In this section I examine whether the theoretical predictions of the model hold for European multinational firms and their majority-owned European affiliates over the sample period (2003 – 2014). At first I analyse how the total debt-to-asset ratio is affected by the tax mechanisms. Then I examine the potential omitted variable biases associated with omitting any of the tax mechanisms from the specification. Furthermore, I assess the economic importance of the estimated coefficients on the tax mechanisms. Finally, I discuss how the total debt-to-asset ratio is affected by the control variables.
The main regression results are presented in Table 3. For each observation, the weighted tax difference and maximum tax difference variables have been constructed. All regressions in the table control for parent, industry and year fixed effects. The R-squared values reported in all regressions are not adjusted for variance explained by the fixed effects variables (parents, industries and years). This implies that the overall effect of fixed effects variables on the fit of the model is not quantified; hence, the reported R-squared values are rather small.
Regression (1) includes only the three tax mechanisms, and the results show that all three coefficients of interest are positive and statistically significant at the one percent level. The significance of the two international debt shifting mechanisms implies that an affiliate’s leverage reflects the overall international tax system faced by the multinational corporation. Furthermore, regression (2) includes also the control variables, which leads to a decrease in the estimated coefficients on the tax mechanisms, as compared to regression (1). Coefficient on the statutory corporate tax rate decreases by 12.1 percentage points, coefficient on the weighted tax difference variable decreases by 6.3 percentage points, and coefficient on the maximum tax difference variable slightly decreases by 0.3 percentage points. This implies that there exists a substantial subsidiary heterogeneity characterizing their leverage decisions, which is captured by the subsidiary-specific control variables. However, after inclusion of the control variables, the estimated coefficients on all tax mechanisms are still statistically significant and positive.
Focusing on regression (2), the estimated size of 𝛽1, 0.164, reflects the effect of host
rate increases by one percentage point, affiliate’s total debt-to-asset ratio increases by 0.164 percentage points. The estimated coefficient is similar to the estimates found by Huizinga et al. (2008, p. 101) and Møen et al. (2011, p. 24), which are 0.184 and 0.197 respectively.34 Furthermore, the two international debt shifting variables capture the effects which apply for multinational firms only. Firstly, the estimated size of 𝛽2, 0.054, reflects the effect of weighted tax difference on affiliate’s total leverage. When the weighted tax difference increases by one percentage point, affiliate’s total debt-to-asset ratio increases by 0.054 percentage points. The estimated coefficient is 6.6 percentage points smaller than the coefficient reported by Huizinga et al. (2008, p. 101), and 22.5 percentage points smaller than the coefficient reported by Møen et al. (2011, p. 24).35 Furthermore, the estimated size of 𝛽
3,
0.051, reflects the effect of maximum tax difference on affiliate’s total leverage. When the maximum tax difference increases by one percentage point, affiliate’s total debt-to-asset ratio increases by 0.051 percentage points. The estimated coefficient is 6.9 percentage points lower than the coefficient reported by Møen et al. (2011, p. 24).
34 Huizinga et al. (2008) refers to the variable as the “domestic” effect, as the coefficient on the
statutory corporate tax rate reflects the impact of taxation on the optimal leverage ratio that applies for both purely domestic firms and multinational firms (p. 95).
35 Huizinga et al. (2008) refers to the variable as the “international debt shifting” effect, as their
specification includes only the weighted tax difference variable as the effect applying for multinational firms only (p. 95). The authors disregard the maximum tax difference variable (internal debt shifting mechanism) in their analysis.
Table 3: Impact of tax mechanisms on total debt-to-asset ratio
The dependent variable in all regressions is the total debt-to-asset ratio. Detailed variable definitions are given in Table 2. Regression (1) includes only the three tax mechanisms as independent variables. Regression (2) adds control variables to the specification. Regressions (3) and (4) examine the omitted variable bias occurring if maximum tax difference or weighted tax difference variables are omitted from the analysis. Regressions (5) to (7) examine the omitted variable bias occurring if two tax mechanisms are omitted from the analysis. The regressions are estimated by the ordinary least squares and include parent, industry and year fixed effects. The sample consists of majority-owned affiliates of European multinational firms over 12 years (2003 – 2014). White’s (1980) heteroskedasticity-robust standard errors are reported in the parentheses. * denotes significance at 10% level, ** denotes significance at 5% level, *** denotes significance at 1% level.
(1) (2) (3) (4) (5) (6) (7)
Statutory tax rate 0.285*** 0.164*** 0.190*** 0.197*** 0.248***
(0.016) (0.016) (0.015) (0.013) (0.010) Weighted tax difference 0.117*** 0.054*** 0.076*** 0.232*** (0.018) (0.017) (0.016) (0.011) Maximum tax difference 0.054*** 0.051*** 0.062*** 0.174*** (0.012) (0.012) (0.011) (0.008)
Fixed asset ratio -0.065*** -0.065*** -0.065*** -0.065*** -0.065*** -0.065***
(0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Log (Sales) 0.027*** 0.027*** 0.027*** 0.027*** 0.028*** 0.028*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Loss carry-forward 0.093*** 0.093*** 0.093*** 0.093*** 0.093*** 0.093*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Profitability -0.044*** -0.044*** -0.044*** -0.044*** -0.044*** -0.044*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Inflation 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** 0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Log (Corruption index) -0.009*** -0.009*** -0.009*** -0.008*** -0.004*** -0.000 (0.002) (0.002) (0.002) (0.002) (0.002) (0.001) Growth opportunities 0.022*** 0.022*** 0.022*** 0.022*** 0.021*** 0.021*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002)
Log (Creditor rights
index) -0.037*** -0.037*** -0.038*** -0.038*** -0.040*** -0.043***
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
Lowest-taxed
affiliates excluded No No No No No No No
Parent, industry, year
fixed effects Yes Yes Yes Yes Yes Yes Yes
Number of
observations 1,039,827 1,039,827 1,039,827 1,039,827 1,039,827 1,039,827 1,039,827 Number of parent
firms 143,405 143,405 143,405 143,405 143,405 143,405 143,405
The tax mechanisms are interrelated both between each other and among all affiliates worldwide that belong to the multinational corporation. Firstly, the tax variables are correlated by construction (each tax mechanism includes the host country corporate tax rate
𝑡𝑝𝑖𝑡), with correlation coefficients approximately 0.4, as observable in Table 4. Furthermore,
a change in the corporate tax rate in a country j affects both the leverage decisions of affiliates which are located there, and the leverage decisions of all other affiliates worldwide that belong to the multinational corporation through the weighted tax difference variable. An increase in a country’s corporate tax rate increases the weighted tax difference of affiliates located there, which consequently leads to higher total debt-to-asset ratios of these affiliates. However, the weighted tax difference decreases for affiliates located in other countries, which leads to lower total debt-to-asset ratios of these affiliates. For an affiliate pi, a change in the host country corporate tax rate, 𝑡𝑝𝑖𝑡, by one percentage point affects its total leverage through all three tax mechanisms; thus, the total effect equals 𝛽1+ 𝛽2(1 − 𝜌𝑝𝑖𝑡) + 𝛽3.36 As
observable from the equation, the total effect on the debt-to-asset ratio decreases in the relative size of affiliate pi, as shifting external debt from a small to a large affiliate constitutes a larger change in the total debt-to-asset ratio of the small affiliate than of the large affiliate.
Table 4: Correlation matrix between tax mechanisms
The tax mechanisms are correlated by construction. This table shows the pairwise correlation estimates between the tax variables. Detailed variable definitions are given in Table 2.
Statutory tax rate Weighted tax
difference Maximum tax difference Statutory tax rate 1
Weighted tax difference 0.4237 1
Maximum tax difference 0.4811 0.2946 1