Foreign acquisition and the performance of New Zealand firms

This paper examines the firm-level determinants of foreign acquisitions of New Zealand companies, and the consequences for both purchased firms and the workers within those firms. We follow a combined propensity score matching and difference-in-difference approach to identify and address endogenous selection of acquisition targets. The results suggest that foreign firms tend to target high-performing New Zealand companies. Acquired firms then exhibit higher growth in average wages and output, relative to similar domestic firms, but do not appear in general to increase their productivity or capital intensity. We find no evidence of differential survival rates for recently acquired foreign firms.


Introduction
Despite its geographic isolation, New Zealand is ranked ninth in the OECD in terms of inward foreign direct investment (FDI) as a percentage of gross domestic product (GDP) (OECD 2010). While a substantial body of international literature shows that foreign-owned firms outperform local firms on a wide range of metrics, and many countries have explicit policies designed to attract foreign investment, public opinion on the value of FDI to the New Zealand economy is divided. Media accounts regularly draw attention to negative aspects of foreign investment, focusing on stories of downsizing by foreign owners, bemoaning the loss of promising New Zealand companies and technologies, and emphasising public fears of a loss of control of New Zealand's natural resources to offshore owners. Anecdotally, however, the owners and directors of New Zealand-based firms extol the benefits of foreign investment, including not only improved access to capital but also access to the new owners' stock of technology, networks and management experience (Simmons 2002). *Corresponding authors. Email: richard.fabling@motu.org.nz; lynda.sanderson@treasury.govt.nz Access to the data used in this study was provided by Statistics New Zealand in accordance with security and confidentiality provisions of the Statistics Act 1975 and the Tax Administration Act 1994. The results in this paper have been confidentialised to protect individual businesses from identification. See an earlier version of this paper, Fabling and Sanderson (2011), for the full disclaimer.
This paper provides an empirical analysis of cross-border mergers and acquisitions (M&A) using data for the population of companies operating in New Zealand between 2000 and 2009. We examine the firm-level factors that attract foreign investors and the performance of firms after foreign acquisition. We build on the existing literature by considering a range of performance measures including outcomes both for acquired firms and for the workers within those firms. In this way, our paper is similar to that of Arnold and Javorcik (2009) who consider post-acquisition performance of Indonesian manufacturing plants and find evidence of both positive selection and post-acquisition performance gains. We provide a small, distant, developed country perspective on this question, while also extending the analysis to include non-manufacturers and by allowing for the effects of foreign acquisitions to differ according to the characteristics of the target firm.
By including a range of outcome measures we consider not only whether firm outcomes improve following foreign acquisition, but also whether the benefits of any performance improvements are shared with workers or, conversely, come at the expense of the local workforce. Firm performance outcomes considered include labour and multi-factor productivity, capital intensity, and gross output. Worker outcomes include total employment and average wages. Finally, we consider whether survival rates differ between domestic and recently acquired firms.
We find that acquired New Zealand firms tend to be larger, pay higher wages, and have higher capital intensity and labour productivity than other domestic firms. Although recently acquired firms appear to increase both average wages and gross output compared with firms that remain in domestic ownership, there is no evidence to suggest that acquired firms increase either their labour or multi-factor productivity performance.
Splitting the sample into initially high-and low-performance firms indicates limited heterogeneity in post-acquisition outcomes. Outcomes for each sub-population broadly mirror the aggregate outcomes outlined above, but suggest that aggregate increases in average wage and gross output may be unevenly distributed, with stronger and more precisely estimated increases in gross output among acquisitions with relatively low initial performance and wage growth concentrated among smaller and more productive acquisition targets. We also see weak evidence of post-acquisition capital deepening, but only among the subsample of initially lower capital intensity acquisitions.
In the next section we outline the existing literature on FDI, focusing on cross-border M&A. Section 3 describes the data and empirical methodology, while Section 4 presents the results. Section 5 summarises the findings and suggests avenues for further work.

Literature review
Theoretical and empirical studies of the determinants of FDI abound, reflecting the complexity of real world investment decisions. The decisions ultimately taken by heterogeneous firms reflect a confluence of firm, industry, and country characteristics. Understanding the factors that drive cross-border acquisitions motivates the choice of explanatory variables in our selection model. We then discuss the literature on plant-and firm-level outcomes following foreign acquisition. 1 A wide range of empirical literature shows that foreign affiliates outperform domestic firms (e.g. Bellak 2004;Greenaway & Kneller 2007). Comparisons between foreignowned firms, locally-owned multinationals and purely domestic firms suggest that the difference is driven by the performance gap between domestic and multi-national enterprises, rather than a 'foreign premium' per se (Bellak 2004;Criscuolo & Martin 2009). It is commonly asserted that in order for foreign firms to be competitive they must have some firm-specific advantages -such as proprietary brands or product lines, high performance production processes or managerial expertise -to compensate for the market-specific knowledge and networks of their local competitors and the additional costs of doing business abroad (Dunning & Lundan 2008;Markusen 1995;Melitz 2003). Empirical evidence suggests a productivity hierarchy in which only the highest productivity firms engage in outward direct investment, while less productive firms export, and the least productive firms retain a purely domestic focus (Greenaway & Kneller 2007).
However, positive self-selection into FDI does not necessarily imply positive selection of acquisition targets. Harris and Robinson (2002) contrast two theories of M&A: managerial discipline and operating efficiency. The theory of managerial discipline suggests that M&A are a form of natural selection, in which inefficient plants are bought out by new owners and undergo some form of managerial change or restructuring to improve their efficiency. In contrast, the operating efficiency theory suggests that M&A occur when the acquiring firm sees a complementarity between its existing operations and those of the target plant. In this case, the acquiring firm will be more likely to target high-performing plants. Post-acquisition performance may decline if there is difficulty assimilating the new plants into the firm's existing operations, or increase if the new parent introduces complementary assets or processes. Harris and Robinson (2002) find support for the operating efficiency hypothesis for cross-border acquisitions, with foreign firms tending to target 'good' plants, but note significant differences between industries and across acquiring firms from different countries. Guadalupe, Kuzmina, and Thomas (2010) take a similar approach, attributing observed positive selection on performance to an imperfect ability on the part of owners to transfer managerial or production technologies to the local subsidiary.
While relative firm-level performance provides one lens with which to view crossborder M&A decisions, industry-, country-, and other firm-specific factors have also been shown to influence the FDI decision. Dunning and Lundan (2008) provide a comprehensive review of the main motivations for offshore investment, using the Ownership-Location-Internalisation (OLI) or 'eclectic' framework (Dunning 1977). Under this framework, the decision to invest abroad is driven by a combination of the firm-specific attributes of the investing firm, which provides it with an advantage over local competitors such as managerial or technological capabilities, reputation and brand ownership (ownership advantages); country-specific motivations such as lower production costs, access to protected markets or favourable tax treatments (location advantages); and the benefits to ownership and internalisation over outsourcing and market-based transactions such as minimising spillovers of proprietory technology or methods to local firms, reducing transactions and contracting costs, and allowing the acquiring firm greater control of management processes and quality control (internalisation advantages).
Empirical support for the concept of ownership and internalisation advantages comes from observed patterns of the industry distribution of investment. Multinationals tend to be more prevalent in industries and firms where intangible assets are important. This includes industries with high levels of product differentiation and advertising, products that are new or technically complex, high R&D intensity, and high shares of professional and technical workers (Markusen 1995). Intangible assets are likely to encourage FDI because of their non-rival nature, 2 which allows firms to duplicate production in several locations. As many intangible assets are only semi-excludable 3 and, in the case of brand names and reputation, there is potential for degradation of the asset, firms with these assets may be less willing to undertake arms-length market transactions. Dunning and Lundan (2008) identify four main types of offshore investment, which they classify as resource-seeking, market-seeking, efficiency-seeking and strategic asset-or capability-seeking. 4 Resource-seeking investments are those designed to access the specific location advantages available in the target country -physical resources such as primary products or manufactured inputs, low-cost labour, or proximity to technological, management or marketing expertise (e.g. research 'listening posts' in advanced countries). Market-seeking investment is designed to increase the firm's reach, providing it with better knowledge about local tastes or the ability to reduce production, transport or transactions costs through proximity. Efficiency-seeking investment allows firms to take advantage of economies of scale and scope and to benefit from risk diversification by concentrating production in a limited number of locations to supply multiple markets while taking advantage of differing factor costs and local supply capabilities. Finally, capability-seeking FDI generally involves M&A with existing firms, as acquirers seek to access specific competitive advantages held by those firms such as technology, market power and distribution channels, or to create R&D synergies or production economies through streamlining and sharing facilities and knowledge.
Given the diversity of FDI motivations, the selection mechanisms and consequences of cross-border M&A remain very much an empirical question. As noted by Harris and Robinson (2002), negative selection on performance may occur if offshore acquirers identify firms that are underperforming and invest with the intention of improving their performance, while positive selection is likely if acquirers seek to integrate the target into their own production system. Meanwhile, offshore owners seeking technological advantages may be inclined towards purchasing small firms, while those seeking marketing networks may target larger organisations (Grimpe & Hussinger 2008).
The summary above suggests that the selection model for target firms should control for measures of firm size and performance, but does not give a clear prediction for the sign of coefficients, suggesting flexible functional forms may be desirable.
In turn, post-acquisition outcomes for both performance and labour markets may depend on the motivation for acquisition. If managerial discipline is considered an issue, acquisition may be followed by a period of restructuring, leading to job losses or a change in the focus of the target firm. In contrast, if foreign owners provide access to new sales opportunities and networks, output and employment may increase.
In keeping with this ambiguity, empirical results have been mixed. The literature examines a range of different outcome metrics, including those related to productivity performance (Arnold & Javorcik 2009), labour market impacts (Almeida 2007), innovation and R&D behaviour (Bertrand 2009), and effects on plant survival (Bandick & G€ org 2010). As well as considering the average effect of all FDI, various authors have also considered differences according to the origin country of the acquiring firm (Chen 2011), whether the acquisition was horizontal or vertical (Conyon, Girma, Thompson, & Wright 2002), and the characteristics of the target firm (Girma 2005).
Broadly speaking, the empirical literature suggests that most FDI is positively selected -that is, that target firms tend to be larger, more productive, and to pay higher wages than firms which remain under domestic ownership (e.g. Guadalupe, Kuzmina, & Thomas 2010;Heyman, Sj€ oholm, & Tingvall 2007). This finding is not unanimous -for example, Conyon et al. (2002) find that foreign acquisitions of UK firms target smaller firms and those with relatively low productivity -but suggests that, in general, at least part of the observed higher performance of foreign-owned firms can be attributed to selection of already-successful targets.
Empirical studies of post-acquisition effects show little consensus. Some studies suggest that this may reflect differences in the characteristics of the acquiring and acquired firms or plants. Arnold and Javorcik (2009) provide a comprehensive study of a outcomes for Indonesian manufacturers following foreign acquisition. Their results indicate both positive selection of acquisition targets and improvements in performance following acquisition, with productivity, output, employment, investment and average wages all increasing relative to non-acquired plants. They also note that foreign acquisition appears to improve connections to the international economy, with increases in both exports and imported inputs. In contrast, using data from the UK, Harris and Robinson (2002) find negative effects on productivity, at least in the first few years after acquisition, which they attribute to difficulties assimilating these plants into the broader organisation.
On the labour market side, Huttunen (2007) finds that foreign acquisition of Finnish plants leads to increases in wage rates for both high-and low-skill worker groups, but as this is accompanied by a fall in the share of highly-skilled workers in employment it does not necessarily translate to an increase in average wage rates. Heyman et al. (2007) find small positive effects on average wages in Swedish target firms but, at the individual level, workers who remain with the newly acquired firm show lower wage growth than those in similar, non-acquired firms.
New Zealand empirical evidence on motivations for foreign acquisitions and the impact on domestic firms is limited. Based on a survey of 516 foreign-owned companies, Scott-Kennel (2004) concludes that market-seeking investment is the dominant motivation for investing in New Zealand, although many foreign-owned firms are also involved in exporting and R&D. This survey also suggests a number of mechanisms through which foreign parents may raise the performance of their local affiliates, including technical assistance, staff training, and provision of information about markets, suppliers and contacts. Cartwright (2001) examines the motivations and activities of foreign multinational enterprises (MNEs) in New Zealand, focusing on the industry groups that account for the greatest turnover. He identifies three main types of foreign investors in New Zealand. The majority of large foreign firms operating in New Zealand seem to be focused on supplying the domestic market, providing distribution and marketing services for their foreign parent. A second group of MNEs are resource-based producers (e.g. forestry and wood-processing, food processing), set up to access the physical resources available in New Zealand. Finally, Cartwright (2001) notes an emerging tendency for foreign firms to target relatively small New Zealand companies with sophisticated capabilities in areas such as electronics, IT and engineering. These abilities may then either be fostered onshore, in conjunction with the MNE network, or the intellectual capital relocated offshore.
In a small scale longitudinal study of successful manufacturing exporters, Gawith (2002) finds that the outcomes that foreign acquired firms experience differ dramatically and depend heavily on the motivation of the acquirer. Among recent acquisitions, there has been a tendency to target firms that can provide strategic assets or capabilities, such as patents and R&D ability. In some cases, the acquired firms expect to see a decline in manufacturing output, as they focus on providing R&D and product development services for the overseas owner. Further, while some New Zealand firms saw foreign acquisition as a means to international expansion, providing access to existing distribution and marketing channels, some later reported that fitting into the new parent companies' networks meant they lost control over their distribution channels and missed out on knowledge of their markets and customers.
Overall, New Zealand research echoes the theoretical ambiguity discussed abovewhile foreign acquisitions generate potential for positive effects on domestic firms, these positive outcomes are not guaranteed and depend heavily on the motivation of the foreign parent.

Empirical strategy
In order to examine the selection and outcomes of foreign acquisitions in New Zealand, we use the prototype Longitudinal Business Database (LBD), a collection of administrative and survey data held by Statistics New Zealand. 5 The LBD is based around the Longitudinal Business Frame, which provides basic information on all economically significant firms in New Zealand from 1999/00 to 2008/09. 6 Foreign ownership is identified through firm responses to the disclosure statement 'Is the company controlled or owned by non-residents?' from annual company tax returns. This limits the population to a total of almost 274,000 limited liability companies over a 10-year period. 7 Table 1 shows the overall level of foreign ownership, employment in foreign firms, and the transition rate from domestic to foreign ownership by industry over the period 2000-2009. Foreign ownership rates differ substantially across industries, with relatively high rates in Mining (B) and Wholesale Trade (F), and low rates in Agriculture, Forestry and Fishing (A), Construction (E), Retail Trade (G) and Personal and Other Services (Q). Across all industries, the tendency for foreign firms to be larger than domestically-owned firms shows through clearly, with foreign ownership associated with larger shares of employment than of firm numbers. Table 2 shows the share of foreign firms, and employment in foreign firms, by year. Falling foreign-ownership rates between 2000 and 2009 reflect stronger growth in the population of domestic firms than foreign-owned firms, through both entry of new firms and corporatisation of existing firms.
We consider six firm outcomes that may be affected by foreign investment. Firm performance measures (labour and multifactor productivity, gross output, and the capital-labour ratio) are derived from Inland Revenue Department and Annual Enterprise Survey data following the method in Fabling and Mar e (forthcoming). 8 Average wages (for employees) and total employment (including working proprietors) are sourced from the Linked Employer-Employee Dataset (LEED), which is based on pay-as-you-earn (PAYE) tax data and other tax records. All outcome variables are expressed as deviations from the industry-year mean and firms are assigned to permanent industries based on employment shares. 9 A full list of the variables used, their definitions and summary statistics for the population included in the selection model can be found in Appendix A. Finally, we consider whether foreign acquisition affects survival, defined as continuing employment in the years following acquisition.
In keeping with the international literature, foreign-owned firms in New Zealand outperform domestic firms on almost all firm outcomes ( Figure 1). They are larger (in terms of both output and employment), more capital intensive, pay higher average wages, and have higher labour productivity. However, Figure 2 suggests that at least part of this difference is due to positive selection of FDI targets. Dividing the population of domestically-owned firms according to their future ownership status -whether or not they will be acquired by a foreign owner in the following year -suggests that pre-acquisition firm characteristics more closely mirror the patterns for foreign-owned firms shown in Figure 1 than those of other non-acquired domestic firms. That is, foreign owners seem to 'cherry pick' high performing firms.
To examine the firm-level factors influencing acquisition and subsequent performance, we follow the recent literature and use a combined difference-in-difference and propensity score matching (PSM) approach. This methodology draws heavily on the programme evaluation literature (e.g. Imbens & Wooldridge 2008) and considers foreign acquisition as a 'treatment'. The basic principle of PSM is that as long as there are no unobserved characteristics that are associated with both the potential outcome and the probability of treatment ('unconfoundedness') and suitable control cases can be found for each treated case ('overlap'), conditioning on the propensity score is sufficient to remove the bias associated with differences in pre-treatment characteristics between the treated and untreated groups. Thus, all systematic differences in outcomes between the treated and control firms are attributable to the treatment. Our implementation of this method closely follows Fabling and Sanderson (2013), who examine New Zealand exporter performance.
The PSM methodology involves two steps. The first step is to establish a suitable control group. A probit model is estimated to determine which pre-acquisition characteristics predict foreign acquisition. The predicted probability of acquisition (propensity score) is calculated and each treated firm is matched to one or more firms which have similar probability of acquisition. 10 We match within industry and preclude self-matches (matching a treated firm to itself in a previous year), using radius matching with a caliper of 0.001. Balancing tests are then carried out to ensure that the treated sample and the matched control group are sufficiently similar on all observed characteristics.
In the second stage, outcomes are compared between treated and control firms. To mitigate any remaining unobserved time-invariant differences between the two groups we follow a difference-in-difference approach, comparing outcomes in terms of changes relative to the pre-acquisition year. Finally, standard errors are estimated by bootstrapping across both stages to account for uncertainty in the matching equation. Figure 3 sets out the timeline for the analysis. Firms are tracked over a six year period. 11 The first two years (t ¼ À1; 0) provide the explanatory variables for the   propensity score matching. t ¼ 1 is the treatment year. We provide estimates of the average treatment effect for the treatment year and the following three years, out to t ¼ 4.
The selection (probit) model is estimated across all firms that can provide the control and treatment variables, and the outcome analysis uses all firms that additionally have a full set of outcome variables for the relevant year. The variables used in the probit equation reflect firm-level factors that may attract prospective foreign buyers. Initial assessments of functional form showed up substantially non-linear relationships between firm performance variables and the probability of acquisition, consistent with the theoretical literature. We therefore include performance variables -log total employment, the capital-labour ratio, labour productivity, average wages and multifactor productivity -as four piecewise linear segments, allowing for both intercepts and slopes to differ across quartiles. This specification reflects a compromise between two objectives -a parsimonious specification of the relationship between firm performance and acquisition, and sufficient flexibility to generate a quality (high R 2 and balanced) match between treated and control firms. 12 In addition to quantiles of performance levels, we include variables capturing recent input dynamics: one-year growth rates of total employment, average wages, and the capital-labour ratio, alongside dummies for firms with missing values of the relevant variable in the previous year (either because the firm was non-employing or because of missing capital data in the previous year). Including dynamics in the selection model (and requiring that matched and control firms balance on this dimension) reduces the potential of matching treated firms to control firms with similar current performance levels but which are on a different growth trajectory, which would confound interpretation of post-acquisition growth differences.
We include a dummy for whether the firm exported in year t ¼ 0, and the intensity of exporting, defined as the share of sales that are exempt from goods and services tax (GST) as a share of total sales. 13 Export activity may affect both the perceived performance of the firm but also its international visibility (e.g. foreign companies may be more likely to notice firms that are already trading in their existing markets). We also include a dummy indicating subsidiary firms of domestic parent companies.
A full set of regional council dummies is included as some geographic locations may be more attractive to foreign owners than others (e.g. cities with an international airport will be more accessible for foreign executives). Finally, we include a full set of year and industry dummies to capture differences over time and across industries.
Balancing tests are performed to ensure that the matching procedure is sufficient to provide a suitable group of control firms against which we can benchmark the post-acquisition performance of the treated group, at least with respect to observable characteristics. There are no significant differences in the mean of each of the outcome and matching variables between the treated firms and the matched controls in either the unbalanced or the balanced panels reported in Table 3. 14 In addition to the full population, we separately consider sub-populations of firms based on their pre-acquisition levels of MFP, employment, capital intensity and average wages. These subsets are chosen because the acquisition of different types of firm may be driven by different motivations on the part of the foreign owner and may therefore exhibit different post-acquisition performance trajectories. We maintain the same matched sample used in the full population across all four splits, and divide the population of treated firms at the median value of the relevant performance measure, so as to have approximately equal numbers of treated firms in each sub-population. 15 Table 3 reports selection equation results for the full population. Panel A gives the estimated coefficients of the five performance variables that are included in piece-wise linear form. Panel B provides the coefficients for the remainder of the matching variables. Significant at: * 10%, ** 5%, *** 1%. F-tests of joint significance passed at 10% level for each piece-wise linear performance metric. Regression also includes a full set of (primarily) two-digit industry and year dummies, and dummies controlling for missing data in the pre-treatment period (unreported).

Results
The inclusion of multiple measures of firm performance makes it difficult to separately interpret the coefficients (e.g. the average wage level captures elements of skill composition that are correlated with measured MFP and labour productivity). Overall, Panel A suggests positive selection of target firms, although often only the top quantile of performance has a coefficient significantly different from the lowest quantile. Exploration of alternative specifications of the model show that the inclusion of multiple correlated measures is indeed affecting the estimated coefficients, 16 but the core patterns remain robust and reiterate the broader impression gained from Figure 1 -foreign acquisition targets appear to be strongly positively selected on almost all performance metrics. The exception is MFP, where further examination suggests a bi-modal impact -after controlling for other observable firm characteristics, foreign targets tend to be concentrated in both extremes of the MFP distribution, perhaps suggesting a role for managerial discipline alongside selection based on operating efficiency (Harris & Robinson 2002).
Foreign acquisition is also positively associated with lagged employment growth, export status and export intensity. Firms are much more likely to be acquired by foreign owners if they are subsidiaries of an existing enterprise group, rather than independent enterprises. Finally, there appears to be a regional element to target selection, with firms that have locations in Auckland, Wellington and Hawkes Bay being more likely to be targeted for acquisition, and firms on the West Coast less likely.
Post-acquisition performance comparisons between acquired firms and matched domestically-owned firms are reported in Table 4. The upper table presents results using the unbalanced panel, in which firms are included in any year for which they have a full set of outcome variables. The lower table restricts attention to firms that have a full set of outcome variables in all four outcome years. Differences between the two populations reflect a combination of firm survival, data availability and right-censoring of outcome years.
Acquired firms exhibit a gradual increase in gross output, relative to similar domestic firms, which becomes statistically significant by t ¼ 3. This increase in output appears to be achieved via relatively strong employment growth, rather than increases in either capital investment or multi-factor productivity. 17 Acquired firms also increase average wages between two and eight percentage points more than similar non-acquired firms. Stronger effects are observed among the balanced panel consistent with, for example, foreign owners investing more in skills or management capabilities where they anticipate a longterm involvement. However, we see no evidence for differential survival rates between domestic and acquired firms (Table 5).
These results present an apparent paradox -that foreign ownership appears to lead to increases in the average wage without any associated increase in labour productivity. Girma and G€ org (2007) discuss a number of possible reasons why foreign firms might choose to pay higher wages for a given level of productivity, including a desire on the part of the owners to minimise labour turnover in order to prevent spillovers of firmspecific information or technologies, or to reduce industrial disputes, particularly during periods of substantial change in working conditions or practices. Foreign owners may also be more likely to source workers and managers from abroad, and these employees may demand higher wages than the New Zealand staff.
Alternatively, there may be issues with the measurement of labour productivity in the case of foreign acquisitions. If foreign owners are able to reduce their taxable profits through the use of transfer pricing (e.g. through reporting higher costs of inputs sourced from the parent company or lower values of outputs going to the parent company) this would reduce measured value added. If the personal income tax system provides fewer opportunities for tax-shifting, this may lead to lower-than-expected productivity, as firms correctly report labour inputs but underestimate value added. Finally, we follow firms over a relatively short time frame and observe a relatively small number of transitions into foreign ownership. If firms experience transition costs in moving to a new management structure, or if value added is more volatile than labour inputs over time, it may be Table 4 Difference-in-difference estimates of average treatment effect.  Difference-in-difference (DID) estimator, from t ¼ 0 to outcome year, applied to matched sample. Radius matching (caliper 0.001, with replacement) with observations pooled across years and matched within two-digit industry (precluding self-matches). Bootstrapped standard errors in brackets (significant at * 10%; ** 5%; *** 1%). Bootstrapping encompasses both probit and DID stages (100 repetitions) and is stratified on treatment and the existence of future outcomes to maintain approximately constant (weighted) population size (N ) across estimates. The table also reports the number of treated firms and the proportion of treated (control) firms dropped because there is no control (treated) firm within the caliper distance. All balancing tests (equivalence of weighted means of matching variables across treated and controls) passed at the 10% level. that a larger dataset over a longer time frame might provide more evidence for changes in firm performance. Differences in acquisition motives or in acquired firm characteristics may alter postacquisition outcomes. For example, if firms are targeted due to perceived underperformance relative to their peers we might expect to see improvements in productivity following acquisition. Similarly, it may be that high-performing firms suffer from dislocation following acquisition and take time to return to normal or, alternatively, that these firms experience large inflows of investment from the new parents that allows them to expand. If there are significant differences between groups, these may average out in aggregate results. Therefore, Table 6 presents outcome comparisons for eight sub-populations, according to whether treated firms are in the top or bottom half of acquired firms in terms of MFP, employment, average wage or capital intensity.
In general, each sub-population mirrors the aggregate picture of higher subsequent growth in average wages, output, and employment in acquired firms. Differences in estimated treatment effects across sub-populations imply some heterogeneity in outcomes. Specifically, post-acquisition growth in average wages is concentrated in smaller firms and those with relatively high initial MFP. Increases in output are concentrated among initially smaller and lower productivity targets, while employment growth is observed primarily among initially high-wage firms. However, few sub-populations display significantly different outcomes between high and low groups. There is only one area in which the sub-population results imply a significant deviation from the broad results found for the full population -post-acquisition capital deepening is observed only in those firms that were relatively capital-shallow prior to foreign acquisition.

Conclusion
This paper examines the impact of foreign direct investment on firm performance and worker outcomes in recently acquired firms. Following recent literature we use combined propensity score matching and difference-in-difference estimation to control for selection effects in acquisition. We find that the main factor underlying observed performance premia for foreign-owned firms in New Zealand is related to positive target selection. Foreign acquisition targets tend to be firms that were already larger, more productive, and more likely to be exporting than their competitors.  Table 4 for notes. Population includes all firms with full set of matching variables at t ¼ 0. Outcome variable is a dummy equal to one if the firm has non-zero employment (including working proprietors) in the relevant outcome year. Balancing tests passed at the 10% level except for the capital-labour ratio in t ¼ 4. Table 6 Difference-in-difference estimates for firms initially above and below median performance for acquired firms.
High MFP (a) Low MFP  We also find limited evidence of positive post-acquisition effects on performance, with recently acquired firms exhibiting stronger growth in average wages, output, and employment than might otherwise be expected. We find no evidence of increased closures in acquired firms. However, these positive effects do not extend to productivity growth, one area where we might have most naturally expected to see benefits associated with foreign investment. Tentative evidence of post-acquisition capital deepening is limited to target firms that were initially relatively capital-shallow (for an acquired firm).

High average wage (b)
Low average wage (c)   Table 4 for notes. Estimated treatment effects for high and low subsamples significantly different from each other at the (y) 5%; (z) 10% level. Balancing tests passed at the 10% level except for specifications as indicated in the table:(a) capital-labour ratio (t ¼ 3; 4); (b) 2005 dummy (t ¼ 2), Canterbury dummy (t ¼ 3); (c) log gross output (t ¼ 1), Auckland dummy (t ¼ 1; 2), Wellington dummy (t ¼ 1), dummies indicating missing lagged variables (t ¼ 3). One firm outcome that has not been considered in the current study is the role of foreign ownership in firm-level export performance. In a survey of firms, Simmons (2002) finds that a key reason why domestic firms pursue foreign investment is in order to access the offshore distribution networks controlled by their new owners. Thus, foreign ownership may lead to expansion of export markets. Future work could restrict the sample to manufacturing firms and focus on the development of further exporting capability, as evidenced by both the value and volume of exports, and by firm-level entry into new markets and products.
14. Based on a two-sided test with significance level of 10%. 15. Minor balancing issues arise for some sub-populations as indicated in table notes. 16. As noted earlier, matching models excluding one or more of these outcomes fail to balance. 17. In robustness tests we exclude control firms that are treated in future years, up to and including the outcome year. These estimates generate qualitatively similar results, with the main difference being that employment growth estimates are statistically significant in t ¼ 3; 4.