather than an inverted U-shaped relationship.

(20) A project is deemed internationally non-mobile if it is considered to be suitable for a location in the UK only. Thus it can reasonably be inferred that internationally mobile projects are most probably undertaken by UK multinationals, or domestic firms that have the potential to be multinationals.

(21) See the official report at http://www.dti.gov.uk/regional/ evaluationRSA91-95.pdf

(22) Apart from distinguishing between domestic exporters, non-exporters and UK multinationals, this is a much more extended sample than the one used by Conyon et al. (2001).

(23) Hence the name, difference-in-differences.

(24) See Dickerson et al. (1997) for econometric evidence from the UK, which supports this view that domestic acquisition activity, is, on average, detrimental to profitability.

REFERENCES

Bartelman, E.J. and Doms, M. (2000), 'Understanding productivity: lessons from longitudinal microdata', Journal of Economic Literature, 38, pp. 569-95.

Benvignati, A.M. (1987), 'Domestic profit advantages of multinational firms', Journal of Business, 60, pp. 449-61.

Bernard, A. and Jensen, J. (1997), 'Exceptional exporters performance: cause, effect or both?', Journal of International Economics, 47, pp. 1-25.

Braunerhjelm, P. (1996), 'The relation between firm-specific intangibles and exports', Economic Letters, 53, pp. 213-19.

Buchinski, M. (1998), 'Recent advances in quantile regression models', Journal of Human Resources, 33, pp. 88-126.

Buckley, P.J. and Casson, M. (1976), The future of the Multinational Enterprise, London, Macmillan.

Caves, R.E. (1971), 'International comparisons: the industrial economics of foreign investment', Economica, 38, pp. 1-27.

Clerides, S., Lach, S. and Tybout, J. (1998), 'Is learning by exporting important? Micro-dynamic evidence from Columbia, Mexico and Morocco', Quarterly Journal of Economics, CXIII, pp. 903-48.

Cohen, W.M. and Levinthal, D.A. (1989), 'Innovation and learning: the two faces of R&D', Economic Journal, 99, pp. 569-96.

Conover, W.J. (1999), Practical Nonparametric Statistics, New York, John Wiley & Son.

Conyon, M.J., Girma, S., Thompson, S. and Wright, P.W (2001), 'The productivity and wage effects of foreign acquisition in the United Kingdom', Journal of Industrial Economics, L, pp. 85-102.

Criscuolo, C. and Martin, R. (2002), 'Multinationals, foreign ownership and productivity in UK businesses', Mimeo, Centre for Research into Business Activity, ONS.

Delgano, M.A., Farinas, J.C. and Sonia, R (2002), 'Firm productivity and export markets: a non-parametric approach', Journal of International Economics, 57, pp. 397-422.

Dickerson, A.P., Gibson, H.E. and Tsakalotos, E. (1997), 'The impact of acquisition on company performance: evidence from a large panel of UK firms', Oxford Economic Papers, 49, pp. 344-61.

Doms, M.E. and. Jensen, J.B. (1998), 'Comparing wages, skills, and productivity between domestically and foreign-owned manufacturing establishments in the United States', in Baldwin, R., Lipsey, R. and Richardson, J.D. (eds), Geography and Ownership as Bases for Economic Accounting, Chicago, Chicago University Press.

Dunning, J.H. (1977), 'Trade, location of economic activity and MNE: a search for an eclectic approach', in Ohlin, B., Hesselborn, P.O. and Wijkman, P.M. (eds), The international allocation of economic activity, London, Macmillan.

Girma, S., Greenaway, D. and Kneller, R. (2002), 'Exporting and productivity: a microeconometric analysis of matched firms', GEP Working Paper, University of Nottingham.

Girma, S., Greenaway, D. and Wakelin, K. (2001), 'Who benefits from foreign direct investment in the UK?', Scottish Journal of Political Economy, 48, pp. 119-33.

Girma, S., Kneller, R. and Pisu, M. (2003), 'Exports versus FDI: an empirical test', mimeo, University of Leicester.

Grant, R.M. (1987), 'Multinationality and performance among British manufacturing companies', Journal of International Business Studies, 18, pp. 79-89.

Grant, R.M., Jammine, J.P. and Thomas, H. (1988), 'Diversity, diversification and profitability among British manufacturing companies', Academy of Management Journal, 40, pp. 767-98.

Greniger, J.M., Beamish, P.W. and da Costas, R.C. (1989), 'Diversification strategy and internalization: implication for MNE performance', Strategic Management Journal, 10, pp. 109-19.

Griffith, R. and Simpson, H. (2002), 'Characteristics of foreign-owned firms in British manufacturing', in Blundell, R., Card D. and Freeman, R. (eds), Creating a Premier League Economy, Chicago, Chicago University Press.

Gomes, L. and Ramaswamy, K. (1999), 'An empirical examination of the form of the relationship between multinationality and performance', Journal of International Business Studies, 30, pp. 173-88.

Han, K.C and Lee, S.H. (1998), 'Multinationality and firm performance', Multinational Business Review, 6, pp. 63-70.

Harris, R. and Robinson, C. (2002), 'The effect of foreign acquisitions on total factor productivity: plant-level evidence from UK manufacturing, 1987-1992', The Review of Economics and Statistics, 84, pp. 562-8.

Hausman, J. and McFadden, D. (1984), 'Specification tests for the multinomial logit model', Econometrica, 52, pp. 1219-40.

Head, K., Ries, J. and Swenson, D. (1995), 'Agglomeration benefits and location choice: evidence from Japanese manufacturing investments in the United States', Journal of International Economics, 39, pp. 223-247.

Heckman, J.J. (1978), 'Dummy endogenous variables in a simultaneous equation system', Econometrica, 46, pp. 931-59.

Hitt, M.A., Hoskisson, R.E. and Kim, H. (1997), 'International diversification: effects on innovation and firm performance in product diversified firms', Academy of Management Journal, 40, pp. 767-98.

Hymer, S.H. (1976), The International Operations of National Firms: A Study of Direct Foreign Investment, Boston, MIT Press.

Jensen, M. (1986), 'Agency costs of free cash flow, corporate finance and takeovers', American Economic Review: Papers and Proceedings, 76, pp. 323-9.

Kindleberger, C.P. (1969), American Business Abroad: Six Lectures on Direct Investment, New Haven, Yale University Press.

Koenker, R. and Bassett, G. (1978), 'Regression quantiles', Econometrica, 50, pp. 43-61.

Kogut, B. (1985), 'Designing global strategies: profiting from operational flexibility', Sloan Management Review, 26, pp. 27-38.

Kumar, M.S. (1984), Growth Acquisition and Investment, Cambridge, Cambridge University Press.

Mathur, I., Singh, M. and Kinberly, C.G. (2001), 'The evidence from Canadian firms on multinational diversification and performance', The Quarterly Review of Economics and Finance, 41, pp. 561-78.

Melitz, M.J (2000),'Estimating firm-level productivity in differentiated product industries', mimeo, Harvard University.

Meyer, B. (1994), 'Natural and quasi-experiments in economics', NBER Technical Working Paper No. 170.

Palepu, K.G. (1986), 'Predicting takeover targets: a methodological and empirical analysis', Journal of Accounting and Economics, 8, pp. 3-35.

Rugman, A.M. (1974), 'Motivation for foreign investment: the market imperfections and risk diversification hypotheses', Journal of World Trade Law, 9, pp. 567-673.

Shapiro, D. (1983), 'Entry, exit and the theory of multinational corporation, in Kindleberger, C. and Audretsch, D. (eds), The Multinational Corporation in the 1980s, Cambridge, Mass.

Shleifer, A. and Summers, L.H. (1988), 'Breach of trust in hostile takeovers', in Auerbach, A.J. (ed.), Corporate Takeovers: Causes and Consequence, Chicago, NBER.

Shleifer, A. and Vishny, R. (1988), 'Value maximization and the acquisition process', Journal of Economic Perspectives, 2, pp. 7-20.

Vella, F and Verbeek, M. (1999), 'Estimating and interpreting models with endogenous treatment effects', Journal of Business and Economic Statistics, 17, pp. 473-8.

Sourafel Girma, Department of Economics, University of Leicester. E-mail: sg116@leicester.ac.uk. I wish to thank an anonymous referee for some useful comments on an earlier draft of this paper. I also wish to acknowledge the support of Martin Weale.