## ch1

## Source details

**Canonical URL:** [ch1](https://www.imf.org/-/media/files/publications/fiscal-monitor/2026/october/english/ch1.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/fiscal-monitor/2026/october/english/ch1.pdf.md)
- [Structured JSON version](/-/media/files/publications/fiscal-monitor/2026/october/english/ch1.pdf.json)

---

### Distortions index construction and validation
- The Distortions Index is a composite measure combining: (i) the level of marginal effective tax rates (METRs), (ii) the dispersion of METRs across assets, (iii) VAT C-efficiency, (iv) one minus the share of revenue from property and sales taxes, and (v) the labor tax wedge.
- Principal component details:
  - The first factor has an eigenvalue of 1.82 and explains 36.5 percent of the total variance; the first two factors together account for 58.1 percent.
  - Four indicators load positively on the common factor; the labor wedge exhibits a small negative loading and high uniqueness.
- Validation evidence:
  - Germany’s 2008 corporate tax reform is associated with a pronounced decline in the index by slightly more than two standard deviations.
  - India’s increase in 2017 may reflect transitional distortions following the introduction of the Goods and Services Tax (GST) on July 1, 2017.
- Construction note:
  - The index is constructed with country-specific PCAs; the relationship between components and the index may differ across countries.

### Impact on GDP per capita — Local Projections specification
- Estimated model:
  - Lny_{c,t+h} − Lny_{c,t−1} = β_h DistortionIndex_{c,t} + γ X_{c,t} + μ_t + λ_c + ε_{c,t+h}. (Eq.1.2)
- Estimation features:
  - Dependent variable: GDP per capita (y_{c,t}).
  - X_{c,t} includes: a lag of the dependent variable and the distortions index; measures of physical and human capital; controls for changes in the real exchange rate, the current account relative to GDP, the debt-to-GDP ratio, the inflation rate, and the policy rate of the central bank.
  - Controls for statutory levels of top corporate income taxes, the top PIT rate, and the standard VAT rate.
  - Time fixed effects (μ_t) and country fixed effects (λ_c) included.
  - Local projections estimated for horizons h = 0,1,2,3,4 (four-year horizon).
  - Baseline graphical results plot response to a one-standard-deviation reduction in the Distortions Index.

### Estimated effects on output
- Main quantitative result:
  - A reduction in the Distortions Index by one standard deviation has significant and persistent effects on output.
  - Three years after a one-standard-deviation reduction in the index, GDP per capita is about 3.5% higher than in the counterfactual world without the change in distortions.
- Interpretation:
  - The estimate is interpreted as the average association between reducing tax distortions and GDP per capita across countries.

### Mapping index changes to component magnitudes
- METRs and dispersion:
  - A one standard deviation increase in the index corresponds to a rise in METRs of about 11 percentage points relative to its sample mean of 17.1 percent.
  - Increasing the dispersion (standard deviation) of METRs across assets by 8.9 percentage points results in a one-unit increase in the index, equivalent to roughly one-half of a standard deviation of the underlying variable.
- VAT C-efficiency and revenue share:
  - A decrease in VAT C-efficiency of about 40 percentage points would generate a comparable change in the index, starting from a sample average of 49 percent.
  - A decrease in the distortions share (one minus the share of revenue from property and sales taxes) of 13 percentage points would generate a comparable change in the index, starting from a sample average of 63 percent.
- Labor wedge:
  - Because the labor wedge loads negatively, a reduction in the labor wedge would raise the Distortions Index; this relationship is not statistically significant at conventional significance levels.

### Robustness and sensitivity checks
- Alternative local projections specifications:
  - Regression on changes in the Distortions Index:
    - Model: Lny_{c,t+h} − Lny_{c,t−1} = β_h ΔDistortionIndex_{c,t} + γ X_{c,t} + μ_t + λ_c + ε_{c,t+h}. (Eq.1.3)
    - Responses are slightly smaller in magnitude but exhibit a very similar pattern.
  - Pooled PCA:
    - Results very similar to main specification; confidence bands slightly wider.
  - Using real GDP (rather than GDP per capita):
    - Results slightly smaller in magnitude; qualitative conclusions identical.
  - Controlling for the tax-to-GDP ratio:
    - Does not affect the results; confidence bands overlap with baseline at most horizons.
- Leave-one-out estimation:
  - Results highly robust to exclusion of individual components.
  - Omitting the labor wedge yields estimates nearly identical to the baseline results.
  - Omitting METR-related variables yields somewhat smaller estimated responses than the baseline.

### Interpretation and key conclusion
- Higher tax distortions are associated with weaker output, with effects that build gradually over time.
- Distortions related to corporate taxation (as captured by METR measures) play a particularly important role.
- The index captures deviations of country-specific distortions from global averages rather than absolute levels.

### Online Annex 1.2 — VAT exemptions as input taxation: measurement and welfare implications
- Objective and measurement steps:
  - Explain how exemptions and non-registration increase the price of intermediate inputs and compute production tax wedges for 26 countries.
  - Compute production and consumption distortions and use World Bank Enterprise Surveys to illustrate firm-level effects.
- Measurement framework:
  - Define t_i = ln(p_i / p_i^*) as the log increase in firm i’s price relative to the no-distortion benchmark.
  - Distortions add across input-output network:
    - t_i = (1−δ_i) Σ_j α_ij τ_jj + Σ_j α_ij t_jj for i = 1,...,N. (Eq.2.1)
  - System solution:
    - t = (1−S)^−1 (1−D) S τ. (Eq.2.2)
  - Decomposition:
    - Direct_i = (1−δ_i) α_i Σ_j α_ij τ_jj
    - Cascading_i = (1/α_i) Σ_j α_ij t_jj. (Eq.2.3)
- Welfare implications — production and consumption:
  - Production loss (up to second order):
    - Production Loss = (σ_M / 2) Σ_i ν_i α_i (1−α_i) R_i^2, with R = t/α and σ_M = 0.5 used in simulations.
  - Consumer loss (second-order CES approximation):
    - Consumer Loss = (σ_D / 2) Σ ω_i [ (τ̂_i + t_i) − (τ̂ + t) ]^2, with σ = 1 used for consumption analyses.
  - Full input VAT recoverability sets R = 0 and t = 0, eliminating production distortions and collapsing consumption loss to dispersion of statutory final rates.
- Empirical mapping and firm-level tests:
  - Sector wedge = weighted average of commodity wedges using input-output tables.
  - World Bank Enterprise Surveys used to test three margins: cost-to-sales, input mix, and exporter status.
  - Key regression coefficients (selected):
    - Wedge R (combined): Cost-to-sales 2.64*** (0.97); Input share 0.62** (0.26); Exporter −1.07*** (0.33).
    - Cascading: Cost-to-sales 4.30** (2.14); Input share 1.97*** (0.75); Exporter −2.13*** (0.79).
    - Direct: Cost-to-sales 1.54 (1.43); Input share −0.02 (0.32); Exporter −0.41 (0.53).
  - Observations across columns: 34,151; 34,151; 34,216; 34,216; 39,769; 39,769.
- Interpretation:
  - Effective input-tax wedge associated with higher cost-to-sales ratio; cascading component drives firm-level effects and exporter margin.

### Online Annex 1.3 — Cost of capital, METRs, AETRs, firm- and sector-level regressions
- Definitions and formulas (symbols preserved):
  - Net cost of capital:
    - ρ = ( (r + δ) (1 − τ_A) ) / (1 − τ) − δ − [ α (1 − τ_φ) τ_i ] / [ (1 − τ) (1 + π) ]. (Eq.3.1)
  - Gross cost of capital:
    - R ≡ ρ + δ. (Eq.3.2)
  - METR:
    - METR = (ρ − r) / r. (Eq.3.3)
    - Alternative normalization: METR = (ρ − r) / ρ. (Eq.3.4)
  - AETR:
    - AETR formula under Devereux–Griffith assumptions preserved as Eq.3.5; computed for a project with a 20 percent real pre-tax rate of return.
- Data and baseline assumptions:
  - Dataset origins: Mc Auliffe, Thunecke, and Wamser (2023), extended through 2025.
  - Real interest rate and inflation assumptions for METR calculations: real interest rate = 3 percent; inflation rate = 2 percent; applied uniformly across countries and years.
  - Economic depreciation rates (OECD (2023)):
    - Buildings and structures: 3.3%
    - Machinery: 12.6%
    - Office equipment: 15.5%
    - Computers: 37.0%
    - Vehicles: 20.2%
    - Intangible assets: 15.0%
  - Computation tools: Stata-based effective tax rate routine (ssc install etr) developed by Hebous and Mengistu (2026).
  - Sample coverage notes:
    - METR averages: sample of 186 countries (uniform real interest rate case) and 90 countries (country-specific real interest rate case).
    - Firm-level regressions sample: about 2.5 million firms from 28 advanced and 10 emerging market economies over 2001–2020.
    - Allocative efficiency sample: 38 advanced and emerging market economies; baseline unbalanced panel consists of 379 country-sectors over 2001–2020.

### Firm-level investment regressions: data, specification, and key results
- Data and summary statistics (units preserved exactly; observations = 15,356,963 where indicated):
  - Log Fixed Assets: Mean 11.73; Median 11.68; Standard Deviation 2.36; 25th Percentile 10.21; 75th Percentile 13.19; Observations 15,356,963.
  - METR (Percent): Mean 21.76; Median 24.45; Standard Deviation 12.83; 25th Percentile 10.60; 75th Percentile 31.68; Observations 15,356,963.
  - Log Sales: Mean 13.69; Median 13.50; Standard Deviation 1.86; 25th Percentile 12.42; 75th Percentile 14.77; Observations 15,356,963.
  - Sales growth (Percent): Mean 3.86; Median 0.14; Standard Deviation 35.13; 25th Percentile −12.39; 75th Percentile 13.57; Observations 15,356,963.
  - Number of employees: Mean 62.11; Median 16.00; Standard Deviation 1685.36; 25th Percentile 3.00; 75th Percentile 17.00; Observations 15,356,963.
  - Long-term debt share (Percent of total assets): Mean 17.61; Median 7.62; Standard Deviation 70.22; 25th Percentile 0.00; 75th Percentile 24.34; Observations 15,356,963.
- Regression specification:
  - Local projections estimating Δ^h ln y_{i,c,t} on ΔMETR_{c,s,t} with two lags of ΔMETR and outcome, firm controls, firm, country-sector, and sector-year fixed effects; horizons h ∈ {0,1,...,4}; standard errors clustered at firm and country-sector levels.
- Key empirical findings:
  - A 1 percentage point increase in the METR is associated with about a 0.2 percent decline in firms’ fixed assets within a year and 0.4 percent after 3 years.
  - Estimated response lines and 90 percent confidence intervals illustrated in Online Annex Figure 1.3.2.

### Allocative efficiency, productivity, and factor responses
- Coverage and outcomes:
  - Country sample: 38 advanced and emerging market economies; baseline unbalanced panel of 379 country-sectors over 2001–2020.
  - Outcomes: firm-level real productivity (TFPQ), capital, labor, and sector-level allocative efficiency.
- Results:
  - A 1 percentage point increase in the METR is followed by a decline in real productivity in the first two years.
  - Capital inputs for the median firm show a short-run decline before recovering.
  - Labor inputs decline more briefly, concentrated in the year of the METR change.
  - Allocative efficiency changes little in the short run but tends to deteriorate over the medium term.
- Graphical presentation: coefficient estimates and 90 percent confidence intervals shown in Online Annex Figure 1.3.3.

### Elasticities of investment to cost of capital and aggregation
- Converted short-run (one-year) firm-level elasticities of investment w.r.t. the gross cost of capital (preserved exactly):
  - −4.27 (Chodorow-Reich and others 2024)
  - −4.7 (Curtis and others 2021)
  - −7.3 (Zwick and Mahon 2017)
  - −9.4 (from the chapter’s above regressions)
  - −10.7 (Maffini, Xing, and Devereux 2019)
- Aggregate mapping and theoretical formulas (symbols preserved):
  - Firm-level long-run elasticity (neoclassical theory) equation preserved as Eq.3.8.
  - Economy-wide long-run elasticity equation preserved as Eq.3.9.
  - Under assumptions ε ≈ ∞ and α* + β* = 1, dlog(K)/dlog(R) = − η / (1 − α).
  - Using η in range 0.7 to 1 and α ≈ 1/3 implies aggregate long-run capital elasticity ≈ −1.04 to −1.5.

### Simulated macro effects and sensitivity (GDP and capital responses)
- Conservative empirical elasticity used in simulations: −4.27 (Chodorow-Reich and others 2024).
- Online Annex Table 1.3.2 — Investment and capital responses to the cost of capital, by income group (numeric values preserved exactly):
  - Advanced Economies:
    - Cost of capital before tax: 10.28
    - Cost of capital after tax: 10.72
    - Short-run investment response (%): −18.4
    - Long run macro response of capital (constant elasticity of substitution = 0.7): −4.50
    - Long run macro response of capital (constant elasticity of substitution = 1): −6.43
  - Emerging Market Economies:
    - Cost of capital before tax: 10.31
    - Cost of capital after tax: 10.85
    - Short-run investment response (%): −22.4
    - Long run macro response of capital (constant elasticity of substitution = 0.7): −5.48
    - Long run macro response of capital (constant elasticity of substitution = 1): −7.83
  - Low Income Developing Countries:
    - Cost of capital before tax: 10.55
    - Cost of capital after tax: 11.12
    - Short-run investment response (%): −23.4
    - Long run macro response of capital (constant elasticity of substitution = 0.7): −5.72
    - Long run macro response of capital (constant elasticity of substitution = 1): −8.17
- Baseline simulation assumptions for Figure 1.4, panel 2: α = 1/3 and η = 1.
- Sensitivity example:
  - Holding α = 0.33 and lowering η to 0.5 reduces estimated GDP gains from 2.1 to 1.06 percent in advanced economies and from 2.7 to 1.35 percent in LIDCs.

### Online Annex 1.4 — Optimal labor taxation: model, calibration, and reform simulation findings
- Model and calibration highlights:
  - Social welfare maximization and sufficient-statistic representation for optimal marginal tax schedule preserved (Eq.4.3).
  - Elasticity assumptions preserved exactly:
    - Intensive labor supply elasticity = 0.2.
    - Average extensive margin elasticity = 0.25.
  - Country coverage for calibrations:
    - EMDEs: Brazil, Côte d'Ivoire, Georgia, Peru, Poland, Vietnam.
    - AEs: Canada, Germany, Spain, United Kingdom, Israel, Japan, Sweden.
- Reform simulations — group-average effects (unweighted across sample countries):
  - EMDEs — Current / Reform (group averages):
    - In-work tax credit — Maximum amount (percent of median income): 17 / 17
    - Income cutoff (percent of median income): 43 / 59
    - Share of exempt workers (percent): 16 / 24
    - Marginal effective tax rates (percent):
      - 10th percentile: 25 / 21
      - 50th percentile: 16 / 23
      - 90th percentile: 26 / 21
    - Earnings gains concentrated in lower half of distribution, peaking at around 7.5 percent near the bottom decile.
  - AEs — Current / Reform (group averages):
    - In-work tax credit — Maximum amount (percent of median income): 24 / 27
    - Income cutoff (percent of median income): 51 / 56
    - Share of exempt workers (percent): 20 / 23
    - Marginal effective tax rates (percent):
      - 10th percentile: 47 → 51
      - 50th percentile: 36 → 42
      - 90th percentile: 44 → 36
    - Earnings gains largest at very bottom and very top, reaching about 6 to 8 percent at both ends; middle-income earners experience small losses.
- Reform effects on labor-supply disincentives:
  - EMDEs: Reform lowers hours-of-work disincentives for low and high incomes.
  - AEs: Reform lowers entry-to-work disincentives for low and high incomes.

### Online Annex 1.5 — Tax compliance costs, tax complexity, and firm performance
- Objective:
  - Assess whether high-quality tax administration and simpler tax frameworks disproportionately benefit small and young firms.
- Key measures:
  - Tax Administration Quality Index (TAQI): range 1 to 4, higher values indicate lower compliance costs; coverage: 63 countries and 72 country-survey year observations over 2013-2025.
  - Tax complexity: measured by TAXCOMP (corporate income tax complexity index).
- Empirical strategy:
  - Baseline specifications: Equation (5.1), (5.2), (5.3) with country-industry-survey year fixed effects. FirmTypeiscv indicates small (≤ 20 employees) or young (≤ 9 years) firms.
- Main results:
  - Better tax administration (higher TAQI) is associated with stronger performance among small and young firms.
  - Better tax administration lowers compliance burdens and improves perceptions of tax administration and business environment for small and young firms.
  - Greater tax framework complexity (TAXCOMP) is associated with weaker relative performance among small and young firms.
- Quantitative magnitudes:
  - Moving from the 25th to the 75th percentile of TAQI:
    - Narrows labor-productivity gap from about 29 to 17 percent for small firms.
    - Narrows labor-productivity gap from about 22 to 12 percent for young firms.
  - Select coefficient highlights (Online Annex Table references preserved as presented):
    - Small × TAQI: 0.211*** (LABPRO); 1.935*** (LABPROGR); 3.103*** (ANSALES GR).
    - Young × TAQI: 0.162*** (LABPRO); 2.127*** (LABPROGR); 3.227*** (ANSALES GR).
    - Small × TAQI (HRCOMP, ADMBIGOBS, REFOBS, INFORM): -0.139**; -0.151***; -0.102***; -0.038** respectively.
    - Small × TAXCOMP: -1.961*** (LABPRO); -18.933** (LABPROGR); -19.433** (ANSALES GR).
- Heterogeneity and complementarity:
  - Larger TAQI effects in low-income developing countries where performance gaps are greatest.
  - Administrative quality and legal simplicity are complementary and jointly important for small and young firms.

*Source: Fiscal Monitor, Chapter 1, IMF | October 2026.*

### 0.59 to 0.72, the labor wedge is loaded negatively. However, a simple regression of the distortions index on the labor

### ch1 - 0.59 to 0.72, the labor wedge is loaded negatively. However, a simple regression of the distortions index on the labor

### Distortions index construction and validation
- The Distortions Index is a composite measure combining: (i) the level of marginal effective tax rates (METRs), (ii) the dispersion of METRs across assets, (iii) VAT C-efficiency, (iv) one minus the share of revenue from property and sales taxes, and (v) the labor tax wedge.
- The first factor has an eigenvalue of 1.82 and explains 36.5 percent of the total variance; the first two factors together account for 58.1 percent.
- Four indicators load positively on the common factor; the labor wedge exhibits a small negative loading and high uniqueness, suggesting it shares little common variation with the other measures but does not materially alter the underlying factor structure.
- Validation exercises comparing large changes in the index with contemporaneous tax policy developments indicate the index effectively identifies changes in tax systems:
  - Germany’s 2008 corporate tax reform is associated with a pronounced decline in the index by slightly more than two standard deviations.
  - The increase observed for India in 2017 may reflect transitional distortions following the introduction of the Goods and Services Tax (GST) on July 1, 2017.

### Impact on GDP per capita — Local Projections specification
- Estimated model:
  - Lny_{c,t+h} − Lny_{c,t−1} = β_h DistortionIndex_{c,t} + γ X_{c,t} + μ_t + λ_c + ε_{c,t+h}. (Eq.1.2)
- Key features of the estimation:
  - Dependent variable: GDP per capita (y_{c,t}).
  - X_{c,t} includes a lag of the dependent variable and the distortions index, measures of physical and human capital, and controls for changes in the real exchange rate, the current account relative to GDP, the debt-to-GDP ratio, the inflation rate, and the policy rate of the central bank.
  - Controls for statutory levels of top corporate income taxes, the top PIT rate, and the standard VAT rate are included.
  - Time fixed effects (μ_t) and country fixed effects (λ_c) are included.
  - Local projections estimated for horizons h = 0,1,2,3,4 (four-year horizon).
  - Baseline graphical results plot response to a one-standard-deviation reduction in the Distortions Index.

### Estimated effects on output
- A reduction in the Distortions Index by one standard deviation has significant and persistent effects on output.
- Three years after a one-standard-deviation reduction in the index, GDP per capita is about 3.5% higher than in the counterfactual world without the change in distortions.
- The estimate is interpreted as the average association between reducing tax distortions and GDP per capita across countries.

### Mapping index changes to component magnitudes
- For METRs:
  - A one standard deviation increase in the index corresponds to a rise in METRs of about 11 percentage points relative to its sample mean of 17.1 percent.
  - Increasing the dispersion (standard deviation) of METRs across assets by 8.9 percentage points results in a one-unit increase in the index, equivalent to roughly one-half of a standard deviation of the underlying variable.
- For VAT C-efficiency and revenue share:
  - A decrease in VAT C-efficiency of about 40 percentage points would generate a comparable change in the index, starting from a sample average of 49 percent.
  - A decrease in the distortions share (one minus the share of revenue from property and sales taxes) of 13 percentage points would generate a comparable change in the index, starting from a sample average of 63 percent.
- Because the labor wedge loads negatively, a reduction in the labor wedge would raise the Distortions Index, although this relationship is not statistically significant at conventional significance levels.
- The index is constructed with country-specific PCAs; the relationship between components and the index may differ across countries, and the numbers above represent the average relationship.

### Robustness and sensitivity checks
- Alternative local projections specifications:
  - Regression on changes in the Distortions Index:
    - Model: Lny_{c,t+h} − Lny_{c,t−1} = β_h ΔDistortionIndex_{c,t} + γ X_{c,t} + μ_t + λ_c + ε_{c,t+h}. (Eq.1.3)
    - Responses are slightly smaller in magnitude but exhibit a very similar pattern.
  - Pooled PCA (index constructed using pooled PCA across all countries and years):
    - Results very similar to main specification; confidence bands slightly wider, suggesting country-specific factor loadings improve precision.
  - Using real GDP (rather than GDP per capita) as outcome:
    - Results slightly smaller in magnitude; qualitative conclusions identical.
  - Controlling for the tax-to-GDP ratio:
    - Does not affect the results; confidence bands overlap with baseline at most horizons.
- Leave-one-out estimation (omitting one component at a time from the index):
  - Results are highly robust to exclusion of individual components.
  - Omitting the labor wedge yields estimates nearly identical to the baseline results, indicating the negative loading of the labor wedge does not materially affect estimation.
  - Omitting METR-related variables (panels excluding METR measures) yields somewhat smaller estimated responses than the baseline, suggesting METR-related distortions are important drivers of the baseline relationship.

### Interpretation and key conclusion
- Higher tax distortions are associated with weaker output, with effects that build gradually over time.
- Distortions related to corporate taxation (as captured by METR measures) play a particularly important role.
- The index captures deviations of country-specific distortions from global averages rather than absolute levels; thus, estimates reflect the association of relative tax distortions with economic output.

*Source: Fiscal Monitor, Chapter 1, IMF | October 2026.*

### References

### ch1 - References

### Cited studies and data sources
- Bachas, Pierre; Matthew Fisher-Post; Anders Jensen; Gabriel Zucman. 2026. “Globalization, Capital Taxation, and Development Evidence from a Macro-Historical Database.” American Economic Journal: Applied Economics.
- Baretti, Christian; Doina Maria Radulescu; Michael Stimmelmayr. "Die Unternehmensteuerreform 2008: Deutschlands Antwort auf die Globalisierung-oder doch ein Stückwerk?." ifo Schnelldienst 61.02 (2008): 30-38.
- Beer, Sebastian, and Shafik Hebous. Forthcoming. “VAT Exemptions as Input Taxation.” IMF Working Paper, International Monetary Fund, Washington, DC.
- Broda, Christian; David E. Weinstein. 2006. “Globalization and the Gains from Variety.” Quarterly Journal of Economics 121 (2): 541–585.
- Comin, Diego A.; Danial Lashkari; Martí Mestieri. 2021. “Structural Change with Long-Run Income and Price Effects.” Econometrica 89 (1): 311–374.
- Feenstra, Robert C.; Robert Inklaar; Marcel Timmer. 2015. “The Next Generation of the Penn World Table.” American Economic Review 105 (10): 3150–3182.
- Feenstra, Robert C.; Philip Luck; Maurice Obstfeld; Katheryn N. Russ. 2018. “In Search of the Armington Elasticity.” Review of Economics and Statistics 100 (1): 135–150.
- Herrendorf, Berthold; Richard Rogerson; Ákos Valentinyi. 2013. “Two Perspectives on Preferences and Structural Transformation.” American Economic Review 103 (7): 2752–2789.
- Johansson, Åsa; Christopher Heady; Jens Mathias Arnold; Bert Brys; Laura Vartia. 2008. “Taxation and Economic Growth.” OECD Economics Department Working Paper 620.
- McDaniel, Cara. 2011. “Forces Shaping Hours Worked in the OECD, 1960–2004.” American Economic Journal: Macroeconomics 3 (4): 27–52.
- Müller, Karsten; Chenzi Xu; Mohamed Lehbib; Ziliang Chen. 2025. “The Global Macro Database: A New International Macroeconomic Dataset (Version 2026-03).” NBER Working Paper No. 33714.
- Prescott, Edward C. 2004. “Why do Americans Work so Much More than Europeans?.” Federal Reserve Bank of Minneapolis Quarterly Review 28 (1): 2–13.
- World Bank. 2018. Implementation of India’s Goods and Services Tax: Design and International Comparison. Washington, DC: World Bank Group.

### Online Annex 1.2 — Value-Added Tax (VAT) Exemptions as Input Taxation: objectives and measurement
- Objective:
  - Provide details on measurement of VAT-related production and consumption distortions arising from exemptions and non-registration and assess economic implications.
  - Analysis proceeds in three steps: (1) explain how exemptions and non-registration increase the price of intermediate inputs and compute production tax wedges for 26 countries; (2) explain how wedges reduce welfare by distorting production and consumption and present equations used to quantify effects; (3) use World Bank Enterprise Surveys to illustrate firm-level effects.
- Sample: 26 countries (Armenia, Belgium, Cyprus, Dominican Republic, El Salvador, Estonia, Ethiopia, Georgia, Honduras, Hungary, India, Kenya, Lithuania, Moldova, Mongolia, Morocco, Namibia, Paraguay, Philippines, Romania, Rwanda, South Africa, Trinidad and Tobago, Uganda, Uzbekistan, Zimbabwe).
- Measurement of VAT wedges:
  - Define t_i = ln(p_i / p_i^*) as the log increase in firm i’s price relative to the no-distortion benchmark.
  - Distortions add across input-output network:
    - t_i = (1−δ_i) Σ_j α_ij τ_jj + Σ_j α_ij t_jj for i = 1,...,N. (Eq.2.1)
      - τ_j statutory tax rate applied on input j.
      - δ_i share of input VAT the firm can recover (δ_i = 1 if registered).
      - α_ij share of input j in firm i’s cost; α_i = Σ_j α_ij intermediate input share.
      - When commodity/firm j is exempt: τ_j = δ_j = 0.
  - System solution:
    - t = (1−S)^−1 (1−D) S τ. (Eq.2.2)
      - S: matrix of input cost shares (elements α_ij).
      - D: diagonal matrix collecting commodity-specific recoverability.
  - Decomposition into direct and cascading components:
    - Direct_i = (1−δ_i) α_i Σ_j α_ij τ_jj
    - Cascading_i = (1/α_i) Σ_j α_ij t_jj. (Eq.2.3)
  - Data and calibration:
    - Statutory VAT rates (τ) from tax legislation.
    - Recoverability parameters (δ) calibrated using registration thresholds and sector-specific treatment.
    - Granular input-output tables used to quantify input requirements per unit of output.
  - Interpretation:
    - When all firms can recover input VAT (D = 1), producer price increase equals zero.
    - Producer prices are more upward biased in industries with large intermediate input shares (row sums of S).

### Online Annex 1.2 — Welfare implications: production and consumption
- Production distortions:
  - Production described by nested CES with outer nest between labor and intermediates and inner nest among N intermediates.
  - Optimal ratio of intermediates to labor proportional to relative price: M/L ~ (P/w)^−σ_M.
  - Materials price faced by firms: P = P^* (1+R) with R = t/α (per unit proportional increase).
  - Production loss (up to second order):
    - Production Loss = (σ_M / 2) Σ_i ν_i α_i (1−α_i) R_i^2,
      - ν_i denotes supply share of commodity i in total output.
  - Key drivers: higher σ_M, α_i near 1/2, larger unrecoverable input VAT (R).
  - Full input VAT recoverability sets R = 0 and eliminates production distortions.
  - Simulations use σ_M = 0.5.
  - Production distortions presented in percent of GDP by multiplying country-specific production losses by output over value added.
- Consumption distortions:
  - Consumer price q_i = p_i (1+τ_i), with p_i = p_i^* e^{t_i}.
  - Effective tax deviation: Δ ln q = τ̂_i + t_i where τ̂_i = ln(1+τ_i).
  - Consumer loss (compensating variation), second-order approximation with CES preferences parameter σ_D:
    - Consumer Loss = (σ_D / 2) Σ ω_i [ (τ̂_i + t_i) − (τ̂ + t) ]^2,
      - (τ̂ + t)̄ = Σ ω_i (τ̂_i + t_i).
      - ω_i expenditure share of households on commodity i.
  - Welfare losses increase with substitution elasticity σ_D and quadratic in tax rate dispersion.
  - Full input recoverability sets t = 0 and collapses consumption loss to dispersion of statutory final rates.
  - Simulations use σ = 1 (Cobb-Douglas preferences) for consumption analyses.
  - Consumer distortions presented in percent of GDP by multiplying consumer losses by share of final consumption expenditure in GDP.

### Online Annex 1.2 — Empirical analysis and estimation results
- Mapping from commodity-level wedges to country-sector measures using input-output tables: sector wedge = weighted average of commodity wedges with weights equal to commodity’s share in sectoral output.
- Three empirical margins tested using World Bank Enterprise Surveys:
  - Average production costs: regress ln(production costs/sales) on sector-, country-, and year-fixed effects and constructed cost increase. Theory predicts coefficient = 1.
  - Input mix: regress firm-level intermediate input shares on effective tax rate on intermediates. Complementarity implies increase in expenditure share on intermediates (positive coefficient); substitutability implies negative coefficient.
  - Exporters: linear probability model for exporter indicator (>10 percent of sales exported) on input-tax wedge. Expect higher wedge to reduce exporting probability since exporters cannot pass unrecovered VAT forward.
- Estimation results (Online Annex Table 1.2.1):
  - Effective input-tax wedge associated with higher cost-to-sales ratio; estimated coefficient statistically indistinguishable from 1 (column 1).
  - Decomposition: effect concentrated in cascading component (column 2); direct component and intermediate-share control not statistically significant.
  - Input-mix regressions: higher wedges associated with larger intermediate cost share (columns 3–4); cascading component carries the effect.
    - Interpretation: expenditure share of intermediates rises when intermediates become more expensive, consistent with complementarity between materials and labor.
  - Exporter regressions: higher input-tax cost wedges reduce probability of exporting (columns 5–6).
  - Table estimates (selected coefficients and standard errors):
    - Wedge R (combined): Cost-to-sales 2.64*** (0.97); Input share 0.62** (0.26); Exporter −1.07*** (0.33).
    - Cascading: Cost-to-sales 4.30** (2.14); Input share 1.97*** (0.75); Exporter −2.13*** (0.79).
    - Direct: Cost-to-sales 1.54 (1.43); Input share −0.02 (0.32); Exporter −0.41 (0.53).
    - Intermediate share α_sc controls: 0.21 (0.19) in relevant specifications; 0.06 (0.05) in others.
    - Observations: 34,151; 34,151; 34,216; 34,216; 39,769; 39,769 across columns (1)–(6).
  - Robustness and interpretation:
    - Results consistent across outcomes; cascading component drives effects.
    - Direct component insignificant, likely due to limited variation after fixed effects.
    - Caution: tax wedges reflect policy choices and estimates—especially exporter result—could be affected by reverse causality.

### Online Annex 1.3 — Cost of Capital, Allocative Efficiency, and Investment: framework and implementation
- Purpose:
  - Compute forward-looking effective tax rates (ETRs), conduct firm- and sector-level regressions of ETR effects on investment and allocative efficiency, and assess aggregate investment effects of cost-of-capital changes.
- Derivation and definitions (following Hebous and Mengistu 2026; King and Fullerton 1984; Devereux and Griffith 2003):
  - Net cost of capital:
    - ρ = ( (r + δ) (1 − τ_A) ) / (1 − τ) − δ − [ α (1 − τ_φ) τ_i ] / [ (1 − τ) (1 + π) ]. (Eq.3.1)
      - r = real required rate of return net of tax.
      - i = nominal interest rate on debt.
      - α = share of investment financed with debt.
      - φ (φ) = depreciation allowance in year of investment (including any initial allowance).
      - δ = rate of economic depreciation.
      - τ statutory corporate income tax rate.
      - A = present value of depreciation allowances per unit of investment.
  - Gross cost of capital:
    - R ≡ ρ + δ. (Eq.3.2)
  - Marginal effective tax rate (METR):
    - METR = (ρ − r) / r. (Eq.3.3) — normalization: percentage increase in pre-tax rate required for marginal investment relative to no-tax benchmark.
    - Alternative normalization: METR = (ρ − r) / ρ. (Eq.3.4)
    - Interpretation: higher METR indicates larger distortion to marginal investment; METR = 0 implies tax neutrality at the margin.
  - Average effective tax rate (AETR):
    - AETR = (pre-tax economic rent − post-tax economic rent) / net present value of pre-tax income stream.
    - Under Devereux–Griffith assumptions:
      - AETR = τ [ 1 + δ/p ] − τ A / (r + δ/p) − [ α (1 − τ_φ) τ_i / p (1 + π) ]. (Eq.3.5)
      - p denotes real pre-tax rate of return on the investment project.
      - AETR computed for a project with a 20 percent real pre-tax rate of return.
    - AETR captures tax burden on infra-marginal returns; relevant for location decisions and large discrete investments.
- Empirical implementation:
  - Cost of capital, METR, and AETR computed using asset-specific tax depreciation provisions, asset weights, and financing structures (debt and equity shares) at the country–sector–year level for period 2000– (period continuation provided in source).

*International Monetary Fund, Fiscal Monitor Chapter 1, IMF | October 2026.*

### 2025. Data on tax depreciation rules for 2000–2020 and financing structures (i.e., debt and equity shares at the sector

### ch1 - 2025. Data on tax depreciation rules for 2000–2020 and financing structures (i.e., debt and equity shares at the sector level)

### Data sources and baseline assumptions
- Dataset origins: Mc Auliffe, Thunecke, and Wamser (2023), extended through 2025 using national tax legislations and corporate tax guides published by Ernst & Young, the International Bureau of Fiscal Documentation, and PricewaterhouseCoopers.
- Real interest rate and inflation assumptions for METR calculations: real interest rate = 3 percent; inflation rate = 2 percent; applied uniformly across countries and years to focus on tax system implications for ETRs.
- Alternative macroeconomic parameter treatment: country-specific five-year moving average inflation and country- and year-specific real interest rates implied by 10-year government bond yields with a 5-percentage point risk premium (used for sensitivity comparisons).
- Economic depreciation rates (following OECD (2023)):
  - Buildings and structures: 3.3%
  - Machinery: 12.6%
  - Office equipment: 15.5%
  - Computers: 37.0%
  - Vehicles: 20.2%
  - Intangible assets: 15.0%
- Computation tools: Stata-based effective tax rate routine (ssc install etr) developed by Hebous and Mengistu (2026).
- Sample coverage note: underlying dataset covers 2000–2025; figures presented are based on the 2020–2025 sample period.
- Country coverage in various analyses:
  - METR averages: sample of 186 countries (uniform real interest rate case) and 90 countries (country-specific real interest rate case, limited by long-term bond yield availability).
  - Firm-level regressions sample: about 2.5 million firms from 28 advanced and 10 emerging market economies over 2001–2020 (approximately 12 percent affiliates of multinational corporations).
  - Allocative efficiency sample: 38 advanced and emerging market economies; baseline unbalanced panel consists of 379 country-sectors over 2001–2020.

### METR definition and sensitivity to macro environment
- METR definition: METR = (ρ − r) / r, where r is the required post-tax rate of return and ρ is the cost of capital; measures percentage increase in the cost of capital for a marginal investment attributable to tax-induced distortions.
- Baseline METR presentation: reported in Figure 1.4, panel 1 (chapter).
- Sensitivity analysis reported (Online Annex Figure 1.3.1):
  - Left-panel specification: inflation = country-specific five-year moving average; real interest rate fixed at 3 percent across countries and years; averages based on 186 countries.
  - Right-panel specification: inflation = same as left; real interest rate = country- and year-specific real yield implied by 10-year government bond yield plus a 5-percentage point risk premium; averages based on 90 countries.
- Rationale for baseline parameters: chosen to facilitate comparisons with other international databases, particularly the OECD’s METR publication; note that average real interest rate in advanced economies during 2020–2025 (including a 5 percentage point risk premium) was approximately 3.5 percent.

### Firm-level investment regressions: data, specification, and key results
- Firm-level data and preprocessing:
  - Source: Bureau van Dijk’s Orbis database.
  - Real denomination: 2015 constant USD using 2-digit NACE-level deflators and World Bank World Development Indicators investment deflators.
  - Sample size for regressions: 15,356,963 observations reported in Online Annex Table 1.3.1 summary statistics.
- Selected summary statistics (from Online Annex Table 1.3.1; units preserved exactly):
  - Log Fixed Assets (logarithm of 2015 constant USD): Mean 11.73; Median 11.68; Standard Deviation 2.36; 25th Percentile 10.21; 75th Percentile 13.19; Observations 15,356,963.
  - METR (Percent): Mean 21.76; Median 24.45; Standard Deviation 12.83; 25th Percentile 10.60; 75th Percentile 31.68; Observations 15,356,963.
  - Log Sales (logarithm of 2015 constant USD): Mean 13.69; Median 13.50; Standard Deviation 1.86; 25th Percentile 12.42; 75th Percentile 14.77; Observations 15,356,963.
  - Sales growth (Percent): Mean 3.86; Median 0.14; Standard Deviation 35.13; 25th Percentile −12.39; 75th Percentile 13.57; Observations 15,356,963.
  - Number of employees (Total number of employees): Mean 62.11; Median 16.00; Standard Deviation 1685.36; 25th Percentile 3.00; 75th Percentile 17.00; Observations 15,356,963.
  - Long-term debt share (Percent of total assets): Mean 17.61; Median 7.62; Standard Deviation 70.22; 25th Percentile 0.00; 75th Percentile 24.34; Observations 15,356,963.
  - Profit margin (Ratio): Mean 0.04; Median 0.03; Standard Deviation 0.12; 25th Percentile 0.01; 75th Percentile 0.08; Observations 15,356,963.
  - Cash flow rate (Ratio): Mean 1.00; Median 0.27; Standard Deviation 3.30; 25th Percentile 0.08; 75th Percentile 0.77; Observations 15,356,963.
- Regression specification:
  - Baseline follows local projections (Jordà 2005) estimating Δ^h ln y_{i,c,t} on ΔMETR_{c,s,t} with two lags of ΔMETR and outcome, firm controls X_{i,c,t}, firm, country-sector, and sector-year fixed effects; horizons h ∈ {0,1,...,4}; standard errors clustered at firm and country-sector levels.
  - Firm controls include lag of log employees, lag of log sales, lag of sales growth, lag of profit margin (EBIT/sales), and cash flow rate (cash flow in year t divided by fixed assets in previous year). Sales growth, profit margin ratio, and cash-flow rate are winsorized at the 1st and 99th percentiles.
- Key regression results (Online Annex Table 1.3.2 and Figure 1.3.2):
  - A 1 percentage point increase in the METR is associated with about a 0.2 percent decline in firms’ fixed assets within a year and 0.4 percent after 3 years.
  - Estimated response lines and 90 percent confidence intervals illustrated in Online Annex Figure 1.3.2.

### Allocative efficiency, productivity, and factor responses
- Data and coverage: country sample of 38 advanced and emerging market economies; firms from all sectors excluding financial services and sectors with large public components; baseline unbalanced panel of 379 country-sectors over 2001–2020.
- Outcome construction: firm-level real productivity (TFPQ) estimated following Hsieh and Klenow (2009); outcomes include productivity, capital, labor, and sector-level allocative efficiency.
- Panel specification: local projections (Jordà 2005) estimating Δ^h ln y_{s,c,t} on ΔMETR_{c,s,t} with lags, country-sector and sector-year fixed effects, clustering at country-sector level.
- Results:
  - A 1 percentage point increase in the METR is followed by a decline in real productivity in the first two years.
  - Capital inputs for the median firm show a similar short-run decline before recovering.
  - Labor inputs decline more briefly, concentrated in the year of the METR change.
  - Allocative efficiency changes little in the short run but tends to deteriorate over the medium term, implying higher tax wedges may gradually worsen allocation of resources across firms.
- Graphical summary: Online Annex Figure 1.3.3 presents coefficient estimates and 90 percent confidence intervals for Productivity (TFPQ), Capital (K), and Allocative Efficiency.

### Elasticities of investment to cost of capital and aggregation issues
- Need for common framework: literature reports diverse specifications and outcomes; all estimates converted into elasticities of investment with respect to the gross cost of capital for comparability.
- Converted short-run (one-year) firm-level elasticities reported:
  - −4.27 (Chodorow-Reich and others 2024)
  - −4.7 (Curtis and others 2021)
  - −7.3 (Zwick and Mahon 2017)
  - −9.4 (from the chapter’s above regressions)
  - −10.7 (Maffini, Xing, and Devereux 2019)
- Aggregate interpretation: combined evidence suggests a short-run firm-level elasticity of investment with respect to the gross cost of capital in the range of approximately −4.3 to −10.7.
- Theoretical mapping from firm-level to long-run elasticities:
  - Firm-level long-run elasticity (neoclassical theory):
    - dlog(K)/dlog(R) = - [1 − β* * (η + (1 − η)/α* + β*)] / (1 − α* − β*)  (equation presented as Eq.3.8 in source; symbols preserved in form).
  - Economy-wide long-run elasticity:
    - dlog(K)/dlog(R) = 1 / [ (1 + ((1 − η) (α* + β*)) / η) * ε/(ε − 1) * α* − 1/η ]  (equation presented as Eq.3.9 in source; symbols preserved in form).
  - Under macro-level assumptions (ε ≈ ∞ and α* + β* = 1), the response simplifies to dlog(K)/dlog(R) = − η / (1 − α).
  - Using η in range 0.7 to 1 and α ≈ 1/3 implies aggregate long-run capital elasticity ≈ −1.04 to −1.5, which is substantially smaller than firm-level micro estimates.
- Parameter references:
  - Elasticity of substitution between capital and labor (η) literature ranges: Oberfield and Raval (2021) find η between 0.5 and 0.7; Bond and Xing (2015) cited for elasticity of substitution; Casey and Horii (2024) report share of reproducible capital = 28.5 percent in a three-factor calibration for the US.

### Simulated macro effects and sensitivity (GDP and capital responses)
- Conservative empirical elasticity used in simulations: −4.27 (Chodorow-Reich and others 2024).
- Online Annex Table 1.3.2 — Investment and capital responses to the cost of capital, by income group (numeric values preserved exactly):
  - Advanced Economies:
    - Cost of capital before tax: 10.28
    - Cost of capital after tax: 10.72
    - Short-run investment response (%): −18.4
    - Long run macro response of capital (constant elasticity of substitution = 0.7): −4.50
    - Long run macro response of capital (constant elasticity of substitution = 1): −6.43
  - Emerging Market Economies:
    - Cost of capital before tax: 10.31
    - Cost of capital after tax: 10.85
    - Short-run investment response (%): −22.4
    - Long run macro response of capital (constant elasticity of substitution = 0.7): −5.48
    - Long run macro response of capital (constant elasticity of substitution = 1): −7.83
  - Low Income Developing Countries:
    - Cost of capital before tax: 10.55
    - Cost of capital after tax: 11.12
    - Short-run investment response (%): −23.4
    - Long run macro response of capital (constant elasticity of substitution = 0.7): −5.72
    - Long run macro response of capital (constant elasticity of substitution = 1): −8.17
- Simulated GDP effects (chapter Figure 1.4, panel 2) and sensitivity (Online Annex Figure 1.3.4):
  - Baseline simulation assumptions for Figure 1.4, panel 2: α = 1/3 and η = 1.
  - Sensitivity: effects are sensitive to η and α and become smaller when either η or α is lower.
  - Numerical example: holding capital share at 33 percent (α = 0.33) while lowering elasticity of substitution to 0.5 reduces estimated GDP gains from 2.1 to 1.06 percent in advanced economies and from 2.7 to 1.35 percent in LIDCs.

*Fiscal Monitor Chapter 1 (IMF | October 2026)*

### References

### References (Chapter 1)

### Bibliographic references
- Selected cited works include:
  - Barro, Robert J., and Jason Furman. 2018. “The Macroeconomic Effects of the 2017 Tax Reform.” Brookings Papers on Economic Activity 49 (1): 257–345.
  - Bond, Stephen, and Jing Xing. 2015. “Corporate Taxation and Capital Accumulation: Evidence from Sectoral Panel Data for 14 OECD Countries.” Journal of Public Economics 130: 15–31.
  - Casey, Gregory, and Ryo Horii. 2024. “A Generalized Uzawa Growth Theorem.” Journal of Political Economy Macroeconomics 2 (2): 336–373.
  - Chodorow-Reich, Gabriel. 2025. “The Neoclassical Theory of Firm Investment and Taxes: A Reassessment.” NBER Working Paper 33922.
  - Chodorow-Reich, Gabriel, Matthew Smith, Owen M. Zidar, and Eric Zwick. 2024. “Tax Policy and Investment in a Global Economy.” NBER Working Paper 32180.
  - Devereux, Michael P., and Rachel Griffith. 2003. “Evaluating Tax Policy for Location Decisions.” International Tax and Public Finance 10 (2): 107–26.
  - Hall, Robert E., and Dale W. Jorgenson. 1967. “Tax Policy and Investment Behavior.” American Economic Review 57 (3): 391–414.
  - Hebous, Shafik, and Andualem Mengistu. 2026. “Forward-Looking Effective Tax Rates under the Global Minimum Corporate Tax.” IMF Economic Review.
  - Hsieh, Chang-Tai, and Peter J. Klenow. 2009. “Misallocation and Manufacturing TFP in China and India.” Quarterly Journal of Economics 124 (4): 1403–48.
  - Jordà, Òscar. 2005. “Estimation and Inference of Impulse Responses by Local Projections.” American Economic Review 95 (1): 161–82.
  - Maffini, Giorgia, Jing Xing, and Michael P. Devereux. 2019. “The Impact of Investment Incentives: Evidence from UK Corporation Tax Returns.” American Economic Journal: Economic Policy 11 (3): 361–89.
  - Oberfield, Ezra, and Devesh Raval. 2021. “Micro Data and Macro Technology.” Econometrica 89 (2): 703–732.
  - Organisation for Economic Co-operation and Development (OECD). 2023. Corporate Tax Statistics 2023. Paris: OECD Publishing.
  - Zwick, Eric, and James Mahon. 2017. “Tax Policy and Heterogeneous Investment Behavior.” American Economic Review 107 (1): 217–48.
- Additional methodological and theoretical references cited include Saez (2001), Piketty and Saez (2013), Bourguignon and Spadaro (2012), Jacquet, Lehmann and Van der Linden (2013), and others relevant to optimal taxation, tax-benefit revealed preferences, and sufficient-statistics approaches.

### Online Annex 1.4 — Optimal Labor Taxation: model, calibration, and reform simulation findings
- Model setup and objectives:
  - Government maximizes social welfare W = ∫_0^∞ ω_z G(u_z) f_z(z) dz, with social weights ω_z and isoelastic G(u_z) = (u_z^{1−γ} − 1)/(1−γ) (Eq.4.2).
  - Individuals choose participation (extensive margin) and hours (intensive margin) in response to net-of-tax wages.
  - Optimal marginal tax schedule derived as a continuous, non-linear function balancing revenue needs, efficiency costs, and redistribution motives.
- Calibration and data inputs:
  - Household survey data from Luxembourg Income Study (LIS); nominal values inflation-adjusted to represent income distributions in 2024.
  - Country coverage:
    - EMDEs: Brazil, Côte d'Ivoire, Georgia, Peru, Poland, Vietnam.
    - AEs: Canada, Germany, Spain, United Kingdom, Israel, Japan, Sweden.
  - Tax and transfer parameters from IMF TAXFIT for Brazil, Côte d'Ivoire, Georgia, Peru, Vietnam; OECD TaxBEN for remaining countries.
  - Marginal effective tax rates (METRs) used to capture labor supply incentives.
- Elasticity assumptions (preserved exactly):
  - Intensive labor supply elasticity = 0.2 (that is, a 10-percent increase in after-tax earnings increases labor supply by 2 percent at the margin).
  - Average extensive margin elasticity = 0.25 (that is, a 10-percent increase in after-tax income increases the propensity to participate in the labor market by 2.5 percent).
  - The extensive margin elasticity is assumed to be the highest for low- and high-income earners.
- Inferred inequality aversion:
  - Estimated inequality aversion implicit in the tax and benefit systems: EMDEs average 0.2; selected AEs average 0.7.
- Optimal marginal tax formula (sufficient-statistic representation):
  - T'(z) = [1 − g̅_z − π̅_z] / [1 − g̅_z − π̅_z + ε_z α_z] (Eq.4.3).
  - Definitions preserved exactly:
    - g̅_z = (1/(1 − F_z(z))) ∫_z^∞ g_y f_z(y) dy with g_y = ω_y G'(u_y)/λ.
    - π̅_z = (1/(1 − F_z(z))) ∫_z^∞ [Γ_y/(1 − Γ_y)] π_y f_z(y) dy.
    - α_z = z f_z(z) / (1 − F_z(z)).
    - ε_z is the intensive elasticity of earnings with respect to the net-of-tax rate.
- Reform simulation design and key aggregate results reporting conventions:
  - Reform derived from Eq.4.3 using estimated densities and calibrated elasticities.
  - Aggregate labor-supply and earnings effects computed from changes in effective marginal and participation tax rates; hours and earnings reported in percent terms; employment in percentage points.
- Reform characterization and distributional effects (group averages are unweighted across sample countries):
  - Reform packages mainly extend in-work support farther up the lower part of the earnings distribution.
  - EMDEs average changes under Reform (group averages):
    - Income cutoff for in-work tax credit rises from 43 to 59 percent of median income.
    - Share of exempt workers increases from 16 to 24 percent.
    - METRs (percent): 10th percentile 25 → 21; 50th percentile 16 → 23; 90th percentile 26 → 21.
    - Earnings gains concentrated in lower half of distribution, peaking at around 7.5 percent near the bottom decile; earnings effects decline and turn slightly negative in the upper half before rising for top earners.
  - AEs average changes under Reform (group averages):
    - Maximum in-work tax credit (percent of median income) rises from 24 to 27.
    - Income cutoff rises from 51 to 56 percent of median income.
    - Share of exempt workers increases from 20 to 23 percent.
    - METRs (percent): 10th percentile 47 → 51; 50th percentile 36 → 42; 90th percentile 44 → 36.
    - Earnings gains largest at the very bottom and the very top of the distribution, reaching about 6 to 8 percent at both ends; middle-income earners experience small losses with a trough around the middle.
  - Reform effect on labor-supply disincentives:
    - EMDEs: Reform lowers hours-of-work disincentives for low and high incomes (panel 1).
    - AEs: Reform lowers entry-to-work disincentives for low and high incomes (panel 2).
- Table of average taxes and benefits: Current and Reform scenarios (group averages)
  - EMDEs — Current / Reform
    - In-work tax credit — Maximum amount (percent of median income): 17 / 17
    - Income cutoff (percent of median income): 43 / 59
    - Share of exempt workers (percent): 16 / 24
    - Marginal effective tax rates (percent):
      - 10th percentile: 25 / 21
      - 50th percentile: 16 / 23
      - 90th percentile: 26 / 21
  - AEs — Current / Reform
    - In-work tax credit — Maximum amount (percent of median income): 24 / 27
    - Income cutoff (percent of median income): 51 / 56
    - Share of exempt workers (percent): 20 / 23
    - Marginal effective tax rates (percent):
      - 10th percentile: 47 / 51
      - 50th percentile: 36 / 42
      - 90th percentile: 44 / 36
- Distributional pattern summary:
  - EMDEs: more hump-shaped profile of earnings changes (largest gains in lower half, peak ~7.5 percent near bottom decile).
  - AEs: more U-shaped profile (gains ~6–8 percent at both bottom and top, small losses for middle-income earners).

### Online Annex 1.5 — Tax Compliance Costs, Tax Complexity, and Firm Performance: objective and data
- Objective:
  - Assess whether high-quality tax administration and simpler tax frameworks disproportionately benefit small and young firms by examining:
    - (i) tax administration quality and firm performance;
    - (ii) tax administration quality, compliance costs, and perceptions of the business environment;
    - (iii) tax framework complexity and firm performance.
- Data and key measurement constructs:
  - Firm-level data: World Bank Enterprise Surveys.
  - Tax Administration Quality Index (TAQI):
    - Constructed from TADAT assessments.
    - Based on taxpayer information services, filing and payment procedures, post-filing processes, transparency and accountability.
    - Range: 1 to 4, with higher values indicating lower compliance costs.
    - Coverage: 63 countries and 72 country-survey year observations over 2013-2025.
    - TADAT scores assumed unchanged for four years after an assessment to maximize coverage.
  - Tax framework complexity:
    - Measured using the Euler and others (2024) corporate income tax complexity index, denoted TAXCOMP, which ranges from 0 to [value truncated in source].

*Fiscal Monitor, Chapter 1. IMF | October 2026.*

### 1. Specifications using this index cover 23 countries over 2015-2024.

### Specifications using this index cover 23 countries over 2015-2024.

### Empirical Strategy
- Three firm-level baseline specifications are estimated:
  - Equation (5.1): Performanceiscv = β₁ FirmTypeiscv + β₂ (FirmTypeiscv × TAQIc) + Xiscv + μcsv + εiscv.
  - Equation (5.2): Costs/Perceptionsiscv = γ₁ FirmTypeiscv + γ₂ (FirmTypeiscv × TAQIc) + Xiscv + μcsv + εiscv.
  - Equation (5.3): Performanceiscv/Percefformancesiscv = θ₁ FirmTypeiscv + θ₂ (FirmTypeiscv × TAXCOMPc) + Xiscv + μcsv + εiscv.
- Identification:
  - All specifications include country-industry-survey year fixed effects; identification comes from comparing small or young firms with larger or older firms within the same country, industry, and survey year.
  - FirmTypeiscv is a dummy indicating a small firm or, in some specifications, a young firm. Small firms have 20 or fewer employees; young firms are nine years old or younger.
  - Firm controls include government ownership, exporter status, and foreign ownership. Subscripts (i), (c), (s), and (v) denote firm, country, industry, and survey year, respectively.
  - Firm performance is measured by labor productivity, labor productivity growth, and annual sales growth.
  - Compliance costs are measured by annual hours spent on tax compliance (HRCOMP). Perceptions include ADMBIGOBS (tax administration as the biggest obstacle), REFOBS (VAT refund difficulties), and INFORM (competition from informal firms).
  - The World Bank Enterprise Surveys consist of repeated cross-sections; firm fixed effects cannot be included.
- Robustness checks:
  - Interact FirmTypeiscv with alternative country-level variables, including governance and tax complexity, and use alternative definitions of small and young firms (Online Annex Table 1.5.3 and Online Annex Table 1.5.4).
  - Results remain robust and statistically significant across these specifications.
- Sample coverage:
  - The sample includes Albania, Armenia, Belarus, Belgium, Benin, Bhutan, Bosnia and Herzegovina, Botswana, Bulgaria, Cambodia, Chad, Colombia, Congo, Democratic Republic of the, Côte d'Ivoire, Czech Republic, Dominican Republic, Ecuador, Egypt, The Gambia, Georgia, Ghana, Greece, Guatemala, Indonesia, Iraq, Jamaica, Jordan, Kazakhstan, Kenya, Kyrgyz Republic, Lebanon, Lesotho, Liberia, Madagascar, Malaysia, Mali, Malta, Moldova, Morocco, Mozambique, Nepal, Niger, North Macedonia, Pakistan, Papua New Guinea, Paraguay, Peru, Philippines, Romania, Rwanda, Serbia, Sierra Leone, Slovak Republic, South Africa, Spain, Tajikistan, Tanzania, Tonga, Trinidad and Tobago, Tunisia, Ukraine, Uzbekistan, and Zambia.
  - One specification explicitly notes the sample includes 23 countries (Online Annex Table 1.5.3).

### Results
- Three consistent empirical findings:
  - Better tax administration (higher TAQI) is associated with stronger performance among small and young firms.
  - Better tax administration lowers compliance burdens and improves perceptions of the tax administration and the business environment for small and young firms.
  - Greater tax framework complexity (TAXCOMP) is associated with weaker relative performance among small and young firms.
- Quantitative magnitudes and comparisons:
  - TAQI interactions (small × TAQI and young × TAQI) are positive and statistically significant across performance measures (labor productivity, labor-productivity growth, total factor productivity). Moving from the 25th to the 75th percentile of TAQI:
    - Narrows the labor-productivity gap from about 29 to 17 percent for small firms.
    - Narrows the labor-productivity gap from about 22 to 12 percent for young firms.
  - TAQI effect on perceptions and compliance (Online Annex Table 1.5.2):
    - Small × TAQI and Young × TAQI interaction terms are negative and statistically significant for HRCOMP, ADMBIGOBS, REFOBS, and INFORM: as TAQI improves, small and young firms spend less time on tax compliance, are less likely to identify tax administration as their main obstacle, report fewer VAT refund difficulties, and perceive less informal competition.
  - TAXCOMP interactions (Online Annex Table 1.5.3):
    - Small × TAXCOMP and Young × TAXCOMP interaction terms are negative in all specifications and statistically significant in most cases, indicating higher tax complexity widens performance gaps.
    - Moving from high to low tax complexity (holding TAQI at its mean) reduces the small-firm labor-productivity gap from about 28 percent to 4 percent and narrows the young-firm gap from about 22 percent to 12 percent.
    - Lower complexity is also associated with higher labor-productivity growth, stronger sales growth, and better total factor productivity outcomes for small and young firms.
- Heterogeneity:
  - High-quality tax administration improves firm performance in both emerging markets and low-income developing countries, with larger effects in low-income developing countries where performance gaps are greatest (Online Annex Table 1.5.4).
- Complementarity:
  - Administrative quality and legal simplicity are complementary: better administration reduces the cost of interacting with the tax authority, while simpler tax rules make compliance and planning less burdensome. Both channels matter most for firms with fewer internal tax resources.

### Empirical Tables (select highlights)
- Online Annex Table 1.5.1 (Equation 5.1; OLS; country × industry × survey FE; robust SE clustered at country-industry-survey level):
  - Small × TAQI coefficient estimates (columns using Small × TAQI): 0.211*** (LABPRO), 1.935*** (LABPROGR), 3.103*** (ANSALES GR).
  - Young × TAQI coefficient estimates (columns using Young × TAQI): 0.162*** (LABPRO), 2.127*** (LABPROGR), 3.227*** (ANSALES GR).
  - Observations reported: 300882, 621326, 820300882621326820 (table formatting preserves counts as presented).
- Online Annex Table 1.5.2 (Equation 5.2; OLS for HRCOMP; conditional logit marginal effects for binary outcomes; clustering as above):
  - Small × TAQI: -0.139** (HRCOMP), -0.151*** (ADMBIGOBS), -0.102*** (REFOBS), -0.038** (INFORM).
  - Young × TAQI: -0.118** (HRCOMP), -0.251*** (ADMBIGOBS), -0.064*** (REFOBS), -0.040** (INFORM).
  - Observations examples: 12677, 17844, 77303, 103 (as presented in table).
- Online Annex Table 1.5.3 (Equation 5.3; OLS; sample includes 23 countries; clustering as above):
  - Small × TAXCOMP: -1.961*** (LABPRO), -18.933** (LABPROGR), -19.433** (ANSALES GR).
  - Young × TAXCOMP: -0.816* (LABPRO), -18.043** (LABPROGR), -23.978** (ANSALES GR).
  - Small × TAQI (in same specification): 0.189*** (LABPRO), 1.044* (LABPROGR), 2.264*** (ANSALES GR).
  - Observations reported: 157581, 388814, 252 (as presented).
- Online Annex Table 1.5.4 (robustness checks; dependent variable LABPRO (logs); country × industry × survey FE):
  - Alternative specifications by income groups and alternative firm definitions produce consistent positive Small × TAQI and Young × TAQI coefficients (examples: Small × TAQI: 0.070, 0.195***, 0.346***, 0.160***; Young × TAQI: -0.015, 0.039, 0.298***, 0.108***).
  - Columns show results for AEs, EMs, LIDCs, alternative size/age definitions, and institutional quality controls.
  - Observations examples: 26042, 60418, 89118, 89185 (as presented).

### Interpretation and Policy Implications (inferred from results)
- Improving tax administration quality (TAQI) can narrow productivity and growth gaps between small/young firms and larger/older firms by reducing compliance frictions and improving perceptions of tax administration.
- Reducing tax framework complexity (TAXCOMP) supports small and young firms’ relative productivity and growth; simpler rules make compliance and planning less burdensome.
- Policy emphasis on both administrative reforms (to raise TAQI) and legal simplification (to lower TAXCOMP) is likely to deliver the largest gains for firms with fewer internal tax resources and in low-income developing countries where performance gaps are greatest.

---


_Source: https://www.imf.org/-/media/files/publications/fiscal-monitor/2026/october/english/ch1.pdf_