Almeida The Labor

Journal of International Economics 72 (2007) 75 96 www.elsevier.
com/locate/econbase
The labor market effects of foreign owned firms

Rita Almeida
The World Bank, United States Received 15 March 2004; received in revised form 28 September 2006; accepted 31 October 2006
Abstract Foreign firms have a more educated workforce and pay higher wages than domestic firms even after controlling for worker quality, at a given moment in time. This does not imply that foreign ownership improves the labor market outcomes of the workers since foreign investment may be guided by unobservable firm and worker characteristics correlated with schooling or wages. This paper asks whether foreign investors acquire firms with high human capital or wages, or whether foreign acquisition improves these outcomes. Using a matched employeremployee data set, I find that foreign acquisitions of domestic firms have small effects on the human capital and on average wages of the acquired firms. Instead, foreign investors cherry pick those domestic firms that are already very similar to the group of existing foreign firms. 2006 Elsevier B.V. All rights reserved.
Keywords: Foreign direct investment; Acquisitions; Employment; Wage structure; Panel data JEL classification: C31; F23; J31
1. Introduction A large empirical literature documents that, in the cross-section, foreign firms are larger, more productive, more capital intensive, pay higher wages and have a more skilled workforce than
I am very grateful to Antonio Ciccone for raising my interest in the topic and for his guidance and to Pedro Carneiro for very valuable comments. I am also very grateful to the editor and two anonymous referees for many detailed comments that substantially improved the paper. I am grateful to the Department of Statistics of the Portuguese Ministry of Employment for access to the data. I thank Koen Debacker, Jaume Garcia, Maia Gell, Adriana Kugler and Jos Mata for several suggestions. I benefited from comments made at the Spie, UPF, Simposio de Analisis Economico, UNL, CEPR conference Globalization and labor markets(Bergen), World Bank and IIOC (Chicago). All the errors are my own. I gratefully acknowledge the financial support of Fundao para a Cincia e Tecnologia. 1818 H Street, NW, MSN MC3-301, Washington, DC 20433, USA. E-mail address: ralmeida@worldbank.org. 0022-1996/$ - see front matter 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.jinteco.2006.10.001
76
R. Almeida / Journal of International Economics 72 (2007) 7596
domestic firms. This suggests that foreign investors may have a positive effect on the welfare of the workforce of the host economy. This argument has been used to justify regional or national industrial policies to attract and secure foreign investment. But, while cross-sectional differences are large, they are not necessarily causal. For example, they may arise because foreigners acquire domestic firms that already have a more educated workforce and pay higher wages than the average firm, because foreign investment leads to an increase in the demand for skills and average wages of the acquired firms, or both. Disentangling correlation from causality is crucial for understanding the welfare effects of foreign investment in the host economy and for designing appropriate policies. In this paper, I analyze the interaction between foreign ownership and labor market outcomes using a Portuguese matched employeremployee data set during the nineties. The paper addresses the following questions: (1) Do foreigners acquire domestic firms with high human capital and that pay higher wages even after controlling for worker quality? (2) Does foreign ownership improve the labor market outcomes of the acquired firms? To quantify the effects of foreign ownership, I analyze the evolution in total employment, human capital of the workforce and average wages conditional on worker quality following a foreign acquisition of a domestic firm.1 The main findings of the paper can be summarized as follows. Most of the large cross sectional differences between foreign and domestic firms are explained by foreigners cherry picking the domestic firms. In the years prior to the acquisition, firms that will become acquired are already larger, employ a more educated workforce and pay higher hourly wages for a given worker quality (both to low and high educated workers) than the average domestic firm in the sector. Moreover, before the foreign acquisition firms are already very similar to the group of existing foreign firms. Consistently with the few theories of foreign acquisitions, these findings strongly suggest that foreign acquisitions are not random. When foreign firms enter a new country and/or a new product market, by acquiring an existing firm it is more likely that they choose firms with relatively high productivity and with technological characteristics that are similar to their own (Hennart and Park, 1993; Buckley and Casson, 1998). Otherwise, foreigners would face very high costs of adapting the technology, changing the workforce and gaining experience in the host country. In Portugal the costs of adjusting the workforce are particularly high since it has one of the most restrictive employment protection regulations in the world. I also find evidence that the change in the nationality of the ownership has little effect on different labor market outcomes of the acquired firms. By comparing the same firms before and after the acquisition, I find that following the acquisition the size of the firm increases but that there is no significant change in the human capital of the average worker in the firm.2 Moreover, there is evidence that following the acquisition average wages increase slightly in manufacturing firms but these results are in stark contrast with the large cross sectional estimates of the foreign wage premium. Suggestive evidence shows that this might not be specific to foreign acquisitions since wages go up by similar magnitudes following domestic acquisitions. The Portuguese case is interesting because it combines two important features. First, Portugal had a permissive legal framework for the operation of foreign firms that translated into substantial
1 Other papers have used foreign acquisitions to quantify the foreign wage premium (e.g., Aitken et al., 1996; Conyon et al., 2002a; Lipsey and Sjoholm, 2002). See Navaretti and Venables (2004) for a survey. 2 I find that employment in the firm increases on average by 14% following the foreign acquisition. However, this evidence should be interpreted cautiously since there is also evidence of selection of foreign investment into firms where employment was growing faster prior to the acquisition.
77
amounts of foreign direct investment (FDI) in the late eighties and nineties. Before becoming an European Union member in 1986, the amount of FDI in Portugal had never reached 1% of GDP (3.2% of total investment) and in the beginning of the nineties this influx had more than tripled to 3.2% of GDP (12% of total investment in 1990).3 Second, there exists a census of all the firms operating in Portugal covering the nineties. For each firm in this data set, foreign participation can be traced over time so that the identification of foreign acquisitions of domestic firms is possible. Moreover, since it is a matched employeremployee data set, workforce characteristics can also be traced over time making this data particularly well suited for the analysis of the effects of foreign ownership on labor market outcomes. One of the contributions of the paper is to provide a systematic evaluation of the impact of foreign investment on different labor market outcomes of workers in the host country. While there is some evidence for the UK that foreigners buy domestic firms with above average productivity and wages, little is known about the human capital characteristics of firms that are acquired by foreigners (e.g., Harris and Robinson, 2002; Girma and Grg, 2003a,b; Conyon et al., 2002a).4 There is also evidence that following a foreign acquisition average wages increase slightly but it is unclear whether this is the result of worker reallocation and changes in the firm's human capital or as a result of increases in labor productivity (e.g., Conyon et al., 2002a; Girma and Grg, 2003a,b). The main reason for this lack of knowledge is the difficulty in obtaining datasets that simultaneously have information on characteristics of the firm and of its workforce over time. Using a sample of Indonesian manufacturing firms Lipsey and Sjoholm (2002) focus on the evolution of employment and average wages within broad skill groups following a foreign acquisition. 5 The Portuguese data has the advantage of covering both manufacturing and non-manufacturing sectors and to have, for each year, detailed information on all wage earners in the private sector. Since each individual is uniquely matched with a firm, several workforce characteristics are observed for each firm. Interestingly, my findings for foreign acquisitions in Portugal are closer to those found for the UK and are quite different from those found by Lipsey and Sjoholm (2002) for Indonesia. Lipsey and Sjoholm (2002) do not find evidence of selection of foreign investment into high wage firms and show that following the acquisition average wages increase substantially, possibly due to increases in productivity. 6 They also show that the employment of blue collar workers and average wages within occupations increase after the acquisition. The stark contrast with my findings suggest that the effects of foreign acquisitions in the labor markets in developed and developing countries can be quite different. My paper is also related with the literature that investigates the foreign wage premium (see Table A.1 in the appendix for a summary). If foreigners are in a disadvantaged position to attract the best workers, they would have to pay higher wages than comparable domestic firms with
3 The most important sources of FDI into Portugal is the European Union and OECD, with 76% and 89% of the FDI in 1992. 4 Harris and Robinson (2002) and Girma and Grg (2003a,b) find evidence of cherry picking of foreign investment into UK firms with a high level and growth of productivity. Conyon et al. (2002a) also find evidence that acquired firms pay above average wages. 5 Table A1 shows that Lipsey and Sjoholm (2002) have information on average wages for blue and white collar workers but they do not have information on schooling or other worker characteristics. Lipsey and Sjoholm (2004) have information on schooling but it is only available in 1996. 6 Arnold and Javorcik (2005) find evidence for Indonesian plants that foreign ownership leads to an increase in productivity in the acquired plants by approximately 30% over the three years following the acquisition. In contrast, Harris and Robinson (2002) find that for the UK, average firm productivity falls in the short run following a foreign acquisition.
78
similar worker quality.7 I am also motivated by the evaluation of the effects of acquisitions on labor reallocation. Ownership changes are often associated with the real-location of resources from inefficient to efficient firms. This can cause the displacement of jobs and lower wages and it is a source of preoccupation for workers and labor unions (e.g., Harris and Robinson, 2002). Furthermore, firm acquisitions may change the allocation of monopoly rents between firms and workers which can also lead to wage changes (McGuckin and Nguyen, 2001). Finally, my work also relates to the literature that studies the profile of firms acquired by foreigners, and the motives for a foreign acquisition of a domestic firm. The paper proceeds as follows. In the next section, I describe the data and the sample used. Section 3 presents evidence on how much better foreign-owned firms look at any point in time and asks whether foreigners select the best domestic firms to be acquired. Section 4 analyzes the effects of foreign acquisitions on different labor market outcomes. Section 5 discusses the robustness of the results. Section 6 concludes. 2. Data The data set used is a based on a mandatory yearly survey conducted by the Portuguese Ministry of Employment, Quadros do Pessoal. The survey covers all the firms with wage-earners in Portugal excluding the public sector. It is a matched employeremployee data where firms can be traced over time and each period workers can be matched to firms.8 The survey collects firm information on the ownership of the capital (foreign, domestic private and public), five digit-ISIC sector, region, size and age of the firm. It also reports detailed worker characteristics like gender, age, schooling, tenure on the job, monthly hours worked (normal and extra hours) and gross monthly wage. To construct the sample, I identified all the firms that changed the nationality of their ownership during the period covered 19911998.9 I consider a firm to be foreign owned if the share of foreign capital is larger than 10%.10 I selected all the firms that are continuously owned by domestic and foreign owners. I kept only those firms that report information on key variables in the paper like wages, hours worked and schooling. I identify 1381 firms that change the nationality of their ownership between 1991 and 1998, of which 688 are foreign acquisitions (domestic firms acquired by foreigner firms) and 505 domestic acquisitions (foreign firms
7 There are different reasons for foreigners paying higher wages (e.g., Navaretti and Venables, 2004). First, workers could have a preference for domestic firms. This could happen if, for example, foreign firms are perceived to be more volatile employers. There is little evidence supporting this claim, though. Second, foreigners might be more likely to minimize turnover if they invest more in training or want to avoid the leakage of their technological advantages. Third, a foreigners might be interested in attracting the best workers. 8 By law the information collected is sent to the Ministry of Employment and latter is made available to all the workers in the firm. Since the data is administrative and it is made available it tends to be very reliable. The survey has been conducted since 1982 and is organized in firm, plant and individual files. The firm level data covers 148,044 firms in 1991 and 224,456 in 1998. The individual files cover approximately 2 million workers each year. 9 Some observations report missing information on the ownership structure of the capital. I tested the robustness of the results to excluding these firms and to make different assumptions on the nationality of the capital. Firms observed just once are assumed to be domestic firms, given its similarity in size, education and average wages to the average domestic firm in the same sector. Firms observed more than once, are assumed that they kept either their last ownership structure or an average of the past ownership's structure (more than 95% of these firms are assumed to be domestic). Since I suspect measurement error problems for those firms with more than one ownership change, these firms were excluded from the sample (less than 3% of the acquisitions). 10 The choice of the 10% threshold is not restrictive since 95% (75%) of the firms with positive foreign ownership have more than 10% (50%) of foreign capital and 50% of the firms with a foreign participation have a full foreign ownership.
79
acquired by domestic firms).11 About 90% of the acquisitions in the sample are observed for three or more periods.12 Since I am interested in analyzing the effects of foreign ownership on firm's characteristics of the workforce, I aggregate these at the firm level. This aggregation is possible because each worker has a unique firm identifier. I consider a worker to be low educated if it has nine or less years of schooling and a worker to be high educated if it has more than nine years of schooling.13 Gross monthly wages are computed summing up monthly labor earnings for regular and extra work hours. Hourly wages are gross wages divided by total monthly hours worked. Firm average wages are computed excluding extreme values for hourly wages.14 The price cost margin is defined as total sales minus total labor costs divided by total sales.15 The consumer price index comes from National Department of Statistics. The level of aggregation used for sector composition was two-digit ISIC sector (total 46 sectors). 3. Evidence on selection of foreign investment In this section I document the differences between foreign and domestic owned firms and examine whether firms that become acquired by foreigners have an average worker with a very different quality than the typical domestic firm. The latter is important to understand the motivations for foreign entry. Moreover, there is almost no evidence in the literature on the human capital profile of the average worker in acquired firms. This is important since foreign firms are often associated with a better technology and a more intensive use of skilled labor than domestic firms. However, this need not translate into an increase in the employment of skilled workers, as long as foreigners acquire firms that already have the desired worker profile. Looking at the theoretical and empirical literature on mergers and acquisitions it is difficult to extract strong predictions about the profile of the acquired firm and its post acquisition performance. This literature has identified two motives for acquisitions independently of the nationality of the acquiring firm. The neoclassical approach views acquisitions as a form of natural selection motivated by profit maximizing managers which result in the replacement of inefficient firms (Lichtenberg and Seigel, 1987, 1990).16 Alternatively, acquisitions could be made by managers motivated to acquire operating efficiency (McGuckin and Nguyen, 1995). In
11 Domestic firms are firms with at least 90% of national capital each year, foreign firms are firms with at least 10% of foreign capital each year and acquisitions are firms that changed once their ownership nationality over the sample period. In the final sample there are 194,560 domestic firms, 2350 foreign firms and 1381 acquisitions. In the manufacturing sample, there are 40,873 domestic firms, 701 foreign firms and 205 foreign acquisitions of domestic firms. 12 81% and 72% of the foreign acquisitions are observed one year before and one year after the acquisition, respectively. About 50% and 52% of the foreign acquisitions are observed two years before and two years after the acquisition, respectively. 13 I prefer the aggregation by education groups rather than by occupational groups since it is less ambiguous and also less likely to be affected by ownership changes. In the worker level data there is a positive correlation between education and skills: 84% of the high educated are skilled workers and 40% of the low educated are unskilled workers. 14 Workers with implausibly low earnings (hourly wage lower than 50% of the minimum wage) or implausibly high earnings were excluded from the sample. There were less than 1% of the workers in each years in these cases. 15 Unfortunately, the Portuguese data does not collect sufficient information to compute firm's productivity. It collects information on total sales but there is no information on raw materials or stock of capital in the firm. Approximately 9% of the firms in the sample report zero sales and hence do not have a price cost margin. I exclude from the analysis those firms with values below percentile 1 and above percentile 99 of the distribution for price cost margin. 16 This theory also predicts that firms will perform better after the acquisition since they are disciplined by managers with cost savings policies.
80
this case, firms with high levels of productivity are more likely to change ownership.17 The empirical literature also shows different results. Lichtenberg and Seigel (1987, 1990) and Conyon et al. (2002b) find that firms with lower productivity or profits are more likely to be acquired by domestic firms, in the US and UK, respectively. Ravenscraft and Scherer (1987), Matsusaka (1993) and McGuckin and Nguyen (1995) find for the US that the more profitable and high productivity firms are more likely to change ownership.18 The theory looking at the motivations of foreign acquisitions is more scarce but less ambiguous. Markusen (1995) shows that foreign firms prefer to establish capacity in the host country, in opposition to export or technology licensing, whenever firms have specific advantages that overcome the costs of entry. Caves (1996) and Pfaffermayr (1999) relate these advantages to economies of scale, brand names, management know-how and other types of advantages that can be exploited in different locations without incurring in additional costs. The literature on the motivations of the different entry modes argues that foreign firms are more likely to acquire an existing firm rather than start a new firm if they have little experience in producing in the host country and learning costs are high, if they are entering a new product market or if there are high costs of increasing competition (Hennart and Park, 1993; Buckley and Casson, 1998). Moreover, conditional on choosing to acquire, foreign firms are more likely to buy domestic firms with relatively high productivity and with technological characteristics that are similar to their own. Otherwise, foreigners firms would face very high costs of adapting the technology, changing the workforce or of gaining experience in the host country. In Portugal, the costs of adapting the workforce are particularly important since it has one of the most restrictive employment protection regulations in the world. The legislation on collective dismissals imposes a long, complex and very costly dismissal process on employers. In particular, the employer must justify a worker's dismissal. The employer must also notify the impending dismissal to the commission of workers as well as the Ministry of Employment. The dismissal will only become official after the commission issued an appraisal and after the worker is given the opportunity to dispute the allegations made. Dismissed workers must receive a written notice at least 60 days prior to the dismissal. These workers then have certain benefits, which include time off to look for a new job, a financial compensation and a special right to resign. Except in the case of a disciplinary dismissal (serious infraction by the worker), the employer must pay the employee one monthly wage per year worked, with a minimum of three monthly payments. No maximum ceiling is established in the law for the severance payments. However, Portugal has a relatively flexible wage policy. The minimum wage is set annually at the national level and collective bargaining agreements establish the starting wage for each occupation. These wage floors are established at relatively low levels so that, in practice, employers have some flexibility in setting the wages. (Bover et al., 1998, discuss in detail the Portuguese labor regulation). Together, these arguments suggest that in my sample, when foreigners choose to enter the market by acquiring an existing firm, they tend to select those firms that already have many of the characteristics that are desirable to them.
17 Theoretical models also yield conflicting predictions for firm performance in the period following the acquisition. When acquisitions are motivated by managerial discipline, some models predict that firm performance improves following the acquisition. When acquisitions are motivated by the acquisition of operating efficiency, models predict that firm performance remains unchanged or decreases slightly following the acquisition. A negative performance could be more likely to occur with foreign acquisitions, where assimilation problems could potentially be larger. 18 Ravenscraft and Scherer (1987) and Matsusaka (1993) show that firms and had little or no gain in productivity in the post-acquisition period. Conyon et al. (2002b) look specifically at the evolution of employment following a change in ownership. They find that acquisitions are generally followed by reductions in employment and output resulting in an increased efficiency in the use of labor.
81
The empirical analysis of foreign acquisitions for developed countries is focused on the UK and finds that acquiring foreign firms cherry pick domestic firms. However, nothing is known about the human capital of firms acquired by foreigners. For example, in the procedure of selection of foreign investment into firms with more educated workers, these firms would pay higher average wages in the pre acquisition period but would not necessarily pay higher wages conditional on worker's schooling. Alternatively, if selection of foreign investment is also based on unobservable firm characteristics (e.g., technological or organizational advantages) average wages for a given worker quality would already be higher in the pre acquisition period. One way of testing for selection of foreign investment, is to compare firms that will become acquired by foreigners, existing foreign firms and domestic owned firms that will not be acquired in the near future.19 I estimate the following equation for different outcomes: yjt bAj dFj
S X s 1
gs Ds
T X t 1
gt Dt jt
3:1
where yjt is the outcome of interest for firm j in period t (employment, average schooling, age, tenure, share of females in the firm, average hourly wage and pricecost margin), Aj is a dummy variable that assumes the value one in the years before the acquisition for those firms that will be acquired by foreigners, Fj is a dummy variable that assumes the value one if firm j is foreign owned in year t and Ds is a two digit sector dummy and Dt is a time dummy for year t. The foreign premium , represents the average difference in outcome y between foreign and domestic firms in the same sector of activity and year. If (future) foreign acquisitions are already better performers relatively to domestic firms in the years prior to the acquisition, should be large and statistically significant. Moreover, if future foreign acquisitions are very similar to existing foreign firms, I expect to see to be very similar to . On the other hand, if foreigners do not cherry pick domestic firms, should be small and different from . However, one would expect foreign firms to pay higher hourly wages simply because their workforce has a different human capital composition than the workforce in domestic firms. Mincer (1974), and a large literature that followed, shows that differences in education, age, tenure and gender are associated with significant differences in hourly wages. Hence, when the dependent variable in Eq. (3.1) is average hourly wages, I also control for differences across firms in average education, age, tenure and gender of the workers, so that the wage premiums captured by and exclude these composition effects.20 Eq. (3.1) is estimated by least squares using all the firms. Standard errors are clustered at the firm level. The estimates are reported in columns (1) to (10) of Table 1. The table also reports the p-values for the test: = . I start by examining how foreign owned firms differ from the average firm in the same sector. In the sample, foreign firms are approximately twice the size of a domestic firms, but do not have more average hours of work per worker. The average years of schooling is almost two years higher than in a domestic firm of the same sector. The workforce in foreign owned firms is, on average, less than 1 year younger and the share of females is 3 percentage points higher than the workforce in domestic firms.
An alternative way to test for selection is to estimate a probit model for the event of being acquired. This approach has been used by Clerides et al. (1998) to document selection of exporting firms. Using this approach the sign of the effect of each variable on the probability of being acquired would be the same as the sign in Table 1 (not reported). 20 The estimated model is identical to the one reported in Eq. (3.1) where the dependent variable is the logarithm of real hourly wages per employee in firm j. The only difference is that I also control for Xjt, a vector of workforce characteristics including average years of schooling, average age, average job tenure and share of females in the workforce. A similar model is used to estimate the foreign wage premium for each education group. The only difference is that average wages and worker characteristics are relative the each education group.
19
82
Table 1 Differences between future foreign acquisitions and foreign owned firms relatively to domestic firms Age workforce (4) (5) (6) (7) (8) Tenure workforce Females/ Average total workers hourly wage Low educated High educated Price-cost hourly wage hourly wage margin (9) (10)
Employment Hours work/ Years of total workers schooling (2) (3)
(1)
Future foreign acquisitions
Foreign firms
0.85 [0.062] 1.05 [0.030] Yes Yes No 0.75 647,984 0.06 647,984 0.40 647,984 0.11 647,984 0.06 647,984 0.47 0.00 0.00 0.00 0.25 0.00 647,984 0.43 0.00
0.0000 [0.004] 0.0003 [0.002] Yes Yes No
1.65 [0.097] 1.93 [0.043] Yes Yes No
1.37 [0.253] 0.66 [0.120] Yes Yes No 0.02 [0.010] 0.03 [0.004] Yes Yes No 0.32 [0.014] 0.40 [0.006] Yes Yes Yes
1.07 [0.220] 0.01 [0.108] Yes Yes No
0.24 [0.013] 0.32 [0.006] Yes Yes Yes
0.40 [0.017] 0.48 [0.008] Yes Yes Yes 0.00 616,491 0.39 256,937 0.39
0.001 [0.001] 0.000 [0.000] Yes Yes No 0.04 579,288 0.17
Sector dummies included? Time dummies included? Av. schooling, age, tenure, sh. Females included? F test: Future foreign acquisitions = 0.00 Foreign firms (P value) N 647,984 R squared 0.22
The numbers in column (1) represent the coefficients of a regression of log employment on a dummy variable if a firm will become acquired by foreigners and on a dummy variable if the firm is foreign owned. Controls for two digit sector dummies and year dummies included ( (3.1)in text). Standard errors are clustered at the firm level. N is the sample size in each regression. P values are for the test that the coefficient on the future foreign acquisitions is equal to that of foreign owned firms. The coefficients for employment, hours of work, average hourly wages (average, low and high) are be interpreted as percentage differences (log dependent variable). E.g., 0.85 in column (1) implies that future foreign acquisitions are 85 percentage points larger than the average domestic firm in the years prior to the acquisition. 1.6 in column (3) implies that the average years of schooling in firms acquired by foreigners is 1.6 years higher than in the average domestic firm in the sector.
83
These differences in the workforce characteristics, if not accounted for, translate into differences in hourly wages across ownership types.21 Columns (8) and (9) of Table 1 report the foreign wage premium for each education group after controlling for differences across firms in average schooling, age, tenure and gender for each group respectively: it is 32% for the low education workers and 48% for the high education workers.22 This finding is consistent with Lipsey and Sjoholm's (2004) result of a foreign wage premium which is increasing with workers' skill. The estimates for reported in the table show that firms that in the future will become acquired by foreigners are already very different from the average domestic firm in the sector in the years prior to the acquisition. Future foreign acquisitions are 85% larger than domestic firms but do not differ in the average hours of work per employee. Two years before the acquisition they already have a more educated workforce since their average number of years of schooling is 1.6 years above the one in existing domestic firms. The workforce in these firms is also younger, less tenured and with more females than the workforce in domestic firms. Controlling for the differences in these human capital characteristics, I find evidence of selection of foreign investment into higher wage firms. Firms that will become acquired pay 32% higher average hourly wages than domestic firms. This premium holds for each education group since hourly wages are 24% and 40% higher for low and high educated workers, respectively. These wage premiums are below the average wage differences between foreign and domestic owned firms (32% for low educated and 48% for the high educated). The price cost margin is 0.1 percentage points lower than in the average domestic firm before the acquisition. However, one should be cautious when interpreting these findings since I cannot control for differences across firms in input prices. Finally, firms that will become acquired by foreigners are not statistically identical to foreign owned firms in the years prior to the acquisition, even though the differences are small in magnitude (the large size of the sample leads to very small standard errors). In sum, the results presented in this section, together with a large empirical literature, confirm that there are substantial differences between foreign and domestic firms. Foreign firms have a younger and more educated workforce than domestic firms. They also pay higher average wages, controlling for the quality of the workforce. Not accounting for differences in worker's quality leads us to overestimate the foreign wage premium by approximately one third. The findings also show that foreigners do not randomly pick domestic firms to be acquired. Several years prior to the acquisition, these firms outperform the typical domestic firm in the sector in terms of size, worker quality and hourly average wages. Moreover, I find evidence that selection on unobservable firm characteristics is also important. Controlling for observed worker quality, in the years prior to the foreign acquisition, the average hourly wages in acquired firms are much higher than in the average domestic firm. 4. Does foreign ownership improve labor market outcomes? Since foreigners select the best domestic firms to be acquired and I do not observe all the variables foreigners select on, cross sectional estimates of the causal effects of foreign ownership on labor market outcomes are biased. Take, for example, the case of the foreign wage premium and assume that foreigners buy firms with more educated workers. If unobserved worker ability were independent of foreign acquisitions and positively correlated with schooling, then least
In appendix A, available at http://econ.worldbank.org/staff/ralmeida, I show how the foreign hourly wage premium varies once I control for differences across firms in several workforce characteristics and when I allow the foreign premium to vary across sectors. Throughout the paper I assume a nationally competitive labor market so that workers move freely across regions. Relaxing this assumption leads to foreign wage premiums that are very similar. 22 The number of observations for the hourly wage of high educated workers is smaller than in the other regressions since 60% of the firms do not report having high educated workers.
21
84
squares estimates for the foreign wage premium may be downward biased (if foreign acquisitions and worker's schooling are positively correlated). Alternatively, unobservable technological characteristics of the firm or unobservable organizational capital which are positively correlated with the foreign acquisition may imply that the cross sectional estimates are biased upwards. Considering the motives for firm acquisition reviewed in the previous section, one would expect to observe at least some changes in the acquired firm due to technological or organizational restructuring of the firm. For example, plant closings can lead to large employment reductions. Technological change can lead to an increased use of more skilled workers and changes in the human capital, or it can lead to more on-the-job training with possibly different effects in the wages of different workers. All this depends on the characteristics of labor markets, on the importance of foreign acquisitions in the economy and on the directions of changes in the labor demand of acquired firms. For example, productivity increases can lead to shifts in the labor demand and, as a result, wages and employment may change. If foreign acquisitions correspond to a small part of employment in the sector and if labor markets are competitive, wages will not increase and there will be only increases in the employment of the firm. Alternatively, wages may increase if foreign acquisitions have a strong impact on aggregate labor demand or if rigidities in the labor market imply that firms face an upward sloping labor supply curve (e.g., due to monopsony). In this section, I present reduced form estimates of the total effect of foreign acquisitions on a variety of labor market variables. The relationship I estimate is not a structural labor demand equation (e.g., as in Hammermesh, 1993). However, it can be derived from it in some cases.23 In order to obtain unbiased estimates of the effect of foreign ownership on different labor market outcomes, I need to take into account the possibility that foreign acquisitions are endogenous, i.e., they might occur as a reaction to some kind of unobserved firm specific shock. The findings in the previous section showed that foreign acquisitions are not random and that foreigners acquire firms that systematically differ from the average domestic firm in the sector. If foreign acquisitions are driven mainly by (unobserved) permanent differences across firms, this endogeneity problem can potentially be addressed by examining the same firm over time and to compare the outcomes of interest in the period before and after the foreign acquisition. Econometrically, this would amount to estimating the following model with firm fixed effects: yjt kj dFjt
T X t 1
gt Dt jt
4:1
where all notation is as above and j is a firm fixed effect. The error term, jt, is assumed to be uncorrelated across firms and time. The estimation of this equation requires longitudinal data for each firm. The identification of is done exclusively by firms that change the nationality of their ownership, i.e., that switch from a domestic to a foreign ownership. Estimating Eq. (4.1) only for the sample of foreign acquisitions could yield biased estimates of . In particular, it would be identical to comparing outcome yj in the pre and post-acquisition period without taking into account other aggregate shocks contemporaneous to the acquisition (Meyer, 1994). Therefore, I include in the analysis a control group of domestic firms that identify the aggregate shocks in the economy captured by t . Domestic firms are a good control group for the
In particular, under the assumption that labor supply curves faced by the firms are not affected by the nationality of their ownership, the reduced form effects are all due to changes in labor demand. Otherwise, they are the combination of labor supply and labor demand changes. Lichtenberg and Seigel (1992), Brown and Medoff (1988) and McGuckin and Nguyen (2001) use a specification (for employment) that is closer to this approach.
23
85
sample of acquired firms if both types of firm are growing at similar rates in the years prior to the acquisition (and hence are likely to be affected similarly by aggregate shocks). In other words, this approach would be invalid if firms which are acquired grow at systematically different rates than the average domestic firm. This evidence would suggest that foreign acquisitions would be correlated with some firm unobserved shock in growth rates and hence the beforeafter comparison would be invalid. Even though there is no formal way to test this claim, I present some evidence that this problem is not very important for several outcomes of interest. To do this I estimate a model similar to Eq. (3.1) but having the growth rate of the outcome yjt as dependent variable: lnyjt lnyjt1 bAj dFj
S X s 1
gs Ds
T X t 1
gt Dt ejt
4:2
where all the notation is as in Eq. (3.1). This equation is estimated for all firms acquired by foreigners and for existing domestic and foreign owned firms. In this specification, measures the percentage point difference in the growth rate between firms that will be acquired by foreigners over the following years and domestic firms in the same sector. Analogously, measures the percentage point difference in the growth rate in existent foreign owned firms and domestic firms in the same sector. The results of estimating Eq. (4.2) for several outcomes with least squares are presented in columns (1) to (10) of Table 2. The evidence suggests that acquired firms are growing in the preacquisition period at rates that are not statistically different from the rates of growth in domestic firms in the same sector. The only exception is for the size of the firm, average tenure of the workers and the price cost margin. In these cases, there is evidence that firms acquired by foreigners were growing at higher rates prior to the acquisition than domestic firms (7 percentage points above for employment, 0.03 years above for average tenure of the workforce and 1 percentage point above for the price-cost margin). Existing foreign firms tend to grow at a different rate than the rates acquired firms are growing in the pre-acquisition period, even though the quantitative differences are small.24 I interpret these findings as suggestive that for most of the outcomes of interest foreign acquisitions are a reaction to a permanent firm specific shock (and hence a first difference approach would yield unbiased estimates for the effect of foreign acquisitions) but that foreigners tend to acquire firms whose employment, tenure and price cost margin is growing above the average for the sector. The results of estimating Eq. (4.1) for foreign acquisitions and for existing domestic firms are reported in Table 3.25 Panel A reports the results when the sample includes firms in manufacturing and non-manufacturing sectors. The findings suggest that following an acquisition there is an increase of 14.5% in the total size of the firm. I do not find evidence of substantial restructuring of the workforce in the acquired firm. There are no significant changes in the average education of the workforce following the acquisition, measured by the average years of schooling. I also do not find evidence that other demographic characteristics of the workforce (like age, tenure or gender) change substantially.26 All this evidence strongly suggests that the profile of the average worker in the firm is
24 For most of the outcomes, I accept the null hypothesis that = . The exceptions include total employment, average hourly wage of high educated workers and the pricecost margin. Because firms could anticipate the effects of the acquisition, I tested the robustness of the results to including only growth rates two and three years before the acquisition for those firms that become acquired. The results are very similar (not reported). 25 Since the differences in growth rates between acquired and foreign owned firms are quantitatively small (Table 2), I also test the robustness of the estimates in Table 3 to the inclusion of foreign owned firms as a control group. I do not find significant differences in the outcomes relatively to those reported in Table 3 (not reported). 26 The coefficients for average age, tenure on-the-job and share of females in the workforce are statistically strong though quantitatively small.
86
Table 2 Differences in the growth of outcomes between future foreign acquisitions and foreign owned firms relatively to domestic firms Years of schooling (3) (4) (5) (6) (7) (8) Age workforce Tenure workforce Females / total workers Average hourly wage Low educated hourly wage High educated hourly wage (9) Price-cost margin (10)
Employment
Hours work per employee
(1)
(2)
Future foreign acquisitions
Foreign firms
0.07 [0.016] 0.02 [0.005] Yes Yes No 0.68 0.11 0.07 0.52 0.35 0.17
0.000 [0.003] 0.01 [0.001] Yes Yes No
0.005 [0.005] 0.003 [0.001] Yes Yes No 0.03 [0.019] 0.002 [0.005] Yes Yes No 0.002 [0.017] 0.01 [0.003] Yes Yes No 0.005 [0.006] 0.004 [0.002] Yes Yes Yes
0.001 [0.003] 0.01 [0.001] Yes Yes No
0.001 [0.006] 0.01 [0.001] Yes Yes Yes
0.01 [0.010] 0.01 [0.002] Yes Yes Yes 0.04
0.01 [0.005] 0.002 [0.001] Yes Yes No 0.05
Sector dummies included? Time dummies included? Growth av. schooling, age, tenure and sh. females F test: Future foreign acquisitions = Foreign firms (P value) N R squared 384,802 0.00 384,802 0.00 384,802 0.01 384,802 0.00 384,802 0.02
0.00
0.11
384,802 0.01
384,802 0.01
363,223 0.04
139,094 0.08
344,439 0.00
The numbers in column (1) represent the coefficients of a regression of employment growth on a dummy variable if a firm will become acquired by foreigners and on a dummy variable if the firm is foreign owned. Controls for two digit sector dummies and year dummies are included (Eq. (4.2) in the text). In columns (7) to (9) controls for the growth in average schooling, age, tenure and females are included. Standard errors are clustered at the firm level. N is the sample size in each regression. P values are for the test that the coefficient on the future foreign acquisitions is equal to that of foreign owned firms. The coefficients for employment, hours of work and average hourly wages (average, low and high) are interpreted as percentage differences (log dependent variable). E.g., 0.07 in column (1) implies that future foreign acquisitions grow 7 percentage points above the average domestic firm in the sector two years before the acquisition takes place. 0.005 in column (3) implies that the growth in the average years of schooling in firms acquired by foreigners is 0.05 percentage points lower than in the average domestic firm in the sector.
87
not significantly affected. Moreover, after controlling for the human capital characteristics of the workforce, average hourly wages remain constant following the acquisition. This result is robust within education groups. Column (10) reports an increase in the price-cost margin by 0.4 percent-age points for the whole sample, suggesting that productivity increases following the foreign acquisition. To keep the comparability with papers in the literature, panel B reports the results of estimating Eq. (4.1) only for manufacturing firms. I find evidence that employment increases by 10.1% and that there are very small changes in the human capital composition of the average worker in the firm.27 Average hourly wages increase by 1.7% following the acquisition (reflecting an increase in average wages and of hours of work by 2.6% and 1.2%, respectively). Hourly wages for the high educated workers increase by 3.5% following the acquisition and average hourly wages for the low educated workers remain unchanged. In sum, my findings show that following the foreign acquisition there are no substantial changes in the human capital of the workforce or on the average wages of acquired firms. There is also evidence that firms increase total employment following the acquisition, even though this estimate is likely to be upward biased since acquired firms were growing at faster rates in the pre-acquistion period. For manufacturing firms, there is evidence of an increase in average hourly wages for the high educated workers, after controlling for changes in the quality of the workforce. I interpret these findings as consistent with a model where foreign firms are relatively risk averse and enter a new market where learning and adjustment costs are high. Foreigners are more likely to buy the best performers since they are already similar to their own characteristics. Otherwise, they would face very high costs of adapting the technology, the workforce or by gaining experience in the host country. Furthermore, acquired firms only account for a very small share of total employment. As a result, changes in the labor demand of these firms will not affect the equilibrium wages in the economy and most of the observed changes will be on the employment margin. Foreign investment may lead to an increase in the productivity of the labor force and to a rise in labor demand but, if wages are fixed, employment will increase. In the manufacturing sector foreign acquisition leads to an increase in wages of the more educated workers. This could happen if foreign firms provide more training to these workers than domestic firms, making them different from other educated workers in the rest of the economy, and therefore paying them higher wages. Alternatively, if there are rigidities in the labor market for educated workers, it is also possible that foreign acquisitions lead to higher wages due to an increase in labor demand, even in the absence of training. The most credible solution to the endogeneity problem of foreign acquisitions would be to find an external instrument to the foreign acquisition. This instrument had to be uncorrelated with all the outcomes of interest except through the acquisition itself but such an instrument does not exist in the data. Alternatively, I could estimate a VAR model and test whether foreign ownership Granger causes firm characteristics. Unfortunately, the panel dimension of the data is not sufficiently long to allow me to estimate a specification without serial correlation in the error terms.28 5. Robustness tests One possible concern is that the findings in Table 3 are driven by the 10% threshold imposed to define foreign ownership. Mansfield and Romeo (1980) argue that the transfer of foreign technology
The only quantitatively important change is associated with a decrease in the share of females in 1.4 percentage points. 28 Details on the results using this approach are available in appendix B at http://econ.worldbank.org/staff/ralmeida. There, I also discuss other robustness tests for the main findings in the paper.
27
88
Table 3 Fixed effects estimates of the effects of foreign acquisitions using as control group domestic firms Years of schooling (3) (4) (5) (6) (7) (8) (9) Age workforce Tenure workforce Females / total workers Hourly wage Low educated hourly wage High educated hourly wage Price-cost margin (10)
Employment
Hours work per employee
(1)
(2)
A. All Sample 0.008 [0.060] Yes No 0.001 [0.010] Yes Yes 608,551 0.86 638,896 0.90 638,896 0.88 638,896 0.91 638,896 0.93 638,896 0.85 0.76 [0.157] Yes No 0.32 [0.108] Yes No 0.004 [0.006] Yes No 0.004 [0.009] Yes Yes 0.008 [0.016] Yes Yes 248,671 0.85 0.004 [0.001] Yes No 571,605 0.72
Foreign ownership
Time dummies included? Av. schooling, age, tenure, sh. females N R squared
0.145 [0.035] Yes No
0.017 [0.004] Yes No
638,896 0.94
638,896 0.68
B. Manufacturing 0.022 [0.050] Yes No 133,985 0.85 133,985 0.91 133,985 0.93 133,985 0.97 0.773 [0.171] Yes No 0.465 [0.106] Yes No 0.014 [0.005] Yes No 0.017 [0.009] Yes Yes 133,985 0.85 0.009 [0.008] Yes Yes 132,775 0.87 0.035 [0.017] Yes Yes 49,965 0.85 0.005 [0.001] Yes No 122,987 0.72
Foreign ownership
Time dummies included? Av. schooling, age, tenure, sh. females N R squared
0.101 [0.018] Yes No
0.012 [0.007] Yes No
133,985 0.96
133,985 0.63
The numbers in column (1) represent the coefficients of a regression of log employment at the firm level on a dummy variable that equals to one if the firm is foreign owned, controlling for year and firm time invariant effects (Eq. (4.1) in the text). In columns (7) to (9) the specification includes as explanatory variable average years of schooling, age, tenure and share females in the workforce. Standard errors are clustered at the firm level. N is the sample size in each regression. The coefficients for employment, hours of work and average hourly wages (average, low and high) are interpreted as percentage differences (log dependent variable). E.g., 0.14 in column (1) implies that following a foreign acquisition the employment in acquired firms is on average 14 percentage points higher than the average employment in a domestic firm. Panel A includes the estimates for all the firms in the sample and Panel B includes only manufacturing firms.
89
is greater in fully owned foreign firms, specially when competitive advantage of the foreign firm is based on intangible assets. Larger technological transfers from foreign firms increase productivity and can possibly increase the use of more educated workers or average wages in the acquired firm. In this case, the choice of the 10% threshold could be biasing the results against the finding on any effect on wages and workforce composition. I reestimate Eq. (4.1) considering foreign firms only those firms with full foreign ownership. Given that most of the firms in the sample have majority participations the results in Table 3 do not change significantly (results are available upon request). A different concern, specific to the effect of foreign ownership on hourly wages, relates to possibility that the foreign acquisition changes the composition of the workforce in some unobservable workforce characteristics (e.g. worker's ability). In Section 2 I showed that foreign and domestic firms differ significantly with respect to the human capital profile of the workforce. This heterogeneity is likely to be much more important when comparing different firms rather than when comparing the same firms over time (which is my approach in Eq. (4.1)). In fact, if the workforce were to be constant over time, the unobserved ability of the workforce would be perfectly captured by a firm time invariant effect. However, with a changing workforce this needs not be the case. Moreover, depending on the correlation between ability and the foreign acquisition the sign of the bias in Table 3 is unclear. To address this concern, I exploit the fact that in the data firms and workers are matched in each point in time and that both can be traced over time.29 The best way to account for the bias would be to compute average wages for those workers that have remained in the firm throughout the entire period the firm is observed.30 This approach ensures that the set of workers remains constant over time and one can fully isolate the effect of the foreign acquisition from the effect of omitted ability. The problem with this approach is that there are very few firms with at least one worker that complies with this criteria, mostly because of large measurement error in worker identifiers. In particular, the total number of acquired firms for which there is information on wages is reduced from a total of 1381 to approximately 360 firms. An alternative to this approach which simultaneously mitigates the omitted ability problem and maximizes the information available over time, is to keep the composition of the workforce fixed only before and after the acquisition. To do this I compute average wages for a moving window of workers across three consecutive periods. For existing firms, I identify the group of workers that are currently employed in the firm and that were employed in the previous year as well as in the following year. For new firms I identify the group of workers that are currently employed in the firm and that will remain employed in the following year. For example, consider a domestic firm that exists for the period 19931997 and that is acquired in 1995. The (moving) average wages obtained for this firm in 1994 and 1995 (used to identify the effect of the foreign acquisition) use information for the same set of workers except for those workers that were in the firm in 1993 but are no longer there in 1996 and viceversa. Using this criteria, I am able to compute the average wages for 1, 346 acquired firms in the sample, which is approximately 95% of the total number of acquisitions in the data.31
The Portuguese data is a matched employer-employee data set. For a review of the litertaure using matched employer employee see Abowd and Kramarz (1999). 30 The worker identifier in this data is the worker's social security number. This identifier is much less reliable than the firm identifier because the Ministry of Labor, specially in the first years of the sample, did not check the validity of the codes provided. I have excluded workers without a unique identifier within a firm when tracing the workers over time. This criteria could still present some problems (e.g., workers with non-valid social security numbers might remain in the sample as long as they only show up once within each firm) but it maximizes the number of workers for which there is information. The share of workers that are not unique according to this definition is between 12% (1991) and 7% (1998). 31 Using this criterion, between 42% (1998) and 32% (1992) of the workers in the sample in each period are used to compute average wages.
29
90
Table 4 Estimates of the effects of foreign acquisitions on wages only for workers that remain in the firm (stayers) Estimation method: All Low Educated OLS (1) A. All sample Foreign ownership Sector dummies included? Time dummies included? Av. schooling, age, tenure, sh. females N R squared B. Manufacturing Foreign ownership Sector dummies included? Time dummies included? Av. schooling, age, tenure, sh. females N R squared 0.23 [0.018] Yes Yes Yes 123,715 0.42 0.18 [0.017] Yes Yes Yes 122,031 0.42 0.32 [0.026] Yes Yes Yes 37,633 0.36 0.028 [0.013] No Yes Yes 123,715 0.85 0.022 [0.012] No Yes Ye 122,031 0.87 0.043 [0.023] No Yes Yes 37,633 0.87 0.36 [0.014] Yes Yes Yes 550,288 0.40 0.27 [0.013] Yes Yes Yes 514,406 0.37 0.43 [0.017] Yes Yes Yes 182,036 0.36 0.009 [0.009] No Yes Yes 550,288 0.86 0.015 [0.009] No Yes Yes 514,406 0.86 0.025 [0.015] No Yes Yes 182,036 0.88 (2) (3) (4) High Educated All Low Educated Fixed effects (5) (6) High Educated
Table reports the foreign wage premium of all the workers that remain in the same firm throughout 3 consecutive periods. The table reports the least squares estimates of Eq. (4.1) in the text ignoring firm fixed effects in columns (1) to (3) and including them in columns (4) to (6). Columns (1) and (4) report the results for all the workers, columns (2) and (5) for all the low educated workers and columns (3) and (6) for the high educated workers. Standard errors are clustered at the firm level. N is the sample size in each regression. Panel A reports the results for the whole sample and Panel B includes only manufacturing firms.
Table 4 report the results of estimating Eq. (4.1) for the average wages for this group of workers. Columns (1) to (3) report the results with least squares ignoring the firm fixed effects (cross section) and columns (4) to (6) report the results with firm fixed effects. As before, the least squares estimates of the effect of foreign ownership are always above the firm fixed effects estimates. This finding suggests that the cherry picking of the foreign investors of domestic firms is robust to keeping the workforce composition constant. The main difference of keeping the composition fixed (relatively to Table 3) is that there are slightly larger effects of foreign ownership for average wages in manufacturing (panel B). In particular, average wages following the acquisition increase by 2.2% for the low educated workers and 4.3% for the high educated workers.32 However, the magnitude of these effects are very small when compared to the least square estimates. These results suggest that the firm expands following the foreign acquisition but that the average quality of the new entrant is slightly below the quality of the average worker. This is consistent with foreigners buying the domestic firms with the best workers and hiring the immediately next to the best when they are expanding in the years following the acquisition (Table 3 shows that average schooling, age and tenure decrease slightly following the acquisition).
I thank the editor for one possible explanation for why there could be a higher increase in average wages for the high educated workers. If following the acquisition total rents increase and the high educated workers have higher bargaining power (due to the accumulation of firm specific skills), then their wage adjustment could be higher. Unfortunately my data does not allow me to test this hypothesis.
32
91
One concern that could still remain with these findings relates to the whether they could be attributed to the foreign ownership or if they can attributed to the acquisition itself, independently of the nationality. To the extent that domestic firms also cherry pick the acquired firms, my results for the selection and for the evolution following the acquisition would be driven by the acquisition itself and not by the foreign acquisitions. As discussed in Section 4 it is difficult to extract strong predictions about the post acquisition firm performance from the extensive theoretical literature on mergers and acquisitions. Unfortunately, in the Portuguese data I cannot identify domestic acquisitions of domestic firms even though I can identify the group of domestic acquisitions of foreign owned firms.33 To address this concern, I estimate the following model: yjt kj bFAjt hDAjt
T X t 1
gt Dt jt
5:1
where FAjt is a dummy variable that assumes the value one in the periods following a foreign acquisition of a domestic firm, DAjt is a dummy variable that assumes the value one in the periods following a domestic acquisition of a foreign firm and Dt are year dummies.34 The results are reported in Table 5. The findings suggest that the effects on employment are different for foreign and domestic acquisitions. While employment in the firm increases between 10% and 15% following a foreign acquisition, in the domestic acquisitions, there is a 5% reduction in total employment (which fails to be statistically significant but is in line with Conyon et al., 2002b). Moreover, there is also evidence that the effects on average wages are common to the two types of ownership changes. The estimates in columns (7) to (9) show that, after keeping the composition of the workforce constant, average wages in acquired firms increase by a very similar magnitude, independently of the nationality of the acquiring firm (even though I never reject that the increase in wages in slightly higher in the foreign acquisitions). Perhaps surprisingly, there is evidence that average years of schooling decrease following a domestic acquisition. This pattern is puzzling even though the magnitude of the effect is very small (average schooling in the sample is 6.3 years and the standard deviation of 2.3 years). A further investigation shows that this effect is not driven by the evolution in the years of schooling in the existent domestic firms. One possible explanation for this can be that the new workers in firms that become acquired by domestics are less educated than the existing workers in these firms. Actually, the average years of schooling of an European worker is higher than for a Portuguese worker. Since approximately 75% of the Portuguese FDI in this period came from the E.U., these changes could be simply reflecting the exit of some foreign workers following the domestic acquisition. On the contrary, when a domestic firm is acquired by a foreign multinational there is no evidence that foreigners change significantly the average schooling of the firm, perhaps because this is already very close to the one in fully foreign firms. Finally, another concern with the estimates in Table 3 is that they are capturing the effects of other firm practices that change simultaneously with the nationality of the ownership. Since foreign activities tend to take place in bundles, those firms that are acquired by foreigners are likely to simultaneously become exporters or importers of some of their intermediate inputs (e.g. Tybout,
33 Foreign acquisitions of domestic firms (or foreign acquisitions of domestic firms) are identified when the share of foreign ownership changes from being less or equal to 10% to being more than 10% (and vice versa). Given that domestic acquisitions do not necessarily translate into changes in the share private domestic capital, they cannot be identified in the data set. 34 A similar analysis than the one reported in Table 1 (not reported) suggests that foreign owned firms that will be acquired by domestics are smaller in size, pay lower average hourly wages and have a less educated workforce than the average foreign firm in the same sector. Still, these firms always outperform the typical domestic firm in the sector. Again, there are no significant differences in the growth of the outcomes of interest in the pre acquisition period.
92
Table 5 Estimates of the effects of foreign and domestic acquisitions using as control group domestic firms Employment Years of schooling All workers (1) (2) (3) (4) (5) (6) (7) (8) Stayers (9) Price-cost margin Hourly wage Low educated High educated Hourly hourly wage hourly wage wage Low educated High educated hourly wage hourly wage
A. All sample 0.007 [0.060] 0.175 [0.062] Yes No 0.018 0.004 [0.001] 0.000 [0.001] Yes No 0.001 573,938 0.72 641,475 610,889 0.86 0.86 250,914 0.86 552,693 0.86 516,469 0.86 0.001 [0.010] 0.002 [0.010] Yes Yes 0.98 0.008 [0.016] 0.014 [0.019] Yes Yes 0.64 0.009 [0.009] 0.008 [0.011] Yes Yes 0.484 0.015 [0.009] 0.016 [0.009] Yes Yes 0.065 641,475 0.90 0.004 [0.009] 0.003 [0.011] Yes Yes 0.88 0.025 [0.015] 0.009 [0.016] Yes Yes 0.232 183,984 0.88
0.145 [0.035] Domestic acquisitions 0.016 [0.037] Time dummies included? Yes Av. schooling, age, tenure, sh. Females included? No F test: foreign acquisitions = domestic 0.000 acquisitions (P value) N 641,475 R squared 0.94
Foreign acquisitions
B. Manufacturing 0.021 [0.104] 0.173 [0.088] Yes No 0.14 0.005 [0.002] 0.000 [0.002] Yes No 0.03 123,679 0.72 134,736 133,518 0.85 0.87 0.017 [0.013] 0.010 [0.015] Yes Yes 0.36 0.009 [0.012] 0.007 [0.012] Yes Yes 0.67 134,736 0.85 0.034 [0.025] 0.000 [0.036] Yes Yes 0.38 50,621 0.85 0.028 [0.013] 0.015 [0.012] Yes Yes 0.039 124,437 0.85 0.022 [0.012] 0.017 [0.011] Yes Yes 0.062 122,736 0.87 0.043 [0.023] 0.040 [0.029] Yes Yes 0.071 38,207 0.87
0.102 [0.066] Domestic acquisitions 0.050 [0.042] Time dummies included? Yes Av. schooling, age, tenure, sh. Females included? No F test: foreign acquisitions = domestic 0.15 acquisitions (P value) N 134,736 R squared 0.96
Foreign acquisitions
The numbers in column (1) represent the coefficients of a regression of log employment at the firm level on a dummy variable that equals to one if the firm is acquired by foreigners, on a dummy that equals one if the firms is acquired by domestics and controls for year and firm time invariant effects (Eq. (5.1) in the text). Columns (1) to (6) report the results for all the workers in the sample while columns (7) to (9) report the results for the stayers. All the wage equations include as explanatory variables average years of schooling, age, tenure and share females in the workforce. Standard errors are clustered at the firm level. N is the sample size in each regression. The coefficients for employment and average hourly wages (average, low and high) are interpreted as percentage differences (log dependent variable). Standard errors are clustered at the firm level. Panel A includes all the sectors in the sample and Panel B includes only manufacturing sectors.
93
2000; Bernard et al., 2005). Since the Portuguese data set does not collect any information on exports it is not possible to check empirically how controlling for these activities would affect the findings. However, given the findings in the literature on exporting, my results for the post acquisition performance should not be too much affected by the exporting activity even if it takes place simultaneously with the foreign acquisition.35 For example, Clerides et al. (1998) and Bernard and Jensen (1999) find evidence that the best performers self select into exporting activities and no evidence of significant changes in productivity or wages after the firm becomes an exporter. The evidence of the effects of importing intermediate inputs are much less studied but several papers find evidence of a positive correlation between importing intermediate inputs and firm performance (e.g. Tybout and Westbrook, 1995). Assuming that the effects on productivity are quantitatively important and if they translate into high wage growth, my estimates for the changes in average wages could be overestimated. Since I find no or small effects in average hourly wages, this would imply that foreign acquisition are associated with a decrease in productivity and wages. This finding would be consistent with a short run decrease in the firm's productivity following a foreign acquisition, which could be caused by assimilation problems when entering into a new market (e.g. Harris and Robinson, 2002; Ravenscraft and Scherer, 1987). 6. Conclusion Foreign firms have a more educated workforce and pay higher wages than domestic firms even after controlling for worker quality, at a given moment in time. This does not imply that foreign ownership improves the labor market outcomes of the workers since foreign investment may be guided by unobservable firm and worker characteristics correlated with schooling or wages. Using a matched employeremployee data set of Portuguese firms during the nineties, I illustrate the importance of selection of foreign investment to high wage and high human capital firms and to isolate the effect of foreign ownership on several labor market outcomes. Existing empirical evidence for European countries is scarce and, apart from evidence for the UK, not much is known about the impact of foreign acquisitions on the labor markets outcomes. Portugal is an interesting case, as in the late 1980s and 90s there was a permissive legal framework for the operation of foreign firms that translated into generous amounts of FDI. Moreover, Portugal is a developed country although at the tail of the income distribution within the European Union. Therefore, the findings in the paper may be particularly important to understand the effects of foreign direct investment in the labor markets of the transition economies as they will soon join the European Union. My findings suggest that most of the cross sectional differences between foreign and domestic firms are driven by foreigners cherry picking domestic firms to be acquired. Acquired firms have a more educated workforce, and pay higher wages for a given workforce quality. Moreover, these firms are already much more similar to the group of existing foreign firms and, following the foreign acquisition, there are no significant changes in the human capital of the workforce. There is evidence that the size of the firm increases and that the hourly wages for those workers that remain in the firm increases slightly following the acquisition by foreigners. Nevertheless, the increase in average wages are modest (between 2% for the low educated and 4% for the high educated workers) and well below the cross sectional estimates of the foreign wage premium. Suggestive evidence shows that it might not be specific to foreign acquisitions since wages go up by similar magnitudes following domestic acquisitions.
35 Bernard and Jensen (1999) find that employment growth is higher for exporters. Therefore, the estimated increase in employment following the acquisition is likely to be biased upwards for those firms that export (in addition to the positive bias caused by selection of foreign investment into high growth firms).
94
Appendix A.
Table A1
Number of acquisitions Firm level sector, region, assets, size, age N.A. 25% skilled 17% unskilled Worker level Cross section Independent variables Foreign wage premium Panel data 14% skilled 9.3% unskilled
A summary of the literature on foreign wage premium using firm level data
Country / period
Data / sample coverage
Dependent variable
Panel of firms
Log average firm wage N.A. by skill group
Aitken, Harrison and Lipsey (1996) Venezuela, 19771989 sector N.A. 9.50%
Encusta Industrial. Manufacturing 0.40%
Girma, Greenaway Panel of firms. and Wakelin (1999) UK, 19911996 OneSource. Manufacturing N.A.
Log average firm wage N.A.
Panel of firms.
Log average firm wage 129 foreign acquisitions assets, sector average wages and 331 domestic acquisitions average regional-sector N.A. wage, average wage in complement skill group, capital sector, region, energy per worker, inputs per worker, size, public ownership sector, region, energy per worker, inputs per worker, size, public ownership sector, size, age, public ownership. Education, gender by skill group
3.30%
Conyon, Girma, Thompson and Wright (2002) UK, 19891994
OneSource. manufacturing 2%4.9% skilled 2% 2.8% unskilled 12% blue collor, 22% white collar
Girma and Gorg (2003)
Panel of firms.
Log average firm wage 346 foreign by skill group acquisitions
U.K. 19801994
ARD. Electronics and food
Lipsey and Sjoholm (2004)
Cross section of firms.
Log average firm wage by skill group
Indonesia, 1996
Census Manufacturing N.A. 28% blue collar 41% white collar 17% blue collar 33% white collar 31% low educated Education, gender, 46% high educated age, tenure, hours worked by education group 0.1% low educated 0.8% high educated
Lipsey and Sjoholm (2002)
Panel of firms.
Log average firm wage 1045 foreign by skill group acquisitions 1243 domestic acquisitions
Indonesia, 19751999
Census Manufacturing
Almeida (2004)
Panel of firms (MEED at firm level)
Log average firm wage 688 foreign by education group acquisitions 505 domestic acquisitions
Portugal, 19911998
Census Manufacturing and Nonmanufacturing
Note: Table reports a summary of the main papers that have studied the foreign wage premium using micro data. The first evidence that foreign firms were associated with higher wages came from studies using industry level data (see e.g. Feliciano and Lipsey, 1999; Aitken, Harrison and Lipsey, 1996). As data sets become available, foreign wage premiums have been estimated using plant or firm level information.
95
Appendix B. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.jinteco.2006.10.001. References
Abowd, J., Kramarz, F., 1999. The analysis of labor markets using matched employeremployee data. In: Ashenfelter, O., Card, D. (Eds.), Handbook of Labor Economics, vol. 3B. North-Holland, Amsterdam, pp. 26292710. Aitken, B., Harrison, A., Lipsey, R., 1996. Wages and foreign ownership: a comparative study of Mexico, Venezuela and the United States. Journal of International Economics 40, 345371. Almeida, R., 2004. The labor market effects of foreign-owned firms. World Bank Policy Research Working Paper, vol. 3300. Arnold, J., Javorcik, B., 2005. Gifted kids or pushy parents? Foreign acquisitions and plant performance in Indonesia. World Bank Policy Research Working Paper, vol. 3597. Bernard, A., Jensen, J., 1999. Exceptional exporter performance: cause, effect, or both? Journal of International Economics 47, 125. Bernard, A., Jensen, J., Schott, P., 2005. Importers, exporters, and multinationals: a portrait of firms in the U.S. that trade goods. National Bureau of Economic Research Working Paper, vol. 11404. Bover, O., Garcia-Perea, P., Portugal, P., 1998. A comparative study of the Portuguese and Spanish labour markets. Banco de Espana Working Paper, vol. 9807. Brown, C., Medoff, J., 1988. The impact of firm acquisition on labor. In: Auerbach, A. (Ed.), Corporate Take-overs: Causes and Consequences. University of Chicago Press, Chicago, pp. 925. Buckley, P., Casson, M., 1998. Analyzing foreign market entry strategies: extending the internalization approach. Journal of International Business Studies 29, 539562. Caves, R., 1996. Multinational Enterprise and Economic Analysis. Cambridge University Press, Cambridge. Clerides, S., Lach, S., Tybout, J., 1998. Is learning by exporting important? Micro-dynamic evidence From Colombia, Mexico and Morocco. Quarterly Journal of Economics 113, 903947. Conyon, M., Girma, S., Thompson, S., Wright, P., 2002a. The productivity and wage effects of foreign acquisition in the United Kingdom. Journal of Industrial Economics 50, 85102. Conyon, M., Girma, S., Thompson, S., Wright, P., 2002b. Merger activity and firm employment. European Economic Review 46, 3149. Feliciano, Z., Lipsey, R., 1999. Foreign ownership and wages in the United States. National Bureau of Economic Research Working Paper, vol. 6923. Girma, S., Grg, H., 2003a. Multinational's productivity advantage: scale or technology. mimeo. Nothingham University. Girma, S., Grg, H., 2003b. Evaluating the causal effects of foreign acquisition on domestic skilled and unskilled wages. IZA Discussion Paper, vol. 903. Girma, S., Greenaway, D., Wakelin, K., 1999. Wages, productivity and foreign ownership in UK manufacturing. University of Nottingham Research Paper, vol. 99/14. Hammermesh, D., 1993. Labor Demand. Princeton University Press, USA. Harris, R., Robinson, C., 2002. The effect of foreign acquisitions on total factor productivity: plant level evidence from UK manufacturing: 19871992. Review of Economic Studies 84, 562568. Hennart, J., Park, Y., 1993. Greenfield vs. acquisition: the strategy of Japanese investors in the United States. Management Science 39, 10541070. Lichtenberg, F., Seigel, D., 1987. Productivity and changes in ownership of manufacturing plants. Brookings Papers on Economic Activity 3, 643673. Lichtenberg, F., Seigel, D., 1990. The effects of leveraged buyouts on productivity and related aspects of firm behavior. Journal of Financial Economics 27, 165194. Lichtenberg, F., Seigel, D., 1992. Takeovers and corporate overhead. In: Lichtenberg, F. (Ed.), Corporate Takeovers and Productivity. MIT Press, Cambridge, Massachusetts, pp. 4567. Lipsey, R., Sjoholm, F., 2002. Foreign firms and Indonesian manufacturing wages: an analysis with panel data. National Bureau of Economic Research Working Paper, vol. 9417. Lipsey, R., Sjoholm, F., 2004. Foreign direct investment, education and wages in Indonesian manufacturing. Journal of Development Economics 73, 415422.
96
Matsusaka, J., 1993. Target profits and managerial discipline during the conglomerate merger wave. Journal of Industrial Economics 41, 179189. Mincer, J., 1974. Schooling, Experience and Earnings. National Bureau of Economic Research, New York. Mansfield, E., Romeo, A., 1980. Technology transfer to overseas subsidiaries by US-based firms. Quarterly Journal of Economics 95, 737750. Markusen, J., 1995. The boundaries of multinational enterprises and the theory of international trade. Journal of Economic Perspectives 9, 169189. McGuckin, R., Nguyen, S., 1995. On productivity and plant ownership change: new evidence from the longitudinal research database. Rand Journal of Economics 26, 257276. McGuckin, R., Nguyen, S., 2001. The impact of ownership changes on labor: a view from labor markets. International Journal of Industrial Organization 19, 739762. Meyer, B., 1994. Natural and quasi-experiments in economics. National Bureau of Economic Research Technical Working Paper, vol. 170. Navaretti, B., Venables, A., 2004. Multinational Firms in the World Economy. Princeton University Press, New Jersey. Ravenscraft, D., Scherer, F., 1987. Mergers, Sell-Offs and Economic Efficiency. Brookings Institution, Washington, DC. Pfaffermayr, M., 1999. Ownership advantages, foreign production and productivity: evidence from Austrian manufacturing firms. Review of Industrial Organization 15, 379396. Tybout, J., 2000. Manufacturing firms in developing countries: how well do they do and why? Journal of Economic Literature 38, 1144. Tybout, J., Westbrook, D., 1995. Trade liberalization and the dimensions of efficiency change in Mexican manufacturing industries. Journal of International Economics 39, 5378.

Almeida The Labor

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Almeida The Labor

Uploaded by

Copyright:

Available Formats

Journal of International Economics 72 (2007) 75 96 www.elsevier.

The labor market effects of foreign owned firms

R. Almeida / Journal of International Economics 72 (2007) 7596

R. Almeida / Journal of International Economics 72 (2007) 7596

R. Almeida / Journal of International Economics 72 (2007) 7596

R. Almeida / Journal of International Economics 72 (2007) 7596

R. Almeida / Journal of International Economics 72 (2007) 7596

R. Almeida / Journal of International Economics 72 (2007) 7596

Employment Hours work/ Years of total workers schooling (2) (3)

Future foreign acquisitions

0.0000 [0.004] 0.0003 [0.002] Yes Yes No

1.65 [0.097] 1.93 [0.043] Yes Yes No

1.07 [0.220] 0.01 [0.108] Yes Yes No

0.24 [0.013] 0.32 [0.006] Yes Yes Yes

0.001 [0.001] 0.000 [0.000] Yes Yes No 0.04 579,288 0.17

R. Almeida / Journal of International Economics 72 (2007) 7596

R. Almeida / Journal of International Economics 72 (2007) 7596

R. Almeida / Journal of International Economics 72 (2007) 7596

R. Almeida / Journal of International Economics 72 (2007) 7596

Hours work per employee

Future foreign acquisitions

0.000 [0.003] 0.01 [0.001] Yes Yes No

0.001 [0.003] 0.01 [0.001] Yes Yes No

0.001 [0.006] 0.01 [0.001] Yes Yes Yes

0.01 [0.010] 0.01 [0.002] Yes Yes Yes 0.04

0.01 [0.005] 0.002 [0.001] Yes Yes No 0.05

R. Almeida / Journal of International Economics 72 (2007) 7596

R. Almeida / Journal of International Economics 72 (2007) 7596

Hours work per employee

0.145 [0.035] Yes No

0.017 [0.004] Yes No

R. Almeida / Journal of International Economics 72 (2007) 7596

0.101 [0.018] Yes No

0.012 [0.007] Yes No

R. Almeida / Journal of International Economics 72 (2007) 7596

R. Almeida / Journal of International Economics 72 (2007) 7596

R. Almeida / Journal of International Economics 72 (2007) 7596

R. Almeida / Journal of International Economics 72 (2007) 7596

R. Almeida / Journal of International Economics 72 (2007) 7596

Data / sample coverage

Log average firm wage N.A. by skill group

Encusta Industrial. Manufacturing 0.40%

Log average firm wage N.A.

Conyon, Girma, Thompson and Wright (2002) UK, 19891994

Girma and Gorg (2003)

Log average firm wage 346 foreign by skill group acquisitions

ARD. Electronics and food

Lipsey and Sjoholm (2004)

Cross section of firms.

Log average firm wage by skill group

R. Almeida / Journal of International Economics 72 (2007) 7596

Lipsey and Sjoholm (2002)

Panel of firms (MEED at firm level)

Census Manufacturing and Nonmanufacturing

R. Almeida / Journal of International Economics 72 (2007) 7596

R. Almeida / Journal of International Economics 72 (2007) 7596

You might also like