## wpiea2025089-print-pdf

## Source details

**Canonical URL:** [wpiea2025089-print-pdf](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025089-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025089-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025089-print-pdf.pdf.json)

---

### Key questions and scope
- Investigates whether progress in firm digitalization contributes to tax collection, through which channels, and whether synergies exist between firm digitalization and GovTech.
- Uses a comprehensive cross-country database combining business and government digitalization measures with country and firm-level corporate tax payments.
- Firm digitalization proxied by the World Bank Digital Adoption Index (DAI) business sub-index (available for years 2014 and 2016). Government digitalization captured by the United Nations E-Governance index.

### Principal findings (country- and firm-level)
- Positive relationship between firm digitalization and domestic tax revenues:
  - Countries with higher levels of business digital adoption have larger tax-to-GDP ratios.
  - A one-standard-deviation increase in firm digitalization is associated with an increase in tax revenues-to-GDP by up to 3 percentage points (country-level estimates).
  - Firm-level: a one-standard-deviation increase in firm digitalization is associated with an increase in the ratio of tax payments to corporate turnover by up to 0.12-0.32 percentage points.
  - Scaled by sample means: about a fifth of average tax revenues-to-GDP ratio (country-level) and by 6-20 percent of firms’ average tax payment over turnover ratio (firm-level).
- Mechanism: improved compliance appears primary:
  - Stronger relationship for firms deemed riskier in compliance terms—specifically, smaller firms and firms operating in countries with higher levels of informality.
  - Oaxaca-Blinder decomposition: unexplained components including the compliance cost channel account for about 40-70 percent of the observed relationship between digitalization and tax payments in the service sector.
- Synergies with GovTech:
  - Impact of firm digitalization on tax revenues is significantly stronger as governments mature in their use of digital technologies for tax administration—confirming positive synergies between private and public sector digitalization.

### Country-level descriptive patterns and regression framework
- Descriptive correlations:
  - Strong positive correlation between the DAI business sub-index and ratio of non-trade tax revenues (normalized by GDP).
  - Positive correlations between firm digitalization and VAT C-efficiency and CIT productivity.
  - Pairwise correlation: firm digitalization has correlation of 61 percent with non-trade tax revenues, slightly higher than household or government digitalization.
  - All reported correlations are different than zero at 1 percent significance level.
- Regression specification:
  - Outcome RR_c,y is the ratio of non-trade tax revenues over GDP for country c in year y.
  - Main regressor: DAI business sub-index (firm digitalization).
  - Model: RR_c,y = δ_y + β · DAI_c,y + γ · X_c,y + ε_c,y, with year fixed effects δ_y.
- Controls (X_c,y):
  - Structural macro variables: log of GDP per capita and its square, share of agriculture in GDP, oil exporters status.
  - Cyclical macro variables: inflation and GDP growth.
  - Institutional quality: rule of law, accountability, control of corruption, regulatory quality.
  - Human capital: average years of schooling.
  - Two tax policy reform measures from Chang, Gueorguiev, Gavin, and Honda (2020): (i) difference between tax revenues projected in a budget for year and actual tax revenues of preceding year; (ii) difference between budgeted revenues in a year versus the previous one, expressed as share of GDP.
- Identification notes:
  - Linear correlations may be confounded by richer countries having more digitalized businesses and better tax collection, better institutions, structural transformation, or human capital; controls included to mitigate omitted variable bias.
  - Results robust to alternative tax policy controls built from statutory rates.

### Cross-country regression magnitudes and robustness
- Main magnitudes:
  - In the most saturated specification (Table 1, column 4), a one-standard deviation increase in the firm digitalization index (0.19) is associated with an increase of tax revenues-to-GDP of 3 percentage point.
  - Sample average tax revenues-to-GDP ratio is 16 percent; the 3 percentage point effect equals about a fifth of that average.
  - Illustrative 2016 averages: non-trade tax-to-GDP ratio 11 percent in low-income countries and 25 percent in advanced economies; average firm digitalization index 0.39 in low-income countries and 0.83 in advanced economies. Coefficient implies if low-income countries had advanced-economy firm digitalization, expected tax-to-GDP would increase by about 7 percentage point (reaching 18 percent of GDP).
- Robustness:
  - Results robust to alternative specifications, lagged controls, inclusion of lagged tax revenues, and alternative measures of tax policy (forgone revenues due to tax expenditures) though sample size shrinks.
  - Correlation with VAT C-efficiency and CIT productivity is positive but not statistically significant when all controls included (Table AI.1).
- Caveats:
  - Firm digitalization measure available only for two close years (2014 and 2016), so variation is mainly cross-sectional; causal interpretation limited without reliable instrument.

### Firm-level data, measurement, and regression design
- Data:
  - Orbis (Bureau van Dijk) covering about 36 million non-financial firms, headquartered in 56 countries (31 AEs & 25 EMDEs), over 2011–2020.
- Firm digitalization ("Digital Intensity"):
  - Measured by share of intermediate inputs sourced from three ICT intensive sectors (manufacturing of computer, electronic and optical equipment; telecommunication; and IT and other information services) using OECD harmonized Input-Output Tables (2021 edition).
- Outcome variables y_{i,c,t}:
  - Binary indicator equal to one if firm records positive tax payments (extensive margin; excludes firms operating at loss with negative EBIT).
  - Log of tax payments (intensive margin).
  - Tax payment normalized by operating revenues (turnovers) (intensive margin).
- Estimation:
  - Two-way fixed effects with country μ_c and year τ_t, country-specific linear trend μ_c · τ_t; controls include firm age, number of employees, total assets, real GDP growth, inflation. Standard errors clustered at country and year level.

### Firm-level empirical results (selected)
- Advanced Economies (Table 3, columns 1–3):
  - Digital intensity coefficients:
    - Tax paid (binary): 0.408∗∗∗ (std. err. 0.053).
    - Log(real tax payment): 2.702∗∗∗ (std. err. 0.186).
    - Tax payment/Turnovers: 0.036∗∗∗ (std. err. 0.004).
  - Observations: 13,282,566; 11,519,808; 11,519,808 respectively. Adj. R2: 0.147; 0.598; 0.138. Number of countries: 23.
- Emerging Market and Developing Economies (Table 3, columns 4–6):
  - Digital intensity coefficients:
    - Tax paid (binary): 0.075∗∗∗ (std. err. 0.019).
    - Log(real tax payment): 3.490∗∗∗ (std. err. 0.214).
    - Tax payment/Turnovers: 0.081∗∗∗ (std. err. 0.006).
  - Observations: 15,316,855; 15,836,281; 15,836,281 respectively. Adj. R2: 0.0992; 0.625; 0.119. Number of countries: 31.
- Interpretation:
  - Firm digitalization associated with higher probability of paying taxes (extensive margin), with larger marginal impact in EMDEs where non-compliance is more prevalent.
  - Larger tax payments on the intensive margin in both AEs and EMDEs.
  - Firm-level one-standard-deviation increases imply tax payments/operating revenue increases of 0.12 percentage points in AEs and 0.32 percentage points in EMDEs (equal to 6 percent and 20 percent of firms’ average tax payment over turnover ratio respectively).
- Caveats:
  - Intensive margin may be underestimated due to profit shifting by high-tech multinationals.
  - Extensive margin estimates should be interpreted with caution because Orbis treats negative tax payments as nonpayment.

### Heterogeneity, interactions with GovTech, and mechanisms
- Interaction framework (Equation (3)):
  - Adds interaction term between business ICT (D) and heterogeneity variable H (firm size, informality, GovTech).
- GovTech complementarity:
  - Extensive margin (EMDEs): as ICT expenditure per taxpayer transitions from 25th to 75th percentile, a one-standard-deviation increase in digital intensity leads to about an 11 percent increase in probability of tax payments (calculation: 2.79 × 0.04 × 100 = 11.2 percent).
  - Intensive margin (AEs): analogous transition implies about a 1.5 percent increase in tax payments relative to operating revenues (calculation: 0.48 × 0.032 × 100 = 1.5 percent).
  - Marginal effects of firm digitalization are statistically different from zero only when E‑governance index is above 0.5 (about half the sample).
- Firm size and informality:
  - Effects larger for small firms (defined as less than five employees) and in countries with high business informality (business informality index from Elgin et al. (2021); high informality when index is above global average).
  - In EMDEs with high informality, firm digitalization more significantly enhances tax payments by detecting liabilities and reducing compliance costs.
- Oaxaca-Blinder decomposition:
  - In the service sector, unexplained components (including compliance cost channel) explain 40-70 percent of larger tax payments for digital firms relative to non-digital firms.
  - In manufacturing, positive impact mostly explained by observable differences between digital and less-digital firms.

### Annex II — heterogeneous regression highlights (selected coefficients and samples)
- Panel A — Advance Economies (selected):
  - Baseline Digital intensity coefficients:
    - (3) Tax paid: 0.096*** (std. err. 0.027)
    - (4) Tax payment/Revenue: 0.044*** (std. err. 0.004)
    - (6) Tax payment/Revenue: 0.103*** (std. err. 0.011)
  - Digital x ICT investment/taxpayers:
    - Column (4) coefficient: 0.019*** (std. err. 0.006)
  - Observations range across columns from 2,672,091 to 15,836,281; number of countries reported between 17 and 31.
- Panel B — EMDEs (selected):
  - Baseline Digital intensity coefficients:
    - (3) Tax paid: 0.363*** (std. err. 0.058)
    - (4) Tax payment/Revenue: 0.027*** (std. err. 0.003)
    - (5) Tax paid: 0.392*** (std. err. 0.044)
  - Digital x ICT investment/taxpayers:
    - Column (2) coefficient: 0.107*** (std. err. 0.037)
  - Observations range across columns from 4,394,066 to 13,282,576; number of countries reported between 16 and 25.
- All annex regressions include firm controls, macro controls, country FE, year FE, and country-specific trend; standard errors clustered at country and year level. Significance notation: ∗ p< 0.1, ∗∗ p< 0.05, ∗∗∗ p< 0.01.

### Quantitative illustrative estimates and cost benchmark
- Country-level illustrative: one-standard-deviation increase in firm digitalization associated with increase in tax revenues-to-GDP by up to 3 percentage points.
- Firm-level illustrative: one-standard-deviation increase in firm digitalization associated with increase in ratio of tax payments to corporate turnover by up to 0.12-0.32 percentage points.
- Relative scale: country-level increase about a fifth of average tax revenues-to-GDP; firm-level increase 6-20 percent of firms’ average tax payment over turnover ratio.
- Policy-relevant projection: if low-income countries achieved advanced-economy firm digitalization (contingent on sufficient GovTech), tax-to-GDP gap nearly halved by illustrative back-of-the-envelope.
- Cost benchmark: provision of universal 4G cellular broadband access estimated to cost at most $418 billion (about 0.45 percent of global GDP) (Oughton et al. (2023)).

### Conclusions and policy recommendations
- Conclusions:
  - Positive relationship between firm digitalization and tax revenues is robust across country- and firm-level analyses, though causal certainty is limited without a highly credible instrument.
  - Compliance channel is important—effects stronger for smaller firms and in high-informality environments.
  - Significant synergy exists between firm digitalization and GovTech: tax gains from firm digitalization are larger where revenue administrations invest in ICT, and vice versa.
- Policy recommendations:
  - Adopt a dual approach: promote both GovTech and firm digitalization to maximize tax collection benefits through complementarities.
  - Coordinate public and private digital investments to lower fiscal “net” cost and increase returns from digitalization, especially in fiscally constrained countries.
  - Supportive measures: universal broadband connectivity and incentives for firms to accelerate ICT investments.
  - Parallel reforms required: institutional and legal frameworks, organizational strength, and staff competency at tax administrations alongside IT investments to ensure effective adoption and revenue gains.

*Source: IMF Working Paper — Leveraging Digital Technologies in Boosting Tax Collection (Country-level Evidence; Chapter 15; Annex II).*

### 1. Country-level Evidence ..............................................................................................

### 1. Country-level Evidence

### Key questions and scope
- Investigates whether progress in firm digitalization contributes to tax collection, through which channels, and whether synergies exist between firm digitalization and GovTech.
- Uses a comprehensive cross-country database combining business and government digitalization measures with country and firm-level corporate tax payments.
- Firm digitalization is proxied by the World Bank Digital Adoption Index (DAI) business sub-index (available for years 2014 and 2016). Government digitalization is captured by the United Nations E-Governance index.

### Principal findings (from country- and firm-level analyses)
- Positive relationship between firm digitalization and domestic tax revenues:
  - Countries with higher levels of business digital adoption have larger tax-to-GDP ratios.
  - A one-standard-deviation increase in firm digitalization is associated with an increase in tax revenues-to-GDP by up to 3 percentage points (country-level estimates).
  - Firm-level estimates: a one-standard-deviation increase in firm digitalization is associated with an increase in the ratio of tax payments to corporate turnover by up to 0.12-0.32 percentage points.
  - When scaled by sample means, effects are: about a fifth of average tax revenues-to-GDP ratio (country-level) and by 6-20 percent of firms’ average tax payment over turnover ratio (firm-level).
- Mechanism: improved compliance appears to be the primary channel:
  - The firm-level analysis finds the relationship between firm digitalization and taxes paid is stronger for firms deemed riskier in compliance terms—specifically, smaller firms and firms operating in countries with higher levels of informality.
  - Oaxaca-Blinder decomposition indicates unexplained components including the compliance cost channel account for about 40-70 percent of the observed relationship between digitalization and tax payments in the service sector.
- Synergies with GovTech:
  - The impact of firm digitalization on tax revenues is estimated to be significantly stronger as governments mature in their use of digital technologies for tax administration, confirming significant positive synergies between private and public sector digitalization.

### Country-level descriptive patterns and correlations
- Binned scatter plots and correlations show:
  - Strong positive correlation between the DAI business sub-index and the ratio of non-trade tax revenues (normalized by GDP).
  - Positive correlations between firm digitalization and VAT C-efficiency and CIT productivity.
  - Pairwise correlation: firm digitalization has correlation of 61 percent with non-trade tax revenues, slightly higher than household or government digitalization.
  - All reported correlations are different than zero at 1 percent significance level.

### Country-level regression framework
- Outcome variable:
  - RR_c,y is the ratio of non-trade tax revenues over GDP for country c in year y.
- Main regressor:
  - DAI business sub-index (firm digitalization).
- Model specification:
  - RR_c,y = δ_y + β · DAI_c,y + γ · X_c,y + ε_c,y, with year fixed effects δ_y.
- Controls (X_c,y) include:
  - Structural macroeconomic variables: log of GDP per capita and its square, share of agriculture in GDP, oil exporters status.
  - Cyclical macroeconomic variables: inflation and GDP growth.
  - Quality of institutions: rule of law, accountability, control of corruption, regulatory quality.
  - Human capital: average years of schooling.
  - Two measures of tax policy reforms from Chang, Gueorguiev, Gavin, and Honda (2020): (i) the difference between the tax revenues projected in a budget for year and the actual tax revenues of the preceding year and (ii) the difference between the budgeted revenues in a year versus the previous one, expressed as share of GDP.
- Notes on identification:
  - The linear correlations may be confounded by richer countries having more digitalized businesses and better tax collection, better institutions, structural transformation, or human capital availability; the regression includes controls to mitigate omitted variable bias.
  - Results are robust to alternative tax policy controls built from statutory rates (reference to Appendix Table AI.4).

### Data and measurement notes
- DAI business sub-index is the average of four normalized indicators: the percentage of businesses with websites, the number of secure servers, the speed of download, and 3G coverage in the country (World Bank 2016).
- The tax revenue measure excludes taxes collected on imports (such as import duties) and includes VAT, CIT, and PIT.
- VAT C-efficiency is defined as the ratio between VAT revenues over the theoretical VAT revenues—consumption multiplied by the standard VAT rate.
- CIT productivity captures the efficiency of corporate income tax collection analogously.

### Policy implications (country-level)
- Findings support a dual approach promoting both GovTech and firm digitalization to enhance tax revenues.
- Investing in digitalization is argued to pay off by increasing future tax revenues; fiscal costs of digital investments are likely to be offset by future tax revenue increases.
- Policies for universal broadband and ICT investments, scaling up ICT expenditure by tax administrations, and ensuring adequate staffing and expertise among tax officials are highlighted as complementary measures.

*Source: IMF Working Paper — Leveraging Digital Technologies in Boosting Tax Collection (Country-level Evidence section).*

### 15. The regression analysis

### 15. The regression analysis

### Cross‑country regression findings
- Business digitalization is statistically significantly associated with higher tax revenues over GDP in cross-sectional regressions (Table 1, columns (1) to (4)), controlling for macroeconomic (structural, cyclical, and human capital), institutional quality, and tax policy factors.
- Business digitalization has stronger predictive power for tax revenues than E-government and household digitalization.

### Magnitude and economic significance
- In the most saturated specification (Table 1, column 4), a one-standard deviation increase in the firm digitalization index (0.19) is associated with an increase of tax revenues-to-GDP of 3 percentage point.
- The sample average tax revenues-to-GDP ratio is 16 percent; the 3 percentage point effect equals about a fifth of that average.
- Illustrative back-of-the-envelope comparisons using 2016 averages:
  - Average non-trade tax-to-GDP ratio: 11 percent in low-income countries and 25 percent in advanced economies.
  - Average firm digitalization index: 0.39 in low-income countries and 0.83 in advanced economies.
  - The coefficient in column (4) implies that if low-income countries had the same firm digitalization as advanced economies, the expected tax-to-GDP ratio would increase by about 7 percentage point (reaching 18 percent of GDP), reducing the difference between the two groups by about half.
- Note: the back-of-the-envelope calculation is illustrative and not a policy counterfactual.

### Robustness checks and caveats
- Results are robust to multiple alternative empirical specifications:
  - Including only one sample.
  - Including lagged controls to allow for lagged impact of business cycle fluctuations on tax collection.
  - Including lagged tax revenues as controls (Table AI.2).
- Results broadly robust to inclusion of forgone revenues due to tax expenditures as an alternative measure of tax policy (sample size shrinks significantly).
- Correlation of firm digitalization with VAT C‑efficiency and CIT‑productivity is positive but not statistically significant when all controls are included (Table AI.1).
- Limitations:
  - Not possible to control for all potential sources of cross‑country differences; firm digitalization measure is available only for two close years, so variation is mainly cross‑sectional.
  - Adding country fixed effects in cross‑country exercise would not produce meaningful results.
  - Causal interpretation is limited without a reliable instrument.

### Synergies between business digitalization and E‑Government (GovTech)
- The relationship between firm digitalization and tax revenues is stronger when E‑Government is high (Table 1, Figure 2).
- Regression allowing coefficients to vary by E‑governance median (Table 1, columns (5)–(8)):
  - After macro and institutional controls, firm digitalization coefficient is significantly different from zero only within the high E‑governance sample; smaller within the low E‑governance sample.
- Interaction specification (Equation (1)) estimates country‑year specific marginal “effects”:
  - Marginal effect of firm digitalization = β + θ · E‑government.
  - Marginal effect of E‑government = α + θ · firm digitalization.
  - Estimated interaction coefficient θ is positive and statistically significant (Table AI.3), indicating synergies.
  - Marginal effects of firm digitalization are statistically different from zero only when the E‑governance index is above 0.5 (about half the sample).

### Robustness of GovTech proxies
- Synergy results are robust to alternative government digitalization proxies:
  - United Nations E‑Governance index (baseline).
  - GovTech maturity index from WB.
  - Digital connectivity (Alper and Miktus, 2019).
  - DAI Government subindex from WB.
  - Digitalization of revenue administration measures from ISORA.
- Graphical illustrations in Annex Figure AI.1.

### Country heterogeneity and policy implications
- Figure 4 classifies countries by E‑Government and firm digitalization, by income groups:
  - Red diamonds: both e‑government and firm digitalization marginal “effects” positive and statistically significant at 95 percent confidence.
  - Orange/pink triangles: only one of the two marginal effects is positive and statistically significant at 95 percent confidence.
  - Black dots: neither marginal effect statistically different from zero.
- Countries with low levels of both business and government digitalization may fail to reap fiscal benefits unless improvements occur on both fronts.
- Low‑income countries, on average, collect a small share of GDP in taxes and have non‑significant marginal effects of both business and GovTech, suggesting need for a dual approach.

### Firm‑level evidence: data and measurement
- Firm‑level analysis uses Orbis (Bureau van Dijk) covering about 36 million non‑financial firms, headquartered in 56 countries (31 AEs & 25 EMDEs), over 2011–2020.
- Firm digitalization (“Digital Intensity”) measured by share of intermediate inputs sourced from three ICT intensive sectors (manufacturing of computer, electronic and optical equipment; telecommunication; and IT and other information services) using OECD harmonized Input‑Output Tables (2021 edition).
- Mapping from Orbis NACE to OECD industry classification at 2‑digit level.

### Descriptive statistics (Orbis, Table 2)
- Panel A: Advanced Economies (N and statistics shown)
  - Tax paid: N = 15,323,618; Mean = 0.930; Std. Dev = 0.256; P10 = 1; P25 = 1; P50 = 1; P75 = 1; P90 = 1.
  - Log (real tax payments): N = 16,163,919; Mean = 8.970; Std. Dev = 2.300; P10 = 6.330; P25 = 7.410; P50 = 8.857; P75 = 10.34; P90 = 11.93.
  - Tax payments/Turnovers: N = 15,842,290; Mean = 0.021; Std. Dev = 0.027; P10 = 0.001; P25 = 0.004; P50 = 0.011; P75 = 0.026; P90 = 0.053.
  - Digital intensity: N = 19,404,315; Mean = 0.024; Std. Dev = 0.032; P10 = 0.007; P25 = 0.010; P50 = 0.017; P75 = 0.025; P90 = 0.039.
  - Firm age: N = 19,395,973; Mean = 18.19; Std. Dev = 15.30; P10 = 4; P25 = 7; P50 = 14; P75 = 25; P90 = 38.
  - Number of employees: N = 19,404,653; Mean = 25.41; Std. Dev = 78.05; P10 = 1; P25 = 2; P50 = 5; P75 = 14; P90 = 45.
- Panel B: Emerging Market and Developing Economies (N and statistics shown)
  - Tax paid: N = 13,339,673; Mean = 0.741; Std. Dev = 0.438; P10 = 0; P25 = 0; P50 = 1; P75 = 1; P90 = 1.
  - Log(real tax payments): N = 11,794,137; Mean = 7.088; Std. Dev = 2.166; P10 = 4.515; P25 = 5.742; P50 = 6.935; P75 = 8.295; P90 = 9.725.
  - Tax payments/Turnovers: N = 11,565,269; Mean = 0.016; Std. Dev = 0.018; P10 = 0.001; P25 = 0.003; P50 = 0.010; P75 = 0.024; P90 = 0.037.
  - Digital intensity: N = 17,349,182; Mean = 0.026; Std. Dev = 0.040; P10 = 0.004; P25 = 0.007; P50 = 0.015; P75 = 0.029; P90 = 0.056.
  - Firm age: N = 17,271,671; Mean = 10.14; Std. Dev = 7.894; P10 = 2; P25 = 4; P50 = 8; P75 = 14; P90 = 21.
  - Number of employees: N = 17,349,214; Mean = 21.24; Std. Dev = 60.71; P10 = 1; P25 = 2; P50 = 5; P75 = 15; P90 = 39.

### Firm‑level regression design
- Two‑way fixed effect estimating equation (Equation (2)):
  - Outcome y_{i,c,t} is one of:
    - Binary indicator equal to one if firm records positive tax payments (extensive margin; excludes firms operating at loss with negative EBIT).
    - Log of tax payments (intensive margin).
    - Tax payment normalized by operating revenues (turnovers) (intensive margin).
  - Key regressor: D_{c,s,t} = sector digital intensity for country c and sector s.
  - Controls X_{i,c,s,t}: firm characteristics (firm age, number of employees, total assets) and macro conditions (real GDP growth, inflation).
  - Fixed effects: country μ_c and year τ_t; country‑specific linear trend μ_c · τ_t included.
  - Standard errors clustered at country and year level.

### Firm‑level results (Table 3)
- Advanced Economies (columns 1–3):
  - Digital intensity coefficients:
    - Tax paid (binary): 0.408∗∗∗ (standard error 0.053).
    - Log(real tax payment): 2.702∗∗∗ (standard error 0.186).
    - Tax payment/Turnovers: 0.036∗∗∗ (standard error 0.004).
  - Observations: 13,282,566; 11,519,808; 11,519,808 respectively.
  - Adj. R2: 0.147; 0.598; 0.138 respectively.
  - Number of countries: 23.
- Emerging Market and Developing Economies (columns 4–6):
  - Digital intensity coefficients:
    - Tax paid (binary): 0.075∗∗∗ (standard error 0.019).
    - Log(real tax payment): 3.490∗∗∗ (standard error 0.214).
    - Tax payment/Turnovers: 0.081∗∗∗ (standard error 0.006).
  - Observations: 15,316,855; 15,836,281; 15,836,281 respectively.
  - Adj. R2: 0.0992; 0.625; 0.119 respectively.
  - Number of countries: 31.
- All regressions include firm controls, macro controls, country FE, year FE; country‑specific trends included.
- Significance: ∗ p< 0.1, ∗∗ p< 0.05, ∗∗∗ p< 0.01.

### Interpretation and comparison across levels
- Firm digitalization is associated with:
  - Higher probability of paying taxes (extensive margin), with larger marginal impact in EMDEs where non‑compliance is more prevalent.
  - Larger tax payments on the intensive margin in both AEs and EMDEs.
- Comparison of magnitudes:
  - Country‑level (Table 1, column 3): one‑standard‑deviation increase in firm digitalization associated with increase in tax revenues/GDP by 3 percentage points in countries with high GovTech (not significant in low GovTech).
  - Firm‑level (Table 3): one‑standard‑deviation increase in digital intensity associated with increase in tax payments/operating revenue ratio of 0.12 percentage points in AEs and 0.32 percentage points in EMDEs, equal to 6 percent and 20 percent of firms’ average tax payment over turnover ratio respectively.
  - Both approaches yield estimates of the same order of magnitude; firm‑level estimates are smaller, reflecting benefits of granular controls for country‑level confounders.
- Caveats:
  - At the intensive margin, effect may be underestimated because high‑tech multinationals may shift profits to avoid taxes.
  - Extensive margin estimates should be interpreted with caution due to Orbis reporting limitations (negative tax payments treated as nonpayment).

*Source: IMF Working Paper — Chapter 15: The regression analysis.*

### 27. We explore heterogeneity in the digitalization-tax relationship to shed some light on the

### 27. We explore heterogeneity in the digitalization-tax relationship to shed some light on the

### Heterogeneity analysis: model and variables
- Estimation extends baseline specification to include interaction term 훽3 between business ICT (D) and heterogeneity variable H (firm size, informality, GovTech):
  - Equation (3): 푦푖cci = 훼 + 훽1∙퐷cci + 훽2∙퐻cci + 훽3∙퐷cci∙퐻cci + 훾∙푋cci + 휇c + 휏i + 휇c∙휏i + 휀cci
- Heterogeneity dimensions:
  - Firm size: small firms defined as less than five employees (median size in sample).
  - Informality: business informality index from Elgin et al. (2021); high informality when index is above global average.
  - GovTech (tax administration digitalization): ICT capacity measured via ISORA data on annual ICT expenditure by each country’s revenue agency per corporate taxpayer.

### Synergy between GovTech and firm digitalization (interactive effects)
- Extensive margin (probability of firms paying taxes):
  - In EMDEs, as ICT expenditure per taxpayer transitions from the 25th to 75th percentile, a one-standard-deviation increase in digital intensity leads to about an 11 percent increase in the probability of firms’ tax payments.
  - Calculation cited: 2.79 (75th-to-25th range in ICT expenditure per taxpayers (in log) for EMDEs) × 0.04 (one standard-deviation of digital intensity) × 100 = 11.2 percent.
- Intensive margin (tax payments relative to operating revenues):
  - In AEs, as ICT expenditure per taxpayer transitions from the 25th to 75th percentile, a one-standard-deviation increase in digital intensity leads to about a 1.5 percent increase in tax payments (relative to operating revenues).
  - Calculation cited: 0.48 (75th-to-25th range in ICT expenditure per taxpayers (in log) for AEs) × 0.032 (one standard-deviation of digital intensity) × 100 = 1.5 percent.
- Interpretation:
  - Strong complementarity between firm digitalization and revenue agency ICT expenditure in EMDEs at the extensive margin.
  - Intensive-margin gains manifest clearly in AEs; impact appears insignificant in EMDEs, suggesting weaker institutional foundations (e.g., digital public infrastructure, digital ID).

### Compliance channel: heterogeneous effects by firm size and informality
- Effects summarized (Figure 7 findings):
  - Firm digitalization effect on tax payments is larger for small firms and in countries with high business informality.
  - Extensive margin: limited overall impact; marginally stronger impact for smaller firms in EMDEs.
  - Intensive margin: stronger efficiency gains from smaller firms in AEs; EMDEs show smaller but positive and significant gains from smaller firms.
  - In EMDEs with high informality, firm digitalization more significantly enhances tax payments by detecting liabilities and reducing compliance costs.

### Decomposition of mean tax payments differentials (Oaxaca-Blinder)
- Approach:
  - Use Earnings Before Interest and Tax (EBIT) and firm characteristics from Orbis to estimate CIT liability and calculate gap between actual and potential tax payments.
  - Classic linearly separable OB decomposition separates mean tax payment differential ∆T into “explained component” (observable characteristics) and “unexplained component” (slope differences, proxied as compliance effect).
- Key OB result:
  - In the service sector, unexplained components (including compliance cost channel) explain 40-70 percent of larger tax payments for digital firms relative to non-digital firms.
  - In manufacturing, the positive impact of firm digitalization on tax payments is mostly explained by observable differences between digital and less-digital firms.
  - The contribution of unexplained component is more pronounced in services where tax compliance is generally weaker than in manufacturing.

### Quantitative magnitudes and illustrative estimates
- Country-level illustrative estimate: a one-standard-deviation increase in firm digitalization is associated with an increase in tax revenues-to-GDP by up to 3 percentage points (based on country-level estimates).
- Firm-level illustrative estimates: a one-standard-deviation increase in firm digitalization is associated with an increase in the ratio of tax payments to corporate turnover by up to 0.12-0.32 percentage points.
- Relative scale:
  - Country-level: increase by about a fifth of average tax revenues-to-GDP ratio.
  - Firm-level: increase by 6-20 percent of firms’ average tax payment over turnover ratio.
- Policy-relevant projection:
  - If low-income countries could achieve the level of firm digitalization seen in advanced economies, the gap in the tax-to-GDP ratio between these groups would be nearly halved (contingent on sufficient GovTech).
- Cost benchmark:
  - Provision of universal 4G cellular broadband access estimated to cost at most $418 billion (about 0.45 percent of global GDP) according to Oughton et al. (2023).

### Conclusions and policy implications
- Main conclusions:
  - Positive relationship between firm digitalization and tax revenues is robust across country- and firm-level analyses, though causal certainty is limited due to lack of a highly credible instrumental variable.
  - Compliance channel is an important mechanism—effects are stronger for smaller firms and in high-informality environments.
  - Significant synergy exists between firm digitalization and GovTech: tax gains from firm digitalization are larger where revenue administrations invest in ICT, and vice versa.
- Policy recommendations:
  - Adopt a dual approach: promote both GovTech and firm digitalization to maximize tax collection benefits through complementarities.
  - Coordinate public and private digital investments to lower fiscal “net” cost and increase returns from digitalization, especially in fiscally constrained countries.
  - Supportive measures include universal broadband connectivity and incentives for firms to accelerate ICT investments.
  - Parallel reforms in institutional and legal frameworks, organizational strength, and staff competency at tax administrations are necessary alongside IT investments to ensure effective adoption and revenue gains.

*IMF WORKING PAPERS Leveraging Digital Technologies in Boosting Tax Collection*

### Annex II. Heterogeneous Impact of Firm Digitalization: Regression

### Annex II. Heterogeneous Impact of Firm Digitalization: Regression

### Panel A — Advance Economies: Firm Tax Payments and Digital Intensity
- Baseline Digital intensity coefficients (columns (1)–(8)):
  - (1) Tax paid: 0.144 (std. err. 0.126)
  - (2) Tax payment/Revenue: -0.006 (std. err. 0.023)
  - (3) Tax paid: 0.096*** (std. err. 0.027)
  - (4) Tax payment/Revenue: 0.044*** (std. err. 0.004)
  - (5) Tax paid: 0.082*** (std. err. 0.024)
  - (6) Tax payment/Revenue: 0.103*** (std. err. 0.011)
  - (7) Tax paid: 0.026 (std. err. 0.046)
  - (8) Tax payment/Revenue: 0.051*** (std. err. 0.003)

- Interaction terms (where included):
  - Digital x ICT investment/taxpayers:
    - Column (2) coefficient: -0.016 (std. err. 0.031)
    - Column (4) coefficient: 0.019*** (std. err. 0.006)
  - Digital x Small firm:
    - Column (3) coefficient: -0.033* (std. err. 0.020)
    - Column (5) coefficient: 0.064*** (std. err. 0.012)
  - Digital x High informality:
    - Column (4) coefficient: -0.180* (std. err. 0.104)
    - Column (6) coefficient: -0.041*** (std. err. 0.011)
  - Digital x Service sector:
    - Column (5) coefficient: 0.045 (std. err. 0.037)
    - Column (8) coefficient: 0.031*** (std. err. 0.007)

- Sample and model statistics by column:
  - Observations:
    - (1) 8,064,843
    - (2) 8,025,253
    - (3) 15,316,855
    - (4) 15,836,281
    - (5) 2,672,091
    - (6) 2,812,446
    - (7) 15,316,855
    - (8) 15,836,281
  - Adj. R2:
    - (1) 0.104
    - (2) 0.0922
    - (3) 0.0979
    - (4) 0.120
    - (5) 0.125
    - (6) 0.142
    - (7) 0.0980
    - (8) 0.119
  - Number of countries:
    - (1) 29
    - (2) 29
    - (3) 31
    - (4) 31
    - (5) 17
    - (6) 17
    - (7) 31
    - (8) 31

- Model specifications (all columns): Firm controls YES; Macro controls YES; Country FE YES; Year FE YES; Country-specific trend YES.
- Standard errors are clustered at country and year level.
- Significance notation: ∗ p< 0.1, ∗∗ p< 0.05, ∗∗∗ p< 0.01

### Panel B — Emerging Market and Developing Economies: Firm Tax Payments and Digital Intensity
- Baseline Digital intensity coefficients (columns (1)–(8)):
  - (1) Tax paid: 0.068 (std. err. 0.179)
  - (2) Tax payment/Revenue: 0.056*** (std. err. 0.018)
  - (3) Tax paid: 0.363*** (std. err. 0.058)
  - (4) Tax payment/Revenue: 0.027*** (std. err. 0.003)
  - (5) Tax paid: 0.392*** (std. err. 0.044)
  - (6) Tax payment/Revenue: 0.028*** (std. err. 0.003)
  - (7) Tax paid: 0.419*** (std. err. 0.105)
  - (8) Tax payment/Revenue: 0.019*** (std. err. 0.002)

- Interaction terms (where included):
  - Digital x ICT investment/taxpayers:
    - Column (2) coefficient: 0.107*** (std. err. 0.037)
    - Column (4) coefficient: -0.003 (std. err. 0.004)
  - Digital x Small firm:
    - Column (3) coefficient: 0.101* (std. err. 0.052)
    - Column (5) coefficient: 0.020** (std. err. 0.009)
  - Digital x High informality:
    - Column (4) coefficient: 0.326 (std. err. 0.202)
    - Column (6) coefficient: 0.084*** (std. err. 0.020)
  - Digital x Service sector:
    - Column (5) coefficient: -0.029 (std. err. 0.136)
    - Column (8) coefficient: 0.021*** (std. err. 0.005)

- Sample and model statistics by column:
  - Observations:
    - (1) 6,164,644
    - (2) 4,394,066
    - (3) 13,282,576
    - (4) 11,519,818
    - (5) 12,528,443
    - (6) 10,748,683
    - (7) 13,282,576
    - (8) 11,519,818
  - Adj. R2:
    - (1) 0.0648
    - (2) 0.147
    - (3) 0.140
    - (4) 0.135
    - (5) 0.135
    - (6) 0.139
    - (7) 0.141
    - (8) 0.134
  - Number of countries:
    - (1) 16
    - (2) 16
    - (3) 25
    - (4) 25
    - (5) 24
    - (6) 24
    - (7) 25
    - (8) 25

- Model specifications (all columns): Firm controls YES; Macro controls YES; Country FE YES; Year FE YES; Country-specific trend YES.
- Standard errors are clustered at country and year level.
- Significance notation: ∗ p< 0.1, ∗∗ p< 0.05, ∗∗∗ p< 0.01

*Source: Annex II. Heterogeneous Impact of Firm Digitalization: Regression, Table AII.1 (Panels A and B).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025089-print-pdf.pdf_