Firm performance following foreign acquisitions in Norway: Evidence of profit shifting?**
Article Category: Article
Online veröffentlicht: 22. Apr. 2024
Seitenbereich: 1 - 21
Eingereicht: 18. Mai 2020
Akzeptiert: 26. Sept. 2022
DOI: https://doi.org/10.2478/ntaxj-2023-0001
Schlüsselwörter
© 2023 J. Tropina Bakke, published by Sciendo
This work is licensed under the Creative Commons Attribution 4.0 International License.
Multinational corporations can minimize their tax payments by shifting profits between affiliates in different countries. The incentives and opportunities for profit shifting arise because of corporate tax differences between the countries and mismatches in international tax law. The result is corporate tax revenue losses for high-tax countries and an uneven playing field between multinational and national firms.
The problem of tax-motivated profit shifting—also referred to as base erosion and profit shifting (BEPS)— has received a lot of attention in recent decades. While tax authorities and journalists uncover individual cases of tax planning regularly,1 providing empirical evidence of profit shifting on a larger scale is challenging. Despite the increasing number of papers in this field, there is still a lot of uncertainty concerning the magnitude of the problem. The OECD (2015) has estimated the size of the corporate tax revenue losses due to profit shifting globally to be between USD 100 and 240 billion or four to ten percent of global corporate tax revenues, while new evidence from Tørsløv et al. (2021) indicates that as much as 40% of the foreign multinational profits (around USD 600 billion) is being shifted to tax havens.
The complexity of the problem and data limitations are the factors that make the analysis of BEPS difficult. Since no data source or method alone can provide a definite answer, using different methods and data sources to analyze companies from different countries and different periods of time is essential to obtain a better insight into the scope of the BEPS problem.
This paper contributes to existing empirical literature on tax motivated profit shifting by looking for evidence of BEPS among Norwegian multinational companies. It uses firm-level accounting data to analyze changes in the profitability of the Norwegian firms that have been acquired by foreign investors as compared with the profitability of similar national non-acquired firms. As profit-shifting activities impact profitability directly, the idea is to see whether the observed changes in profitability are consistent with the profit shifting activities of the firms in Norway. Since only foreign acquired firms have the incentives and opportunity to engage in tax-motivated profit shifting activities, the underlying assumption is that the observed differences in profitability development between these two types of firms can at least partially be attributed to profit shifting activities.
The results in this paper show lower profitability growth for the foreign-acquired firms following acquisition as compared with similar domestic firms, which is consistent with profit shifting behavior by the acquired firms.
These results rest on two important assumptions. The first one concerns accounting for selection into foreign ownership. Selection into foreign ownership means that firms that are being acquired may differ systematically in terms of both observable and unobservable factors from those that are not. These factors may also affect profitability differences, making it more challenging to attribute the observed differences in profitability to foreign acquisition. Accounting for this selection is the main underlying assumption for the identification strategy used in this paper. I have used propensity score matching to control for selection on observables when comparing the profitability of foreign acquired firms and similar national firms in Norway. To account for the possibility that selection bias can also be due to time invariant unobservable factors, propensity score matching is combined with the difference-in-difference estimator, as proposed by Blundell and Costa-Dias (2002).
Second, it is assumed that the changes in profitability for the acquired firms following foreign acquisition are due to profit shifting. In order to find more support for this evidence, I analyze whether the factors that drive the profitability differences are related to profit shifting and look closer at the changes in the real economic performance of the firms acquired. I find that the main driver of the lower profitability is a significant increase in the share of material costs, which is one of the most interesting items from a transfer pricing perspective. The results indicate further that the observed change in profitability is not associated with similar changes in other economic performance measures (as measured by total assets turnover, the current ratio, and interest coverage), which also supports the hypothesis that profit shifting may be driving the changes in profitability we observe. Providing this type of evidence by linking the profitability differences closer to profit shifting is one of the most important contributions of this paper.
This paper uses data for the period 1994-2005, which means that it describes the situation in Norway prior to many of the important changes in national and international legislation aiming to combat tax-motivated profit shifting being introduced. For example, stricter transfer price regulations were introduced in Norway in 2008, three years after the last year with data in this study. The thin capitalization rules were introduced in 2014 and then extended and tightened in 2019. The country-by-country reporting came into force in 2016. All of these measures were meant to limit the profit shifting activities of the multinational firms that this paper aims to capture. Thus, the results of this paper can serve as a benchmark for further analysis of the effectiveness of the new rules aiming to combat profit shifting behaviors of the multinational firms. If the profitability differentials documented here are due to the profit shifting activities of the multinational firms, and the measures introduced have been effective, one would expect that the replication of the same type of analysis of the data for later years would show that the profitability differential between foreign acquired and national firms has been reduced or evened out. This is confirmed by Hopland et al. (2019), who extend the analysis in this current paper in several respects and look at a longer data panel of Norwegian firms from 1993–2012. They show that the changes in profitability of the firms that became multinational became much less pronounced after the introduction of stricter transfer pricing regulations in 2007/2008, which supports the claim that the profitability differences between national and multinational firms we observe are due to profit shifting activities. This result is also consistent with the results of other empirical studies of profit shifting using different approaches that indicate that the extent of profit shifting seems to decrease over time—e.g., Lohse and Riedel (2013).
The rest of the paper is organized as follows. An overview of the existing literature is presented in Section 2. Section 3 describes the data sources, the sample, and the variables used in the analysis. Section 4 presents estimation method and the main estimation results. Section 5 concludes by summarizing the main results.
The question of whether profit shifting is a systematic and extensive problem is interesting and important from a policy perspective, but, as stated above, providing empirical evidence of profit shifting is challenging for tax administrations and researchers alike. The main reason for this is that it is difficult to know whether the financial numbers we observe are a result of real economic operations or tax planning activities on the part of the multinational firms. Without a thorough understanding of the value creation process in each of the corporations and insight into details of the internal transactions of each particular firm, it is thus practically impossible to provide direct evidence of profit shifting.
Despite the fact that we do not have a clear answer to the question of the magnitude of profit shifting, an extensive strand of empirical papers on profit shifting, using different data sources and methods, have contributed greatly to the understanding of the challenges of profit shifting, data limitations, and the approximate size of the problem.
Several of these papers give an excellent summary of the papers in this field, e.g., OECD (2015); Riedel (2018), and Dharmapala (2014, 2019). They provide a good overview of the studies, explain the divergence in the results in different studies, and discuss potential weaknesses and strengths of different approaches and possibilities for new research. They also show that there has been a great development in the literature with researchers testing different techniques, data sources, and identification strategies, thereby providing a better understanding of the profit shifting problem.
The general trend is that all the empirical papers in the field report evidence consistent with profit shifting, while the magnitude estimates vary among studies depending on the method/data sources used (e.g., earlier studies using macro-level data report higher estimates than later studies using firm-level data, which can be explained by the fact that firm-level data allow for controlling for firm and industry fixed effects), and they depend on the period studied (with more recent studies documenting smaller effects indicating the effectiveness of the BEPS counteractive measures).
To place this current paper in the larger framework of the extensive existing literature requires an explanation of the distinction between the empirical studies following the direct and indirect approaches. Studies following the direct approach normally focus on one profit shifting channel at a time, of which transfer pricing and debt shifting are the most common. Direct studies investigating evidence of transfer price manipulations typically use internal trade data. These studies analyze whether the prices set on the goods and services traded between the affiliates of the multinational companies (internal or transfer prices) are being artificially over- or understated as compared with market prices (arm’s length prices), or whether internal prices are sensitive to tax rates in a manner consistent with profit shifting. Swenson (2001); Clausing (2003); Bernard et al. (2006) do that type of analysis using international trade data from the U.S.; Vicard (2015) and Davies et al. (2018) use French import and export data; Cristea and Nguyen (2016) use data on Danish exports. All of these papers find evidence consistent with transfer price manipulations in the multinational firms. In the case of debt shifting, interest payments on internal debt and internal leverage are in focus— for example, Desai et al. (2004) provide evidence of debt shifting using data on foreign affiliates of American multinational corporations; Buettner and Wamser (2013) and Møen et al. (2019) use data on German multinational companies.
Other empirical studies on profit shifting, including this one, follow a so-called indirect approach. The general idea of the indirect approach is to look for evidence of profit shifting in some bottom-line measures, for example, taxable profits or taxes, which would reflect all profit shifting activities. Some of the studies in this field use country-level data, while others, such as this paper, analyze firm-level data. Some papers provide evidence of profit shifting by looking at the tax sensitivity of taxable income of multinational firms, such as Hines and Rice (1994), who use macro-level data, and Huizinga and Laevan (2008), who use firm-level data. Others exploit the variation coming from productivity shocks, such as Dharmapala and Riedel (2013), or tax reforms such as the introduction or strengthening of thin capitalization rules, as in Buettner et al. (2012). Still others (such as this one) use a corporate twin approach to compare similar multinational and national firms. Some of the studies use the corporate twin approach to compare tax payments between similar multinational and national firms, such as Egger et al. (2010). The results of our study are most closely related to the corporate twin approach studies looking at the differences in taxable income between multinational and national firms, such as Grubert et al. (1993) and Grubert (1998) for the US, Langli and Saudagaran (2004) for Norway, Belz et al. (2013) for a sample of European firms, and Oyelere and Emmanuel (1998, 2002) and Bilicka (2019) for the UK. The results of my paper are consistent with the results reported in all these papers showing that foreign-controlled firms are systematically less profitable than comparable national firms, which is consistent with tax-motivated profit shifting behavior of multinational firms. These results are also consistent with Tørsløv et al. (2021), who use macroeconomic data and find among other things that in tax havens foreign firms are much more profitable than local firms, while the opposite is true for high-tax countries.
The main advantage of this indirect approach is that it can uncover profit shifting by all the different profit shifting channels. The main limitation of these studies is that the observed changes and differences in taxable income and taxes can capture more than just profit shifting activities. To address this limitation, most of the studies in this field use different approaches to show that changes in bottom line measures between different types of firms or over time can be attributed to tax-motivated profit shifting.
To investigate to what extent the reduction in profitability in the acquired firms in Norway following the acquisition can be attributed to profit shifting activities, I carried out further analysis to study what is driving the reported changes in profitability. I did that by decomposing the profitability measures and studying some of the items more closely. I found that acquired firms outperformed non-acquired firms in terms of sales and asset growth, and investigated further which particular items on the cost side may be causing the reduction in profitability. I looked more closely at what happens to the share of material costs, labor costs, and other operating costs following the acquisition. The results indicate that the negative profitability differences are mainly driven by the significant increase in the share of the material costs for the acquired firms. Material costs include payments for the raw materials, intermediary goods, and royalty and trademark internal payments, as long as they can be charged on any specific product or service. Thus, it is one of the most interesting items from a transfer pricing perspective. Even though we are not able to distinguish between the costs of materials traded internally and externally, it is known that the share of intra-firm trade in total trade of affiliates tends to be considerable in many countries (we do not have the numbers for Norway, but the share of internal exports in all affiliate exports is estimated to be 64% for neighboring Sweden (Lanz and Miroudot, 2011)). The significant differences in material costs between acquired and non-acquired firms are also consistent with the result in Oyelere and Emmanuel (2002), who also find that the lower profitability of the foreign-controlled firms in the UK is partially due to higher trading expenses, which they argue may be an indicator of transfer price manipulations by the suppliers, who are most likely to be related affiliates in other countries.
Furthermore, I have studied whether the change in profitability for the acquired firms follows their real economic performance, as measured by alternative financial ratios reflecting dimensions of performance other than profitability. In particular, I looked at how the changes in total assets turnover, the current ratio, and interest coverage relate to the development of profitability. A few studies look at the effects of foreign acquisitions on both profitability measures as well as other performance measures together (Fukao et al., 2008; Chen, 2011; Chari et al., 2012). Most of them find positive effects of foreign acquisitions on both efficiency and profitability measures of the acquired firms. To study changes in the productivity and efficiency of the firms, Fukao et al. (2008); Chen (2011); Chari et al. (2012) study indicators such as total factor productivity, employment, fixed assets, and sales. I found, in line with Fukao et al. (2008), that sales, as mentioned above, increase more following the acquisition for the acquired than for the non-acquired firms. Furthermore, the effect of foreign acquisitions on fixed assets (property, plant, and equipment) is found to be positive. While I do not have data to estimate total factor productivity, I used total assets turnover as a partial proxy. The results indicate that the total assets turnover improves following the foreign acquisitions, but the change is not statistically significant. Additionally, I looked at two other performance measures, current ratio (reflecting liquidity of the firms) and interest coverage (reflecting solvency of the firms), but found no statistically significant changes following the foreign acquisitions in these two measures of performance. These results may be a further indication that the observed change in profitability is not associated with similar changes in other performance measures of the firms. Below, I discuss the way in which profit shifting activities may affect these performance measures, and the implications of this for the results.
The sample for the analysis is constructed by combining three different data sets. The first data set, provided by Dun & Bradstreet (D & B), contains data on the financial statistics of all companies registered in Norway. The second data set, SIFON, provides data on foreign ownership in Norwegian companies (inward FDI). Finally, I used data on direct investments of Norwegian companies abroad (outward FDI), from the Survey of Outward Foreign Direct Investment (Utenland-soppgaven), collected by the Norwegian tax authorities. These three data sets are merged by unique company IDs, administered by the governmental Brønnøysund Register Center.
The initial sample consists of limited liability firms that either remained domestically owned (purely domestic) or became foreign controlled during the period considered (foreign acquired). A firm is considered foreign acquired if its total foreign ownership share changed in the period studied from below 50% to above. As the focus of the paper is on what happens to profitability of the foreign acquired firms, the sample only includes firms in Norway that do not have affiliates abroad. The sample is further restricted to include firms that belong to one of the following industries: manufacturing, wholesale trade, retail trade, transport, real estate, or R&D/business activities.
To avoid the potential “noise” from a large number of small companies, among which the share of multinational companies is very small, the sample is further limited to companies with average total assets or sales of more than NOK 1,000,000.2 Furthermore, observations with extreme values for the key variables are excluded.3
The final sample consists of 629,919 observations over 12 years for 103,073 Norwegian firms. Table 1 shows the number of observations in the sample over time. Around 99% of the observations in the sample are for purely domestic firms. The rest are for firms that were subject to foreign acquisition during the period studied. In total, there are 8,796 observations of the foreign-acquired firms. The total number of acquisitions during the period studied in the sample is 1,040. Most of the acquired firms belong to the wholesale industry (28%), where they constitute two percent of all firms in the industry. Around 20% of all acquisitions are firms involved in R&D/business activities, followed by the firms in manufacturing and real estate industries, with 17% and 13% of acquisitions, respectively.
Number of observations over time
| Inward FDI | |||||
|---|---|---|---|---|---|
| Year | Total | Purely domestic | pre-acquisition | acquisition | post-acquisition |
| 1994 | 40,225 | 39,754 | 392 | 79 | 0 |
| 1995 | 42,876 | 42,346 | 428 | 30 | 72 |
| 1996 | 45,722 | 45,163 | 423 | 32 | 104 |
| 1997 | 48,895 | 48,273 | 468 | 27 | 127 |
| 1998 | 51.741 | 51,041 | 513 | 34 | 153 |
| 1999 | 53,185 | 52,435 | 500 | 68 | 182 |
| 2000 | 54,790 | 53,988 | 484 | 88 | 230 |
| 2001 | 55,982 | 55,139 | 414 | 124 | 305 |
| 2002 | 56,292 | 55,423 | 343 | 119 | 407 |
| 2003 | 58,120 | 57,237 | 268 | 110 | 505 |
| 2004 | 60,168 | 59,265 | 134 | 172 | 597 |
| 2005 | 61,923 | 61,059 | 0 | 157 | 707 |
| Total | 629,919 | 621,123 | 4,367 | 1,040 | 3,389 |
Several profitability measures are studied to ensure that the choice of profitability measure does not affect the results. Return on sales is measured as a ratio of taxable income to sales (total operating income) and reflects the profit-generating capacity of all aspects of the firm, including its capital structure. As such, it does not account for differences in leverage or the capital intensity of the firms. From a profit shifting perspective, this means that it can reflect both transfer price manipulations and debt shifting activities. Return on total assets is a ratio of taxable income to total assets and is a measure for how profitable the company is in relation to its total assets. The operating margin, calculated as a ratio of earnings before interest, taxes, depreciation, and amortization (EBITDA) to sales would, among other things, eliminate distortions from various depreciation methods employed following the acquisition, as well as from the differences in capital structure of the firms. However, unlike return on sales, it will not reflect debt shifting activities, if any.
In addition to profitability, I looked at three alternative measures that are meant to reflect other aspects of the firms’ economic performance (productive efficiency, solvency, and liquidity). I used Total Assets Turnover as a proxy for productive efficiency. It is calculated as the ratio of sales to the value of assets, and reflects the efficiency with which a firm is using its assets to generate sales revenue. The Current Ratio is the ratio of current assets to current liabilities and measures the ability of the firm to pay off its current or short-term liabilities with current assets. Firms with a large amount of current assets (such as cash and cash equivalents) are better positioned to pay off current liabilities at their due dates without having to sell off long-term revenue-generating assets. Interest Coverage is calculated as the ratio of earnings before interest and taxes (EBIT) to interest expense. It measures the firm’s solvency, and reflects how easily it can pay interest expenses on outstanding debt.
To examine what drives the differences in profitability between the acquired and non-acquired firms, I looked at how different cost items from the income statement change following a foreign acquisition. Material costs (of goods sold) cover direct material expenses incurred by firms (raw materials), when they purchase goods for resale or buy parts for assembling the goods sold, as well as royalty fees and license payments for the sold goods. This cost item reflects, among other things, transfer prices paid to the related parties in these transactions. Labor costs cover wages and salaries, as well as payroll tax, pension costs, and other employee benefits. Other operating costs are calculated by subtracting material costs and labor costs from the total operating costs. To see whether any of these costs in particular may be driving the profitability differences, I scaled each cost by operating income. I then looked at the differences in development in each of those for the acquired firms from before to after the acquisition and for the matched non-acquired firms.
In addition, I studied the corresponding development in sales, total assets, fixed assets, and leverage. Sales are defined as operating income from the income statement. Total assets are defined as the sum of debt and equity in the balance sheet. Fixed assets cover property, plant, and equipment. The leverage variable is constructed as a share of debt to total assets as reported in the balance sheet.
A number of variables are used as covariates when we estimate the probability of the firms becoming foreign acquired. In addition to the variables described above (operating income, fixed assets, total assets, leverage), age is used and is constructed as the difference between the current year and the year of incorporation of a firm. Market share is also used and is measured as the share of sales of a firm to the total sales in the industry, where industry affiliation is based on three-digit NACE codes.
The summary statistics for the outcome variables and the covariates described above are provided in Table 2. As expected, foreign acquisition targets are much larger in size than the rest of the firms, both in terms of sales and assets, and also possess much higher market shares than domestic firms. Age does not seem to be a determining factor for acquisitions, based on the summary statistics. While foreign acquisition targets do not differ much from the rest of the firms in terms of capital structure, purely domestic firms have comparatively higher shares of fixed assets in their total assets. Tangibility decreases even further in foreign targets following the acquisition. This may reflect disinvestment as a part of the restructuring activities undertaken by the new owners. At the same time, the acquired firms are shown to grow considerably in terms of total assets following the acquisition.
Summary statistics for the main covariates
| Inward FDI | |||||
|---|---|---|---|---|---|
| Variable | Total | Purely domestic | pre-acq. | acq. | post-acq. |
| Return on sales | 0.0997 | 0.1002 | 0.0591 | 0.0604 | 0.0587 |
| Return on total assets | 0.0884 | 0.0885 | 0.0818 | 0.0647 | 0.0771 |
| Operating profit margin (EBITDA) | 0.2114 | 0.2126 | 0.1442 | 0.1254 | 0.1109 |
| Total assets turnover | 2.136 | 2.136 | 2.140 | 2.069 | 2.207 |
| Current ratio | 1.744 | 1.745 | 1.669 | 1.603 | 1.789 |
| Interest coverage | 27.25 | 27.14 | 30.99 | 33.44 | 42.09 |
| Sales | 13,720 | 13,107 | 38,040 | 62,109 | 79,792 |
| Total assets | 11,208 | 10,795 | 31,246 | 48,001 | 49,806 |
| Fixed assets | 4,104 | 4,048 | 8,650 | 11,850 | 6,207 |
| Labor costs | 2,262 | 2,146 | 7,603 | 11,635 | 13,650 |
| Material costs | 7,977 | 7,642 | 20,294 | 30,328 | 46,557 |
| Labor cost share | 0.2059 | 0.2057 | 0.2281 | 0.2222 | 0.2112 |
| Material cost share | 0.3546 | 0.3537 | 0.3899 | 0.4126 | 0.4550 |
| Other operating costs share | 0.2832 | 0.2833 | 0.2815 | 0.2778 | 0.2559 |
| Tangibility | 0.2619 | 0.2634 | 0.1940 | 0.1411 | 0.1111 |
| Leverage | 0.7791 | 0.7794 | 0.7652 | 0.7823 | 0.7381 |
| Market share | 0.0031 | 0.0029 | 0.0105 | 0.0133 | 0.0170 |
| Loss-making | 0.2460 | 0.2458 | 0.2512 | 0.3163 | 0.2606 |
| Age | 11.72 | 11.71 | 11.57 | 10.23 | 13.41 |
| Obs. | 629,919 | 621,123 | 4,367 | 1,040 | 3,389 |
The summary statistics show that firms that were foreign acquired were less profitable than purely domestic firms before the acquisition. Profitability decreases even further following acquisition, and in turn, profitability differences between the acquired and non-acquired firms increase further after acquisition. Similarly, the share of firms with losses is slightly larger among the acquired firms as compared with purely domestic firms.
This may indicate that foreign investors choose the more poorly performing firms for acquisition (lemon picking). The share of loss-making firms (defined as a share of the firms with non-positive taxable income) increases slightly following acquisition, reflecting the fact that foreign investors do not always manage to improve the performance of the acquired “lemon” firms and some of the acquisitions fail. It is therefore important to test whether the decrease in profitability following acquisition may be driven by the failed acquisitions rather than profit shifting.
Turnover, current ratio, and interest coverage all increase for the firms acquired following acquisition, indicating that acquired firms seem to become slightly more effective in the post-acquisition period.
The main empirical challenge in this paper is to estimate the causal effects of foreign acquisitions on the performance of acquired firms. This is a typical treatment evaluation problem, where foreign acquisition can be considered a treatment, and the different performance measures to be studied constitute the outcomes. This problem can be considered within a counterfactual framework, also known as the potential outcome model (Roy, 1951; Rubin, 1974). In this framework, a treatment (T) defines the existence of two possible states: the treatment exposure state and the no-treatment or control state. We can never observe potential outcomes in both treated and control states for each of the units, or firms, so counterfactuals for each of the states will always be missing. This is a fundamental challenge of counterfactual models. The average treatment effects can, however, still be calculated if we can approximate the missing counterfactual outcomes. The average treatment effect on the treated (ATT), which is focused on in this paper, requires that we only approximate the counterfactual for the treated units. Outcomes of non-acquired domestic firms (non-treated units) are used to construct the counterfactual in this paper.
The average treatment effect can formally be expressed as follows:
where
If the selection expression (last term in the brackets above) is not equal to zero, the estimate of the average treatment effect on the treated units will be subject to selection bias. In this case, we cannot say with certainty whether the effect we estimate is due to the treatment or to systematic differences between the treated and the control units. In the context of this study selection bias may arise from the fact that foreign acquisition targets are not chosen randomly and may differ from the non-acquired firms both in terms of observable and non-observable characteristics. To account for this potential bias, while estimating the effect of foreign acquisitions, I use propensity score matching combined with the difference-in-difference estimator. Propensity score matching allows us to find a match for each of the acquired firms among the non-acquired domestic Norwegian firms, based on similar pre-acquisition characteristics. As matching on many different characteristics is difficult in practice, the matching is done based on propensity scores. This is a common method proposed by Rosenbaum and Rubin (1983) to reduce the dimensionality problem. The propensity score is an estimated predicted probability of becoming treated, conditional on the observed characteristics,
In the context of this study, it represents the probability of becoming acquired, so that the counterfactual group to the acquired firms will be composed of the non-acquired firms with similar probabilities of becoming acquired.
Identification of the average treatment effect on the treated units using propensity score matching requires that two assumptions are satisfied (Rosenbaum and Rubin, 1983). The conditional independence assumption states that, based conditionally on all the firm characteristics (
The common support assumption is another important assumption behind the propensity score matching estimator, which ensures that treatment status cannot be perfectly predicted by any of the covariates, so that matches for both treatment and control group observations can be found: 0 <
The common support restriction implies that if the support of covariates does not overlap for the acquired and non-acquired firms, matching should be performed over the region of common support. Restricting matching to the common support region is advised (Heckman et al., 1997) and implies in practice that some of the units would be dropped to ensure that we do not use units with propensity scores that are too far off, which thus ensures that we compare only comparable units.
If selection depends not only on observable, but also on unobservable factors, the conditional independence assumption described above can be problematic, and the simple propensity score matching estimator may be both biased and inefficient. It may, for example, be argued that foreign takeover targets possess some type of intangible assets, such as unique production processes, that are not easily reflected in accounting figures, but that affect both their attractiveness for foreign investors and their performance. As my data are in panel format, this issue can be addressed by combining the PSM estimator with the difference-in-difference (DiD) estimator, as proposed by Blundell and Costa-Dias (2002). This implies that instead of comparing the post-acquisition performance of the acquired and the matched non-acquired firms, development in performance over time (from the pre- to the post-acquisition periods) for both groups of firms is compared, and the PSM DiD estimator is as follows:
where
where
In addition to looking at year-by-year changes in performance following the acquisition, I also looked at changes in average performance for the four years following the acquisition as compared with average performance three years prior to acquisition.
The propensity score is estimated using probit regression on the following model:
where the dependent variable
To select the covariates for the propensity score estimations, I followed the existing literature, using matching methods to estimate the performance effects of foreign acquisitions (Bandick and Karpaty, 2010; Chen, 2011; Chari et al., 2012; Fukao et al., 2008; Arnold and Javorcik, 2009).
To control for size, I included a logarithm of total assets for the firm. Dummy variables for each of the yearly quantile age groups were also included. Furthermore, I controlled for the tangibility of the assets. The share of tangible assets is meant to capture the composition of target firms’ assets. Profitability in terms of returns on sales is also included in the estimation to account for the size of the firms in terms of sales and for the possibility that foreign investors systematically choose more or less profitable firms as their targets. A variable for a firm’s market share in terms of sales is included to account for the possibility that firms with a larger market share are more attractive foreign acquisition targets.5
The results of the probit model are presented in Table 3. The results show that larger firms with a higher market share are more attractive targets for foreign acquirers. In terms of profitability, measured by return on sales, the foreign acquirers seem to have preference for less profitable firms, which is consistent with the descriptive statistics in Table 2. In terms of age, younger firms are shown to be more likely to be acquired than older firms.
propensity score estimation — probit results
| Coefficient | Standard Error | ||
|---|---|---|---|
| Return on sales | −0.4463*** | 0.000 | |
| Tangibility | −0.3929*** | 0.000 | |
| Market share | 0.5742** | 0.044 | |
| Ln(Total Assets) | 0.1765*** | 0.000 | |
| Age Q2 | −0.1496*** | 0.000 | |
| Age Q3 | −0.1676*** | 0.000 | |
| Age Q4 | −0.2554*** | 0.000 | |
| Year dummy variables | Yes | ||
| Industry dummy variables | Yes | ||
| Region dummy variables | Yes | ||
| Pseudo R-squared | 0.1160 | ||
| LR chi-squared | 992.08 | ||
| Observations | 389,749 | ||
| Treated | 568 | ||
| Non-treated | 389,181 |
The covariates are averaged over three years prior to acquisition (except for age). The coefficients reported are in log-odds form.
To ensure that we find matches for each of the treated firms from a comparison group consisting of firms with similar propensity scores, a common support restriction is imposed by excluding observations from both groups that have a probability of being acquired that are higher than the 99th percentile and lower than the first percentile of the propensity score distribution in the acquired group.6 This implies that in total, 86,170 observations are dropped from the analysis, of which ten are acquired firms (around 1% of all acquired firms). The number of acquisitions for which the propensity score could be estimated and that survived the common support restriction is 558.
To test the quality of the matching procedure, I looked at the balancing of the covariates by doing a number of tests to establish that the distribution of covariates is comparable between the acquired and the non-acquired firms (see Appendix A for details).
The matching was completed using the one-to-one nearest-neighbor method. This method requires that the absolute distance between the propensity scores of treated and control units is minimized, so that the outcome of the control (non-acquired) firm that has the propensity score closest to the treated (foreign-acquired) firm is used as a prediction of the missing potential outcome for that treated firm. The advantage of this method is that it allows bias to be minimized by avoiding use of more, but poorer, matches. Each of the non-acquired firms is allowed to be used as a match for more than one acquired firm, which can also minimize the propensity score distance between the acquired and the matched non-acquired firms, and thus reduce bias. The following matching algorithm is used7: first, the observations are sorted based on the estimated propensity score. Next, the closest neighbor is chosen to each of the treated units whether it is above or below the treated unit in the sorted list. If the neighbors from above and below are equally good, the match is drawn randomly from the two. Finally, the DiD estimates of the average treatment effect on the treated units are calculated by comparing the differences between the two groups before and after the foreign acquisition period.
The robustness of the main results is tested using other matching methods (kernel and radius methods) in the sensitivity tests section (see Appendix B).
The estimates for the three different profitability measures are reported in Table 4. The estimates are DiD estimates showing the differences in the change in profitability for each of the four post-acquisition years (the first five rows in the table) compared with average profitability over the three years prior to acquisition, for the acquired and matched non-acquired firms. The average estimate in the last row is the difference between average profitability in the four years after the acquisition as compared with the average profitability three years prior to the acquisition for the acquired and matched non-acquired firms.
ATT matching estimates, profitability measures
| ATT | |||
|---|---|---|---|
| Taxable income/Sales (1) | EBITDA/Sales (2) | Total assets (3) | |
| Differences | Benchmark: 3 years average prior to acquisition | ||
| The year of acquisition | |||
| (−0.474) | (−1.160) | (−0.893) | |
| 1 year after acquisition | |||
| (−1.508) | (−3.387) | (−1.613) | |
| 2 years after acquisition | |||
| (−1.982) | (−3.889) | (−2.019) | |
| 3 years after acquisition | |||
| (−2.003) | (−2.706) | (−1.196) | |
| 4 years after acquisition | |||
| (−3.421) | (−4.683) | (−1.771) | |
| (−3.996) | (−5.490) | (−2.636) | |
The matching estimates in this table are based on nearest neighbor 1-to-1 matching with replacement. The coefficients in bold are significant at least at the 10% significance level. The
For example, the average operating margin estimate of −0.018 in the last row of column 3 in Table 4 means that on average, the change in average profitability four years after the acquisition as compared to average profitability three years prior to acquisition for acquired firms is 1.8 percentage points lower than the change in the corresponding period for the matched non-acquired firms. This corresponds to a reduction in profitability for the acquired firms of about 14%. All the estimates reported are negative, which means the performance of acquired firms in Norway is reduced after acquisition. The size of the estimates tends to increase in absolute value in the later years following the acquisition.8 The estimates remain negative—economically and mostly statistically significant—regardless of the profitability measure used but are the largest in absolute value when we look at return on total assets.
The coefficients for the operating margin (EBITDA) in relation to sales (column 3 in Table 4), which account for differences in the firms’ leverage, are negative, economically significant, and mostly statistically significant. The fact that the coefficients are smaller in size as compared with return on sales (based on taxable income) may indicate that debt shifting may play a limited role but does not seem to be the main driving force for the result of negative profitability differential.
The estimates based on taxable income (columns 1 and 2 in Table 4) are comparable in size with the estimates in earlier related studies that use Norwegian data but other estimation methods, such as Langli and Saudagaran (2004), who use OLS, and Tropina (2010), who uses panel data methods. The results are also consistent with, and comparable in size to, the estimates in Belz et al. (2013), who also use propensity score matching in combination with the DiD estimation and find significantly negative effects of foreign acquisitions on profitability when the foreign acquirer is tax aggressive.
As a robustness test, I have tested whether the profitability differences may be driven by loss-making firms, which do not have the same type of incentives to involve in profit-shifting. The results indicate that even though loss-making firms are affecting the main results, this is not the group of firms driving the results. I have also tested whether the results are more pronounced for the firms that have a higher foreign ownership share than the default threshold of 50%, since the costs of engaging in profit shifting can be expected to be lower, and the opportunity to do so can be expected to be higher for these firms. The profitability differential was indeed larger in absolute value for those firms as compared with the rest of the acquired firms. The results of both these tests support the hypothesis that profit shifting may be the driving force behind the observed profitability differentials. See Appendix B for the results of these tests.
In this section, I look more closely at the factors that may be driving the profitability differences between the acquired and the non-acquired firms. The results in Table 5 show that both taxable income, sales and total assets are growing faster for the acquired firms than for the matched non-acquired firms. The growth rates for the sales and total assets (that are used as scales in profitability measures above) are generally higher for the total assets and sales as compared with the growth rate for taxable income. This indicates that while the foreign-acquired firms do not seem to have any trouble generating sales income and growing by accumulating more total assets, their taxable income is not growing as fast as that of the corresponding non-acquired firms. The effect on the leverage ratios is on average negative, but there is considerable variation across the years (column 5 of Table 5). This is also an indication that the lower profitability differential is not driven by the potential debt shifting activities by the multinational firms, which is also consistent with the findings by Belz et al. (2013).
ATT matching estimates, other performance indicators
| ATT | |||||
|---|---|---|---|---|---|
| Sales (1) | Taxable Income (2) | Total Assets (3) | Fixed Assets (4) | Leverage Ratio (5) | |
| Differences | Benchmark: 3 years average prior to acquisition | ||||
| The year of acquisition | |||||
| (2.527) | (2.464) | (4.822) | (1.007) | ||
| 1 year after acquisition | |||||
| (6.273) | (2.559) | (6.093) | (2.585) | (−.418) | |
| 2 years after acquisition | 0.031 | ||||
| (6.311) | (3.631) | (5.872) | (.422) | (−1.815) | |
| 3 years after acquisition | |||||
| (5.981) | (2.850) | (6.221) | (3.368) | (−3.987) | |
| 4 years after acquisition | |||||
| (5.382) | (2.312) | (6.138) | (4.315) | (1.249) | |
| (8.450) | (2.377) | (8.712) | (3.090) | (−3.041) | |
The matching estimates in this table are based on nearest neighbor 1-to-1 matching with replacement. The coefficients in bold are significant at least at the 10% significance level. All the variables (except leverage) are included in the estimations in log-form. The
Furthermore, I have investigated whether any of the cost items in particular may be driving the negative profitability differential result. With the data available, I can examine more closely what happens to material costs, labor costs, and other operating costs scaled by the firms’ operating income. Material costs are interesting in the context of this study, because they include transfer prices paid for the materials and intermediate goods bought from the related parties. These can be artificially manipulated to shift profits in or out of Norway, after firms become foreign acquired. Labor costs are also included in the estimations in order to test whether the lower profitability growth of the foreign-acquired firms is a result of the relatively higher wages following acquisitions.9
The results in column 1 of Table 6 indicate that following an acquisition, material costs grow significantly more for the acquired firms as compared with the corresponding non-acquired firms. This result is consistent with the results of Oyelere and Emmanuel (2002) in the UK, who also find that trading expenses are partially responsible for the underperformance of the foreign-controlled firms in their sample. The labor costs share and the share of other operating costs, on the other hand, go down following the acquisition, but the decrease is mostly not statistically significant (columns 2 and 3 in Table 6).
ATT matching estimates, cost items
| ATT | |||
|---|---|---|---|
| Material costs of Goods sold /Sales (1) | Labor costs/Sales (2) | Other operating costs /Sales (3) | |
| Differences | Benchmark: 3 years average prior to acquisition | ||
| The year of acquisition | 0.006 | −0.005 | 0.003 |
| (0.963) | (−1.112) | (0.446) | |
| 1 year after acquisition | −0.010 | ||
| (4.487) | (−2.685) | (−1.366) | |
| 2 years after acquisition | −0.005 | 0.000 | |
| (2.213) | (−0.877) | (0.014) | |
| 3 years after acquisition | −0.003 | ||
| (2.765) | (−1.901) | (−.330) | |
| 4 years after acquisition | −0.002 | −0.009 | |
| (2.710) | (−0.256) | (−0.966) | |
| 0.000 | −0.000 | ||
| (2.682) | (0.038) | (−0.020) | |
The matching estimates in this table are based on nearest neighbor 1-to-1 matching with replacement. The coefficients in bold are significant at least at the 10% significance level. The
As the material costs of goods sold include, but are not limited to, the materials traded internally, this does not serve as direct evidence of profit shifting but is an additional piece of evidence that may indicate profit shifting by means of transfer pricing.
In this section, I will test whether the negative changes in profitability reported in Table 4 are accompanied by corresponding changes in other performance measures. Total assets turnover is an indicator of the productive efficiency of firms, and reflects the efficiency with which firms are using their assets. The results in column 1 of Table 7 show that the estimates for the turnover differential between acquired and non-acquired firms vary in sign, from being negative right after the acquisition to becoming positive in the later years, but are not statistically significant. This indicates that foreign-acquired firms do not seem to significantly underperform relative to non-acquired firms in terms of efficiency. It can be argued that this measure may be affected by profit-shifting activities itself through sales, which can potentially include transfer prices on internal sales. If the profits are shifted by artificially lowering export sales prices, this would affect the total asset turnover negatively, meaning that the turnover would go down following the acquisition all else equal. The results in Table 5 show no indication of decreasing sales following the acquisition. There is, however, an indication that the profits may be shifted by the inflated material costs (Table 6). If foreign-acquired firms in Norway indeed shift profits out of Norway by artificially inflating the costs of internal goods traded, this efficiency measure should not be affected.
ATT matching estimates, performance measures
| ATT | |||
|---|---|---|---|
| Change from year before acquisition to | Total asset turnover (1) | Current ratio (2) | Interest coverage (3) |
| Differences | Benchmark: 3 years average prior to acquisition | ||
| The year of acquisition | −0.078 | 1.554 | |
| (−2.20) | (−1.047) | (0.360) | |
| 1 year after acquisition | −0.034 | 0.001 | 2.415 |
| (−0.719) | (0.016) | (0.438) | |
| 2 years after acquisition | 0.055 | 0.038 | −2.584 |
| (1.003) | (0.344) | (−0.477) | |
| 3 years after acquisition | 0.029 | 0.165 | 0.290 |
| (0.476) | (1.367) | (0.045) | |
| 4 years after acquisition | 0.042 | 0.099 | −9.364 |
| (0.619) | (0.723) | (−1.484) | |
| −0.031 | 0.074 | −4.521 | |
| (−0.632) | (0.794) | (0.949) | |
The matching estimates in this table are based on nearest neighbor 1-to-1 matching with replacement. The coefficients in bold are significant at least at the 10% significance level. The
The DiD estimates for the current ratio are reported in column 2 in Table 7. This is a liquidity measure, which reflects the ability of the firm to pay off current liabilities with current assets. The estimated results show that the estimates for the current ratio are positive for all the years following the acquisition but not statistically significant.
Interest coverage, which can be considered an indicator of the solvency of the firms, reflects how easily a firm can cover its interest expenses. This performance measure can be argued to be affected by profit-shifting activities, both by means of transfer price manipulations (through the EBIT in the nominator) and by debt shifting (through financial expenses in the denominator). If acquired firms used both types of manipulations with the goal of shifting profits out of Norway, the numerator of this ratio would be understated, while the denominator would be overstated. This means that the effect of foreign acquisition on the interest coverage in the presence of profit shifting would be negative. The results in column 3 of Table 7 show that the estimate varies in size and sign for the different years of acquisition, on average being negative, but is never statistically significant.
To sum this topic up, while the effect of foreign acquisition on profitability is negative, we do not observe a corresponding significant negative effect of foreign acquisition on other performance measures. While this discrepancy is not a direct proof of profit shifting activities, it is consistent with the hypothesis that profit shifting drives the profitability differential rather than real performance changes.
This study investigates changes in the profitability of foreign acquired firms following acquisitions in Norway during the period 1994–2005. Using propensity score matching combined with DiD estimation, I have found that the average effect of foreign acquisitions on the profitability of acquired firms in Norway is negative. This result is consistent with earlier evidence from Norway (Langli and Saudagaran, 2004; Tropina, 2010) and may be indicative of profit shifting behavior. The sign of the aggregate net effect indicates that if profit shifting is really the driving factor behind this profitability differential, the net flow of profits is out of Norway, meaning that foreign-acquired firms may not be contributing their fair share of taxes in Norway.
One of the main contributions of this paper is that it studies more closely what is driving these profitability differences. I find that foreign-acquired firms grow more quickly following the acquisition as compared with their domestic counterparts in terms of both total assets and operating revenue. While they do not seem to have problems growing and generating sales, there seem to be certain expenses that limit their profit-generating capacity. My results indicate that while there are no significant differences in the changes of labor cost share and the share of other operating costs between the acquired and the non-acquired firms, the share of material costs of goods sold increases more for the acquired firms than for the non-acquired firms following the acquisitions. This cost item is interesting, because it includes transfer prices paid for the materials and intermediate goods purchased from related parties. I also compare the change in profitability from before acquisition to after acquisition for the acquired and corresponding non-acquired firms to the corresponding change in other performance measures, namely, solvency and liquidity, which reflect productive efficiency. I have found basically no statistically significant differences between foreign acquired and non-acquired firms in terms of these performance measures. Thus, even though the evidence provided in this paper is indirect, the findings are consistent with the fact that profit shifting by means of transfer pricing may be common among the number of foreign acquired firms in Norway.
It is also important to keep in mind that this paper describes the situation in Norway in the period before many of the important changes in national and international legislation aiming to combat tax-motivated profit shifting had been introduced (e.g., stricter transfer price regulations, the thin capitalization rules, and the country-by-country reporting). All these measures were meant to limit the profit shifting activities of the multinational firms that this paper aims to explore. If the profitability differentials documented in this paper are due to the profit shifting activities of the multinational firms and the measures introduced have been effective, the differentials would likely be lower if measured today. This is confirmed by Hopland et al. (2019), who look at a longer data panel of the Norwegian firms from 1993 to 2012, and show that the profitability changes among the firms that become multinational became much less pronounced after the introduction of stricter transfer pricing regulations in 2007/2008. This supports the claim that the profitability differences between national and multinational firms that we observe in this paper are due to profit shifting activities, and these estimates can potentially serve as a benchmark for further analysis of the effects of other BEPS measures.