Paying more to hire the best? Foreign firms, wages, and worker mobility.

Martins, Pedro S.

I. INTRODUCTION

As globalization evolves, there is greater interest in its labor-market implications. One dimension of this question concerns the role of foreign firms in terms of their remuneration of host-economy workers. While earlier cross-sectional evidence suggests that foreign firms offer more generous pay levels than their domestic counterparts (Aitken et al. 1996; Feenstra and Hanson 1997), some of these results have been questioned in recent research based on longitudinal worker-level data (Martins 2004; Heyman et al. 2007; Andrews et al. 2007). Moreover, firm-level studies based on foreign acquisitions of domestic firms find mixed results. (1)

The main problem in research about the foreign-ownership wage differential concerns unobserved heterogeneity across workers employed in either domestic or foreign firms. If firm affiliation is correlated with other factors that may affect wages but that are not controlled for, then estimates will be biased. While some research that aims to tackle this issue considers the case of acquisitions (when the same workers can be observed under the two types of firms), here we approach the unobserved heterogeneity challenge from an original perspective--worker mobility.

Specifically, we draw upon a longitudinal census of the Portuguese labor market in order to consider virtually all spells of interfirm worker mobility over a long period of time (1991-2000). Such mobility spells allow us to observe the same workers when employed by domestic and foreign firms, in order to disentangle the wage policy of the firm from the heterogeneity of the workforce. We also consider "selection" effects, namely the unexplained differences in the wages of workers who are to experience a movement to a different firm, before such movement occurs. These unexplained wage differences are likely to capture additional skills not measured in the data but that are observable by those workers' current employers such as ability, motivation, etc.

On the other hand, the "wage policy" effect, which is more directly related to the goal of this paper, concerns differences in remuneration experienced by workers who engage in interfirm mobility, as they move between firms. Such differences in remuneration practices across firms are predicted by noncompetitive models of the labor market, namely efficiency wages and search models. Moreover, the scope for such "wage policy" differences is also supported by abundant empirical evidence, including that of rent sharing, discrimination, cohort effects, and other evidence of firm heterogeneity in general (Abowd et al. 1999; Bartelsman and Doms 2000).

As far as we know, our paper is the first to conduct a systematic analysis of interfirm worker mobility drawing on census data. These population data are particularly important for our purpose as the analysis of even large samples would dramatically diminish one's ability to follow workers over time. Furthermore, we are the first to conduct such an extensive analysis in the context of the foreign-ownership wage differential literature. (2) Finally, our results may also be useful in terms of reconciling some contrasting evidence for different countries; and in terms of shedding light on the role of worker mobility as a channel of productivity spillovers from foreign to domestic firms (Fosfuri et al. 2001 ; Javorcik 2004).

Unlike earlier research based on foreign acquisitions, our paper finds very strong evidence of a sizeable, positive "wage policy" effect for foreign firms. However, "selection" effects are also present, but at a much smaller scale. These results are robust to a number of checks, including the consideration of the case of displaced workers and an analysis of the wage growth patterns of movers when in hiring firms.

The structure of the paper is as follows: Section II describes the data; Sections III and IV present the results and the robustness analysis; finally, Section V concludes.

II. DATA

This paper draws on a particularly rich annual census of all firms in Portugal that employ at least one worker in any year--Quadros de Pessoal (Personnel Records). This census is administered by the Ministry of Employment, which requests information about a large set of variables concerning the firm, its establishments (if any) and also about each one of all the firms' employees. (3)

Crucially for the purpose of this paper, the list of variables available in the data includes unique identifiers for both firms and employees. These variables allow us to follow workers over time and, in particular, as they move between (foreign and domestic) firms. The set of variables at the firm level includes industry (five digits), region (three digits), size (number of workers), age, foreign-ownership percentage, sales, and equity. Moreover, at the worker level, the variables include education, age, gender, tenure (in months), occupation (five digits), wages, hours, job level (two digits), and promotions.

There are a total of five wage variables (base pay, overtime pay, tenure-related pay, bonus, and a residual category) and two hours variables (normal time and overtime). The hourly wage measure we use throughout in this paper is defined as the sum of all five wage variables above divided by the sum of the two hour variables. This hourly wage is then deflated using the Consumer Price Index.

We then use the foreign-ownership variable to characterize firms as either foreign or domestic owned. Specifically, we define a firm to be foreign owned when foreign investors control at least 50% of its voting rights. (4) Moreover, while we do not have information about domestic multinationals, we know that their number was particularly small over this period (less than 100). (5)

While the census has been ongoing since 1982, in this paper we use data from 1991 to 2000. (6) This is also a period in which foreign direct investment (FDI) into the Portuguese economy grew considerably, which may be explained, at least in part, by the accession to the European Community in 1986--see Figure 1 for the evolution of FDI inflows and outflows from Portugal from 1985.

[FIGURE 1 OMITTED]

We constructed our main mobility data sets by matching each annual file of all employees (and their firms) from 1992 to 2000 with the equivalent file for the previous year (Each year corresponds to a snapshot of the firms and their workers in the census month: March, up to 1993, and October, from 1994). Workers are matched over each pair of years based on their personal identification number (and also using their gender and year and month of birth variables as further checks). Moreover, by comparing the firm identifier of each worker over the two subsequent years, the worker can be classified as either a "stayer" or as a "mover." (7)Spurious movers--when the worker's firm identifier is different between t and t + 1 but the date of entry into the firm does not change in a consistent way (most likely because of mergers or movements across firms that belong to the same holding)--are dropped.

Moreover, as we acknowledge that many movers between firms will not necessarily be present in the data immediately in the following year's census month, we also consider movers between years t and t + 2. However, this case of course only applies when the individual's identifier is not present in the data in year t + 1. In this case, the date of entry into the firm in year t + 2 is required to be consistent with some spell outside the Quadros de Pessoal data during year t + 1, which implies that the individual was unemployed, inactive, or worked outside the coverage of the data (informal sector or public servants) in, at least, some period during year t + 1 (including that year's census month).

Finally, we also consider all other workers who can be defined as "stayers." These are workers present in years t and t + 1 at the same firm. However, according to our classification, a "stayer" between years t and t + 1 can then, of course, become a "mover" between years t + 1 and t + 2, for instance. The only "stayers" whom we disregard are those employed in firms involved in acquisitions over the years in which the acquisition takes place. A subset of these workers involved in acquisitions are examined in the work of Martins (2004).

Table 1 presents averages and standard deviations of the resulting data set, which corresponds to a total of more than 10 million observations. A total of 7.4% of all observations are movers from domestic to another domestic firm. Movers from foreign to domestic firms are 0.5% of the total while 0.6% move in the opposite direction. Only 0.1% move between different foreign firms. The remaining 91.4% of the data are "stayers." (8)

In addition, in Tables 2 and 3, we present statistics for each subsample of movers, according to their specific path. In particular, we separately describe the workers who move from foreign to domestic firms (about 55,000 workers), from domestic to foreign firms (about 67,000 workers), between domestic firms (over 800,000 workers), and between foreign firms (about 14,000 workers). Perhaps the most remarkable difference amongst the four groups of movers concerns their wage growth. They range from 22% in the case of movers from domestic to foreign firms to -6.3% in the case of movers from foreign to domestic firms. In the case of movers between domestic firms or movers between foreign firms, the average wage growth levels are similar: 6.4% and 5.8%, respectively.

This descriptive evidence suggests strongly that foreign firms do offer more generous wage policies. However, this difference may be explained by a large number of potential differences across the four groups of movers. The following sections will therefore examine this preliminary result in considerable detail.

III. RESULTS

A. Wage Levels

Following the earlier discussion, the main equation we consider in our empirical analysis is:

[w.sub.it] = [[beta].sub.1]D[F.sub.it] + [[beta].sub.2]D[F.sub.it] [[beta].sub.3]D[D.sub.it] + [[beta].sub.4]F [F.sub.it] + [X'.sub.it][[beta].sub.5] [F'.sub.it][[beta].sub.6] + [[gamma].sub.t] + [[epsilon].sub.it],

where [w.sub.it] represents the logarithm of the real wage of worker i in year t, X are worker controls (schooling, quadratics in tenure and experience, and gender), F are firm controls concerning the characteristics of the firm that employs worker i in year t (log firm size--measured by the number of workers, two-digit industry and region dummies, and a foreign-firm dummy), and [[gamma].sub.t] are year fixed effects. D [F.sub.it] is a dummy variable taking value 1 if a worker is employed by a domestic firm in year t (and in year t's job spell) (9) and will in the following job spell be employed by a foreign firm. Similarly, [FD.sub.it] is a dummy variable taking value 1 if a worker is employed by a foreign firm in year t (and in year t's job spell) and will in the following job spell be employed by a domestic firm. More formally,

(2) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

where s denotes the spell of employment of the individual. (10) Similarly, we have the following definitions for the remaining dummy variables:

(3) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

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

and, finally,

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

Given the motivation of this paper, [[beta].sub.1] and [[beta].sub.2] are the main parameters of interest. Their coefficients indicate the average difference in wages for workers who subsequently move from foreign to domestic firms or from domestic to foreign firms, respectively, in comparison to workers who stay in the same firm. Moreover, [[beta].sub.3] and 134 indicate the average difference in wages for workers who move from a domestic firm to another domestic firm or from a foreign firm to another foreign firm, respectively, before they move.

The first column of Table 4 presents the results for the estimation of Equation (1). We find that foreign firms pay on average about 13.6% more, a result consistent with those from other countries when using similar specifications. More importantly, we find that workers in domestic firms who will have a subsequent spell at a foreign firm are already paid (2.3%) more before they move. There is also evidence that workers in foreign firms whose subsequent employment spell is at another foreign firm are already paid substantially more (about 7.1% more) than similar workers in foreign firms but that will stay on in their current firm. On the other hand, workers who are employed in foreign firms, but employed at domestic firms in subsequent years, do not earn a significantly different wage than those who stay in their current firm. Finally, workers who move from a domestic firm to another domestic firm are on average less well paid (-0.9%) before they move.

As mentioned earlier, we also consider different versions of Equation 1. First, we allow for firm unobserved heterogeneity by including firm fixed effects ([[eta].sub.j]) in that equation:

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

This model allows for systematic differences across different categories of movers in terms of the wage policies of their firms. For instance, movers from domestic to foreign firms may tend to be employed in low-wage domestic firms. In that case, the domestic-to-foreign dummy variable coefficient may be spuriously high if no controls for firm-specific heterogeneity are included.

Moreover, specification (6) is also attractive as it can be interpreted as presenting within-firm evidence about the differences of each type of mover with respect to their colleagues at the same firm. In other words, we can compare the wages of each type of "mover-to-be" (domestic to domestic or domestic to foreign, for instance) with the wages of their colleagues who will stay at the same firms. For the benefit of robustness, we also consider extended versions of the model in Equation (6) by considering firm-year dummies ([[eta].sub.j] * [[gamma].sub.t]), instead of including separately firm and year dummies ([[eta].sub.j] + [[gamma].sub.t).

Our results, presented in columns B and C of Table 4 are consistent with the findings without controls for firm unobserved heterogeneity reported in column A. We find that workers who will move to foreign firms (regardless of being employed in domestic or foreign firms) are already receiving significantly higher wages even before they move, even when compared with their colleagues in the same firm or in the same firm-year. Movers from domestic to foreign firms earn about 0.8% more than stayers, while movers from foreign to other foreign firms earn about 1.7% to 2.2% more. On the other hand, movers from foreign to domestic firms again do not earn statistically different wages than their colleagues at foreign firms.

Finally, we examine longitudinal variation in each worker's wage, by including worker-specific fixed effects ([[alpha].sub.i]) and allowing for worker time-invariant unobserved heterogeneity:

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

This specification provides evidence about possible differences in the path of wages for workers who change their firm, as we allow for individual-specific heterogeneity. It is important to underline that the mobility dummies ([DF.sub.it], [FD.sub.it], etc.) refer to the employment spell that predates the movement between firms. This implies that the interpretation of the results is symmetrical to the more common case of dummy variables that are switched on after some occurrence. In other words, in our specification, a negative coefficient for a type of worker who moves, for instance, from a foreign to a domestic firm corresponds to an increase in wages as the worker is employed by the domestic firm.

Moreover, for the benefit of robustness, we consider first a version of Equation (7) without firm controls ([F.sub.it])--see column D. Such specification has the advantage of not partialling out any differences in wages that may result from workers moving, say, from "high-wage" to "low-wage" firms. If such differences in firm attributes are driven by compensating differentials, then it will be appropriate to control for their role in wages. However, if those differences are instead created by noncompetitive forces (e.g., rent sharing), then one should not control for them. (11) By presenting the results from both approaches (first without and then with firm-level control variables), we instead obtain what can be argued to be lower and upper bounds of the wage effects of different types of mobility between firms.

The last two columns of Table 4 indicate that all mover dummy estimates are negative. However, one should also take into account that movers between foreign and domestic firms will also gain or lose the wage premium associated to foreign ownership. This means that while domestic-to-foreign movers gain a total average wage increase of approximately 18% (9.5% + 8.5%)--see column D--movers in the opposite direction have an average wage change of -8.4% (0.08%-8.5%). On the other hand, movers from one domestic firm to another or from one foreign firm to another gain respectively 4.8% and 4.1% as they switch firms of the same ownership type. Furthermore, we find that some results are indeed attenuated when controlling for firm characteristics, but not to a very large extent. In this specification (see column E), movers from foreign to domestic firms take a pay cut of 3.2% (1.1%-4.3%), while movers from domestic to foreign firms gain a pay increase of 10.2% (5.9% + 4.3%). Movers between domestic (foreign) firms gain a wage increase of 3.1% (3.4%).

According to the earlier discussion, these findings are important evidence of more generous wage policies offered by foreign firms. On average, workers who move from a domestic to a foreign firm are more qualified (in terms of their wage residuals) than those who do not move at all from their domestic firms--a result that supports the existence of a "selection effect." However, when such workers switch to a foreign firm, they receive a very considerable pay increase. This finding supports the case that, on top of the selection effect, there is also a "wage policy effect." The latter result also suggests that a large number of such workers move voluntarily.

On the other hand, the wages of movers from foreign to domestic firms present very different characteristics. First, they tend to be (marginally) less well paid in the foreign firms from which they leave, either in a standard cross-section analysis or when comparing those movers-to-be with their colleagues who do not move. Second, movers from foreign to domestic firms take a considerable wage cut upon mobility. The contrast in the results for domestic-to-foreign and foreign-to-domestic movers give further support to the view that foreign firms offer more generous wage policies.

B. Wage Growth

One concern about the results presented earlier is that they may mask a trade-off between wage levels and wage growth. For instance, workers may accept lower starting wages (as they seem to do when moving from foreign to domestic firms) in exchange of steeper wage profiles at their new firms (Pakes and Nitzan, 1983). It is therefore important to investigate more deeply what happens after workers start their new jobs.

We conduct this analysis by estimating wage growth equations that allow for different wage growth levels depending on the type of between-firm mobility. We essentially adopt the wage equations described earlier (particularly the specifications presented in columns A and B of Table 4) but considering wage growth instead as the dependent variable:

(8) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

where [DELTA] [w.sub.it+1] = [w.sub.it+i] - [w.sub.it] is the wage growth of worker i between years t + 1 and t.

As discussed earlier, we also consider models with firm fixed effects ([[eta].sub.j]), in order to allow for firm-specific wage growth patterns:

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

Moreover, we also compute wage growth over different periods: between the second year in which the worker is in the "new" firm (domestic or foreign, depending on the specific mobility path) and the first year at that firm, between the third and the second year, and between the fourth and the third year (t + 2, t + 3 or t + 4). The comparison group for each analysis corresponds to a 25% sample of all workers who have also stayed on in their firms for at least 2, 3, or 4 yr (the same as the main group), but who have not moved between firms over the period.

Our results, presented in Table 5, systematically indicate higher wage growth for workers who move from domestic to foreign firms than for workers who move from foreign to domestic firms. For instance, in column A, we find that the former experience average wage growth of 4.6%, while for the latter average wage growth is only 2.5%. Very similar results are obtained for the specification with firm fixed effects and for wage growth comparisons over the third or fourth years, although the gaps in wage growth tend to fall with time. (12)

It is also interesting to notice that the inclusion of firm fixed effects in these wage growth equations dramatically increases the fit of the model, suggesting that there are very clear differences across firms not only in wage levels, as seen before, but also in wage profiles. However, the coefficients of the mobility dummies hardly change when such fixed effects are introduced.

In conclusion, we find that the differences between the two main types of movers obtained from the initial analysis in Section III(A) are actually strengthened, not weakened, when we consider wage growth patterns. Movers from domestic to foreign firms benefit not only from higher wage increases upon switching firms but also sustained higher wage growth levels after that, at least over the second and third years at their new firms.

C. Displacement

As mentioned earlier, this paper seeks to provide evidence about the impact of different types of mobility. This goal may not be rigorously achieved with observational data as ours, even when considering our extensive set of control variables. Intuitively, the wages earned by workers who do not move between firms may not provide an appropriate counterfactual for the wages of workers who move (if they had not moved). So, for instance, while we find that workers who move from domestic to foreign firms experience a very large wage increase with respect to workers who do not move, the former group of workers would perhaps also have experienced a similarly large wage increase if they had stayed at their original domestic firm. In this case, the effect of the domestic-to-foreign mobility type would be overestimated.

In order to provide complementary evidence that may be less affected by the endogeneity of interfirm mobility, we conduct an analysis based on displaced workers (Jacobson et al.

1993). The displaced workers we consider are derived from two groups. The first group of workers are those who leave firms that undergo mass layoffs but still stay in business. "Mass layoffs" are defined as a net job creation rate of -40% or less, provided the firm employs that year at least 20 workers. The second group of displaced workers are those who are employed in firms that go bankrupt, defined here as firms whose identifiers do not appear again in the data. (13) Given the motivation of our analysis, all displaced workers are required to be observed again in the data in year t + 1 or t + 2.

We also consider a sample of 25% of all workers who have never switched firms (our "control" group). All movers between firms that have moved outside of the context of a displacement (as defined above) are dropped from our sample. Table 6 presents some descriptive statistics of the sample of displaced workers. These correspond to approximately 183,000 observations or about 18.6% of all movers. Amongst other results, we find a large percentage of industry switchers (46.3%), which is, however, smaller than in the set of all movers. It is also interesting to notice that there is a greater share of domestic-to-foreign movers in the displaced sample than in the set of all movers.

We find that our earlier results based on all workers hold in the displaced movers subsampie--see Table 7. For instance, across specifications A, B, and C, we find that displaced movers from domestic to foreign firms are systematically better paid (at the domestic firms) compared with workers who do not move to foreign firms (a significant premium ranging from 1.7% to 4.9%). However, displaced workers who move from foreign to domestic firms earn lower wages than their colleagues at the foreign firms. In this case, we find a negative and significant premium ranging from -3.8% to -7.6%, except when not considering firm heterogeneity, when the premium is insignificant.

Recall that all such foreign-to-domestic coefficients were significant when considering the entire sample (Table 4).

Moreover, displaced movers from foreign to domestic firms also undergo considerable pay cuts, from -7.7% (-2.1% -5.6%) to -5.6% (-2.6% -3%). On the other hand, displaced movers from domestic to foreign firms do still enjoy a considerable increase in their pay, from 14.5% (8.9% + 5.6%) to 5.8% (2.8% + 3%).

Overall, these findings reinforce the earlier results on different patterns of wages across different types of movers. Workers who move from foreign to domestic firms are paid less (or, at least, not more) than their colleagues before they move (the "selection effect"). On top of that, these workers subsequently take pay cuts when they move (the "wage policy effect"). On the other hand, workers who move from domestic to foreign firms are already paid more than their colleagues. These displaced movers from domestic to foreign firms then go on to earn considerable pay increases at their new firms.

IV. ROBUSTNESS

In this section, we consider the robustness of our results to different subsamples of our data set. We start by considering the specific case of "high-skill" industries. Our motivation is twofold: First, "high-skill" industries are more prevalent in developed countries. The analysis of such industries may therefore shed light on the international differences regarding the evidence on the foreign-firm wage differential as mentioned earlier (Heyman et al. 2007; Andrews et al. 2007; Martins and Esteves 2008).

A second aspect, although related to the first, is that one may argue that there is less scope for large wage discrepancies between domestic and foreign firms in "high-skill" industries than in "low-skill" industries. In fact, the latter type of industries may allow for greater scope in terms of wage dispersion between firms and, in particular, between domestic and foreign firms, given the low wages typically paid there, especially when compared to the wages for similar jobs in the home country of the foreign firms.

Finally, we examine wage differences for the entire data set but adopting the point of view of differences and changes in wages after the spell of mobility, again for different types of mobility. This allows us to assess the wage increases upon mobility when controlling for the characteristics of the hiring firm (in our benchmark results we controlled instead for the characteristics of the firm the worker was leaving). Moreover, this approach also allows us to investigate the "ranking" of the new workers in terms of their colleagues at the hiring firms.

A. High-Skill Industries

We analyze high-skill industries by selecting only industries which exhibit particularly high levels of worker schooling. In particular, we construct our sample by considering the average schooling of all workers in each firm and in each year. We then calculate the mean of that average schooling per firm-year across all firms in each industry over the 10 yr in our sample. Finally, we select only firms in industries whose average schooling is in the top third of the distribution of schooling across all firms.

The results are presented in Table 8. Again, the findings are consistent with the earlier analysis, namely that there is a negative "selection effect" regarding the workers in foreign firms that are hired by domestic firms, although the finding is reverted in the analysis within each firm. Moreover, while such movers do still take pay cuts at domestic firms, it is noticeable that the magnitude of these pay cuts is considerably smaller than in previous analysis. On the other hand, we still find the same pattern as to the wage level and wage growth differences for workers who move from domestic to foreign firms. (14)

Overall, these findings lend further strength to the main results from Section III, especially in terms of the positive "selection" and "wage policy" effects concerning movers from domestic to foreign firms. However, the evidence concerning movers from foreign to domestic firms for some definitions of "high-skill" industries presents some differences, as the wage cuts for those workers tend to be smaller and there is some evidence of positive selection.

As mentioned earlier, this difference between the main results and those regarding "high-skill" industries may help one reconcile the contrasting results about the role of foreign firms in developed and developing economies in research using worker-level longitudinal data (Heyman et al. 2007; Andrews et al. 2007; Martins and Esteves 2008). The first two papers, covering the case of Sweden and Germany, respectively, find little evidence of wage differences between foreign and domestic firms, unlike in the case of the third paper, which considers the case of Brazil. To the extent that, in developing economies, foreign firms are more likely to be located in "low-skills" industries, then the scope for foreign firms to pay higher wages is greater. Our results, comparing different sectors of the same economy, are consistent with that hypothesis.

B. "After-Mobility" Analysis

In our final robustness analysis, we reexamine our earlier, benchmark results from the point of view of the firm to which a worker moves to, rather than from the point of view of the firm from which a worker leaves. This complementary perspective on the wage consequences of interfirm worker mobility serves two purposes. The first is to confirm the size of the wage changes following a movement to a different firm. This is important as at least part of the large wage increases documented for movers from domestic to foreign firms (and vice-versa) could be driven by the characteristics of foreign firms, particularly those that hire those movers. The second purpose of this analysis is to contrast the "selection effects" before and after the worker moves between firms.

We conduct our analysis by estimating the following equation:

(10) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

where [w.sub.it] represents the logarithm of the average real wage of worker i in year t, X are worker controls (schooling, quadratics in tenure and experience, and gender), F are firm controls (log firm size--measured by the number of workers, industry and region dummies, and a foreign-firm dummy) that now concern the firm to which the worker moves, and [[gamma].sub.t] are year fixed effects.

The mobility dummies are defined in a similar way as before, except that, as mentioned earlier, the point of view is now based on the period after the mobility took place. For instance,

(11) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

where s is the worker-firm spell, as defined earlier.

The results from different specifications based on Equation (10) are presented in Table 9. The results suggest a very strong "selection effect" as regards to movers from foreign to domestic firms. These workers tend to be very generously placed along the wage distribution of their firms, as the premium ranges between 4.6% and 12.8% (columns A to C). On the other hand, movers from domestic to foreign firms are, on average, slightly below similar workers pay levels (a negative premium of between -1.2% and -0.5%). Movers between domestic or between foreign firms tend to be well rewarded, particularly the latter.

With respect to wage growth, our evidence from Table 9 is very consistent with earlier findings. Wage growth of movers from foreign to domestic firms is negative, particularly when not controlling for firm characteristics (columns D and E). On the other hand, movers from domestic to foreign firms experience massive wage increases (from 18.9% to 10.7%). Movers between domestic or between foreign firms also experience wage gains, particularly the latter.

V. CONCLUSIONS

This paper provides comprehensive empirical evidence about the wage consequences of worker mobility between domestic and foreign firms. Using detailed matched employer-employee panel data from Portugal (covering both the manufacturing and services sectors), we trace virtually all spells of interfirm worker mobility in the country over a period of 10 yr (1991-2000).

Our results indicate that movements from domestic to foreign firms translate into considerable and robust average pay increases, of more than 10% in many cases. This pay increase is consistent with a "wage policy effect"--greater "generosity" in the remuneration practices of foreign firms vis-a-vis their domestic counterparts. On the other hand, there is also a "selection effect," although typically much smaller. This latter effect arises as foreign firms hire workers that are, on average, already better remunerated in their domestic firms than "similar" workers, even when conducting such comparison within each worker's firm. Moreover, our results for domestic firms are largely symmetric to those for foreign firms. For instance, movers from foreign to domestic firms earn, on average, a large pay cut when they move, a finding that lends further support to the "wage policy" effect documented earlier.

Our results also prove to be very robust across different specifications and samples. This is particularly the case for the subset of displaced workers, whose mobility can be argued to be less subject to endogeneity concerns. However, we find that both the "wage policy" and the "selection" effects are somewhat weaker in the specific case of mobility of workers from foreign to domestic firms in some "high-skill" sectors. This result may help explain the apparent negative relationship between the foreign-firm wage premium and economic development (Heyman et al. 2007: Andrews et al. 2007; Martins and Esteves 2008) to the extent that high-skill sectors are more prevalent in developed economies.

Overall, our findings for the "wage policy" and "selection" effects can be easily reconciled. Foreign firms can attract the "best" workers as they offer them large wage increases. Such wage increases will presumably more than compensate for the costs involved in the mobility process and therefore increase welfare in the host country.

Our results may also inform the debate about the productivity spillovers of foreign firms (Javorcik 2004) and, specifically, the role of worker mobility in those spillovers (Fosfuri et al. 2001; Balsvik 2006; Pesola 2007). Indeed, we find that domestic firms tend to hire "below-average" workers from foreign firms who take, on average, pay cuts (which is consistent with involuntary mobility). These results suggest that worker mobility is unlikely to be a major source of productivity spillovers from foreign to domestic firms. In fact, our findings, including the result that foreign firms attract some of the "best" workers in domestic firms, suggest that, if any, productivity spillovers from worker mobility occur from domestic to foreign firms.

Finally, the stark contrast between our strong results from an analysis of worker mobility and the mixed findings from the approach based on foreign acquisitions also deserves attention. One possible way to reconcile the two sets of findings is that acquisitions involve only a subset of a wider range of foreign-firm involvement in host labor markets. For instance, such approach necessarily disregards all opportunities for wage increases that domestic workers have once foreign firms set up in their countries. While it is possible and even likely that some domestic firms also employ generous wage policies, our results clearly indicate that domestic workers find it beneficial to have foreign firms as potential employers of their labor.

ABBREVIATION

FDI: Foreign Direct Investment

doi: 10.1111/j.1465-7295.2010.00301.x

REFERENCES

Abowd, J., F. Kramarz, and D. Margolis. "High Wage Workers and High Wage Firms." Econometrica, 67(2), 1999, 251-334.

Aitken, B., A. Harrison, and R. Lipsey. "Wages and Foreign Ownership: A Comparative Study of Mexico, Venezuela, and the United States." Journal of International Economics, 40(3-4), 1996, 345-71.

Almeida, R. "The Labor Market Effects of Foreign Owned Firms." Journal of International Economics, 72(1), 2007, 75-96.

Andrews, M., L., Bellman, T. Schank, and R. Upward. "The Takeover and Selection Effects of Foreign Ownership in Germany: An Analysis Using Linked Worker-Firm Data." GEP, University of Nottingham, Research Paper 2007/08. 2007.

Balsvik, R. "Is Mobility of Labour a Channel for Spillovers from Multinationals to Local Domestic Firms?" Norwegian School of Economics and Business Administration Discussion Paper 25. 2006.

Bartelsman, E. J., and M. Doms. "Understanding Productivity: Lessons from Longitudinal Microdata." Journal of Economic Literature, 38(3), 2000, 569-94.

Bjelland, M., B. Fallick, J. Haltiwanger, and E. McEntarfer. "Employer-to-employer Flows in the United States: Estimates Using Linked Employer-Employee Data," Discussion Series 2007-30, Board of Governors of the Federal Reserve System. 2007.

Davis, S. J., J. C. Haltiwanger, and S. Schuh. Job Creation and Destruction. Cambridge, MA: MIT Press, 1996.

Feenstra, R. C., and G. H. Hanson. "Foreign Direct Investment and Relative Wages: Evidence from Mexico's

Maquiladoras." Journal of International Economics, 42(3-4), 1997, 371-93.

Fosfuri, A., M. Motta, and T. Ronde. "Foreign Direct Investment and Spillovers through Workers' Mobility." Journal of International Economics, 53(1), 2001, 205-22.

Girma, S., and H. Gorg. "Evaluating the Causal Effects of Foreign Acquisition on Domestic Skilled and Unskilled Wages." Journal of International Economics, 72(1), 2007, 97-112.

Heyman, F., F. Sjoholm, and P. G. Tingvall. "Is There Really a Foreign Ownership Wage Premium? Evidence from Matched Employer-Employee Data." Journal of International Economics, 73(2), 2007, 355-76.

Huttunen, K. "The Effect of Foreign Acquisition on Employment and Wages: Evidence from Finnish Establishments." The Review of Economics and Statistics, 89(3), 2007, 497-509.

Jacobson, L. S., R. J. LaLonde, and D. G. Sullivan. "Earnings Losses of Displaced Workers." American Economic Review, 83(4), 1993, 685-709.

Javorcik, B. S. "Does Foreign Direct Investment Increase the Productivity of Domestic Firms? In Search of Spillovers through Backward Linkages." American Economic Review, 94(3), 2004, 605-27.

Martins, P. S. "Do Foreign Firms Really Pay Higher Wages'? Evidence from Different Estimators," IZA Discussion Paper 1388. 2004.

--. "Heterogeneity in Real Wage Cyclicality." Scottish Journal of Political Economy, 54(5), 2007, 684-98.

--. "Paying More to Hire the Best? Foreign Firms, Wages and Worker Mobility." IZA Discussion Papers 3607. 2008.

--. "Dismissals for Cause: The Difference That Just Eight Paragraphs Can Make." Journal of Labor Economics, 27(2), 2009, 257-79.

Martins, P. S., and L. Esteves. "Foreign Ownership, Employment and Wages in Brazil: Evidence from Acquisitions, Divestments and Job Movers." IZA Discussion Paper 3542. 2008.

Pakes, A., and S. Nitzan. "Optimum Contracts for Research Personnel, Research Employment, and the Establishment of Rival Enterprises." Journal of Labor Economics, 1(4), 1983, 345-65.

Pesola, H. "Foreign Ownership, Labour Mobility and Wages." Helsinki School of Economics Discussion Paper 175, 2007.

(1.) Huttunen (2007) finds positive effects on wages from foreign acquisitions, Girma and Gorg (2007) find positive effects for some types of acquisitions but no effects for other types, and Almeida (2007) finds small or no effects. See Andrews et al. (2007) also for a thorough survey of the literature.

(2.) See Martins and Esteves (2008) for a different analysis of worker interfirm mobility, based on the case of Brazil. See also Bjelland et al. (2007) for recent evidence of interfirm mobility in the United States.

(3.) The census is designed to check compliance with employment laws. It also serves general statistical purposes. Firms that do not fill in the census questionnaire correctly are subject to penalties that are perceived to have ensured high levels of data quality.

(4.) Strictly speaking, this threshold is neither a necessary nor a sufficient condition for a firm to be controlled by a foreign investor. However, we believe 50% is the optimal level in terms of separating firms with a large enough foreign presence from the remaining firms. In any case, our results are not sensitive to a definition based on a threshold of 10% of voting rights.

(5.) This enables us to consider that foreign firms are virtually the same as multinational firms, sidestepping the debate about whether it is multinationality or foreign ownership that matters (Heyman et al. 2007).

(6.) Although it is possible to consider a longer period, worker-level data have not been made available by the Ministry of Employment for 1990 and 2001. Therefore, the analysis of a longer period would introduce time gaps in our study.

(7.) See Martins (2007) and Martins (2009) for other examples of, respectively, worker and firm longitudinal analysis based on the same data set.

(8.) Amongst other results, almost 40% of all workers are female, the average tenure is 8.7 yr, and 7.8% of workers are employed by foreign firms. The average net job creation rate (weighted by firm size) is 3.7%. (The net job creation rate is defined as NJC= [L.sub.t] - [L.sub.t-1]/0.5([L.sub.t] + [L.sub.t-]), in which [L.sub.t] denotes the number of workers in period t [Davis et al. 1996].)

(9.) A job spell is defined as a set of all years in which a worker is continuously employed by the same firm.

(10.) We do not need to impose extra conditions such as [[eta].sub.is] [not equal to] [[eta].sub.i,s-1], where [[eta].sub.is] denotes the firm employing worker i in spell s, when [FD.sub.is] = 1, indicating that the worker moves to a different firm, as we have ruled out firm acquisitions, given our sample design. However, the effect of the foreign-firm dummy variable can be estimated out of (a very small number of) two-period movers who return to the same firm after that firm has undergone a domestic or foreign acquisition.

(11.) For instance, if the positive relationship between firm size and wages is driven by rent sharing, then when controlling for firm size, the domestic-to-foreign mobility coefficient would wrongly fail to pick up the wage increase related to rent sharing.

(12.) For instance, column F of Table 5 indicates that there are virtually no differences in wage growth between workers who moved from foreign to domestic firms or from domestic to foreign firms when they reach their fourth year in their new firms. However, in the fourth year, there may be important selection issues, as the sample drops to almost one-third of its original size in the second year.

(13.) Very occasionally, data for (smaller) firms exhibit gaps in some years. We conduct our analysis making sure these gaps are not regarded as displacements.

(14.) Moreover, we also replicated our analysis with different alternative measures of skill, including worker experience, tenure, and wages (Martins 2008) with very similar results.

PEDRO S. MARTINS *

* I thank, without implicating, Erling Barth, Andrew Bernard, John Earle, Gabor Kezdi, Mikael Lindahl, Lisa Lynch, David de Meza, Claudio Piga, conference/seminar participants at SOFI (Stockholm), FIEF (Stockholm), ISF (Oslo), Central European University (Budapest), ZEW (Mannheim), IPEA (Brasilia), EALE (Prague), CEG-IST (Lisbon), Queen Mary (University of London), Loughborough University and the European Commission, and two referees and the editor (P. Arcidiacono) for their useful comments on this and other versions of this research. Support from the ESRC (RES-062-23-0546) and Banco de Portugal is gratefully acknowledged.

Martins: School of Business and Management, Queen Mary, University of London, Mile End Road, London El 4NS, United Kingdom; CEG-IST, Lisbon, Portugal. Phone +44/0 2078827472, Fax +44/0 2078823615, E-mail p.martins@qmul.ac.uk

TABLE 1
Descriptive Statistics--All Workers

Variable Mean (Std. Dev.) N

Log hourly pay 1.383 (0.612) 10,419,465
Schooling 6.892 (3.689) 11,302,053
Female 0.39 (0.488) 11,552,228
Tenure 8.705 (8.599) 11,552,228
Experience 23.892 (12.318) 11,302,053
Foreign firm 0.078 (0.268) 11,552,228
For-to-Dom 0.005 (0.069) 11,552,228
Dom-to-For 0.006 (0.076) 11,552,228
Dom-to-Dom 0.074 (0.261) 11,552,228
For-to-For 0.001 (0.035) 11,552,228
Log firm size 4.469 (2.316) 11,552,228
Net job creation rate 0.037 (0.259) 10,697,558

Notes: "Foreign" is a dummy taking value I if the firm-year is
foreign owned (and value 0 otherwise). "Dom-to-For" is a dummy taking
value 1 if the worker moves in that period from a domestic firm to a
foreign firm (and value 0 otherwise), that is, if in the next period
the worker will be at a foreign firm. "Dom-to-Dom" is a dummy taking
value 1 if the worker moves from a domestic firm to another domestic
firm (and value 0 otherwise). "For-to-Dom" is a dummy taking value 1
if the worker moves from a foreign firm to a domestic firm (and value
0 otherwise). "For-to-For" is a dummy taking value 1 if the worker
moves from a foreign firm to another foreign firm (and value 0
otherwise).

TABLE 2
Descriptive Statistics--All Workers Moving from Foreign to Domestic
Firms (left) or between Foreign Firms (right), While at Foreign Firm

Variable Mean (St.Dev.) N

Log hourly pay 1.573 (0.664) 51,434
Wage growth -0.063 (0.551) 48,085
Schooling 8.734 (4.103) 52,954
Female 0.398 (0.49) 54,565
Tenure 2.89 (4.284) 54,565
Experience 15.173 (9.743) 52,954
Foreign firm 1 (0) 54,565
Log firm size 5.769 (1.709) 54,565
Net job creation rate 0.12 (0.383) 51,058
Industry switcher 0.668 (0.471) 54,564
Year gap 1.28 (0.449) 54,565
1992 0.071 (0.257) 54,565
1993 0.1 (0.3) 54,565
1994 0.075 (0.264) 54,565
1995 0.099 (0.298) 54,565
1996 0.107 (0.309) 54,565
1997 0.144 (0.351) 54,565
1998 0.167 (0.373) 54,565
1999 0.14 (0.347) 54,565

Variable Mean (St.Dev.) N

Log hourly pay 1.741 (0.702) 13,588
Wage growth 0.058 (0.529) 13,238
Schooling 9.896 (4.189) 13,838
Female 0.426 (0.495) 14,232
Tenure 3.789 (5.579) 14,232
Experience 14.086 (9.388) 13,838
Foreign firm 1 (0) 14232
Log firm size 5.781 (1.74) 14,232
Net job creation rate 0.116 (0.418) 13,346
Industry switcher 0.576 (0.494) 14232
Year gap 1.281 (0.45) 14,232
1992 0.062 (0.242) 14,232
1993 0.093 (0.29) 14,232
1994 0.067 (0.251) 14,232
1995 0.096 (0.294) 14,232
1996 0.116 (0.32) 14,232
1997 0.192 (0.394) 14,232
1998 0.166 (0.372) 14,232
1999 0.141 (0.348) 14,232

Notes: See description of variables in notes of Table 1. "Wage
growth" denotes the difference in the logarithm of the hourly wage
between years t + 1 and t. "Industry switcher" is a dummy variable
taking value 1 if the worker is in a different two-digit industry in
year t + 1 when compared with year t. "Year gap" is the year
difference between the 2 yr over which the worker moves between firms
(by design, this is between one and two).

TABLE 3
Descriptive Statistics-All Workers Moving from Domestic to Foreign
Firms (left) or between Domestic Firms (right), While at Domestic Firm

Variable Mean (St.Dev.) N

Log hourly pay 1.334 (0.642) 61,811
Wage growth 0.217 (0.526) 59,583
Schooling 8.314 (3.961) 64,371
Female 0.442 (0.497) 66,609
Tenure 2.8 (4.413) 66,609
Experience 14.592 (9.638) 64,371
Foreign firm 0 (0) 66,609
Log firm size 4.437 (2.079) 66,609
Net job creation rate 0.099 (0.386) 59,346
Industry switcher 0.684 (0.465) 66,607
Year gap 1.252 (0.434) 66,609
1992 0.097 (0.296) 66,609
1993 0.114 (0.317) 66,609
1994 0.085 (0.279) 66,609
1995 0.091 (0.288) 66,609
1996 0.099 (0.299) 66,609
1997 0.124 (0.33) 66,609
1998 0.148 (0.355) 66,609
1999 0.166 (0.372) 66,609

Variable Mean (St.Dev.) N

Log hourly pay 1.174 (0.55) 776,666
Wage growth 0.064 (0.507) 726,018
Schooling 6.821 (3.471) 821,424
Female 0.368 (0.482) 849,294
Tenure 3.132 (4.494) 849,294
Experience 17.871 (10.747) 821,424
Foreign firm 0 (0) 849,294
Log firm size 3.656 (1.934) 849,294
Net job creation rate 0.072 (0.368) 744,665
Industry switcher 0.529 (0.499) 849,293
Year gap 1.263 (0.44) 849,294
1992 0.103 (0.304) 849,294
1993 0.118 (0.323) 849,294
1994 0.087 (0.282) 849,294
1995 0.098 (0.297) 849,294
1996 0.105 (0.306) 849,294
1997 0.127 (0.333) 849294
1998 0.14 (0.347) 849,294
1999 0.124 (0.329) 849,294

Note: See description of variables in notes of Tables 1.

TABLE 4
Wage Equations-All Workers

 A B C

Foreign firm 0.136 -0.002
 (.001) *** (.002)
For-to-Dom 0.002 -0.004 -0.002
 (.003) (.002) (.002)
Dom-to-For 0.023 0.008 0.008
 (.002) *** (.002) *** (.002) ***
Dom-to-Dom -0.009 0.005 0.007
 (.0006) *** (.0006) *** (.0007) ***
For-to-For 0.071 0.017 0.022
 (.005) *** (.004) *** (.004) ***
Worker controls x x x
Firm controls x x x
Worker fixed effects
Firm fixed effects x
Firm-year fixed effects x
Obs. 42,85,462 4,285,467 4,285,467
[R.sup.2] 0.579 0.713 0.77

 D E

Foreign firm 0.085 0.043
 (.001) *** (.001) ***
For-to-Dom -0.0008 -0.011
 (.002) (.002) ***
Dom-to-For -0.095 -0.059
 (.002) *** (.002) ***
Dom-to-Dom -0.048 -0.031
 (.0006) *** (.0006) ***
For-to-For -0.041 -0.034
 (.003) *** (.003) ***
Worker controls x x
Firm controls x
Worker fixed effects x x
Firm fixed effects
Firm-year fixed effects
Obs. 4,285,467 4,285,462
[R.sup.2] 0.874 0.879

Notes: Dependent variable: log real hourly wage. Worker-level
controls are schooling, experience and its square, tenure and its
square, and a female dummy variable. Firm-level controls are region
and industry dummies and firm size. "Foreign firm" is a dummy taking
value 1 if the firm-year is foreign owned (and value 0 otherwise).
"Dom-to-For" is a dummy taking value I for workers who are in the
current employment spell at a domestic firm and will in the next
employment spell be at a foreign firm (and value 0 otherwise). "For-
to-For" is a dummy taking value I for workers who are in the current
employment spell at a foreign firm and will in the next employment
spell be at a foreign firm (and value 0 otherwise). "For-to-Dom" is a
dummy taking value 1 for workers who are in the current employment
spell at a foreign firm and will in the next employment spell be at a
domestic firm (and value 0 otherwise). "Dom-to-Dom" is a dummy taking
value I for workers who are in the current employment spell at a
domestic firm and will in the next employment spell be at a different
domestic firm (and value 0 otherwise). "For-to-For" is a dummy taking
value 1 for workers who are in the current employment spell at a
foreign firm and will in the next employment spell be at a different
foreign firm (and value 0 otherwise). All specifications include year
dummies. Robust standard errors, clustered at the worker level.
Significance levels: *: .10; **: .05; ***: .01.

TABLE 5
Wage Growth Equations-Only Workers Who Stay in t + 2, t + 3, or t + 4
in Same Firm as in t + l

 t + 2

Wage Growth in: A B

Foreign firm 0.002 0.0002
 (.001) ** (.003)
For-to-Dom 0.025 0.020
 (.003) *** (.004) ***
Dom-to-For 0.046 0.046
 (.003) *** (.003) ***
Dom-to-Dom 0.014 0.015
 (.0008) *** (.001) ***
For-to-For 0.042 0.040
 (.006) *** (.006) ***
Worker controls x x
Firm controls x x
Firm fixed effects x
Obs. 1,343,810 1,343,813
[R.sup.2] 0.007 0.119

 t + 3

Wage Growth in: C D

Foreign firm 0.003 0.018
 (.001) ** (.004) ***
For-to-Dom 0.007 0.008
 (.003) ** (.004) *
Dom-to-For 0.021 0.022
 (.003) *** (.003) ***
Dom-to-Dom 0.009 0.009
 (.0009) *** (.001) ***
For-to-For 0.030 0.029
 (.006) *** (.007) ***
Worker controls x x
Firm controls x x
Firm fixed effects x
Obs. 839,529 839,532
[R.sup.2] 0.005 0.153

 t + 4

Wage Growth in: E F

Foreign firm -0.002 -0.015
 (.002) (.005) ***
For-to-Dom 0.005 0.005
 (.004) (.005)
Dom-to-For 0.014 0.018
 (.003) *** (.004) ***
Dom-to-Dom 0.005 0.008
 (.001) *** (.002) ***
For-to-For 0.009 0.004
 (.007) (.009)
Worker controls x x
Firm controls x x
Firm fixed effects x
Obs. 552,900 552,902
[R.sup.2] 0.005 0.186

Notes: Dependent variable: growth of the real hourly wage. Worker-
level controls are schooling, experience and its square, tenure and
its square, and a female dummy variable. Firm-level controls are
region and industry dummies and firm size. "Foreign firm" is a dummy
taking value 1 if the firm-year is foreign owned (and value 0
otherwise). "Dom-to-For" is a dummy taking value 1 for workers who
are in the current employment spell at a domestic firm and will in
the next employment spell be at a foreign firm (and value 0
otherwise). See notes to Table 4 for descriptions of remaining
variables. All specifications include year dummies. Robust standard
errors, clustered at the worker level. Significance levels: *: .10;
**: .05; ***: .01.

TABLE 6
Descriptive Statistics-Only Movers Who Are Displaced

Variable Mean (Std. Dev.) N

Log hourly pay 1.157 (0.562) 163,652
Wage growth 0.065 (0.493) 152,025
Schooling 6.685 (3.432) 177,144
Female 0.401 (0.49) 183,638
Tenure 3.686 (5.059) 183,638
Experience 19.29 (11.176) 177,144
Foreign firm 0.048 (0.213) 183.638
For-to-Dour 0.038 (0.192) 183,638
Dom-to-For 0.056 (0.229) 183,638
Dom-to-Dom 0.897 (0.304) 183,638
For-to-For 0.009 (0.096) 183,638
Log firm size 3.165 (1.956) 183,638
Net job creation rate 0.024 (0.443) 149,013
Industry switcher 0.463 (0.499) 183,636
Year gap 1.269 (0.443) 183,638
1992 0.099 (0.298) 183,638
1993 0.157 (0.364) 183,638
1994 0.082 (0.274) 183.638
1995 0.097 (0.296) 183,638
1996 0.099 (0.299) 183,638
1997 0.119 (0.324) 183,638
1998 0.131 (0.338) 183,638
1999 0.136 (0.342) 183.638

Notes: "Displaced movers" are those that left firms that leave the
data or that left firms that were undergoing major downsizing (net
job creation of -40% or less). See description of variables in notes
to Tables 1 and 2.

TABLE 7
Wage Equations--Only Movers Included Are Those Displaced

 A B C

Foreign firm 0.140 0.0001
 (.001) *** (.002)
For-to-Dom 0.009 -0.038 -0.076
 (.008) (.008) *** (.015) ***
Dom-to-For 0.049 0.017 0.031
 (.006) *** (.006) *** (.008) ***
Dom-to-Dom -0.0003 0.013 0.022
 (.001) (.002) *** (.005) ***
For-to-For 0.062 -0.003 -0.045
 (.016) *** (.015) (.021) **
Worker controls x x x
Firm controls x x x
Worker fixed effects
Firm fixed effects x
Firm-year fixed effects x
Obs. 2,260,710 2,260,713 2,260,713
[R.sup.2] 0.621 0.752 0.809

 D E

Foreign firm 0.056 0.030
 (.002) *** (.002) ***
For-to-Dom 0.021 0.026
 (.005) *** (.005) ***
Dom-to-For -0.089 -0.028
 (.004) *** (.004) ***
Dom-to-Dom -0.028 -0.008
 (.001) *** (.001) ***
For-to-For -0.019 0.006
 (.009) ** (.009)
Worker controls x x
Firm controls x
Worker fixed effects x x
Firm fixed effects
Firm-year fixed effects
Obs. 2,260,713 2,260,710
[R.sup.2] 0.933 0.934

Notes: "Displaced movers" are those who left firms that leave the
data or who left firms that were undergoing major downsizing (see
more details in the main text). Dependent variable: log real hourly
wage. Worker-level controls are schooling, experience and its square,
tenure and its square, and a female dummy variable. Firm-level
controls are region and industry dummies and firm size. "Foreign
firm" is a dummy taking value 1 if the firm-year is foreign owned
(and value 0 otherwise). "Dom-to-For" is a dummy taking value 1 for
workers who are in the current employment spell at a domestic firm
and will in the next employment spell be at a foreign firm (and value
0 otherwise). See notes to Table 4 for descriptions of remaining
variables. All specifications include year dummies. Robust standard
errors, clustered at the worker level. Significance levels: *: .10;
**: .05; ***: .01.

TABLE 8
Wage Equations-Only Workers from "High-Skill" Industries

 A B C

Foreign firm 0.167 -0.003
 (.002) *** (0.003)
For-to-Dom -0.017 0.008 0.007
 (.003) *** (.003) ** (.004) **
Dom-to-For 0.021 0.013 0.011
 (.003) *** (.003) *** (.003) ***
Dom-to-Dom -0.025 0.009 0.011
 (.001) *** (.001) *** (.001) ***
For-to-For 0.057 0.024 0.028
 (.005) *** (.005) *** (.006) ***
Worker controls x x x
Firm controls x x x
Worker fixed effects
Firm fixed effects x
Firm-year fixed effects x
Obs. 1,534,477 1,534.482 1,534.482
[R.sup.]2 0.581 0.722 0.776

 D E

Foreign firm 0.063 0.041
 (.002) *** (.002) ***
For-to-Dom -0.027 -0.037
 (.003) *** (.003) ***
Dom-to-For -0.117 -0.090
 (.003) *** (.003) ***
Dom-to-Dom -0.082 -0.060
 (.001) *** (.001) ***
For-to-For -0.053 -0.050
 (.004) *** (.004) ***
Worker controls x x
Firm controls x
Worker fixed effects x x
Firm fixed effects
Firm-year fixed effects
Obs. 1,534.482 1,534,477
[R.sup.]2 0.903 0.906

Notes: "High-skill" industries defined as those at the top third of
the schooling distribution. Dependent variable: log real hourly wage.
Worker-level controls are schooling, experience and its square,
tenure and its square, and a female dummy variable. Firm-level
controls are region and industry dummies and firm size. "Foreign
firm" is a dummy taking value I if the firm-year is foreign owned
(and value 0 otherwise). "Dom-to-For" is a dummy taking value 1 for
workers that are in the current employment spell at a domestic firm
and will in the next employment spell be at a foreign firm (and value
0 otherwise). See notes to Table 4 for descriptions of remaining
variables. All specifications include year dummies. Robust standard
errors, clustered at the worker level. Significance levels: *: .10;
**: .05: ***: .01.

TABLE 9
Wage Equations--All Workers, Variables Measured "After" Moving to
New Firm

 A B C

Foreign firm 0.157 -0.009
 (.001) *** (.003) ***
FD 0.128 0.052 0.046
 (.002) *** (.002) *** (.003) ***
DF -0.005 -0.015 -0.012
 (.002) ** (.002) *** (.002) ***
DD 0.047 0.016 0.013
 (.0007) *** (.0007) *** (.0008) ***
FF 0.125 0.064 0.062
 (.005) *** (.005) *** (.005) ***
Worker controls x x x
Firm controls x x x
Worker fixed effects
Firm fixed effects x
Firm-year fixed effects x
Obs. 4,183,008 4,183,009 4,183,009
[R.sup.2] 0.562 0.704 0.764

 D E

Foreign firm 0.100 0.054
 (.002) *** (.002) ***
FD 0.039 0.042
 (.002) *** (.002) ***
DF 0.089 0.053
 (.002) *** (.002) ***
DD 0.040 0.037
 (.0007) *** (.0007) ***
FF 0.066 0.053
 (.003) *** (.003) ***
Worker controls x x
Firm controls x
Worker fixed effects x x
Firm fixed effects
Firm-year fixed effects
Obs. 4,183,009 4,183,008
[R.sup.2] 0.88 0.884

Notes: Dependent variable: log real hourly wage. Worker-level
controls are schooling, experience and its square, tenure and its
square, and a female dummy variable. Firm-level controls are region
and industry dummies and firm size. "Foreign firm" is a dummy taking
value I if the firm-year is foreign owned (and value 0 otherwise).
"Dom-to-For" is a dummy taking value 1 for workers who are in the
current employment spell at a domestic firm and will in the next
employment spell be at a foreign firm (and value 0 otherwise). See
notes to Table 4 for the descriptions of remaining variables. All
specifications include year dummies. Robust standard errors,
clustered at the worker level. Significance levels: *: .10; **: .05;
***: .01.