The fixed effects specifications in columns 1 to 3 of panel A show that processinnovationis positively and significantly associated with foreign ownership.Column 1a shows that a foreign-acquired firm is 57 percent more likely to haveundertaken a process innovation while foreign owned, relative to a firm that staysdomestic. This estimate is robust to controlling for industry trends and laggedfirm characteristics (columns 2a and 3a). Column 4a shows that the coefficientestimateon the lead indicator for acquisition is not significantly different from zero.Furthermore, it is significantly smaller than the coefficient of interest (as shownby the p-value of 0.048). Column 5a presents the propensity score reweightedregressionsthat allow us to control for time-varying selection. The coefficient 0.611is similar to earlier columns and also highly significant, implying that firms undertakemore process innovation upon acquisition.31Turning to the second and third panels of Table 3, the estimated coefficients incolumn 1 reveal that product innovation and the assimilation of foreign technologiesalso increase after acquisition. However, the point estimates fall, and the standarderrors are larger with further controls and in the propensity score estimation.32These results are the average over all types of acquired firms, and as we will see inSection IVC, there is evidence that these two types of innovation increase significantlyin firms that export through the foreign parent after acquisition.Table 4 shows the effect of foreign ownership on the disaggregated measures ofprocess innovation. We distinguish between firms that report to have invested only innew machines (panel B), only in new methods of organizing production (panel C),or in both simultaneously (panel A). The results reveal some interesting contrasts.While foreign-acquired firms are not significantly more likely to only introduce newmachinery or only introduce new ways of organizing production, the simultaneousintroduction of new machinery and new organizational processes is significantlyassociated with foreign acquisition (Teece 1977). Panel A shows this result. Boththe fixed effects specifications of columns 1a to 4a and the propensity score estimationof column 5a show that upon acquisition, firms are more likely to introduce newmachines and new organizational methods simultaneously.33 This is an interestingresult since we might have expected that foreign firms would also be more likelyto introduce either type of process innovation independently. The findings are consistentwith the complementarities found by Black and Lynch (2001), Bresnahan,Brynjolfsson, and Hitt (2002), and Bartel, Ichniowski, and Shaw (2007) betweendifferent types of technology upgrading. Since firms appear to introduce both typesof innovations jointly, it is important to allow for the effect of both actions whenquantifying the multinational productivity advantage.3431 The results shown in the first differences specifications in Appendix Table A1 reveal that this increase occursonly one year after acquisition, with further increases in the second and third years.32 Note that the variable indicating the assimilation of foreign technologies is available only every four years,reducing the number of observations in these specifications and, thus, reducing the power of the fixed effects resultssince we have, at most, five observations within a firm for this variable.33 The first differences specification in Appendix Table A1 reveals that, for the subset of firms in that specification,the firms that are acquired are more likely to undertake both types of process innovation prior to acquisition,but are also even more likely to undertake both types of process innovation simultaneously after acquisition.34 All of our results are robust to the analysis of firms that report no change in reporting unit throughout the timethey are in the sample, as well as restricting the sample to firms that report no changes in the number of establishments.This rules out the concern that the definition of the reporting unit changes following acquisition.
C. The Role of Market Access Provided by the Foreign ParentThese findings on increased process innovation following acquisition, togetherwith our positive selection results, are consistent with a world in which multinationalschoose to acquire the most productive firms since that is where the returns to theirinvestment are highest. One explanation for this is that—as proposed in the literatureon the sources of multinational advantage—the foreign firm gives accessto technology at a lower cost ( b i ) than the acquired firm would have faced had itremained under domestic control. However, we highlight an alternative reason for
our findings,based on a key feature of multinationals: they often grant their subsidiariesaccess to a larger global market.Tables 5 and 6 explore whether innovation decisions are related to the fact thatforeign ownership provides access to foreign markets. We regress the innovationvariables on indicator variables for whether the firm exports, and for whether thefirm exports through the foreign parent. Exporting through the foreign parent maymean that the firm is using the parent’s distribution channels and networks to export,or that it sells its goods directly to the foreign parent (as part of a global productionsystem). The base category includes all the channels that were always available tothe domestic firms (exporting through its own means, using a Spanish specializedintermediary, or cooperative export agreements with other firms).Table 5 presents the results for overall process innovation (panel A) and for processinnovation that involves the simultaneous introduction of new machines andnew methods of organizing production (panel B). Column 1a reveals that exportingis positively associated with investment in process innovation, consistent with thefindings of previous studies (Verhoogen 2008; Bustos 2011; Lileeva and Trefler2010; Aw, Roberts, and Xu 2011). This result holds when controlling for foreignownership (column 2a), which is also significant, suggesting that the ownershipmechanismoutlined in this article offers a separate explanation for acquired firms’increased process innovation.Columns 3a to 5a introduce our key variable of interest, showing fixed effectsregressions using process innovation as a dependent variable, where we include theindicator variable for whether the firm exports via the foreign parent. Notably, wefind that starting to export through a foreign parent has a large and significant positivecoefficient. Since this specification also includes the interaction between exportingand being foreign owned, this suggests that it is not exporting while foreign ownedper se, but exporting through the foreign parent, that is associated with innovation.35Since we can distinguish between different types of process innovation, we evaluatethe type that exporters are more likely to undertake. Although exporting is,on average, not significantly associated with the simultaneous introduction of newmachines and new forms of organizing production (column 1b), foreign-ownedfirms are more likely to engage in this type of process innovation (column 2b).Column 3b shows that, similar to the process innovation results in column 3a, innovationis driven mainly by the foreign-owned exporters that export via the foreignparent. In contrast, we find that exporting is significantly associated with the introductionof new machines exclusively, while exporting through a foreign parent isnot (unreported). This reinforces our findings in Table 4, which suggest that foreignownership leads to a specific type of process innovation, involving both newmachines and new methods of organizing production.36Columns 6 through 8 in Table 5 present the propensity score results for the marketaccess channel, allowing us to better control for time-varying selection. Here,we consider the treatment to be starting to export through the foreign parent, and35 Consistent with the idea that foreign firms provide market access to exporting subsidiary firms in our data,Artopoulos, Friel, and Hallak (2011) document that firms with knowledge of business practices in foreign marketsare more successful exporters. We argue that foreign firms can provide that knowledge.36 We find no evidence that exporting through a foreign parent leads to the introduction of new machines or neworganizational practices separately (results unreported).3618 THE AMERICAN ECONOMIC REVIEW December 2012we recalculate the propensity score and the associated weights for each firm asdescribed in Section IVA. Column 6 shows that exporting through a foreign parentis associated with more process innovation (column 6a) and, in particular, withinnovation that involves the simultaneous introduction of new machines and neworganizational practices (column 6b). This result holds when controlling for laggedforeign ownership (column 7), exporting status, and their interaction, and industrytime trends (column 8).Table 6 shows the effect of market access through the foreign parent on productinnovation and the assimilation of foreign technologies. Using both the fixed effectsand the propensity score estimator, we find that exporting via a foreign parent leads
to more product innovation and the assimilation of foreign technologies. Theseresults shed light on those in Table 3, where we found a statistically weaker relationshipbetween foreign ownership and these two variables. Once we distinguishbetween foreign-owned firms that export via a foreign parent and those that do not,we see that those that use the parent as an export channel also invest in new productsand assimilate new foreign technologies.Taken together, these results imply that when firms are acquired by a foreign parent,they increase innovation, especially when the parent firm provides access toexport markets. The observed relationship between market access and innovationactivity offers further support for the mechanism outlined in the model, as it highlightsthe role for market access as a driver of innovation decisions. It also indicatesthat market access can be a sufficient reason for acquisition (even when foreign anddomestic firms face similar variable innovation costs), when larger market accessincreases the potential benefits from investment activity. Furthermore, to the extentthat there is persistence in market access, it provides a rationale for persistent productivitydifferences among firms.D. Exports and WagesFinally, in Table 7, we show other changes that take place within firms as a consequenceof foreign ownership. We study how the share of exports in total sales(panel A), the logarithm of total exports for exporters (panel B) and the logarithmof average firm wage (computed as the total wage bill divided by the number ofemployees, panel C) change with foreign acquisition. Columns 1 through 4 showthe equal-weighted fixed effects specification, and column 5 shows the propensityscore reweighted results.We find that the proportion of exports in total sales increases significantly followingforeign acquisition. The propensity score estimate in column 5a shows thatthe share of exports is, on average, 6.7 percentage points higher in each year foracquired firms than for similar firms that are not acquired. The fact that the salesincrease is disproportionately large in foreign markets is consistent with subsidiarieshaving increased access to these markets after acquisition. We also find that thevolume of exports for exporters is 33 percent higher for exporters under foreignownership (column 5b). Finally, panel C provides some suggestive evidence ofaverage wages increasing upon acquisition, although this is not statistically significant.While this could mean that firms are increasing their wages and/or upgradingthe skill of the workforce, we cannot distinguish between these possibilities withthe available data.V. Foreign Ownership and Productivity EvolutionSection IIIB showed that there is positive selection of target firms by foreignmultinationals; Section IVB established that, upon acquisition, firms upgradetheir technology by doing more process innovation and, in particular, by investingsimultaneously in new machines and new methods to organize production. Now,we investigate the effect of acquisition on firm productivity directly, as well as itsconsequences for the evolution of the distribution of productivity within industries.VOL . 102 NO. 7 guada lupe et al.: innovati on and foreign ownership 3621Under the assumption that the investment activities described above are, indeed,productivity enhancing, we predict that the increased levels of these activities uponacquisition lead to higher productivity for acquired firms. Figures 4 and 5 illustrate ourbasic productivity results. Figure 4 shows the distribution of firm productivity in thebase year, and four years after that, for firms that are domestic in that first year but willbe foreign owned four years later. The distribution is shifted to the right, indicatingthat productivity increased for acquired firms after acquisition over the whole distributionof firm initial productivity. Figure 5 shows the distribution of productivity in thebase year and four years later for firms that remained domestic. While there is a slightincrease in productivity, it is much less pronounced than for foreign-acquired firms.00.10.20.3Density−5 0 5t + 4tln firm sales, domestic in t, foreign in t + 4kernel = epanechnikov, bandwidth = 0.700000.10.20.3Density−5 0 5t + 4tln firm sales, domestic in t, domestic in t + 4kernel = epanechnikov, bandwidth = 0.7000Figure 4. Distribution of Productivity for Acquired Firms, before and after the Foreign AcquisitionNotes: The dashed line shows the empirical probability density function (pdf) of initial productivity (measured bylog sales demeaned by industry) of firms that are domestic at time t but will become foreign owned by time t + 4.The bold line shows the empirical pdf of productivity of these firms at time t + 4 (i.e., after acquisition).Figure 5. Distribution of Productivity for Nonacquired Firms, Change over Four YearsNotes: The dashed line shows the empirical probability density function (pdf) of initial productivity (measured bylog sales demeaned by industry) of firms that are domestic at time t and are still domestic at time t + 4. The boldline shows the empirical pdf of productivity of these firms at time t + 4.3622 THE AMERICAN ECONOMIC REVIEW December 2012Table 8 shows the results of estimating equation (6) with our measures of productivityas the dependent variable. Column 1 in panels A and B (for each productivitymeasure) estimates equation (6) without firm fixed effects; columns 2, 3, 4, and 5progressively add firm fixed effects, industry trends, further selection controls, andlead and contemporaneous indicators of acquisition. Column 6 shows the propensityscore reweighting estimates.The point estimates are much larger in the cross-sectional estimation of column 1relative to any of the other columns that include fixed effects and better control forselection using the propensity score. This reflects the fact that the positive selectionidentified earlier will lead to substantial overestimation of the productivity advantagein cross-sectional analysis (as also demonstrated by Criscuolo and Martin2009)—by as much as three times in the case of labor productivity. However, wealso find that acquisition is significantly positively associated with increased productivity,controlling for selection. The propensity score reweighted specificationsTable 8—Foreign Ownership and Firm Productivityln sales(1a) (2a) (3a) (4a) (5a) (6a)Panel ALag foreign 2.042* 0.165* 0.120** 0.112* 0.0700* 0.182*
(0.161) (0.0621) (0.0599) (0.0582) (0.0421) (0.0540)
Forward foreign −0.0104
Observations 20,671 20,671 20,671 16,867 14,760 17,578
R2 0.169 0.100 0.147 0.275 0.284 0.130
p-value of test lag foreign = forward foreign 0.211
ln labor productivity
(1b) (2b) (3b) (4b) (5b) (6b)
Lag foreign 0.367* 0.126* 0.109 0.0877 0.109** 0.114**(0.0496) (0.0466) (0.0449) (0.0538) (0.0425) (0.0487)Foreign 0.0571(0.0390)Forward foreign −0.0218(0.0425)Observations 20,359 20,359 20,359 16,639 14,567 17,338R2 0.185 0.014 0.031 0.029 0.035 0.016p-value of test lag foreign = forward foreign 0.0119Firm FEs Yes Yes Yes Yes YesIndustry FEs YesIndustry trends Yes Yes YesSelection controls Yes YesPropensity score weighting YesNotes: Foreign is an indicator variable that equals one if the firm has at least 50 percent foreign ownership. ln salesis the natural logarithm of the firm’s real sales. ln labor productivity is the natural logarithm of real value addedper worker. Selection controls include lagged export status, lagged average wage, lagged log capital per employee,lagged log capital. All columns include year fixed effects. Standard errors are clustered by firm.*** Significant at the 1 percent level.** Significant at the 5 percent level.
Significant at the 10 percent level.VOL . 102 NO. 7 guada lupe et al.: innovati on and foreign ownership 3623in column 6 imply that, after acquisition, real sales increase by 18 percent and laborproductivity by 11 percent, on average.37Finally, we discuss the implications of our findings for the evolution of the distributionof productivity within industries. We show that foreign firms are more likelyto acquire the most productive firms within industries (Table 2), and that, uponacquisition, firms innovate (Tables 3 and 4), increasing productivity (Table 8). Thisset of results implies that acquisition activity can lead to an increase in the dispersionof the productivity distribution. This is an important consequence of our earlierfindings since it has implications for the evolution of within-industry productivitydistributions as more foreign firms enter an industry. Under this mechanism, foreignentry does not lead to productivity convergence, but, on the contrary, could leadto further divergence.38 Of course, there could be other reasons (such as spillovereffects or other externalities) why multinational entry may improve less productivefirms’ productivity. However, the direct effect of the foreign acquisition process isan increase in productivity heterogeneity.VI. ConclusionIn this paper, we use rich and detailed data on Spanish manufacturing firmsto establish that foreign firms acquire the best firms within industries (“cherrypicking”)but also invest more in a number of innovation activities upon acquisition.In particular, controlling for the selection effect, firms increase their processinnovation, with the simultaneous introduction of new machines and organizationalpractices. Acquired firms that export through their parent firm also report that theyincrease their product innovation and start to assimilate more foreign technologies.We develop a simple model that illustrates how these two facts can be fundamentallyrelated. The model relies on standard assumptions about production, firmheterogeneity, consumer demand and market competition (Helpman and Krugman1985; Melitz 2003) and incorporates two well-known characteristics of multinationalfirms: multinationals grant access to larger markets and/or have lower technology-implementation costs. Since the incentives for innovation and acquisitionare increasing in initial productivity, the surplus created by the acquisition is alsoincreasing in initial productivity. Therefore, foreign firms find it more profitable toacquire the most productive firms and to innovate more on acquisition.The observed positive selection suggests that there are complementarities betweeninnovation activity and the initial characteristics of the acquired firm. Our results alsosuggest a complementarity between market access and innovation. Taken together,these findings can explain a number of important facts: first, why more productive37 Unlike the measures of innovation activity, the productivity measures are based on reported revenues. Theremay be incentives to change how revenues are reported within a multinational by adjusting transfer prices, affectingdomestic firms once acquired. For example, reported revenues could reflect removal of double marginalization uponintegration. This effect could lead to a decline in revenues, but this is not present in the data. The multinational mayalso face incentives to misreport the location of profits for tax purposes. We expect this problem to be small, givenrelative Spanish tax rates.38 If multinational entry also serves to raise the threshold level of productivity at which firms exit the domesticmarket (as in Helpman, Melitz, and Yeaple 2004), the lowest-productivity surviving firm in the distribution willhave a higher productivity level. This general equilibrium effect will serve to offset the increase in dispersiondescribed above.3624 THE AMERICAN ECONOMIC REVIEW December 2012firms innovate more; second, why foreign firms acquire the most productive firmswithin industries; and third, why foreign-owned firms increase their innovation uponacquisition.39 Our contribution is to illustrate the drivers of the innovation process andto highlight that superior or proprietary technologies from the parent firm are not necessaryto generate the prediction that a given firm finds it optimal to invest more underforeign control than under domestic control.In addition, the observed complementarity between market scale and innovationoffers one explanation for why all firms do not imitate the practices of high productivityfirms in the market and why productivity differences persist. To the best of our knowledge,we are the first to link market scale to the jointly determined acquisition outcomesand innovation incentives. Finally, while we focus on the multinational firm’s acquisitionchoice, the economic mechanism we emphasize should also be relevant for purelydomestic integration decisions when the acquirer facilitates access to larger markets.
Let’s block ads! (Why?)