## _wp12281

## Source details

**Canonical URL:** [_wp12281](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12281.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12281.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12281.pdf.json)

---

### I. Introduction and empirical motivation
- Corporate tax systems allow interest deductibility but not equity returns, creating a "debt bias."
- Empirical literature finds significant debt-bias effects for nonfinancial firms; almost no empirical work focuses specifically on banks.
- This paper studies debt bias in multinational banks using a subsidiary-level dataset covering 558 commercial bank subsidiaries of the 86 largest multinational banks.
- Empirical headline: "taxes matter significantly both through domestic debt bias and international debt shifting."
- The international debt shifting channel appears more robust and tends to be larger than the traditional domestic debt bias, implying significant international spillovers with policy implications for tax competition, anti–base erosion measures, and regulation (including capital requirements).

### II. Theoretical model: setup and comparative statics
- Framework: trade-off theory for a multinational bank with tax bias and convex non-tax costs of debt (financial distress, legal penalties).
- Environment and notation:
  - m countries, one subsidiary per host country i; subsidiary assets A_i; loans L_i; borrows B_i at interest rate r; fixed assets FA_i; equity E_i.
  - Balance: A_i = FA_i + L_i = B_i + E_i.
  - Outside investor required net rate of return = n; outside ownership share k_i.
  - Group total assets A_p = sum_i A_i; asset share s_i = A_i / A_p.
- Objective: parent maximizes sum of post-tax profits of all subsidiaries minus outside investors' share minus total cost from debt finance.
- First-order condition yields subsidiary leverage ratio depending on:
  - Local tax term (coefficient denoted alpha_local): predicted positive effect of local CIT on leverage.
  - International tax-difference term (coefficient denoted alpha_international): predicted positive effect of weighted tax differences between subsidiary country and other subsidiaries (weights = asset shares s_j).
- Role of capital requirements:
  - Parameters capturing capital requirement tightness reduce tax sensitivity; tighter subsidiary or parent capital requirements raise marginal legal cost of debt and make leverage less responsive to tax changes.
- Comparative magnitude:
  - If parent marginal cost of debt > subsidiary marginal cost, coefficient on local tax < coefficient on international tax difference (international effect larger).
  - Reverse holds if parent cost < subsidiary cost.

### III. Empirical approach, data coverage, and variable construction
- Econometric specification: panel regressions of subsidiary total liabilities / total assets on:
  - statutory CIT rate τ_it,
  - an international tax difference variable (asset-weighted average of differences between subsidiary host country tax rate and tax rates of other subsidiaries of same parent),
  - bank-level controls,
  - subsidiary host country-level controls,
  - subsidiary host country fixed effects.
- Measurement of the international tax difference:
  - Asset-weighted average using asset shares q_jt; positive value implies incentive to shift debt into subsidiary host country.
  - Alternative weightings: time-invariant asset weights and leverage weights (used to address endogeneity).
  - Can be rewritten to separate subsidiary tax level from difference with weighted average tax rate of the group.
- Bank-level controls included:
  - Size: book value of bank assets and its square.
  - Profitability: pre-tax return on assets.
  - Growth: total book assets growth.
  - Collateral: proportion of total security assets and non-earning assets.
  - Non-debt tax shields: total non-interest expenses / total assets.
- Subsidiary host country controls:
  - GDP growth, inflation, deposit insurance dummy, minimum capital requirement, financial crisis dummy (Laeven and Valencia (2010)).
- Data source and sample construction:
  - Bankscope database (Bureau Van Dijk); sample from the 100 largest multinational banks.
  - Focus on commercial banks; branches excluded; subsidiary defined as >50 percent owned by parent.
  - Consolidation: 36 percent report consolidated accounts; 64 percent report unconsolidated statements.
  - Sample cleaning removed extreme or implausible values (listed in source).
  - Final sample: 558 subsidiary banks owned by 86 largest commercial banks; parents headquartered in 25 countries; subsidiaries located in 66 host countries; sample spans 1998-2011.

### IV. Key descriptive statistics (exact values preserved)
- Sample size: Obs. 3,905 for many core variables.
- Leverage (percent): Obs. 3905; Mean 86.74; St. Dev. 14.70; Min 0.52; Median 90.70; Max 98.98.
- Short-term leverage (percent): Obs. 3131; Mean 80.55; St. Dev. 17.06; Min 0.09; Median 85.87; Max 98.79.
- CIT rate (percent): Obs. 3905; Mean 30.78; St. Dev. 7.55; Min 10.00; Median 30.00; Max 56.05.
- International tax difference (percent): Obs. 3905; Mean -2.13; St. Dev. 6.53; Min -28.04; Median -0.04; Max 23.65.
- Alt.: Asset-weighted average tax (percent): Obs. 3905; Mean 32.91; St. Dev. 5.74; Min 15.07; Median 33.72; Max 56.05.
- Log total assets: Obs. 3905; Mean 14.86; St. Dev. 2.28; Min 7.94; Median 14.58; Max 21.82.
- Profitability (percent): Obs. 3905; Mean 1.73; St. Dev. 3.81; Min -18.46; Median 1.26; Max 85.81.
- Capital requirement (percent): Obs. 3905; Mean 8.54; St. Dev. 1.13; Min 7.00; Median 8.00; Max 12.00.
- Deposit insurance: Obs. 3905; Mean 0.94; St. Dev. 0.24; Min 0.00; Median 1.00; Max 1.00.
- Financial crises: Obs. 3905; Mean 0.22; St. Dev. 0.42; Min 0.00; Median 0.00; Max 1.00.
- Mean and median total assets of subsidiary banks: Mean USD 2.8 billion; Median USD 2.1 billion.
- Country examples (Leverage, CIT Rate, Intl. Tax Diff.):
  - Austria 90.1 28.0 0.5
  - France 92.1 35.8 1.2
  - Germany 87.2 39.0 3.8
  - India 94.2 35.6 0.2
  - United States 78.7 39.3 2.4
  - Japan 93.0 40.2 0.6

### V. Baseline estimation results — key coefficients and magnitudes (exact values)
- Table 6 Regression (4) "Add Country Variables and Tax Differences" (Observations 3,905; R-squared 0.472):
  - CIT rate: 0.1590*** (t-statistic 3.329) — expected sign (+)
  - International tax difference: 0.1846*** (t-statistic 4.919) — expected sign (+)
  - Lag of log of total assets: 15.0102*** (13.131)
  - Lag of square log of total assets: -0.3941*** (-11.221)
  - Lag of profitability: -1.0738*** (-8.938)
  - Lag of total assets growth: 0.0435*** (5.341)
  - Lag of Collateral: -0.0770*** (-5.180)
  - Lag of non-debt tax credit: -0.0548** (-1.993)
  - GDP growth: 0.1721*** (3.416)
  - Financial crises: -1.1383* (-1.938)
- Interpretation example:
  - Local tax coefficient close to 0.3 in an earlier regression: an increase in the statutory CIT rate by 10 percentage points increases leverage by 3 percentage points (= 0.3 × 10).
  - Using Regression (4): local CIT coefficient = 0.1590 implies a 10 percentage point increase raises leverage by 1.59 percentage points; international tax difference coefficient = 0.1846 implies a 10 percentage point change in the weighted international tax differential raises leverage by 1.846 percentage points.

### VI. Robustness checks and alternative specifications (selected exact results)
- Table 7 (Clustered and Driscoll & Kraay SEs) — Regression (8) Driscoll & Kraay (Observations 4,208; R-squared 0.436):
  - CIT rate: 0.1479** (t-statistic 2.717)
  - International tax difference: 0.1774* (2.073)
- Table 8 (Alternatives):
  - (9) Only Ave. Tax: CIT rate 0.3436*** (6.183)
  - (10) Time Invariant: CIT rate 0.1179** (2.479); International tax difference (time-invariant asset-weights) 0.2469*** (5.543)
  - (11) Leverage Weights: CIT rate 0.1588*** (3.325); International tax difference (leverage-weights) 0.1825*** (4.888)
  - (12) Short Term: CIT rate 0.3074*** (3.453); Deposit insurance -4.7074*** (-3.009); Financial crises -2.3441*** (-2.734)
- Table 9:
  - Quantile regression (13): CIT rate 0.0645*** (2.845); International tax difference 0.0518* (1.830)
  - Trend regression (14): Detrended CIT rate -0.1089 (-1.106); International tax difference 0.2313*** (6.019)
- Table 10 (Subsamples):
  - (15) Unconsolidated (Obs. 2,569; R-squared 0.504): CIT rate 0.1454** (2.315); International tax difference 0.3219*** (5.587).
  - (16) Profitable (Obs. 3,556; R-squared 0.478): CIT rate 0.1739*** (3.669); International tax difference 0.1961*** (5.012).
  - (17) Advanced Economies (Obs. 1,771; R-squared 0.504): CIT rate 0.1210 (1.310); International tax difference 0.4657*** (6.222).
  - (18) Before Crisis (Obs. 2,961; R-squared 0.487): CIT rate 0.1363** (2.494); International tax difference 0.2248*** (5.049)

### VII. Capital tightness heterogeneity (Table 11)
- Capital tightness classification: banks divided into three equal-sized groups by subsidiary capital relative to legal minimum.
  - Capital-abundant banks: equity/total assets > minimum capital requirement by 3 percentage points.
  - Capital-tight banks: equity/total assets < minimum requirement by 1.2 percentage points.
- Results:
  - (19) Abundant Capital (Obs. 1,346; R-squared 0.446): CIT rate 0.2780* (1.861); International tax difference 0.3460*** (4.260).
  - (20) Tight Capital (Obs. 1,350; R-squared 0.585): CIT rate 0.0545*** (5.100); International tax difference -0.0187* (-1.901).
- Interpretation: banks closer to capital requirements are less sensitive to tax changes; taxes exert larger impacts on leverages of capital-abundant banks.

### VIII. Main conclusions and policy implications (preserved phrasing and recommendations)
- Main empirical conclusions:
  - A bank's leverage depends on corporate taxes via two channels:
    - (i) Local host-country CIT rate (traditional debt bias).
    - (ii) International tax difference vis-a-vis other bank subsidiaries of the same group (international debt shifting).
  - Tax effects are statistically significant and large; international debt shifting channel is more robust and often larger than the traditional domestic debt bias.
  - Tax policy induces significant international spillovers through multinational bank behavior.
- Policy concerns and recommendations:
  - International spillovers may intensify incentives for tax competition; strengthens case for international tax coordination.
  - Countries may use measures to remedy international debt shifting, e.g., thin capitalization rules restricting interest deductibility on intracompany loans; such measures generally do not apply to banks and raise the issue of bank-specific regulation or taxation.
  - Consider eliminating debt bias by neutralizing tax treatment of debt and equity, for example by introducing an allowance for corporate equity (application to the banking sector alone is a possibility).
  - Capital requirements interact with taxation: banks close to minimal capital requirements become less responsive to tax changes; regulatory design matters for policy responses to debt bias.

*Source: extracted content from _wp12281 — IMF staff working paper (References and selected sections).*

### References .............................................................................................................

### _wp12281 - References

### I. Introduction: context and empirical motivation
- Corporate tax systems typically allow interest deductibility but not equity returns, creating a "debt bias" (see Auerbach, 2002).
- High indebtedness of firms, while unlikely to have caused the financial crisis, may have increased vulnerability and deepened the crisis.
- Empirical literature finds significant debt-bias effects for nonfinancial firms (examples: Graham, 2003; De Mooij, 2011; Feld and others, 2011).
- Almost no empirical studies focus specifically on debt bias in the banking sector; many studies exclude financial firms or do not distinguish them.
- Keen and De Mooij (2012) analyze banks using unconsolidated statements of over 14,000 commercial banks in 82 countries and find sensitivity of banks' debt to taxation is very similar to non-financial firms. They also find large (systemic) banks are notably less responsive to tax.
- Multinational firms exploit international debt shifting: relatively more debt in high-tax locations and relatively more equity in low-tax locations. Empirical evidence from U.S. and European studies confirms this pattern (Hines and Hubbard, 1990; Collins and Shackelford, 1992; Grubert, 1998; Altshuler and Grubert, 2002; Desai and others, 2003; Mills and Newberry, 2004; Moore and Ruane, 2005; Huizinga and others, 2008; Buettner and Wamser, 2009; Egger and others, 2010).
- Policy response: Several high-tax countries restrict interest deductibility to prevent base erosion (e.g., Buettner and others, 2008).
- This paper combines the two strands by studying debt bias in multinational banks using a subsidiary-level dataset covering 558 commercial bank subsidiaries of the 86 largest multinational banks in the world.
  - Empirical headline: "taxes matter significantly both through domestic debt bias and international debt shifting."
  - The international debt shifting channel appears more robust and tends to be larger than the traditional domestic debt bias, implying significant international spillovers through multinational banks with policy implications for tax competition, anti–base erosion measures, and regulation (including capital requirements).

### II. Theoretical model: setup and comparative statics
- Framework: trade-off theory for a multinational bank with tax bias and convex non-tax costs of debt (financial distress, legal penalties for violating capital requirements).
- Environment and notation (as used in the analysis):
  - Multinational bank operates in m countries with one subsidiary bank in each host country i.
  - Subsidiary i: total assets A_i (assumed given), loans L_i yielding interest l_i, borrows B_i (including deposits and other debts) at interest rate r, fixed assets FA_i, equity E_i.
  - Balance sheet constraint: A_i = FA_i + L_i = B_i + E_i.
  - Subsidiary partly owned by parent and partly by outside investor; outside investor's required net rate of return equals return on alternative investments (n); outside ownership share denoted k_i.
  - Legal minimum capital requirements for subsidiary and parent: parameters denoted (implicitly) and affect costs of violating capital rules.
  - Total cost of debt finance comprises subsidiary cost and parent cost; subsidiary cost is convex in leverage and increasing in subsidiary capital requirement tightness; parent cost is increasing in group leverage and parent capital requirement tightness.
  - Group total assets A_p = sum_i A_i; group total leverage is sum of subsidiaries' debts over total group assets; asset share of subsidiary i in the group denoted s_i = A_i / A_p.
- Objective: parent maximizes sum of post-tax profits of all subsidiaries minus share to outside investors minus total cost from debt finance.
- First-order condition yields an expression for subsidiary leverage ratio showing dependence on:
  - Local tax term: coefficient (denoted alpha_local or similar in text) multiplying the local CIT rate — predicted positive effect.
  - International tax-difference term: coefficient (denoted alpha_international or similar) multiplying weighted tax differences between subsidiary country and other subsidiaries, with weights equal to asset shares s_j of other subsidiaries — predicted positive effect (international debt shifting).
- Role of capital requirements:
  - Parameters capturing capital requirement tightness for subsidiary and parent (denoted by positive terms in the model) reduce tax sensitivity: as capital requirement tightness increases, the marginal legal cost of holding debt rises, making leverage less responsive to tax changes.
- Comparative magnitude insight:
  - If the marginal cost of debt finance is larger for the parent than for the subsidiary (i.e., parent cost parameter > subsidiary cost parameter), then the coefficient on local tax < coefficient on international tax difference; international tax differences then exert a larger effect on subsidiary debt ratios than local tax.
  - Conversely, if debt finance is less costly for the parent than for subsidiaries, the local tax effect could be larger than the international differential effect.

### III. Empirical approach, data coverage, and primary empirical findings (as reported)
- Data: novel subsidiary-level dataset for 558 commercial bank subsidiaries of the 86 largest multinational banks.
- Primary empirical findings (as stated in the text):
  - Taxes matter significantly for subsidiary bank leverage through:
    - Domestic (local) debt bias.
    - International debt shifting.
  - The international debt shifting channel is more robust and tends to be larger than the traditional domestic debt bias.
  - Implication: taxation causes significant international spillovers through multinational banks, with potential implications for tax policy and bank regulation (capital requirements).
- Related empirical literature notes:
  - Huizinga and others (2008) analyze debt bias in multinationals but exclude banks.
  - Cerutti et al. (2007) find multinational banks' entry mode (subsidiary vs branch) responds to local corporate taxes; branches more likely where taxes are higher. The present empirical analysis focuses on subsidiaries and does not consider branches.
  - Keen and De Mooij (2012) found large-sample bank evidence using unconsolidated statements of over 14,000 commercial banks in 82 countries; large banks are less responsive to tax.

### IV. Model predictions (summarized)
- 1. Bank leverage depends positively on the local CIT rate.
- 2. Bank leverage depends positively on the difference between the subsidiary's own country's CIT rate and that of other subsidiary countries, with CIT rates in other subsidiaries weighted by asset shares (international debt shifting).
- 3. The impact of the local tax might be either smaller or larger than the impact of the international tax difference, depending on the marginal cost of financial leverage in the subsidiary versus the parent bank.

*Source: extracted content from _wp12281 - References (PDF chapter/section).*

### 4.      As a host country‟s capital requirement becomes tighter, ceteris paribus, its bank

### 4.      As a host country‟s capital requirement becomes tighter, ceteris paribus, its bank leverage becomes less sensitive to tax changes.

### Methodology
- Panel regressions of the general form regressing subsidiary bank total liability/assets ratio on:
  - statutory corporate income tax (CIT) rate faced by the subsidiary bank (τ_it),
  - an international tax difference variable,
  - bank-level controls,
  - subsidiary host country-level controls,
  - subsidiary host country fixed effects.
- Focus on coefficients on local tax rate and international tax difference: theory predicts β_local > 0 and β_international < 0.
- Expectation: tax elasticity of leverage is larger if a bank holds more capital beyond the capital requirement it faces; banks with higher debt ratios relative to the legal capital requirement will have smaller marginal impacts from tighter capital requirements.

### Measurement of the international tax difference
- International tax difference is the asset-weighted average of differences between the subsidiary host country tax rate and those of other subsidiaries of the same parent (following Huizinga and others (2008)).
- Formula concept (parent has m subsidiaries with asset shares q_jt): a positive value indicates incentive to shift debt into the subsidiary host country; a negative value indicates the opposite.
- Illustrative examples:
  - Example 1 (m = 3, equal asset shares = one third): if τ_A = 10 percent and τ_B = τ_C = 20 percent, international tax difference for A equals -7 percent (subsidiary A is in a low-tax country → incentive to decrease leverage). If τ_A increases to 50 percent, the international tax difference rises to 20 percent (subsidiary A is in a high-tax country → incentive to increase leverage).
  - Example 2 (A has one half of group assets; B and C equal): assume τ_A = 50 percent and τ_B = τ_C = 20 percent; international tax difference for A = 15 percent instead of 20 percent. Larger subsidiary asset share reduces the weighted international tax difference and thus reduces scope for debt shifting.
- Alternative weighting schemes used in regressions: time-invariant asset weights (to address potential endogeneity from endogenous assets) and leverage weights.
- International tax difference can be rewritten to separate the subsidiary tax level from the difference with the weighted average tax rate of the group; regressions estimate combined and separate coefficients accordingly.

### Bank-level controls (included in regressions)
- Size: book value of bank assets and its square (expect positive relationship between bank size and leverage).
- Profitability: pre-tax return on assets (ambiguous theoretical sign; could be positive or negative).
- Growth: total book assets growth used as proxy (effect ambiguous).
- Collateral: proportion of total security assets and non-earning assets out of total assets used as proxy (in non-bank capital structure literature, collateral tends to increase leverage; banking sector regulation may change this relation).
- Non-debt tax shields: total non-interest expenses to total assets ratio; expected negative relation with leverage.

### Subsidiary host country controls
- GDP growth and inflation (high growth expected to facilitate debt finance; inflation effect ambiguous).
- Deposit insurance: 0/1 dummy for existence of deposit insurance (expected positive impact on leverage).
- Minimum capital requirement (expected negative impact on leverage).
- Financial crisis dummy from Laeven and Valencia (2010) (a crisis may increase leverage short-term due to equity declines, but may subsequently reduce leverage).

### Data
- Source: Bankscope database (Bureau Van Dijk); sample drawn from the 100 largest multinational banks.
- Focus: commercial banks; branches excluded. A subsidiary defined as >50 percent owned by the parent bank.
- Consolidation in data: 36 percent of subsidiaries report consolidated accounts; 64 percent report unconsolidated statements.
- Sample construction and cleaning:
  - Dropped inactive subsidiaries.
  - Dropped subsidiaries with leverage ratio larger than 99 percent.
  - Dropped subsidiaries with pre-tax profit-to-asset ratios smaller than -20 percent or larger than 250 percent.
  - Dropped subsidiaries with negative total non-interest expenses.
  - Dropped subsidiaries with non-earning assets-to-total assets ratio larger than 99 percent.
  - Dropped subsidiaries with total assets growth larger than 150 percent.
  - Dropped subsidiaries with effective tax rates smaller than zero.
  - Dropped subsidiaries with missing total assets and CIT rates.
- Final sample:
  - 558 subsidiary banks (both domestic and foreign),
  - owned by 86 largest commercial banks,
  - parents headquartered in 25 countries,
  - subsidiaries located in 66 host countries,
  - sample spans 1998-2011.
- Descriptive highlights:
  - Financial leverage ranges from 69.3 percent in Argentina to 94.9 percent for Spain.
  - CIT rates highest in Japan and lowest in Bosnia and Herzegovina.
  - International tax difference variable indicates largest implied debt levels in Germany, Thailand, and Zambia; lowest implied debt ratios in Albania, Bosnia and Herzegovina, and Ireland.

### Results — Baseline regressions
- Table 6 baseline OLS regressions use subsidiary host country fixed effects.
- Regression (1) (local tax level + bank-level variables) result:
  - Local tax coefficient close to 0.3 and statistically significant.
  - Interpretation provided: an increase in the statutory CIT rate by 10 percentage points will increase the leverage ratio by 3 percentage points (= 0.3 × 10).
- Additional analyses:
  - Subsequent tables (7–10) provide robustness checks with alternative estimators, samples, and specifications.
  - Table 11 partitions observations by capital tightness (results not detailed in the supplied excerpt).

*Source: IMF staff working paper — extracted section on methodology, data, and baseline regression results.*

### 0.26 found for non-financial firms in Huizinga and others (2008), and in the meta analysis of

### _wp12281 - 0.26 found for non-financial firms in Huizinga and others (2008), and in the meta analysis of

### Bank-level determinants of leverage
- Larger banks have higher leverage ratios.
- Higher profitability reduces the leverage ratio.
- Faster growing banks accumulate more debt (bank growth associated with higher leverage).
- The collateral variable has a negative coefficient (consistent with Huizinga and others, 2008): collateral appears to make equity issuance less costly through lower asymmetric information, reducing leverage.
- Non-debt tax shields tend to substitute for the tax benefits of debt financing and reduce leverage (DeAngelo and Masulis, 1980).

### Tax channels and regression results (Table 6 overview)
- Regression (1): local CIT effect measured by coefficient λ1; baseline local tax channel effect sizeable (coefficient reported elsewhere as 0.25 in later text).
- Regression (2): adds international tax difference; international tax difference coefficient λ2 = 0.12 (positive and statistically significant), indicating an international debt shifting effect in addition to the local tax channel.
- Interpretation:
  - Local tax channel (λ1) captures effect of local taxation on subsidiary leverage.
  - International tax difference channel (λ2) captures effect of multinational-group-level international tax differentials on subsidiary leverage.
  - In regression (2) the local tax coefficient is 0.25.
- Regressions (3) and (4):
  - GDP growth enters positively.
  - Most other host country controls are not statistically significant, except a weakly significant negative financial crisis variable in column (4).
  - Regression (3) local CIT coefficient similar to column (1) but slightly smaller.
  - Regression (4): local CIT coefficient reduced to 0.16; international tax differential coefficient = 0.18 (positive and significant).
  - In column (4) local tax coefficient (0.16) is slightly smaller than international tax difference coefficient (0.18); F-test of coefficient equality has p-value of 0.69 (difference not statistically significant).
- Illustration for a hypothetical U.S. multinational bank with 10 equal-size subsidiaries:
  - Tax cut in country A of 10 percentage points:
    - Using regression (3): 10 percent rate reduction reduces subsidiary leverage by 2.5 percent.
    - Using regression (4):
      - Local tax channel effect = 1.59 percentage points reduction.
      - International debt shifting channel effect = 1.62 percentage points reduction (= 0.18 x 10 x 0.9).
      - Overall impact = 3.21 percentage points (= 1.59 + 1.62).
  - If statutory CIT rate declines by 10 percentage points in all countries except A:
    - Regression (3): no effect on leverage (foreign tax rates not included).
    - Regression (4): international tax difference rises by 9 percent (= 0.9 x 10 percent) and increases subsidiary leverage in country A by 1.62 percentage points (9 x 0.18).

### Sample and comparative magnitude
- Mean total assets of subsidiary banks in the data: USD 2.8 billion.
- Median total assets: USD 2.1 billion.
- Comparison: Keen and De Mooij (2012) data: only 5 percent of banks exceed assets size of USD 1.2 billion.
- In column (4) a considerable part of tax effect originates from international debt shifting; for large subsidiary banks the impact on debt is more than half explained by international debt shifting.

### Robustness checks (Tables 7–10 summary)
- Table 7 (regressions (5)-(8)):
  - Correct standard errors by clustering across parent banks, host countries, and subsidiaries: tax coefficients unchanged but standard errors increase; tax coefficients remain significant at 1 percent or 5 percent.
  - Driscoll and Kraay standard errors: local tax variable remains significant at 5 percent; international tax difference loses significance to 10 percent.
- Table 8 (regressions (9)-(12)):
  - Regression (9): exclude local tax rate from second term (international tax difference constructed only from foreign tax rates):
    - CIT coefficient becomes larger (captures both channels).
    - Foreign tax coefficient negative: higher foreign taxes reduce subsidiary leverage.
    - Relationship: λ1' = λ1 + λ2 and λ2' = - λ2.
  - Regression (10): use constant asset weights for weighted average foreign tax rate to reduce endogeneity from endogenous assets:
    - Core tax coefficients remain significant; international tax difference tends to become larger; local tax variable becomes smaller.
  - Regression (11): use leverage shares instead of asset shares for international tax difference: results very similar to column (4) of Table 7.
  - Regression (12): dependent variable = short-term debt (total leverage minus long-term funding):
    - Both tax variables positive; local tax coefficient larger than before → short-term debt more responsive to tax than long-term debt.
    - Deposit insurance variable reduces short-term debt; deposit insurance has an insignificant coefficient for total leverage, suggesting deposit insurance exerts a positive effect on long-term funding.
- Table 9 (regressions (13)-(14)):
  - Regression (13): quantile regression (conditional median) to reduce outlier influence:
    - Tax effects remain statistically significant but magnitudes smaller, implying outliers important for effect sizes in OLS.
  - Regression (14): adds host-country trend variable (captures declining global CIT rates):
    - Local tax variable becomes statistically insignificant; international tax difference remains significant and large.
- Table 10 (regressions (15)-(18)):
  - Regression (15): restrict to subsidiaries with unconsolidated data (observations drop from 3905 to 2569): core tax results similar; international tax difference impact larger.
  - Regression (16): restrict to profitable subsidiaries: coefficients λ1 and λ2 larger than baseline (supporting larger tax effects for profitable banks).
  - Regression (17): restrict to subsidiaries in advanced countries (domestic credit provided by banking sector as share of GDP higher than sample average):
    - λ2 much larger at 0.47 (vs. 0.18 benchmark).
    - No significant effect of local tax rate.
    - Positive impact of inflation on leverage; negative impact of capital requirements.
  - Regression (18): restrict to period before current global recession (exclude 2009–2011):
    - Tax effects similar to baseline, with larger coefficient for international tax difference.
    - Inflation negatively associated with bank leverage in this subsample.

### Extension: Capital Tightness (Table 11)
- Capital tightness measure: subsidiary capital relative to legal capital requirement; banks divided into three equal-sized groups (most abundant capital, intermediate, tightest capital).
- Regressions for capital-abundant vs capital-tight banks (columns (19) and (20)):
  - Capital-abundant banks: local tax coefficient = 0.28; international tax difference coefficient = 0.35 (taxes exert larger impact on leverages of capital-abundant banks).
  - Capital-tight banks: local tax coefficient = 0.05; international tax difference coefficient = -0.02 (statistically insignificant).
- Interpretation: banks closer to capital requirements are less sensitive to tax changes, consistent with theoretical hypothesis (3) and prior findings in Keen and De Mooij (2012).

### Conclusions and policy implications
- Main empirical conclusions:
  - A bank's leverage ratio depends on corporate taxes in two channels:
    - (i) Local tax rate in host country (traditional debt bias).
    - (ii) International tax difference vis-a-vis other bank subsidiaries in the same group (international debt shifting).
  - Tax effects are statistically significant and large; international debt shifting channel is more robust and often larger than traditional debt bias.
  - Tax policy induces significant international spillovers through multinational bank behavior.
- Policy concerns and recommendations:
  - International spillovers may intensify incentives for tax competition, potentially leading to inefficient policies; strengthens case for international tax coordination.
  - Countries may seek measures to remedy international debt shifting, e.g., thin capitalization rules restricting interest deductibility on intracompany loans; such measures generally do not apply to banks and raise the issue of bank-specific regulation or taxation.
  - Consider eliminating debt bias by neutralizing tax treatment of debt and equity, for example by introducing an allowance for corporate equity (applied to banking sector alone is a possibility; examples cited of countries that have done so).
  - Capital requirements interact with taxation: banks closer to minimal capital requirements become less responsive to tax changes; regulatory design matters for policy responses to debt bias.

*Source: IMF working paper text (excerpt)._

### REFERENCES

### _wp12281 - REFERENCES

### References Cited
- Auerbach, A.J., 2002, Taxation and Corporate Financial Policy in Alan J. Auerbach and Martin Feldstein (eds.), Handbook of Public Economics Vol. 3, pp. 1251-92 (Amsterdam, Elsevier North-Holland).
- Albertazzi, U., and L. Gambacorta, 2010, Bank Profitability and Taxation, Journal of Banking and Finance, 34(11), pp. 2801-10.
- Altshuler, R., and H. Grubert, 2002, Repatriation Taxes, Repatriation Strategies, and Multinational Financial Policy, Journal of Public Economics, 87, pp. 73-107.
- Bankscope, online June 4th, 2012 at https://bankscope2.bvdep.com/version-2012713/Search.QuickSearch.serv?_CID=1&context=2QGD7U3339HXXS6
- Buettner, T., and G. Wamser, 2009, Internal Debt and Multinationals‟ Profit Shifting: Empirical Evidence from Firm-level Panel Data, Working Paper 0918, Oxford.
- Cerutti, E., Dell‟Ariccia, G., and M. S. Martinez Peria, 2007, How Banks Go Abroad: Branches or Subsidiaries?, Journal of Banking and Finance, 31, pp. 1669-92.
- Collins, J.H., and D. Shackelford, 1992, Foreign Tax Credit Limitations and Preferred Stock Issuance, Journal of Accounting Research, 30, pp. 103-24.
- DeAngelo, H., and R. Masulis, 1980, Optimal Capital Structure under Corporate and Personal Taxation, Journal of Financial Economics, 8, pp. 3-29.
- de Haan, J., and T. Poghosyan, 2011, Bank Size, Market Concentration, and Bank Earnings Volatility in the U.S., Journal of International Financial Markets, Institutions and Money, 22, pp. 35-54.
- de Mooij, R.A., 2011, Taxes Elasticity of Corporate Debt: A Synthesis of Size and Variations, IMF Working Paper 11/95 (Washington: International Monetary Fund).
- Driscoll, J.C., and A.C. Kraay, 1998, Consistent Covariance Matrix Estimation with Spatially Dependent Panel Data, Review of Economics and Statistics, 80, 549-60.
- Egger, P., W. Eggert, C. Keuschnigg, and H. Winner, 2010, Corporate Taxation, Debt Financing and Foreign-plant Ownership, European Economic Review, 54, pp. 69-107.
- Feld, L.P., J. Heckemeyer, and M. Overesch, 2011, Capital Structure Choice and Company Taxation: A Meta-study, CESifo Working Papers No. 3400.
- Frank, M., and V. Goyal, 2009, Capital Structure Decisions: Which Factors are Reliably Important? Financial Management, 38, pp. 1-37.
- Graham, J.R., 2003, Taxes and Corporate Finance: A Review, Review of Financial Studies, 16, pp. 1075-1129.
- Gropp, R. and F. Heider, 2009, The Determinants of Bank Capital Structure, European Central Bank Working Paper, No. 1096.
- Gropp, R. and J. Vesala, 2001, Deposit Insurance and Moral Hazard: Does the Counterfactual Matter? ECB Working Paper No. 47, March.
- Grubert, H., 1998, Taxes and the Division of Foreign Operating Income among Royalties, Interest, Dividends and Retained Earnings, Journal of Public Economics, 68, 269-290.
- Hines, J.R., and R.G. Hubbard, 1990, Coming Home to America: Dividend Repatriation by U.S. Multinationals, in Razin, A. and R.G. Hubbard (Eds.), Taxation in the Global Economy, (Chicago: University of Chicago).
- Huizinga, H., L. Laeven, and G. Nicodeme, 2008, Capital Structure and International Debt Shifting, Journal of Financial Economics, 88, pp. 80-118.
- IMF, A Fair and Substantial Contribution by the Financial Sector: Final Report for the G-20, June, 2010 (Washington: International Monetary Fund).
- IMF, World Economic Outlook (WEO), online June 4th, 2012 at http://www.imf.org/external/pubs/ft/weo/2012/01/weodata/index.aspx (Washington: International Monetary Fund).
- Keen, M., and R.A. de Mooij, 2012, Debt, Taxes and Banks, IMF Working Paper 12/48 (Washington: International Monetary Fund).
- KPMG, Corporate Tax Survey, 1993-2012.
- Laeven, L., and F. Valencia, 2010, Resolution of Banking Crises: The Good, the Bad, and the Ugly, IMF Working Paper 10/146, Banking Crisis Database (2010 version).
- Mills, L.F., and K.J. Newberry, 2004, Do Foreign Multinationals‟ Tax Incentives Influence Their U.S. Income Reporting and Debt Policy? National Tax Journal, 57, pp. 89-107.
- Mintz, J., and A.J. Weichenrieder, 2009, Indirect Side of Direct Investment: Multinational Company Finance and Taxation (Cambridge, Massachusetts: MIT Press).
- Moore, P.J., and F.P. Ruane, 2005, Taxation and the Financial Structure of Foreign Direct Investment, IIIS Discussion Paper No. 88, Dublin.
- Myers, S., and N. Majluf, 1984, Corporate Financing and Investment Decisions when Firms Have Information that Investors do not Have. Journal of Financial Economics, 13, 187-221.
- OECD Tax Database, available online June 4th, 2012 at http://www.oecd.org/document/60/0,3746,en_2649_34533_1942460_1_1_1_1,00.html
- Rajan, R.G., and L. Zingales, 1995, What Do We Know About Capital Structure? Some Evidence from International Data, Journal of Finance, 50, 1421-60.
- World Bank Regulation Survey, 2000, 2003, 2008 (Washington: The World Bank).

### Technical Appendix — Model Overview
- Setup:
  - Multinational bank operating in m countries, one subsidiary in each host country i with total assets A_i (A_i assumed given).
  - Parent maximizes sum of post-tax profits of all subsidiaries.
- Key model elements (symbols preserved as in source):
  - Subsidiary loans to borrowers with interest rate; interest expenses from debt with interest rate r.
  - Capital requirements influence functions that are positive and increasing in the capital requirement.
  - Subsidiary outside owners possess a share and require return n.
- First-order conditions and comparative statics:
  - Derivations show that leverage (and related choice variables) increase with certain parameters and are negatively related with others.
  - Statement preserved: "Because ... and ... are positive, ... increases with ... and ... . Moreover, since ... is an increasing function of ..., tax‟s impacts on ..., i.e. ... and ..., are negatively related with ... ."

### Figure and Data Sources
- Figure 1. Bank Leverage Histogram
  - Source: Bankscope and authors‟ calculations.
  - Axis labels preserved: Density; Bank leverage.
- Table 1. Variable Source and Construction
  - Core variables and exact constructions (examples preserved verbatim):
    - Leverage = Total liabilities / Total assets (Source: Bankscope)
    - Short-term leverage = Leverage – Long-term funding / Total assets (Source: Bankscope)
    - CIT rate = Statutory tax rate (From KPMG Corporate Tax Survey, OECD Tax Database, and Mintz and Weichenrieder (2009))
    - Effective tax rate = Taxes/ Pre-tax Profit. (Source: Bankscope)
    - Capital requirement = Survey item 3.1 (Minimum total capital-to-assets ratio). Database years and imputation method documented (World Bank Regulation Survey).
    - Deposit insurance = Survey items 8.1 and 8.5; database years and imputation documented (World Bank Regulation Survey).
    - Financial crises = Dummy variable (Banking Crisis Database (2010 version), Laeven, Luc and Fabian Valencia, 2010).
- Note: Definitions of variables are provided in Table 1.

### Key Descriptive Statistics (from Table 2)
- Sample size and central tendencies (exact values preserved):
  - Leverage (percent): Obs. 3905; Mean 86.74; St. Dev. 14.70; Min 0.52; Median 90.70; Max 98.98.
  - Short-term leverage (percent): Obs. 3131; Mean 80.55; St. Dev. 17.06; Min 0.09; Median 85.87; Max 98.79.
  - CIT rate (percent): Obs. 3905; Mean 30.78; St. Dev. 7.55; Min 10.00; Median 30.00; Max 56.05.
  - International tax difference (percent): Obs. 3905; Mean -2.13; St. Dev. 6.53; Min -28.04; Median -0.04; Max 23.65.
  - Alt.: Asset-weighted average tax (percent): Obs. 3905; Mean 32.91; St. Dev. 5.74; Min 15.07; Median 33.72; Max 56.05.
  - Log total assets: Obs. 3905; Mean 14.86; St. Dev. 2.28; Min 7.94; Median 14.58; Max 21.82.
  - Profitability (percent): Obs. 3905; Mean 1.73; St. Dev. 3.81; Min -18.46; Median 1.26; Max 85.81.
  - Capital requirement (percent): Obs. 3905; Mean 8.54; St. Dev. 1.13; Min 7.00; Median 8.00; Max 12.00.
  - Deposit insurance: Obs. 3905; Mean 0.94; St. Dev. 0.24; Min 0.00; Median 1.00; Max 1.00.
  - Financial crises: Obs. 3905; Mean 0.22; St. Dev. 0.42; Min 0.00; Median 0.00; Max 1.00.
- Source: authors‟ calculations.
- Note: Some observations have lower capital than minimum capital requirement because leverage is weighted by total assets instead of risk-weighted assets.

### Correlations (selected exact values from Table 3)
- Leverage correlations:
  - Leverage with CIT rate: -0.0363
  - Leverage with International tax difference: 0.065
  - Leverage with Log total assets: 0.4228
  - Leverage with Profitability: -0.3574
- International tax difference correlations:
  - With CIT rate: 0.6764
  - With Alt.: International tax difference with leverage-weights: 0.9997
- Additional variable correlations are provided in full in Table 3.

### Country and Bank Counts (Table 4)
- Total: No. of Parent Banks 86; No. of Sub. Banks 558 domestic; No. of Sub. Banks 558 foreign.
- Country-level counts provided in Table 4 (per-country parent and subsidiary bank counts).

### Average Financial Leverage and Tax Rates (Table 5) — selected values
- Total (sample average across countries): Leverage 86.7; CIT Rate 30.8; Intl. Tax Diff. -2.1.
- Examples (country-specific exact triples: Leverage, CIT Rate, Intl. Tax Diff.):
  - Austria 90.1 28.0 0.5
  - France 92.1 35.8 1.2
  - Germany 87.2 39.0 3.8
  - India 94.2 35.6 0.2
  - United States 78.7 39.3 2.4
  - Japan 93.0 40.2 0.6

### Baseline Estimation Results (Table 6) — key coefficients (exact values and significance)
- Regression (4) "Add Country Variables and Tax Differences" (Observations 3,905; R-squared 0.472):
  - CIT rate: 0.1590*** (t-statistic 3.329) — expected sign (+)
  - International tax difference: 0.1846*** (t-statistic 4.919) — expected sign (+)
  - Lag of log of total assets: 15.0102*** (13.131) — (+)
  - Lag of square log of total assets: -0.3941*** (-11.221) — (-)
  - Lag of profitability: -1.0738*** (-8.938) — (?)
  - Lag of total assets growth: 0.0435*** (5.341) — (?)
  - Lag of Collateral: -0.0770*** (-5.180) — (?)
  - Lag of non-debt tax credit: -0.0548** (-1.993) — (-)
  - GDP growth: 0.1721*** (3.416) — (+)
  - Financial crises: -1.1383* (-1.938) — (-)
- Note: Robust t-statistics and expected signs in parentheses; *, **, *** denote significance at the 10, 5, and 1 percent level. All regressions use OLS with host country fixed effects.

### Robustness Checks — Standard Errors and Alternatives (Tables 7–9)
- Table 7 (Clustered and Driscoll & Kraay SEs) — Regression (8) Driscoll & Kraay (Observations 4,208; R-squared 0.436):
  - CIT rate: 0.1479** (t-statistic 2.717)
  - International tax difference: 0.1774* (2.073)
  - Lag of log of total assets: 15.1020*** (17.781)
  - Lag of square log of total assets: -0.3927*** (-16.065)
  - Lag of profitability: -0.7520*** (-4.838)
  - Lag of total assets growth: 0.0166*** (3.557)
- Table 8 (Alternatives):
  - (9) Only Ave. Tax: CIT rate 0.3436*** (6.183)
  - (10) Time Invariant: CIT rate 0.1179** (2.479); Alt.: International tax difference with time-invariant asset-weights 0.2469*** (5.543)
  - (11) Leverage Weights: CIT rate 0.1588*** (3.325); Alt.: International tax difference with leverage-weights 0.1825*** (4.888)
  - (12) Short Term: CIT rate 0.3074*** (3.453); Deposit insurance -4.7074*** (-3.009); Financial crises -2.3441*** (-2.734)
- Table 9 (Leverage Skewness and Tax Trend):
  - Quantile regression (13): CIT rate 0.0645*** (2.845); International tax difference 0.0518* (1.830)
  - Trend regression (14): Detrended CIT rate -0.1089 (-1.106); International tax difference 0.2313*** (6.019)

### Subsample and Heterogeneity Results (Tables 10–11)
- Table 10 (Subsamples):
  - (15) Unconsolidated (Obs. 2,569; R-squared 0.504): CIT rate 0.1454** (2.315); International tax difference 0.3219*** (5.587).
  - (16) Profitable (Obs. 3,556; R-squared 0.478): CIT rate 0.1739*** (3.669); International tax difference 0.1961*** (5.012).
  - (17) Advanced Economies (Obs. 1,771; R-squared 0.504): CIT rate 0.1210 (1.310); International tax difference 0.4657*** (6.222).
  - (18) Before Crisis (Obs. 2,961; R-squared 0.487): CIT rate 0.1363** (2.494); International tax difference 0.2248*** (5.049).
- Table 11 (Capital Tightness):
  - (19) Abundant Capital (Obs. 1,346; R-squared 0.446): CIT rate 0.2780* (1.861); International tax difference 0.3460*** (4.260).
  - (20) Tight Capital (Obs. 1,350; R-squared 0.585): CIT rate 0.0545*** (5.100); International tax difference -0.0187* (-1.901).
  - Definition: capital-abundant banks have equity/total assets > minimum capital requirement by 3 percentage points; capital-tight banks have equity/total assets < minimum requirement by 1.2 percentage points. Note on equity weighting by total assets vs. risk-weighted assets.

*Source: authors' calculations and references as listed in the source content.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12281.pdf_