1 Introduction

Interest in corporate tax planning has reached an all-time high, and both the financial media and academics identify cross-jurisdictional income and debt shifting as key sources of tax savings (e.g., Atwood et al. 2012; Beuselinck et al. 2015; Collins et al. 1998; Klassen et al. 1993; Klassen and Laplante 2012; Markle 2015; Newberry and Dhaliwal 2001; Rego 2003). However, designing and implementing cross-jurisdictional shifting schemes can be costly and is not necessarily the single tool available for corporations to avoid paying high taxes. Dyreng et al. (2017), for instance, observe that effective tax rates (ETRs) for U.S. corporations have declined for both multinational and domestic firms over the 25 years from 1988 through 2012. This finding suggests that for domestic firms, too, a wide array of local tax avoidance opportunities exist and these can as well give rise to lower tax bills.

Ample anecdotal and academic evidence suggests that corporations lower their tax bill by focusing on local tax optimization schemes, which do not necessarily require investments in cross-jurisdictional shifting. Local schemes can take various forms, including investments in research and development (R&D) tax credits (Berger 1993; Rao 2016), investments in tax-favorable local assets (Engel et al. 1999), exploiting within-country regional tax differences (Dyreng et al. 2013), and utilizing prior-period loss offsetting. It is generally accepted that local tax avoidance schemes are important for firms and this is also often echoed by the view of tax professionals.Footnote 1

A defining feature of local tax avoidance schemes is that they do not necessarily require investments in cross-jurisdictional shifting. However, some locally designed avoidance opportunities reduce the tax base or the tax rates by so much that multinationals may also respond by shifting income to the respective country. Examples include advanced tax rulings, special purpose entities (e.g., Fox and Luna 2005; Feng et al. 2009), intellectual property box regimes (Bornemann et al. 2020), and allowance for corporate equity (Hebous and Ruf 2017; Zangari 2014), making it challenging to examine the continuum of a corporation’s global versus local tax avoidance strategies empirically. One reason for this is the conceptual difficulty in separating local versus global tax strategies, and another reason relates to the lack of affiliate-level data. In this paper, we make a first attempt to observe and provide more empirical evidence of these complementary tax planning strategies.

In recent years, the tax and accounting literature at-large has gained important new insights through larger-scale parent and affiliate entity data that became available via Amadeus (Europe) and Orbis (Global), both Bureau van Dijk products. These databases bring the advantage that they allow investigations of corporate groups, including information on reporting sub-entities from the same ultimate owners, and it has proven to be useful in several international tax studies including the work of Huizinga and Laeven (2008), Beuselinck et al. (2015), Markle (2015), De Simone (2016), De Simone et al. (2017) and Alexander et al. (2020). We also rely on the Orbis database to identify the global network of European headquartered multinational groups and propose a novel calculation approach to identify the importance of and dynamics in local tax strategies vis-a-vis profit shifting. Our method is based on the observed relationship between the multinational consolidated effective tax rate (ETR) and the sum of the (pretax income-weighted) subsidiary-country ETRs and aims to bring a novel, yet parsimonious, measure that can complement other profit shifting studies in identifying tax optimization strategies.

Our logic develops as follows. Theoretically, and absent consolidation differences, a multinational group-level effective tax rate (ETR) is a weighted function of its subsidiaries’ ETRs. So, if all subsidiaries decrease their ETR by one percentage point, the groups’ overall ETR will also drop by one percent. In such a stylized example, the correlation coefficient between the groups’ ETR and the pre-tax income-weighted subsidiaries’ ETR would be equal to one. However, the correlation coefficient may differ from one if not all subsidiaries are observed. For example, details on tax haven subsidiaries are often undisclosed so that we do not know how much income these subsidiaries report and how much taxes, if any, are due. Therefore, for each multinational (S) that decides to shift part of its profits to unobservable tax havens, one will observe a group-level ETR that is less systematically correlated with the ETR of the traceable subsidiaries. In other words, more profit shifting into unobservable tax havens automatically shifts the positive correlation between S’s consolidated ETR and its observable pre-tax income-weighted subsidiaries’ ETR downward (i.e., toward zero).

Alternatively, an economically comparable multinational (L) can decide not to shift profits to tax havens but rather invest in local tax planning strategies by exploiting subsidiary country tax opportunities and loopholes and still reach a similar ETR as S. Yet, for L all potential tax optimization strategies will also manifest in the observable subsidiary ETR, which then results in a positive correlation coefficient between the consolidated ETR and its respective subsidiary ETRs that is closer to one. In reality, however, we expect multinationals to pursue both global and local tax strategies simultaneously, which would result in a correlation coefficient between zero and one. Consequently, we propose that a group level versus income-weighted subsidiary level ETR correlation can serve as an inverse measure of profit shifting and that a lower correlation coefficient indicates more profit shifting, relative to local tax planning as part of the overall tax planning strategy.

Empirically we proceed as follows. We compile subsidiary-level financial data (154,022 subsidiary-year observations) from the Bureau Van Dijk Orbis database to calculate GAAP ETRs of subsidiaries and group-level data (32,849 multinational group-year observations) to calculate GAAP ETRs of European multinational groups over the period of 2006–2014. Next we regress the group consolidated GAAP ETR on the pre-tax income weighted GAAP ETR for the entirety of its observable subsidiaries. A low association between the pre-tax income-weighted ETR of the subsidiaries and the group-level ETR indicates profit shifting, whereas a high association indicates affiliate entity ETRs being more aligned with the overall group tax planning.

We first document that, for the average multinational, subsidiary local tax planning explains around 13 percent of total multinationals’ tax planning. In line with our expectations, the correlation is considerably higher (43 percent) for subsamples where we observe at least half of the consolidated sales in affiliate entities. We acknowledge that, like in any other study on intra-group profit shifting, results can be influenced by incomplete subsidiary information or because aggregating net income across affiliates can lead to double counting of at least some income in complex group structures (Blouin and Robinson 2020). For these reasons, initial correlation results should not be regarded in isolation. However, our measure does allow for explorations of which set of firms (e.g., by industry or by country) have a higher correlation between the groups’ ETR and subsidiary ETRs, compared to another set of firms. Furthermore, we find that the coefficient more than doubles from around 10 percent in the early observation years to around 21 percent toward the end of the observation period, which is consistent with the decreasing trend in profit shifting over time, as documented in the literature (e.g. Alexander et al. 2020).Footnote 2 In a nutshell, the proposed proxy is a useful initial screening tool for observing cross-sectional or intertemporal profit shifting differences.

Beyond its descriptive appeal, the measure may be particularly helpful to study the effectiveness of regulatory changes initiated to curb a particular form of tax avoidance, such as cross-jurisdictional income shifting. To this end, we validate our measure by using the staggered adoption of transfer pricing documentation requirements across the 27 EU analyzed countries over the period of 2006–2014 as a source of exogenous variation in the enforcement of international tax compliance. In support of prior findings documenting a significant reduction in profit shifting after the installment of stricter documentation requirements (e.g., Beer and Loeprick 2015; Alexander et al. 2020), we observe a significant increase in the correlation coefficient after the transfer pricing documentation requirements adoption. Timing tests, in the spirit of Bertrand and Mullainathan (2003), confirm that the parallel trends assumption holds, and the effect is only observed from the tranfer pricing documentation requirements introduction onwards. This observation, in combination with stable group ETR levels pre compared to post-transfer pricing documentation requirements, not only indicates that multinationals reduce profit shifting but also that multinationals rely significantly more on local tax avoidance to achieve stable group-level ETRs when shifting costs increase.

We next exploit cross-sectional heterogeneity to identify the channels through which the transfer pricing documentation requirements differently affect the importance of local versus global tax policies. In particular, we perform two additional validation tests for subsamples where theoretical motivations would predict local tax avoidance to become more pronounced when shifting costs increase. First, we expect the substitution effect to be more pronounced in groups where a relatively high proportion of reported profits are generated in low-tax countries for various reasons. One reason is that these groups may be better designed to exploit income shifting activities and a sudden increase in shifting costs is expected to rebalance that equilibrium toward relatively less income shifting and more local tax avoidance. Further, this scenario is especially expected to materialize since affiliate managers’ incentive contracts are likely to be redesigned so that local managers are prompted to refocus their tax planning in line with the new group-level optimum (Ortmann and Schindler 2021). Both rationales predict stronger results for groups that realize a large fraction of their results in low-tax rate countries. Our results confirm this rationale.

We also examine the responsiveness of substituting income shifting by local tax avoidance in relation to ETR target pressure. Kim et al. (2019) show that firms invest resources in tax planning until a target ETR is achieved. If the demand for a target ETR is highly inelastic, managers who experience the greatest ETR target pressure may be willing to incur higher nontax costs to achieve this target ETR, causing them to invest relatively more in local tax planning after profit shifting opportunities are curbed. Consistent with this rationale, we observe that groups facing a greater ETR pressure resort to more local tax avoidance post-transfer pricing documentation requirements while this is not observed for firms with low ETR target pressure.

Our findings are consistent with the Scholes-Wolfson framework and suggest that firms pursue tax planning opportunities that optimize after-tax returns (Scholes et al. 2016) in such a way that, when shifting costs increase, multinationals rebalance tax-planning strategies toward local tax planning opportunities. Our study makes a methodological contribution in that it provides a relatively easy-to-interpret proportional measure of subsidiary local tax planning within multinationals and further allows for dynamic interpretations of multinational tax planning strategies over time. Whereas other measures on local tax avoidance may be precise on the exact impact of country- or region-specific tax initiatives, our measure can be generated for the universe of Orbis industrial firms and is available for both publicly listed and unlisted firms.

Further, our study provides international insights into the dynamics of tax planning and adds to recent related work by Lampenius et al. (2021), who decompose tax avoidance into a tax rate and tax base component. We also contribute to the literature that investigates how local and global tax strategies may influence each other and relate to the type of tax system that is applied (Kohlhase and Pierk 2020). Finally, our work provides out-of-sample evidence consistent with the findings of Kim et al. (2019) that corporations—when confronted with out-of-equilibrium optimal tax planning—rebound relatively quickly by adjusting their tax planning strategies.

Our examination method and findings are of importance for regulators, academics, and policymakers, as documenting dynamics between alternate tax strategies and identifying specific tax strategies are empirically very challenging (Wilde and Wilson 2018). Moreover, national governments have increasingly responded to threats of tax base erosion by strengthening transfer pricing regulations and increasing the capacity of tax auditors (Alexander et al. 2020; Lohse and Riedel 2013) as well as with initiatives on country-by-country reporting (Joshi et al. 2020). It is crucial for tax authorities to recognize to what extent multinationals focus on local tax planning strategies and perhaps use these as a substitute for cross-jurisdictional income shifting. Our findings suggest that political decision-makers should not neglect the importance of local tax planning opportunities when debating fair tax rules at the international level.

The remainder of the paper is as follows. Section 2 describes the conceptual components of global versus local tax strategies and relates that to examples and prior academic work. Section 3 elaborates on the hypotheses. Section 4 discusses the research design. Section 5 presents the sample, results, and robustness sections. Section 6 concludes.

2 Local versus global tax planning strategies

An extensive literature shows that multinationals avoid taxes by shifting income from high-tax to low-tax jurisdictions, including tax havens (e.g., Dharmapala 2014; Dharmapala and Riedel 2013; Dyreng and Markle 2016; Huizinga and Laeven 2008). There exists, however, evidence that the corporate tax bill can also be lowered by techniques other than international profit shifting. In fact, Dyreng et al. (2017) document that effective tax rates (ETRs) for U.S. corporations have declined for domestic-only corporations in equal proportions as for multinationals over the period 1988 to 2012. This finding suggests that domestic firms also have an array of tax planning opportunities without having access to lower-taxed offshore locations.Footnote 3 Despite the consensus that both local and global tax strategies can reduce the overall tax burden, there exists very little insight into the firm-specific tactics to use one versus the other strategies. To this end, we describe and position various examples of commonly identified local versus global tax planning strategies. Identification of such local (i.e., within jurisdictions) versus global (i.e., cross-border) tax strategies is essential in understanding the empirical concepts and conceptual logics of our empirical proxy in the following sections.

We structure this overview into three parts: First, we discuss tax-motivated cross-country income shifting, that is, tax planning that we label as “global” tax avoidance and that mainly relates to income shifting via altering transfer prices or inter-company loans. Second, we focus on local tax avoidance strategies (“local”), which reduce the effective tax rate of the respective affiliate but do not necessarily incentivize multinationals to alter their global tax strategies. And, third, we discuss local tax avoidance strategies that are either known or that we consider at least likely, to also affect profit shifting and as such relate to both the local and global components of tax avoidance. For each category, we give a brief conceptual overview followed by several examples.Footnote 4

2.1 Global tax planning opportunities

Global tax planning, in general, refers to tax strategies enabled via access to the international web of countries and tax regulations where global firms (multinationals) are present. Widely recognized global tax avoidance schemes often manifest in actions of relocating assets and risks across affiliate countries or via the manipulation of intra-firm transactions so that income is reported in low-tax countries. The global tax avoidance literature has been extensively reviewed by Hanlon and Heitzman (2010) and, more recently, Wilde and Wilson (2018) and Jacob (2021).

Profit shifting via transfer pricing manipulation

A long series of research in accounting, finance, and economics finds that firms shift income out of high-tax jurisdictions into low-tax ones. One way of doing so is by establishing a pricing system for goods and services between subsidiaries so that profits mainly arise in tax-favorable entities and less so in high-tax affiliates. Although claims are often made about transfer pricing manipulation, research most often centers on income-based tests in combination with tax rate differentials across affiliate countries (e.g., Beuselinck et al. 2015; Collins et al. 1998; Desai and Dharmapala 2009; Dyreng and Markle 2016; Huizinga and Laeven 2008; Klassen et al. 1993).

Debt shifting and tax optimizing interest rates adjustments

Multinationals can also avoid paying high taxes via designing intra-group corporate loan schemes. For example, borrowing from an affiliate in a tax haven increases the profit in the tax haven and allows the affiliate in the high tax country to deduct interest payments from its tax base. Evidence of debt shifting is mentioned in the popular press (e.g., Wells and Houlder 2017) and examined by, for instance, Newberry and Dhaliwal (2001), Mintz (2004), and Blouin et al. (2014). To curb the concentrated use of tax-related debt financing in groups, governments, especially in high-tax countries, have adopted thin capitalization rules (e.g., Buettner et al. 2012), which specify the maximum amount of debt on which interest deductions are eligible.

Relocation of intangible assets

Multinationals can also decide to arrange cross-border transactions on intangible assets, which are by nature more difficult to value and easier to relocate de jure across borders (De Simone et al. 2014; Dischinger and Riedel 2011). In an influential paper, for instance, Grubert (2003) shows that R&D intensive subsidiaries not only have more opportunities for income shifting but also that about half of the income shifted from high-tax to low-tax countries is R&D-related. It is therefore no surprise that a lot of tension on tax avoidance schemes is on intangible-intensive industries and corporations making the taxation of transactions involving intangible assets also one of the OECD’s biggest concerns in international tax cases (e.g., OECD 2014).

Hybrid financial instruments

Hybrid mismatches exploit differences between tax systems to achieve double nontaxation. Conceptually, a hybrid mismatch would occur if, for example, an affiliate in country, A, makes a payment to an affiliate in country, B, and B considers the payment as a dividend (i.e., not taxed), while A considers it as deductible interest.Footnote 5 The combination results in a situation where the multinational realizes a tax advantage that would not have been possible without having access to the global web of countries. Such constructs were, for example, used in Luxembourg, as revealed by the “LuxLeaks” publications of the International Consortium of Investigative Journalists. Huesecken and Overesch (2015) document that corporations that establish equity-backed finance companies in Luxembourg and use these vehicles for intra-company loan transactions can reduce ETRs significantly, compared to nonruling firms. Hardeck and Wittenstein (2018) examine conduit entities established in Luxembourg and similarly conclude that these arrangements resulted in a lower tax burden. In general, hybrid mismatches are a global tax avoidance concern for all countries, as the mere existence reduces the collective tax base of countries. Conceptually, the occurrence and manifestation of hybrid mismatches may be seen as an extreme example of global tax avoidance.

2.2 Local tax planning opportunities

Local, that is, jurisdiction-specific, tax planning refers to tax optimization schemes that do not require investments in cross-jurisdictional profit shifting. In essence, these belong to a set of jurisdiction-specific tax advantage schemes that are generally available for any corporation that is operating on a specific territory or for a corporation that satisfies specific conditions.

Loss Offsetting

Most jurisdictions allow for offsetting losses with either future periods (tax-loss carryforwards) or prior periods (tax-loss carry backwards). In most countries, this tax-loss offsetting is only allowed within the boundaries of local jurisdictions, making this a predominantly local tax avoidance opportunity. In a summary study on local tax attractiveness indices, Keller and Schanz (2013) show that loss offsetting is allowed in a substantial number of the 100 countries they analyze worldwide. However, carry-forward losses are far more commonplace than carry-backward of losses.

Investments in tax-favorable local assets

Governments may establish regulations that make local assets tax favorable. One example is the monthly preferred income stock as examined by Engel et al. (1999). The authors show that corporations make use of the tax benefits of monthly preferred income stock and are willing to bear nontax costs if doing so allows for local tax optimization. Other locally tax-favorable assets may be available and can include tax-exempt municipal bonds (e.g., Drukker et al. 2020).

R&D tax credits

A common government incentive for firms to pursue R&D is R&D tax credits, which can manifest in proportional R&D investments tax credits, tax-accelerated depreciation schemes for R&D, tax exemptions, or a combination of these. The peculiarity and the country-specific details of the system most likely make this a local tax avoidance opportunity and various systems are in place worldwide (e.g., EY 2021). For the United States, Berger (1993) and Rao (2016) show that tax incentives encourage increased R&D investments. While Berger (1993) studies the effect of the tax credit effect on actual R&D spending and the implicit negative shareholder wealth effects of firms not benefiting from the new regulation, Rao (2016) examines both short- and long-run effects of R&D credits on the magnitude and components of research spending.

Within-country optimization schemes

Several countries not only levy taxes at the country level but also at the state or municipality level. In Europe, for example, local taxes are due in Austria, France, Germany, Hungary, Italy, Luxembourg, and Portugal (Fossen and Bach 2008).Footnote 6 For the United States, Dyreng et al. (2013) show that U.S. corporations use one specific U.S. state, namely Delaware, to allocate intangible assets to allow them to reduce their total state tax burden by 15 percent to 24 percent. Further, Gupta and Mills (2002) examine how corporate tax planning opportunities change as firms expand into new U.S. states that do not necessarily have more favorable reporting requirements, compared to the average state, before the expansion. Consistent with their expectations, they find that tax-planning opportunities increase when a firm operates in more than one but fewer than all states and that the minimum state ETR is reached when multistate firms operate in 24 states. Combined, the evidence suggests that local tax avoidance opportunities may also arise because firms exploit discrepancies within a country’s tax jurisdiction.

Similarly, the international evidence is consistent with the view that group tax optimization may also occur within one country or jurisdiction. Gramlich et al. (2004), for instance, show that Japanese keiretsu group members strategically shift income across affiliate companies. Shevlin et al. (2012) document that Chinese domestic groups shift income in response to regional tax incentives and that, in particular, intangible-intensive groups and groups with minimum equity requirements are the most frequent shifters. Beuselinck and Deloof (2014) find that Belgian business groups (holding companies) manage domestic earnings in response to tax incentives and that intra-group transactions facilitate such earnings management.

2.3 Local tax planning opportunities incentivizing cross-country shifting

This final category includes local tax planning opportunities that may either reduce the tax base or the tax rate to such a large extent that multinationals can be incentivized to shift assets or income to the respective country. The combination of local and global tax avoidance concertation can also be labeled “glocal” tax avoidance. Without claiming to be exhaustive in the summarized examples, we highlight four prominent cases of these kinds of tax avoidance schemes.

Special purpose entities

Special purpose entities can be created in one jurisdiction to act as flow-through entities and allow multi-state firms to shift income from high- to low-tax states (Fox and Luna 2005) or to allocate tax benefits to those group partners or investor classes that can best use such benefits. Feng et al. (2009), for instance, discuss how setting up these entities and combining this with synthetic leases may be useful when firms are facing high marginal tax rates because of the more favorable tax effect of capital versus operating lease treatment. Tighter regulations, including the Dodd-Frank Act of 2010 and increased investor scrutiny, however, have reduced special purpose entity tax planning opportunities over recent periods or at least have raised the bar on key compliance so that establishing these entities for tax benefits is potentially less prominent in more recent years.

Intellectual property box regimes

Several countries worldwide allow corporations to apply for so-called patent boxes (intellectual property boxes), which allow for deducting a large proportion of patent income for tax purposes. The use of patent boxes takes different forms and exists in several EU countries, such as Belgium, France, Hungary, Ireland, Italy, Luxembourg, the Netherlands, Spain, and the United Kingdom. Using the Belgian setting, Bornemann et al. (2020) find that firms with patents enjoy 7.2 percent to 7.9 percent lower effective tax rates, with the most significant financial benefits accruing to multinationals, compared to domestic firms.

Allowance for corporate equity

Several countries, including Belgium (since 2006), Brazil (since 1996), Italy (since 1997), Portugal (since 2010), and Turkey (since 2015), allow firms to deduct fictitious interest on corporate equity, thereby providing tax neutral treatment to debt and equity.Footnote 7 Hebous and Ruf (2017) show that favoring the tax neutrality of debt and equity reduces the corporate debt ratio in multinational affiliates and increases intra-group lending but has no effect on production investments. This observation comports with the critique that especially wealthy, capital-intensive groups can benefit from the tax planning opportunity and that economic rents merely flow to shareholders in the form of higher dividends rather than help creating new jobs (Zangari 2014).

Advanced tax rulings

Advanced tax rulings, such as advanced pricing agreements, are jurisdiction-specific administrative approaches that attempt to prevent transfer pricing disputes by determining the arm’s length criteria ex ante rather than disputing ex post. Advanced pricing agreements are designed to benefit both multinationals’ tax resource management as well as a country’s tax collections and should allow for eliminating double taxation, providing more tax certainty, and even increasing reputation through cooperation with tax authorities. Yet, at the heart of such agreements often lies the interpretation of economic nexus. In the European Union, for instance, advanced pricing agreements are commonplace but have been criticized because of examples where they were designed with low tax jurisdictions to avoid paying both local and foreign taxes (e.g., the LuxLeaks scandal (Bowers 2014)). Therefore the existence of bilateral or multilateral advanced pricing agreements may create local tax opportunities that multinationals over time have also designed in combination with a physical or de jure presence in specific jurisdictions.

3 Hypotheses development

An important research stream in the taxation literature examines the dynamics of corporate tax strategies. According to the Scholes-Wolfson framework, a corporate tax strategy refers to a system of decision-making activities and corporate actions with respect to tax planning that maximizes after-tax returns (Scholes et al. 2016). The traditional view on tax planning trade-off decisions has considered the relative importance of tax benefits against the associated nontax costs (e.g., Scholes et al. 1992). The literature that investigated nontax costs of tax planning strategies is extensive and has broadly focused on financial reporting consequences of tax decisions and the effects of agency costs on tax minimization. (See Shackelford and Shevlin (2001) and Hanlon and Heitzman (2010) for extensive overviews.)

One area of corporate tax planning decisions that received little attention so far relates to the dynamics of local versus international tax planning strategies. In fact, while the recent public debate on multinational tax avoidance has often focused on tax avoidance through profit shifting, the evidence presented in Section 2 also clearly illustrates that several other local channels exist through which internationally active firms can reduce their global tax bill. In principle, multinationals are expected to shift as much as possible and to claim local benefits until the relative cost of doing both becomes too high (Scholes et al. 2016). Research suggests that, although tax planning activities have increased over the last several decades, the relative cost of profit shifting may have grown in several jurisdictions worldwide and especially in Europe (Beuselinck et al. 2015; Dharmapala 2014). Alexander et al. (2020) observe a substantial increase in the tax base for a sample of 26 European countries, which they see as a likely reason for the decreasing trend in profit shifting over time. Therefore we first validate our new measure and hypothesize that multinationals reduce their profit shifting when the practice becomes more costly. Empirically, we expect an increase in the correlation between the groups’ ETRs and the weighted subsidiaries’ ETRs over time.

  • Hypothesis 1a: Multinationals reduce intra-group profit shifting when shifting costs increase.

Consistent with the idea that local tax avoidance serves as a counterbalance in times when shifting costs increase, we expect multinationals to rebalance their tax strategies and revert relatively more to local tax strategies, compared to before the cost increase. In other words, the relative net benefit of pursuing local tax strategies increases. Empirically, we expect that, while the correlation between the groups’ ETRs and the subsidiaries’ ETRs increases, the overall ETR does not change as groups rely more on local tax avoidance.

  • Hypothesis 1b: Multinationals increase local tax planning vis-à-vis profit shifting after shifting costs increase.

Although the economic intuition of relying more on local tax avoidance after a positive shock to the cost of global income shifting is intuitive, it is also important to consider the possibilities of multinationals in doing so. Theoretically, if all firms would be fully tax-efficient, then, all else being equal, they would have exhausted local tax planning opportunities until the marginal cost of doing this exceeded the marginal benefit. However, it is likely that cross-sectional variation in tax avoidance practices is observed and that income shifters by default react more strongly to increased shifting costs, compared to firms that were already focusing more on local tax avoidance. We therefore conjecture that firms for which income shifting is relatively more important are more likely to substitute income shifting by local tax avoidance when shifting costs increase.

At the same time, recent studies (Ortmann and Schindler 2021; Klassen and Valle Ruiz 2022) suggest that tax-efficiency updates in income shifting indirectly affect managers’ incentives and that the way in which firms respond to this via adjusting incentive packages signals their tax aggressiveness. The analytical model of Ortmann and Schindler (2021), for instance, predicts that multinationals update incentive contracts of affiliate managers to compensate for any offsetting effect of intangibles shifting. Therefore, if profit shifting gets more expensive, the group is expected to respond by changing the incentive contract of the local manager to pursue local tax avoidance strategies at the subsidiary level. The combined arguments lead to the expectation that multinationals with a relatively high proportion of profits realized in low-tax affiliate countries, that is, the ones that have the opportunities to shift, are the ones that are more likely to respond by substituting income shifting by local tax avoidance after an increase in profit-shifting costs. This combined rationale results in our Hypothesis 2a.

  • Hypothesis 2a: Especially multinationals identified as ex ante income shifters substitute local tax planning for income shifting after profit shifting costs increase.

In addition, the demand for a target ETR may be highly inelastic so that managers are willing to bear higher nontax costs to achieve it. Under these conditions, managers who bear the largest pressure may be willing to incur higher nontax costs to achieve an optimal target ETR. This type of pressure may relate to a firm’s public listing status as well as its historical tax efficiency. Public groups, for instance, are relatively more sensitive to capital market pressures than private groups (Graham et al. 2011) and may therefore respond more to target ETR pressures. Relatedly, Blouin et al. (2012) find that repatriation of foreign earnings as a dividend to the U.S. parent by public firms is more sensitive to the repatriation tax rate than repatriation by private firms.

Also, target ETR pressure may depend on whether a firm is above or below its optimal tax avoidance level. Kim et al. (2019) show that the speed of adjustment toward its target ETR is almost twice as high if a firm is below its optimal level of tax avoidance, compared to when it is above its optimal level.Footnote 8 Combined, this leads to the expectation that multinationals with higher target ETR pressure (proxied by their public listing status and by their ETR level, compared to industry-country peers) are more likely to respond by more local tax avoidance after an increase in profit shifting costs, compared to multinationals with lower target ETR pressure:

  • Hypothesis 2b: Especially multinationals with larger target ETR pressure substitute local tax planning for income shifting after profit shifting costs increase.

4 Research method

4.1 Basic model

Observed multinational tax planning is a result of the trade-off between both cross-jurisdictional income shifting and subsidiary local tax planning. The first contribution of our paper therefore lies in the identification of local tax planning that is contributing to the total tax planning. Although it is both conceptually and empirically challenging to isolate the local versus global components of a corporate tax planning strategy, we propose an estimation method that combines both group-level and subsidiary-level data. In essence, the measurement technique focuses on the association between the multinational group-level tax planning (proxied by ETR) and a weighted subsidiary-level tax planning. If tax planning at the group level is strongly associated with the subsidiary-level tax planning, this indicates local tax planning strategies that contribute to group-level tax avoidance. On the contrary, if there is only a weak or a zero association between group- and subsidiary-level tax planning, this suggests that the multinational realizes its consolidated tax strategy by other means than subsidiary-level local tax strategies. Tax avoidance in this case is likely to result from global income shifting.

To identify the proportion of tax planning that originates from subsidiary local tax planning versus total multinational (group) tax planning, we analyze the relationship between the multinational consolidated effective tax rate (ETR) and the pre-tax income weighted ETR of the domestic and foreign subsidiaries based on unconsolidated data (wSubETR). First, the effective tax rate (ETR) is calculated as GAAP tax expense divided by GAAP pretax income. In our empirical quantification, we start by computing the effective tax rate for each group and each subsidiary. We can perform this type of analysis since our dataset (described in more detail in Section 5.1) allows us to observe unconsolidated (subsidiary-entity) financial statements of domestic and foreign affiliates that are majority-owned by global multinationals. The pretax income that is reported in unconsolidated financial statements is the source-country income that is subject to local tax. Notably, this is the income reported in a country after potential profit shifting into or out of that country. Next we compute the weighted average (by pretax income, PTI) ETR for all subsidiaries of a given multinational to obtain one measure of tax planning of all its subsidiaries in year t. This measure can be interpreted as the weighted local tax planning within jurisdictions where the subsidiaries are located (wSubETRg,t) and where the weight is formed by the level of the subsidiary taxable income.

$$wSubETR_{g,t}=\frac1{\sum_{s=1}^mPTI_{s,t}}\ast\sum\limits_{s=1}^mETR_{s,t}\ast PTI_{s,t}.$$

Next we define the effective tax rate (ETR) of the group based on its consolidated statements. We then regress the ETR of the group (ETRg,t) on the pre-tax income weighted subsidiary ETR (wSubETRg,t) to measure to what extent the parent’s tax planning is associated with the subsidiaries’ tax planning. A coefficient of zero would indicate that there is no association between realized multinational tax planning and local tax planning in subsidiaries. This result of a zero correlation indicates tax planning that is realized via income shifting, as it is not related to any subsidiary country tax planning. A coefficient of one would indicate that the parent’s tax planning is fully captured by the subsidiaries’ local tax planning strategies. A significantly positive coefficient of β1 (with β1 ranging between 0 and + 1) indicates subsidiary local tax planning contributing to x % of multinational total tax planning.Footnote 9 The model of interest goes as follows.

$$ET{R}_{g,t}={\beta }_{0}+{\beta }_{1}*wSubET{R}_{g,t}+control{s}_{g,t}+Fixed\_Effects{ + \varepsilon }_{g,t}.$$

4.2 Global versus local tax avoidance strategies

In Section 2, we discussed local, global, and combined tax avoidance strategies. In this section, we discuss how relying on a different combination of both ends of the spectrum of tax planning opportunities affects the correlation between the consolidated ETR and the observable pre-tax income-weighted subsidiaries’ ETR.

Model (2) above suggests that relying more on profit shifting into unobservable tax havens would decrease the correlation between the consolidated ETR and the observable pre-tax income-weighted subsidiaries’ ETR, and vice versa. In the scenario when profit shifting becomes more costly, for example, with the introduction of transfer pricing documentation requirements, the expectation is that correlation between the consolidated ETR and the observable pre-tax income-weighted subsidiaries’ ETR would increase. In other words, a relatively higher weight on local tax avoidance opportunities, compared to shifting income away, predicts that more tax opportunities are exploited locally. As such, the correlation serves as an inverse measure of profit shifting, and a higher correlation coefficient indicates less shifting.

4.3 Control variables and fixed effects

We insert determinants as control variables that research has identified to be important drivers of corporate ETR (e.g., Chen et al. 2010; Desai and Dharmapala 2009; Gupta and Newberry 1997). First, we control for a firm’s size (SIZE), proxied by the natural logarithm of firm assets. In line with the work of Mills et al. (1998) and Rego (2003), we expect SIZE to relate negatively to ETRs, since large firms are expected to do more effective tax planning. However, in line with the political cost argument of Zimmerman (1982), SIZE may also relate positively to ETRs. Second, we control for a firm’s pretax profitability. Following the arguments of Gupta and Newberry (1997), we expect that, under the condition of stable tax preferences and for a given level of total assets, ETR relates negatively to ROA. This result is also predicted from the perspective that multinationals with higher levels of pre-tax income have more opportunities to reduce their overall tax burdens through tax planning (e.g., Rego 2003). Third, we control for capital intensity (PPE) and interpret this variable as a proxy for a firm’s asset mix. In line with the idea that tax benefits are associated with capital investments, we expect that capital-intensive firms should face lower ETRs (e.g., Gupta and Newberry 1997). Fourth, we control for the level of capitalized intangibles (INTANG), as more intangible firms can benefit from favorable tax treatments for research and development (e.g., patent boxes).Footnote 10

Fifth, we include the leverage ratio (LEV) to control for a firm’s financing policy. The tax codes generally accord differential treatment to the capital structure of firms because interest expenses are deductible for tax purposes, whereas dividends are not, leading to the expectation that firms with higher leverage would have lower ETRs. However, a positive relation between ETRs and leverage is possible if firms with high marginal tax rates are more likely the ones that can attract and use debt financing (Gupta and Newberry 1997). Next we include a dummy, which is coded one if the respective group had a loss in the previous years (LAGLOSS). As tax-loss carryforwards are unobservable but apply in most of the observed institutional settings under study, LAGLOSS captures these to some extent. Since multinationals are also exposed to the economic conditions of the host countries of their affiliations, we furthermore control for the subsidiary asset-weighted change in GDP, asset-weighted inflation, and the asset-weighted tax attractiveness (Schanz et al. 2017). Lastly, our fixed effect structure includes, depending on the model, country-year fixed effects to control for institutional changes such as changes in the corporate tax rates, and firm fixed effects (multinational FE) to absorb time-invariant unobservable firm characteristics.

4.4 Staggered adoption of transfer pricing documentation requirements

One cost component that makes income shifting presumably more costly is the introduction of transfer pricing documentation requirements. Transfer pricing documentation requirements have been increasing in importance, as evidenced by the share of international transactions covered by documentation rules rising from 12% in 2003 to about 80% in 2013 (Beer and Loeprick 2015). Alexander et al. (2020) report that profit shifting has declined by over 50% over the period of 2003–2013 and especially after the introduction of transfer pricing documentation requirements. In line with the logic that regulatory changes in transfer pricing documentation affects tax risk, Mescall and Klassen (2018) consider the documentation requirements as an input variable for transfer pricing risk.

In the main tests that relate to our hypotheses, we augment the model with a treatment effect (TREATg) related to the multinational home-country staggered adoption of transfer pricing documentation requirements. The variable is set equal to zero in all years before the introduction of the requirements and is equal to one from the introduction year onward. The interaction of the treatment effect with the weighted effective tax rates of the subsidiaries (\(TREA{T}_{g,t}*wSubET{R}_{g,t}\)) allows for isolating the effect of increased transfer pricing costs on the local tax planning dynamics. (See the Appendix Table 8 for a detailed country overview of transfer pricing documentation requirements adoption years.) We include multinational fixed effects (firm fixed effects) to account for time-invariant firm characteristics. Including multinational fixed effects makes the model effectively a difference-in-differences research design. This design allows for correct inferences about a true treatment effect in contexts where countries adopt regulations at different times or do not adopt at all (e.g., Simintzi et al. 2015). We include multinational fixed effects in our specification, as this allows us to interpret a firms’ pre-period ETR as the normal level of ETR and abnormal ETR (the interaction term) as the deviation from the pre-period.Footnote 11 The augmented model reads as follows.

$$ \begin{aligned} ET{R}_{g,t} &= {\beta }_{0}+{\beta }_{1}*wSubET{R}_{g,t}+{\beta }_{2}*TREA{T}_{g,t}+{\beta }_{3}*TREA{T}_{g,t}\\ &\quad*wSubET{R}_{g,t}+Fixe{d}_{Effects}+control{s}_{g,t}+{\varepsilon }_{g,t}. \end{aligned}$$

In line with our combined hypotheses H1a and H1b, we predict a significantly positive coefficient for β3, which is the difference-in-differences estimator. A positive coefficient would reflect increased reliance on local tax planning after the introduction of transfer pricing documentation. We additionally run model (3) on subsamples to test the cross-sectional (H2a and H2b) hypotheses.

5 Sample and results

5.1 Sample and descriptive statistics

The sample contains all nonfinancial groups from 27 EU member states and their global subsidiaries from the Bureau van Dijk Orbis database covering the period 2006 to 2014.Footnote 12 This database contains information on the (most recent) ultimate owner of each corporation, which we use to construct corporate groups. We exclude purely domestic (i.e., single-jurisdictional) groups, since these firms can only avoid taxes locally and therefore cannot trade off local tax planning with cross-jurisdictional income shifting. For each EU member state, we download the consolidated parent financial data and the unconsolidated subsidiary-level data. Subsidiaries are defined as such if the parent company directly or indirectly owns at least two-thirds (67%) of the shares.Footnote 13 This search strategy allows us to combine all unique subsidiary observations with their ultimate parent. We exclude observations with missing data on pretax income and total assets and for which we have missing data on control variables, for firm-years with negative pretax income or negative tax expense, firm-years with a tax rate above 100% of pre-tax income, and subsidiaries with net income of exactly zero.

The final subsidiary-level dataset consists of 154,022 subsidiary-year observations from 64 countries from both within and outside the European Union. This sample also corresponds to 32,849 multinational group-year observations headquartered in the European Union. In our sample, we capture a significant portion of the overall groups’ activities. For example, the sum of all subsidiary assets to consolidate assets is around 38%, the sum of all subsidiary pretax income to consolidated pretax income is around 63%, and the sum of all subsidiary tax expense to consolidated tax expense is around 58%.Footnote 14 In Table 1, we provide country-level details on the locations of subsidiaries (rows) and the origins of the respective groups (columns).

Table 1 Location of groups and subsidiaries

For expositional purposes, we separately show the multinational parent/subsidiary observations only for parent countries with at least 1,000 observations and subsidiary countries with at least 500 subsidiary-year observations. The parent countries for which this is the case are, in alphabetical order, Austria, Belgium, Denmark, Finland, France, Germany, Ireland, Luxembourg, Netherlands, Poland, Portugal, Spain, Sweden, and United Kingdom. In the interest of readability, the observations of all other parent countries (N = 12) are pooled in the final column (Other), and all other subsidiary countries (N = 32) are pooled in the final row (Other). As shown in Table 1, we observe most subsidiary locations (rows) in the United Kingdom (GB: 18,720), followed by Spain (ES: 16,397) and France (FR: 15,314). In terms of the multinational parent-origin (columns), we observe that multinationals from Germany (DE) have the most subsidiaries (39,980), followed by the United Kingdom (GB: 21,748) and Sweden (SE: 14,733). Further, a large fraction of the observed subsidiaries is located domestically. For example, the highest fraction of local subsidiaries is observed in the United Kingdom (GB/GB: 10,660).

In Table 2, we observe that the mean (median) weighted subsidiary-level ETR is 23.6% (24.2%) and the interquartile range goes from 17.5% to 29.2%. At the consolidated group level, numbers are very similar, albeit slightly higher: the mean (median) ETR is 28.4% (27.1%), and the interquartile range goes from 20.9% to 33.2%. While such average and median ETRs are consistent with rates reported in research in a U.S. setting (e.g., Dyreng et al. 2017), the top quartile of observed ETRs is substantially higher.

Table 2 Summary statistics

The average (median) group has 4.7 (2.0) subsidiaries (#SUBSg) in the final sample. In terms of profitability (ROAg), groups are on average highly profitable (mean = 9.7%; median = 7.5%). The average group has 9.0% of its balance sheet total in capitalized intangibles, and the maximum level of intangibility is 67.2%. Mean (median) level of PPE is 24.5% (21.1%). The average multinational has total assets of about €128.8 million and financial leverage (including short and long-term debt) of 57.5%. Finally, 5.9% of the observations recorded a negative income in the pre-observation year, and about one quarter (24.9%) of all observations relate to publicly listed groups.

The correlation table (Table 3) provides initial evidence that group-level tax planning, measured as the group-level effective tax rates (ETRg), is positively correlated with the pre-tax income-weighted tax planning of its subsidiaries (wSubETRg). The Pearson correlation between ETRg and wSubETRg is 0.21, and the Spearman rank correlation is 0.15 (both statistically significant at the 1% level). Furthermore, and only considering correlation coefficients larger than |0.05|, Table 3 suggests that the consolidated ETRg relates positively to INTANGg (+ 0.07; p < 0.01), and LEVg (+ 0.12; p < 0.01). At the same time, ETRg relates significantly negatively to ROAg (−0.20; p < 0.01) and the group listing status (PUBLIC: −0.08; p < 0.01).

Table 3 Correlations

5.2 Basic model

In this section, we provide results on the association between the level of subsidiary (local) tax planning and the multinational total tax planning.

Table 4 reports the regression results for the validation analyses of our local tax planning tests. In the first column, we quantify the association between the group-level tax planning (ETRg) and the pretax income-weighted effective tax rate (wSubETRg) within subsidiary affiliate countries. Column (2) is identical but additionally includes year fixed-effects and country-fixed effects, and Column (3) includes multinational fixed effects and country-year fixed effects. In all columns, we observe that group tax planning is positively associated with the subsidiary local tax planning. In particular, we observe in the baseline model that a one percent change in subsidiary tax planning contributes to a 19.1 basis point change in multinational tax planning (12.8 basis points after controlling for multinational fixed effects in Column 3).Footnote 15 Not surprisingly, multinational fixed effects increase the explanatory power of the model substantially. Hence we report results including multinational fixed effects in all further tables.

Table 4 Regression results – local tax avoidance

Next we provide several figures to provide more details about the correlation coefficient in the time series and cross-section. Figure 1 includes a time trend of the coefficient and shows that the coefficient was slightly above 0.10 in 2006 yet increased to around 0.21 in 2014. Regressing the yearly correlation coefficients on a time trend yields an intercept of 0.103 and a statistically significant slope coefficient of 0.012 (t-stat: 5.52). This finding indicates a decreasing trend of profit shifting and comports with the findings of Alexander et al. (2020), who find a downward trend in profiting shifting between 2003 and 2013, further confirming that the correlation between the consolidated ETR and the pretax income weighted subsidiary ETRs successfully captures (the inverse of) profit shifting. In Fig. 2, we plot the coefficient by country and find the strongest coefficient in Belgium and the lowest coefficient in Austria. Finally, in Fig. 3, we find positive coefficients in all industries, but the variation is less pronounced compared to across countries. One exception is real estate activities, an industry with few options to shift income across countries and for which we find the largest coefficient. In line with this graphical evidence, we additionally observe that firms with more profit shifting opportunities (larger firms and intangible asset-intensive firms) have lower coefficients.Footnote 16 Overall these initial cross-sectional examinations corroborate that our measure successfully captures profit shifting; that is, firms with more (less) profit shifting opportunities have a lower (higher) coefficient.

Fig. 1
figure 1

Yearly time trend. The figures show the coefficients of wSubETR, the by pretax income weighted effective tax rates of the groups’ subsidiaries, of separate regressions by year. Regressing the yearly correlation coefficients on a time trend yields an intercept of 0.103 and a statistically significant slope coefficient of 0.012 (t-stat: 5.52)

Fig. 2
figure 2

Coefficient by country. The figures show the coefficients of wSubETR, the by pretax income weighted effective tax rates of the groups’ subsidiaries, of separate regressions by country

Fig. 3
figure 3

Coefficient by industry. The figures show the coefficients of wSubETR, the by pretax income weighted effective tax rates of the groups’ subsidiaries, of separate regressions by industry

5.3 Staggered adoption of transfer pricing documentation requirements

In this section, we formally test Hypothesis H1a and examine whether a plausibly exogenous increase in multinational shifting costs, namely the country-level introduction of transfer pricing documentation requirements, alters the average observed balance between local and global tax planning. Table 5 reports results employing a difference-in-differences methodology that exploits intertemporal variations in transfer pricing documentation, that is, requirements that are assumed to mitigate tax-induced profit shifting costs and opportunities. Recall that TREAT is a dummy variable equal to one at or after the introduction of transfer pricing documentation requirements and zero alternatively.

Table 5 Transfer pricing documentation requirements introduction – local tax avoidance

We observe that the association becomes significantly more pronounced in the post- transfer pricing requirement period (wSubETRg*TREATg =  + 0.058, p < 0.01). In relative terms, local tax planning is around 46% higher compared to the pre-transfer pricing documentation requirement period (0.058/0.127 = 0.457). Results are very similar in column 2, where we introduce time-varying firm-level covariates. Finally, a qualitatively similar result is observed when we additionally introduce multinational country-year fixed effects to control for country-specific tax rate changes (column 3).Footnote 17 These results are in line with H1a and support the conjecture that multinationals reduce profit shifting after shifting costs increase.

To overcome the potential concern that our difference-in-difference design suffers from reverse causality and mistakenly captures the change in multinational behavior prior to the change in regulations, we follow an approach similar to that of Bertrand and Mullainathan (2003) to ensure that our difference-in-differences experiment satisfies the parallel trends assumption. To do so, we perform a timing test where we decompose the introduction of transfer pricing documentation requirements into separate periods for each country. Specifically, we replace the transfer pricing documentation requirement interaction variable with several time-specific interaction variables and activate these variables in specific periods prior to and after the requirement introduction year. For instance, the variable wSubETRg*TREATg (≤ −2) represents the relative importance of local tax planning in the multinational total tax planning two years or earlier before the introduction of transfer pricing documentation requirements. The variable wSubETRg*TREATg (= −1) refers to the same change in weight the year before the introduction of the requirements, etc. If reverse causality were a concern, then we would expect to observe a trend in the local tax planning coefficient prior to the requirement’s introduction.

Results in Table 6 show that the reduction in profit shifting is only observed from the year of introduction of transfer pricing documentation requirements. In fact, the relative importance jumps by about 50 percent (= 0.063/0.123) in the introduction years and manifests with a similar magnitude in later years (≥ + 2). The coefficient for the post-introduction year (+ 1) is positive too, albeit not significant at conventional levels. Results are very similar if we exclude or include time-varying firm-level covariates.Footnote 18

Table 6 Timing effects of transfer pricing documentation requirements on local tax avoidance

5.4 Cross-sectional variance and substitution analyses

Next, in Table 7, we investigate in further detail whether multinationals substitute local tax avoidance for income shifting post transfer pricing documentation requirements (Hypothesis H1b). To this end, we first estimate our model with a TREATg effect only (i.e., without combining this with the subsidiary-level ETRs), to capture whether transfer pricing documentation requirements have impacted average ETR levels. Interestingly, Regressions (1) and (2) of Panel A suggest no significant effect on the average ETR as the coefficient is close to zero and not significant at conventional levels in any of the regression specifications. Results are similar for regressions with and without the insertion of time-varying firm-level covariates. Overall these results suggest that transfer pricing documentation requirements did not result in higher ETRs on average. This finding, in combination with our previous results and other observations that transfer pricing documentation requirements did result in a significant reduction in profit shifting over time (Alexander et al. 2020; Beer and Loeprick 2015), suggests that multinationals may have applied alternative tax planning strategies, allowing them to maintain a target ETR level. This finding also suggests that the demand for a target ETR is inelastic on average.

Table 7 Substitution versus mechanical relation

Our cross-sectional hypotheses predict differences in the relative importance of local tax planning after compared to before transfer pricing documentation requirements by (i) the ex ante likelihood of income shifting (H2a) as well as by the degree of target ETR pressure in the respective groups (H2b). To test H2a, we partition our sample of multinationals into two groups based on the group’s ex ante likelihood to realize tax savings via income shifting to unobserved tax havens versus other groups. This split is warranted given the findings of Klassen et al. (2017) that some corporations focus on tax minimizing strategies in setting transfer prices while others focus on tax compliance. We operationalize this sample split in two ways. First, by labeling multinationals with a group ETR below the income-weighted subsidiary ETR as income shifting suspect firms (ETRg < wSubETRg), whereas other multinationals are labeled as nonshifters (Panel B: column 1 versus column 3). The logic for this partition is that groups falling under this criterion by default are applying tax planning strategies that push their ETR below the sum of the observable parts. Second, we partition by identifying groups that have their summed observed taxable subsidiary income located in low STR countries (Panel B: column 2 versus 4).Footnote 19

In line with our classification logic, we observe higher correlation coefficients between the consolidated ETR and the income-weighted subsidiary ETRs in subsamples identified as nonshifters. Interestingly, we observe that post transfer pricing documentation requirements, multinationals identified as income shifters (Panel B: columns 1 and 2) substitute local tax avoidance for income shifting by a sizeable proportion, while we observe no change for other multinationals (Panel B: columns 3 and 4). For the first identification of shifter groups (i.e., ETRg < wSubETRg), we observe an increase by about 30 percent (0.073/0.267) in local tax avoidance post transfer pricing documentation requirements. For the second identification method (i.e., income-weighted subsidiary-level STRs belonging to the bottom quartile of observations), we even observe an increase in the correlation coefficient of 78 percent (0.088/0.113). These findings confirm our conjecture that, post transfer pricing documentation requirements, especially multinationals identified as ex-ante income shifters are relying more on local tax avoidance schemes after regulations made shifting more costly.

To test hypothesis H2b, we proceed similarly and partition our sample into two buckets based on the expected target ETR pressure. Because capital market pressures are highly relevant for public groups that are mostly absent for private groups (e.g., Ball and Shivakumar 2005; Beuselinck et al. 2021; Burgstahler et al. 2006), we study Eq. (3) separately for public versus private multinationals. Second, we partition the multinationals into two groups where the separation stems from a firm’s historical tax avoidance level, relative to its industry-country peers. Groups that realize ETRs below their industry-country peers may face stronger pressure to keep their ETR advantage compared to less tax-efficient groups. We define all multinationals with ETRs below (above) the sample median as high (low) ETR pressure groups; finding evidence that high target ETR pressure firms are more likely to respond by more local tax avoidance, compared to multinationals with lower target ETR pressure, would be consistent with hypothesis H2b.

We observe that high target ETR pressure firms identified as publicly listed groups as well as multinationals that realize a historically low tax pressure, relative to their industry-country peer groups, respond by increased local tax avoidance post transfer pricing documentation requirements, while we observe a zero effect for the control groups (Panel C: column 1 versus column 3). In particular, for public multinationals, the interaction term is highly significant and nearly doubles (wSubETRg * TREATg = 0.107; p < 0.01) in the post transfer pricing documentation requirement period and remains unaltered for control group observations (PUBLIC = 0). The coefficient also increases sharply by 94% (= 0.046/0.049) in the multinational bucket with historically below-median ETRs and remains stable for the control group. This evidence is consistent with hypothesis H2b. Combined, these cross-sectional tests further confirm that the observed associations between group and subsidiary level ETRs pre versus post transfer tricing documentation requirements are not random movements but rather behave in ways as predicted by theoretical argumentations and economic sense.

5.5 Robustness tests

First, we re-run our main analysis on a sample where we exclude observations of groups with a presence in countries with advanced tax rulings (i.e., Luxembourg and the Netherlands), countries with notional interest deductions (i.e., Belgium and Cyprus), or countries with hybrid mismatches (Ireland). Across all specifications, our results continue to hold, and we find coefficients of similar magnitude. Second, we carefully control for potential effects of unobservable net operating losses. Unlike in the United States, it is impossible to observe net operating losses (NOL) in the European sample. Empirically, we address this concern by controlling for tax-loss offsetting by including a proxy for the previous year’s losses. To further increase understanding of how tax-loss offsetting potentially affects our empirical results, we use the information provided in Appendix B of Kohlhase and Pierk (2021) to split our sample based on whether the respective country allows for unlimited tax-loss carryforward or if tax-loss carryforwards are capped after a specific period. In untabulated results, we find that the correlation remains positive and is highly significant in both subsamples, suggesting that imprecisely observing tax-loss offsetting does not alter our results.

6 Conclusion

The current study proposes a novel calculation to identify the importance of and dynamics in local tax strategies vis-a-vis profit shifting. Our method is based on the observed relationship between the multinational consolidated effective tax rate (ETR) and the sum of the (income-weighted) subsidiary-country ETRs and aims to introduce a measure that can complement existing profit shifting studies in identifying tax optimization strategies.

Our main results show that the consolidated tax planning of the average multinational in our sample relates significantly to the subsidiaries’ tax planning. This finding is consistent with the conjecture that multinationals’ tax planning is partly explained by its affiliate country tax planning and is not originating exclusively from cross-jurisdictional income shifting. We validate our measure by showing it captures decreased profit shifting after the introduction of transfer pricing documentation requirements.

Our findings are consistent with the Scholes-Wolfson framework and suggest that firms pursue tax planning opportunities that optimize after-tax returns (Scholes et al. 2016) in such a way that when shifting costs increase, multinationals rebalance tax-planning strategies toward less costly local tax planning opportunities. Our findings also have important policy implications. First, our study suggests that multinationals trade off both local and global tax planning strategies and that, when the cost structure of the tax strategy components changes, corporations flexibly adjust tax planning strategies toward the relatively less costly technique. Therefore it is important that tax regulators and policy debates also consider within-country tax planning as a potentially important tax planning mechanism. In addition, our findings suggest that in times where the pressure for a fairer tax game is growing and initiatives are launched to curtail profit shifting opportunities (OECD 2013), multinationals can work around regulations via updated tax planning strategies and maintain an ETR relatively close to their target ETR.

Our study has limitations. First, country differences in reporting requirements for private entities may result in incomplete coverage of the web of multinational subsidiaries, which may weaken the validity of our proposed measure. For example, countries like the United States have no reporting requirements for private firms, and data is consequently not available for all legal entities. Second, affiliate-level data does not necessarily allow for precisely determining how much of the consolidated group result is impacted by intercompany reporting and consolidation methods. Blouin and Robinson (2020), for example, show that aggregating net income across affiliates will lead to double counting of at least some income in complex group structures, especially when a directly owned subsidiary itself owns another entity and reports its profit as equity income. Double counting of income in such a way would also bias affiliates’ ETRs downward. For both reasons, we wish to caution researchers that the proposed metric ideally should not be used in isolation. However, in combination with an exogenous shock or regulation change, the proxy can yield interesting insights into the dynamics of local versus global tax planning strategies. Third, we discuss global versus local tax strategies based on the literature and the popular press, but we acknowledge that a clear distinction is not always possible and eventually one also affects the other. Our measure and analyses are only a step toward a better understanding of the dynamics of global and local tax planning dynamics.

Future research might consider changes other than tax law regulations affecting multinational decisions to substitute local tax planning for income shifting or can investigate the governance determinants of local versus global tax planning decisions. Furthermore, our results suggest that there exists considerable variation across countries. Further exploring such country-level variation could be a fruitful area of research, especially for local researchers with advanced knowledge of the specific tax rules.