## wpiea2023226-print-pdf ## Source details **Canonical URL:** [wpiea2023226-print-pdf](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023226-print-pdf.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023226-print-pdf.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023226-print-pdf.pdf.json) --- ### Introduction and scope - Rapid expansion in the use of Information and Communication Technologies (ICT) has changed market structure and transactions and reshaped public service provision. - Since the 1980s, digital technologies have been deployed in upgrading tax collection, social benefit payment, and public procurement. - GovTech agenda in developed countries includes “disruptive technologies” (e.g., artificial intelligence (AI), machine learning, cloud computing) and public data platforms that facilitate innovations for more efficient public service provision. - For developing countries, GovTech adoption offers opportunities to leapfrog the technology frontier and overcome pre-existing barriers in delivering goods and services, particularly to vulnerable groups typically living in rural areas (Gupta et al., 2017). ### Microeconomic evidence on GovTech effects - Digitalized cash transfers help better access and target social assistance to the poor in developing countries (Aiken et al., 2023; Aker et al., 2016; Aker and Blumenstock, 2015; Muralidharan, Niehaus, and Sukhtankar, 2016). - E-procurement is expected to improve the efficiency of public expenditure by reducing corruption and leakage of anti-poverty welfare programs (Banerjee et al., 2020) and by raising the quality of infrastructure investments (Lewis-Faupel et al., 2016). - E-government platforms facilitate citizens’ interaction with government officials, cut administrative tiers, and strengthen monitoring of public service quality (Grossman et al., 2018). ### Macro and inclusion channels - GovTech can strengthen Public Financial Management (PFM) by automating budget cycle, data reconciliation, and payment processes. - Automation of budget processes reduces bureaucratic discretion and can make expenditure policy more inclusive (Zouhar et al., 2021). - Enhanced fiscal transparency (for example, through e-procurement) could reduce bias toward small firms (especially women-led SMEs), potentially creating positive macroeconomic impacts (di Giovanni et al., 2022). ### Distributional concerns and ambiguity - Technological progress is inherently skill-biased and raises concern of a “digital divide” across skill levels, gender, and residential locations (Agrawal and Bütikofer, 2022). - Given the trade-off between efficiency and equity, the distributional impact of GovTech is a priori ambiguous and motivates empirical investigation. ### Data and empirical approach - Uses World Bank’s GovTech Maturity Index (GTMI) (World Bank, 2022) capturing timing and intensity of GovTech adoption across fiscal operations (treasury operation, tax administration, procurement, public investment). - GTMI aggregates GovTech maturity in four areas: - GT1: supporting core government systems; - GT2: improving public service delivery; - GT3: enhancing citizen engagement; - GT4: fostering GovTech enabling factors. - GTMI records establishment year of a dedicated GovTech institution and adoption years for initiatives including: digital budget payments via FMIS, Treasury Single Account (TSA) IT system, e-procurement portal, and PIMS. - As of 2022: number of countries with a dedicated GovTech institution increased from 11 in 2000 to 51 within 10 years and reached 155 countries by 2022. - About half of GovTech institutions are under the Ministry of ICT as of 2022 (World Bank, 2022). - Empirical methods: cross-country panel regression (two-way fixed effects), event-study with Sun and Abraham (2021)’s interaction-weighted estimator, and synthetic control estimation (generalized synthetic control, Xu (2017)). ### Key empirical findings — Budget transparency and governance - GovTech maturity is positively associated with higher budget transparency and better expenditure management (Open Budget Index, PEFA). - Using PEFA sub-indices: a 50 percent increase in digital adoption of core fiscal functions (GT1) is associated with improvement in: - fiscal risk reporting (exceeding 1 standard deviation (SD) from the mean), - public investment and procurement management (impact of about 0.8 SD), - expenditure/cash control. - GT2 shows relatively larger positive association with efficiency of service delivery and budget outturns than GT1/GT3. - GT3 exhibits strong correlation with fiscal risk reporting and procurement management. - Digitalization in core fiscal functions, public service delivery, and citizen engagement associates with higher Corruption Perception Index and Quality of Government (QoG) index. - Positive impact of e-procurement on budget transparency is particularly large for EMDEs (11 points in OBI for EMDEs, Annex Table AIII.1). ### Key empirical findings — Expenditure efficiency and social assistance coverage - Digital payments integrated into FMIS can expand social assistance program coverage to the extremely poor (living below $1.9 a day poverty line) by 7 percent. - Panel regression results in EMDEs: - Introduction of an e-payment system is positively associated with higher SSN coverage for both the extremely poor and the poorest population. - Digitalization of budget payment using the internet (integrated into FMIS) shows potential to expand SSN coverage for the extremely poor by 7 percent. - GT1 and GT2 exhibit clear and significant association with increased efficiency and broader coverage of social assistance programs; the BCR concerning the extremely poor shows weak correlation. ### Heterogeneity and enabling conditions - Positive impact of GovTech on budget implementation depends on: - progress in legislative reforms (e.g., cybersecurity), - robust system and organizational integrations, - coordinated GovTech strategy across ministries, - presence of a centralized GovTech institution. - Positive impact of digital payments on SSN coverage is conditional on GovTech enabling environment: - Positive impact emerges where a dedicated GovTech institution existed at the time of adopting digital treasury operations. - Positive interaction with GTMI enabler index underscores need for social assistance database, secure digital identity, and information security/privacy/data protection. - Digital connectivity is pivotal for reaching the poorest households. - Countries in the high group of institutional maturity experience expanded social assistance coverage compared with bottom quartile counterparts. - The GTMI Enabler Index shows ambiguous correlation with per-capita income; EDAI positively correlates with per-capita income. ### Impact on income inequality — Event-study, cohort, and channels - Synthetic control and event-study analyses indicate: - Centralized GovTech institutions and digital budget payments are associated with increases in pre-tax income shares for the bottom 50th percentile and female workers, primarily via reduction in business informality. - Positive impact of digital budget payments materializes gradually over ten years. - Cohort results: - National GovTech strategy initiation increases female pre-tax income share by 1.6 percentage points over 10 years for longtime adopters. - Launch of digital budget payment via TSA raises pre-tax income share of the bottom 50th percentile by approximately 0.5 percent. - Launch of digitalized PIMS does not show statistically significant inclusion effect. - Heterogeneity by enablers and informality: - Digital budget payments significantly raise income shares for bottom 50th percentile in countries with high digital connectivity; in EMDEs the treatment effect is larger by 0.6 percent when digital connectivity is high. - Female pre-tax income treatment effects of GovTech institutions and TSA are around 2 percent when digital connectivity and GovTech maturity are high. - In EMDEs, digital budget payments can have negative impacts in contexts of weak digital connectivity or low GovTech enabling maturity. - Digital budget payments tend to positively impact bottom-half pre-tax income share when informality is low (especially in AEs); impact on female pre-tax income is more pronounced in high informality contexts. ### Annex VI — Specification tests and dynamic ATE (highlights) - Comparison across models finds: - TWFE (or alternative DID) estimated coefficients tend to be unstable with large standard errors, partly due to violation of parallel trend. - Generalized synthetic control (SC) shows relatively stable performance with tighter confidence intervals and better identification of treatment effect. - Preferred SC specification includes only country fixed effect as it minimizes MSPE. - Pre-treatment fit: actual and counterfactual estimates are closely aligned in pre-treatment period; number of pre-treatment periods sufficiently large to minimize bias. - Main dynamic ATE findings (Figures 10 and 11): - Panel A (bottom 50th percentile): In EMDEs, significant positive effect materializes after 5 years from adoption, accumulating to reach 0.65 percentage points over 10 years. Across all countries, positive distributional effect emerges typically 7-8 years after adoption. - Panel B (female pre-tax income share): Impact increases gradually, reaching about 1.8 percentage points over 10 years; no statistically significant difference between overall sample and EMDEs. - Channels: - Average real income of the poor gradually increases to about 20 percent higher in real terms over 10 years, translating to annual real income gains of about 2 percent. - Informal outputs decline, reaching 3 percentage points lower as a percent of GDP within 10 years. - Reduction in business informality identified as primary channel behind equity benefits; heterogeneous reductions contingent on GovTech enabling maturity and state capacity. ### Annex VI — Exact dynamic ATE comparison table excerpts (Annex Table AVI.1 and AVI.2) - AVI.1. Average Treatment Effect on Pre-tax Income Share of Bottom 50th Percentile (rows are relative years -4 through 10; columns are model specifications): - -4: (1) 0.082 (2) 0.035 (3) 0.034 (4) 0.034 - -3: (1) 0.057 (2) -0.020 (3) -0.026 (4) -0.003 - -2: (1) 0.078 (2) -0.048 ** (3) -0.043 * (4) -0.041 - -1: (1) 0.000 (2) -0.009 (3) -0.020 (4) -0.008 - 0: (1) 0.057 (2) 0.006 (3) 0.003 (4) 0.007 - 1: (1) 0.112 (2) -0.025 (3) 0.002 (4) 0.014 - 2: (1) 0.122 (2) -0.060 (3) -0.027 (4) -0.028 - 3: (1) 0.092 (2) -0.077 * (3) -0.067 (4) -0.014 - 4: (1) 0.226 (2) -0.116 ** (3) -0.024 (4) 0.030 - 5: (1) 0.219 (2) -0.038 (3) 0.106 ** (4) 0.112 - 6: (1) 0.300 (2) -0.031 (3) 0.188 *** (4) 0.145 - 7: (1) 0.264 (2) 0.095 * (3) 0.373 *** (4) 0.277 * - 8: (1) 0.282 (2) 0.163 *** (3) 0.460 *** (4) 0.281 * - 9: (1) 0.314 (2) 0.209 *** (3) 0.535 *** (4) 0.371 ** - 10: (1) 0.610 * (2) 0.258 *** (3) 0.654 *** (4) 0.318 * - Sample: (1) ALL (2) ALL (3) EMDEs (4) ALL - MSPE: ... 0.000085 0.000082 0.000290 - Significance legend: * p < 0.10, ** p < 0.05, *** p < 0.010 - AVI.2. Average Treatment Effect on Pre-tax Female Income Share (rows -4 through 10; columns are model specifications): - -4: (1) -0.531 (2) -0.143 * (3) -0.147 * (4) -0.141 - -3: (1) -0.372 (2) -0.120 *** (3) -0.107 *** (4) -0.103 - -2: (1) -0.340 (2) 0.043 *** (3) 0.048 ** (4) 0.073 - -1: (1) 0.000 (2) 0.065 (3) 0.040 (4) 0.091 - 0: (1) -0.138 (2) 0.177 (3) 0.163 (4) 0.225 - 1: (1) 0.072 (2) 0.321 *** (3) 0.284 ** (4) 0.386 * - 2: (1) 0.128 (2) 0.582 *** (3) 0.500 *** (4) 0.607 ** - 3: (1) -0.090 (2) 0.624 *** (3) 0.578 *** (4) 0.509 * - 4: (1) -0.146 (2) 0.715 *** (3) 0.590 *** (4) 0.503 * - 5: (1) 0.090 (2) 0.956 *** (3) 0.819 *** (4) 0.726 ** - 6: (1) 0.161 (2) 1.131 *** (3) 0.960 *** (4) 0.645 * - 7: (1) 0.422 (2) 1.238 *** (3) 1.094 *** (4) 0.598 - 8: (1) 0.405 (2) 1.405 *** (3) 1.232 *** (4) 0.646 - 9: (1) 0.569 (2) 1.555 *** (3) 1.358 *** (4) 0.687 - 10: (1) 0.765 (2) 1.836 *** (3) 1.554 *** (4) 0.790 - Sample: (1) ALL (2) ALL (3) EMDEs (4) ALL - MSPE: ... 0.000077 0.000080 0.000726 - Significance legend: * p < 0.10, ** p < 0.05, *** p < 0.010 ### Annex III — Selected regression table diagnostics (EMDEs, two-way FE) - Selected summary entries preserved as reported: - Observations22553352252252252681341333333333337 - Number of countries829290909091819290909091 - Adj. R-squared0.48830.42640.43030.44140.46210.44230.45370.29570.31420.31670.31560.3027 - Region FEYesYesYesYesYesYesYesYesYesYesYesYes - Income FEYesYesYesYesYesYesYesYesYesYesYesYes - Year FEYesYesYesYesYesYesYesYesYesYesYesYes - Example coefficient strings preserved as reported (selected raw values): - 38.051*68.158*** (standard errors shown below as (20.269)(17.179)) - 56.349***75.002***77.717***74.642***87.437***76.904***32.45443.406***45.757***43.722***55.818***46.237*** - -3.901***-4.676***-4.854***-4.697***-5.439***-4.832***-2.214*-2.629***-2.773***-2.663***-3.337***-2.804*** - 14.803***15.834***15.377***15.342***14.261***15.728***2.8325.679***5.348***5.198***4.491***6.040*** - -156.778* -241.772***-249.885***-236.535*** -290.477*** -244.395***-114.559-150.651***-159.384*** -150.985*** -203.417***-160.468*** ### Policy implications and recommendations - Continue digitalization in fiscal operations, supported by technical assistance from development partners, to strengthen fiscal transparency and expenditure efficiency. - Address gaps in FMIS function, data exchange capabilities, interoperability between old and new IT systems, and legal and regulatory frameworks when encouraging digital adoption. - In low-income countries with limited digital network coverage and state capacity, mitigate risks of leaving vulnerable populations unattended. - Invest in ICT infrastructure, promote digital skill development, and coordinate inter-agency GovTech initiatives to establish adequate GovTech enabling institutions. - Centralized GovTech coordination agency plays a critical role (though not sufficient) in reaping both efficiency and equity gains. - Cautionary notes: - Analysis relies on timing of GovTech adoption at the extensive margin and cross-sectional intensity information from the 2022 GTMI dataset. - Possible measurement errors in GTMI due to lack of official records should be noted. - Future research should improve measurement of GovTech adoption at the intensive margin over time and combine cross-country analysis with detailed case studies. *Inclusive GovTech: Enhancing Equity and Efficiency Through Public Service Digitalization — Working Paper No. WP/2023/226* ### References ............................................................................................................. ### wpiea2023226-print-pdf - References ### Introduction and scope - Rapid expansion in the use of Information and Communication Technologies (ICT) has changed market structure and transactions and reshaped public service provision. - Since the 1980s, digital technologies have been deployed in upgrading tax collection, social benefit payment, and public procurement. - The GovTech agenda in developed countries includes “disruptive technologies” (e.g., artificial intelligence (AI), machine learning, cloud computing) and public data platforms that facilitate innovations for more efficient public service provision. - For developing countries, GovTech adoption offers opportunities to leapfrog the technology frontier and overcome pre-existing barriers in delivering goods and services, particularly to vulnerable groups typically living in rural areas (Gupta et al., 2017). ### Microeconomic evidence on GovTech effects - Digitalized cash transfers help better access and target social assistance to the poor in developing countries (Aiken et al., 2023; Aker et al., 2016; Aker and Blumenstock, 2015; Muralidharan, Niehaus, and Sukhtankar, 2016). - E-procurement is expected to improve the efficiency of public expenditure by reducing corruption and leakage of anti-poverty welfare programs (Banerjee et al., 2020) and by raising the quality of infrastructure investments (Lewis-Faupel et al., 2016). - E-government platforms facilitate citizens’ interaction with government officials, cut administrative tiers, and strengthen monitoring of public service quality (Grossman et al., 2018). ### Macro and inclusion channels - GovTech can strengthen Public Financial Management (PFM) by automating budget cycle, data reconciliation, and payment processes. - Automation of budget processes reduces bureaucratic discretion and can make expenditure policy more inclusive (Zouhar et al., 2021). - Enhanced fiscal transparency (for example, through e-procurement) could reduce bias toward small firms (especially women-led SMEs), potentially creating positive macroeconomic impacts (di Giovanni et al., 2022). ### Distributional concerns and ambiguous net effects - Technological progress is inherently skill-biased and raises concern of a “digital divide” across skill levels, gender, and residential locations (Agrawal and Bütikofer, 2022). - Given the trade-off between efficiency and equity, the distributional impact of GovTech is a priori ambiguous and motivates empirical investigation. ### Data and empirical approach - The analysis uses World Bank’s GovTech Maturity Index (GTMI) (World Bank, 2022), which provides cross-country variations in both intensity and timing of digital technology adoption across layers of fiscal policy operations (e.g., treasury operation, tax administration, procurement, and public investment). *Source: wpiea2023226-print-pdf - References* ### Annex I of Amaglobeli, de Mooij et al (2023) provides glossary of key concepts and a list of abbreviations and acronyms. ### wpiea2023226-print-pdf - Annex I of Amaglobeli, de Mooij et al (2023) provides glossary of key concepts and a list of abbreviations and acronyms ### Overview - Annex I is identified as the glossary of key concepts and a list of abbreviations and acronyms for Amaglobeli, de Mooij et al (2023). - The main paper analyzes GovTech adoption using the World Bank GTMI data and multiple empirical methods (cross-country panel regression, event-study with Sun and Abraham (2021)’s interaction-weighted estimator, and synthetic control estimation). ### Data and GTMI (GovTech Maturity Index) - GTMI covers GovTech application practices over 198 countries and records timing and intensity of GovTech adoption. - GTMI aggregates GovTech maturity in four key areas: (a) supporting core government systems (GT1), (b) improving public service delivery (GT2), (c) enhancing citizen engagement (GT3), and (d) fostering GovTech enabling factors (GT4) (Annex 1). - GTMI captures establishment year of a dedicated GovTech institution and adoption years for initiatives including: digital budget payments via FMIS, Treasury Single Account (TSA) IT system, e-procurement portal, and PIMS. - As of 2022: number of countries with a dedicated GovTech institution increased from 11 in 2000 to 51 within 10 years and reached 155 countries by 2022. - About half of GovTech institutions are under the Ministry of ICT as of 2022 (World Bank, 2022). ### Key empirical findings — Budget transparency and governance - GovTech maturity is positively associated with higher budget transparency and better expenditure management (Open Budget Index, PEFA). - Using PEFA sub-indices: a 50 percent increase in digital adoption of core fiscal functions (GT1) is associated with improvement in: - fiscal risk reporting (exceeding 1 standard deviation (SD) from the mean), - public investment and procurement management (impact of about 0.8 SD), - expenditure/cash control. - GT2 (government service quality) shows relatively larger positive association with efficiency of service delivery and budget outturns than GT1/GT3. - GT3 (citizen engagement) exhibits strong correlation with fiscal risk reporting and procurement management. - Digitalization in core fiscal functions, public service delivery, and citizen engagement associates with higher Corruption Perception Index and Quality of Government (QoG) index. ### Key empirical findings — Expenditure efficiency and social assistance coverage - Digital payments integrated into FMIS can expand social assistance program coverage to the extremely poor (living below $1.9 a day poverty line) by 7 percent. - Panel regression results in EMDEs: - Introduction of an e-payment system is positively associated with higher SSN coverage for both the extremely poor and the poorest population. - Digitalization of budget payment using the internet (integrated into FMIS) shows potential to expand SSN coverage for the extremely poor by 7 percent. - GT1 and GT2 exhibit clear and significant association with increased efficiency and broader coverage of social assistance programs; the BCR concerning the extremely poor shows weak correlation. - Positive impact of e-procurement on budget transparency is particularly large for EMDEs (11 points in OBI for EMDEs, Annex Table AIII.1). ### Heterogeneity and enabling conditions - Positive impact of GovTech on budget implementation depends on: - progress in legislative reforms (e.g., cybersecurity), - robust system and organizational integrations, - coordinated GovTech strategy across ministries, - presence of a centralized GovTech institution. - The positive impact of digital payments on SSN coverage is conditional on GovTech enabling environment: - Positive impact emerges where a dedicated GovTech institution existed at the time of adopting digital treasury operations. - Positive interaction with GTMI enabler index underscores need for social assistance database, secure digital identity, and information security/privacy/data protection. - Digital connectivity is pivotal for reaching the poorest households. - Countries in the high group of institutional maturity experience expanded social assistance coverage compared with bottom quartile counterparts. - The GTMI Enabler Index shows ambiguous correlation with per-capita income; EDAI (Enhanced Digital Access Index) positively correlates with per-capita income. ### Impact on income inequality — Event-study and cohort results - Synthetic control and event-study analyses indicate: - Centralized GovTech institutions and digital budget payments are associated with increases in pre-tax income shares for the bottom 50th percentile and female workers, primarily via reduction in business informality. - The positive impact of digital budget payments materializes gradually over ten years. - Cohort (recent adopter vs. longtime adopter) results: - National GovTech strategy initiation increases female pre-tax income share by 1.6 percentage points over 10 years for longtime adopters. - Launch of digital budget payment via TSA raises pre-tax income share of the bottom 50th percentile by approximately 0.5 percent. - Launch of digitalized PIMS does not show statistically significant inclusion effect. - Heterogeneity by enablers and informality: - Digital budget payments significantly raise income shares for bottom 50th percentile in countries with high digital connectivity (Panel A); in EMDEs the treatment effect is larger by 0.6 percent when digital connectivity is high. - Female pre-tax income treatment effects of GovTech institutions and TSA are around 2 percent when digital connectivity and GovTech maturity are high. - In EMDEs, digital budget payments can have negative impacts in contexts of weak digital connectivity or low GovTech enabling maturity. - Digital budget payments tend to positively impact bottom-half pre-tax income share when informality is low (especially in AEs); impact on female pre-tax income is more pronounced in high informality contexts. ### Identification and methods notes - Cross-country panel regressions use two-way fixed effects controlling for GDP per capita (linear and squared), rule of law, control of corruption, and regional dummies; Huber-White heteroskedasticity-robust standard errors used. - Event-study employs Sun and Abraham (2021)’s interaction-weighted estimator to unpack cohort-specific treatment effects. - Synthetic control estimation applies generalized synthetic control (Xu, 2017) to estimate dynamic effects allowing for unobservables and unbalanced panels. - Caveats noted: - GTMI relies on self-reported surveys; recorded intensity may not perfectly measure actual quality of PFM operations. - Measurement errors due to lack of official information should be noted for cross-country comparisons. *Source: Annex I, Amaglobeli, de Mooij et al (2023), IMF Working Paper "Inclusive GovTech: Enhancing Efficiency and Equity Through Public Service Digitalization."* ### Annex VI reports the specification test that compares the dynamic average treatment effect (ATE) estimates ### wpiea2023226-print-pdf - Annex VI reports the specification test that compares the dynamic average treatment effect (ATE) estimates ### Methodology and specification test - Compares dynamic average treatment effect (ATE) estimates between a traditional two-way fixed effect (TWFE) OLS model (and alternative DID estimator like Borusyak et al (2023)) and generalized synthetic control (SC). - With the author’s data: - TWFE (or alternative DID) estimated coefficients tend to be unstable with large standard errors, partly due to the violation of parallel trend. - Generalized SC shows relatively stable performance with tighter confidence intervals and provides better identification of the treatment effect. - Preferred SC specification: - Includes only country fixed effect because it minimizes the mean squared prediction error (MSPE). - SC model with both country and year fixed effects is unstable with larger MSPE given limited number of control group. - SCM (Synthetic Control Method) constructs counterfactual trajectory by taking a weighted average of countries without a GovTech such that it closely resembles the treated country; the SC algorithm chooses optimal country weight 푤푤푗∗ to match pre-treatment trajectory. - The specification test is summarized in Annex VI. ### Pre-treatment fit and bias considerations - The actual and counterfactual estimates are closely aligned in the pre-treatment period, illustrating validity of pretreatment fit of the generalized SC model (Figure 10). - In the analytical sample, the number of pre-treatment period is sufficiently large such that the bias of 휏휏횤횤,푖푖� is minimized (Abadie, Diamond, and Hainmueller, 2010). - The gap between actual treated average (bold line) and estimated counterfactual average (dash line) after the treatment year (time 0; dashed vertical line) represents the dynamic ATE of GovTech adoption. ### Main dynamic ATE findings (Figures 10 and 11) - Panel A: Impact on pre-tax income share, the bottom 50th percentile - In EMDEs, a significant positive effect materializes after 5 years from adoption. - Effect accumulates to reach 0.65 percentage points over 10 years. - Across all countries, positive distributional effect emerges with a delay, typically 7-8 years after adoption. - In Figure 10, the pre-tax income share of the bottom 50th percentile starts to grow faster after 7 years of digital payment adoption for treatment group relative to control group. - Panel B: Impact on pre-tax female income share - Adoption of digital budget payments leads to an increase in pre-tax female income share. - Impact gradually increases, reaching about 1.8 percentage points over 10 years. - No statistically significant difference in the treatment effect between the overall sample and EMDEs. - In Figure 10, pre-tax female income share shows clear bifurcation between treatment and control group immediately after technology adoption; the gap continues to expand over time. - Error bars in Figures indicate 95 percent confidence intervals computed using a parametric bootstrap procedure. - Charts include the additive component of country fixed effects which minimizes MSPE; number of unobserved factors automatically selected by cross-validation. ### Heterogeneous treatment effects and complementarities (Figure 12) - The impact on the poor and female workers is larger with robust digital connectivity and GovTech enabling institutions in place. - Sub-sample results: - Countries with a dedicated GovTech institution (green line), high GovTech enabling institutions (red line), and high digital connectivity (purple line) show ATT increasing at a faster rate than baseline over time. - Panel A underscores strong complementarity between a centralized GovTech agency and digital budget payments; consistent with strong positive interactive effect found in Table 3. - Threshold definitions for high groups: - Digital connectivity index above the 25 percentile value (excluding the last miles, below green line in Figure 2) → high digital connectivity group. - GTMI Enabler index above the 40 percentile value (above red line in Figure 2) → high GovTech Enabler. - Including countries on the margin (between the 25th to 40th percentile values of GovTech enabler index, between green and red lines in Figure 2) leaves main results robust. ### Channels behind inclusion effects (Figure 13 and Figure 14) - Event-study analysis: - Average real income of the poor gradually increases and reaches about 20 percent higher in real terms over 10 years (left chart of Figure 13). - This translates to annual real income gains of about 2 percent. - Informal outputs decline, reaching 3 percentage points lower as a percent of GDP within 10 years (right chart of Figure 13) using the Informal Economy database. - Reduction in business informality is identified as a primary channel behind equity benefits of GovTech initiatives. - Heterogeneous effect on informality (Figure 14): - Reduction in informality due to GovTech institutions and digital budget payments is contingent upon GovTech enabling maturity and state capacity. - The most significant reduction in informality occurs when mature GovTech enabling institutions and strong state capacities are present. - Example: India’s Digital Public Infrastructure (“India Stack”) — unique identity (Aadhaar), complementary payments systems (Unified Payments Interface, Aadhaar Payments Bridge, Aadhaar Enabled Payment Service), and data exchange (DigiLocker and Account Aggregator) — helped spur inclusion, market expansion, and fiscal efficiency (as discussed in the text). ### Policy implications and recommendations - Findings support advancing the GovTech agenda but highlight significant heterogeneity in treatment effects, underscoring importance of policy design. - Recommended policy priorities: - Continue digitalization in fiscal operations, supported by technical assistance from development partners, to strengthen fiscal transparency and expenditure efficiency. - Address gaps in FMIS function, data exchange capabilities, interoperability between old and new IT systems, and legal and regulatory frameworks when encouraging digital adoption. - In low-income countries with limited digital network coverage and state capacity, mitigate risks of leaving vulnerable populations unattended. - Invest in ICT infrastructure, promote digital skill development, and coordinate inter-agency GovTech initiatives to establish adequate GovTech enabling institutions. - Centralized GovTech coordination agency plays a critical role (though not sufficient) in reaping both efficiency and equity gains. - Cautionary notes: - Analysis relies on timing of GovTech adoption at the extensive margin and cross-sectional intensity information from the 2022 GTMI dataset. - Possible measurement errors in GTMI due to lack of official records should be noted. - Future research should improve measurement of GovTech adoption at the intensive margin over time and combine cross-country analysis with detailed case studies. *IMF WORKING PAPERS Inclusive GovTech: Enhancing Efficiency and Equity Through Public Service Digitalization* ### Chapter 3 of “Well Spent: How Strong Infrastructure Governance Can End Waste in Public Investment”, ### Chapter 3 — Annexes and Empirical Measures: GovTech, Budget Transparency, and Social Assistance ### A. GovTech Measures (World Bank GTMI) — Definitions and Key Components - GT1: Supporting core government systems (“Core Government systems”). the composite score (normalized) captures the modernization and integration of government systems through cloud services, interoperability framework, and other digital platforms. It covers digital technologies in tax administration, treasury and budget operations, public procurement (e.g., e-procurement), public investment management, and human resource management. - GT2: Enhancing public service delivery (“Public service delivery”). the composite score (normalized) measures the presence of government’s online service portals and the maturity of government public services in tax administration, e-payment, custom administration and so forth. It accounts for the simplicity, transparency, and universal accessibility of government online services. The index also accounts for services accessibility from demand-side, e.g., affordability of digital solutions such as mobile phones and free open-source applications to reach all intended beneficiaries and users. - GT3: Mainstreaming citizen engagement (“Citizen engagement”). the composite score (normalized) captures the progress in citizen’s on-line participation and feedback process in policy decision through the CivicTechtools (e.g., citizen feedback and complaint-handling mechanisms) on the Open Government/data portal. It also accounts for the degree of government’s efforts to achieve greater transparency using accountability tools as well as the government’s responsiveness to public options (e.g., response time, information disclosure policy). - GT4: Fostering GovTech enabling factors (“GovTech enablers"). the index measures a country’s progress in the establishment of legal and regulatory regimes, safeguarding institutions, and national strategy to promote the GovTech agenda. The index is constructed based on 16 key indicators related to the foundations of digitalization. The indicators are related to data governance, citizen’s right-to-information (RTI) law, data protection and privacy law, data-driven service orientation (e.g., digital identification system, digital signature), ITU’s cybersecurity index, citizen’s digital skill level and skill development strategy, public sector innovation strategy, and private sector involvement in GovTech. ### B. Summary Statistics for GTMI Aggregate Indices (Annex Table AI.1) - Panel A. Aggregate GTMI index (N=186 for each series) - GT1 (Core Government System): Mean 0.494; Std dev. 0.224; p25 0.315; p50 0.467; p75 0.668 - GT2 (Public Service delivery): Mean 0.614; Std dev. 0.244; p25 0.434; p50 0.646; p75 0.815 - GT3 (Citizen Engagement): Mean 0.469; Std dev. 0.301; p25 0.196; p50 0.403; p75 0.789 - GT4 (GovTech enablers): Mean 0.536; Std dev. 0.281; p25 0.273; p50 0.544; p75 0.793 - Source: World Bank. ### C. GTMI Sub-Index — Operational Elements and Implementation Notes - Dedicated GovTech institution - Digital government strategy should be typically supported by a dedicated agency to implement newly emerging initiatives and whole-of-government approach, while addressing country-specific challenges in the implementation in alignment with each country’s national priorities, capacity and budget constraint. - The availability of GovTech institutions is not a sufficient condition for the success in digitalization agenda (World Bank, 2021). T he implementation status is mixed and varies widely across countries depending on the country’s capacity and resource constraints, digital inclusion, data privacy, cybersecurity, and other factors. - Treasury Single Account (TSA) IT system - Core PFM systems (e.g., Financial Management Information System (FMIS), budget planning, cash management, commitment control, tax, customs, debt management, and HR management information system) are often not fully interconnected. The data exchange could be automated by adopting web services or application programming interfaces (APIs). - A launch of TSA IT system will improve the interconnectivity and interoperability of existing PFM systems. - The centralized TSA at the central bank will also improve the efficiency of budget financial and treasury operations. - Budget payments to beneficiaries will be more efficient when a TSA IT system is integrated with an FMIS in automating cash payments and back reconciliation. - E-procurement portal - E-procurement platform as directed by a potential buyer (a government agency or firm) would facilitate transactions between buyers and potential sellers of goods and services related to recurrent budgetary and/or capital budget transactions. - E-procurement has potential to make a radical change in improving the efficiency of public procurement, which will likely benefit buyers as well as suppliers and citizens. - Public Investment Management System (PIMS) - Once projects are registered at the Public Investment Management Information System (PIMIS), it streamlines, automates, and supports the management of the public investment throughout the entire project lifecycle - the submission, review, and approval of project proposals, as well as the execution and monitoring of the projects. - The centralization of PIMS is expected to improve efficiency and transparency in implementing infrastructure investments. ### D. Definitions and Data Sources (Selected variables summarized) - GTMI Sub-index indicators and binary coding (0/1) - The presence of a government body focused on GovTech – digital transformation, whole-of-government strategy, services, and so on. (0/1) — Source: World Bank GTMI - TSA IT system fully operational: The centralized TSA (treasury account at Central Bank) is interconnected with FMIS, data exchange is automated, and the IT system fully operational, capturing both revenue and expenditure transactions for more than 75 percent of total transactions. (0/1) — Source: World Bank GTMI - E-procurement portal: E-procumbent portal supports public procurement including the announcement of tenders and publication of contract information (0/1) — Source: World Bank GTMI - Public investment management system operational: The presence of public investment management system (PIMS), which streamlines, automates and supports the management of public investment lifecycle, involving the submission, review, and approval of project proposals as well as the execution and monitoring of the projects (0/1) — Source: World Bank GTMI - Digital accessibility data - Enhanced Digital Access Index (EDAI): The composite digital connectivity index (normalized), summarizing the level of country’s digital connectivity from both demand-side (affordability of access, education level of population, actual internet usage) and supply-side (availability of digital infrastructure, the quality of ICT services). — Source: Alper and Miktus (2019) - Percent of work-age population using internet — Source: International Telecommunication Union (ITU) - Budget transparency measures - Open Budget Index (OBI): The average score of questions on public’s access to timely and comprehensive budget information related to key budget documents (the value ranges from 0 to 100). — Source: International Budget Partnership (IBP)’s Open Budget Survey - Public Participation Index: The measure of public participation in budget decision-making and monitoring — Source: International Budget Partnership (IBP)’s Open Budget Survey ### E. Heterogeneous Impacts — Figures and Notes - Annex II (Figure AII.1) assesses the heterogeneous impact of GovTech Services (GT2) on Budget Transparency between High vs. Low groups: - Note: High vs. Low defined by 40th percentile threshold value of GovTech Enabler Index and Enhanced Digital Connectivity Index. Result using 25th percentile threshold value is similar with smaller impact heterogeneity between two groups. - Annex IV (Figure AIV.1) assesses the heterogeneous impact of GovTech Services (GT2) on Social Assistance between High vs. Low groups: - Note: High vs. Low defined by 25th percentile threshold value of GovTech Enabler Index and Enhanced Digital Connectivity Index. ### F. Regression Evidence — Annex III (Two-way Fixed Effects: EMDEs) - Annex Table AIII.1 reports multiple model specifications (columns 1–12) estimating effects on budget transparency in EMDEs using two-way fixed effects. - Selected reported rows and numbers (as presented) - Observations22553352252252252681341333333333337 - Number of countries829290909091819290909091 - Adj. R-squared0.48830.42640.43030.44140.46210.44230.45370.29570.31420.31670.31560.3027 - Region FEYesYesYesYesYesYesYesYesYesYesYesYes - Income FEYesYesYesYesYesYesYesYesYesYesYesYes - Year FEYesYesYesYesYesYesYesYesYesYesYesYes - Coefficients and significance markers are reported across specifications for variables including: - E-Government Index; PIMS fully adopted; Log (Per Capita GDP_PPP), lagged; Square log (Per Capita GDP_PPP), lagged; CPI inflation, lagged; External debt (%GDP), lagged; Log (Agriculture, %GDP), lagged; e-Procurement adopted with network interface; OBI (IBP) Public Participation (IBP); Establishment of GovTech institution; TSA/FMIS adopted with e-payment; TSA/FMIS adopted with mobile payment. - Example coefficient entries as reported (selected raw values from table): - 38.051*68.158*** (standard errors shown below as (20.269)(17.179)) - 56.349***75.002***77.717***74.642***87.437***76.904***32.45443.406***45.757***43.722***55.818***46.237*** (standard errors shown in the next line) - -3.901***-4.676***-4.854***-4.697***-5.439***-4.832***-2.214*-2.629***-2.773***-2.663***-3.337***-2.804*** (standard errors shown in the next line) - 14.803***15.834***15.377***15.342***14.261***15.728***2.8325.679***5.348***5.198***4.491***6.040*** (standard errors shown in the next line) - -156.778* -241.772***-249.885***-236.535*** -290.477*** -244.395***-114.559-150.651***-159.384*** -150.985*** -203.417***-160.468*** (standard errors shown in the next line) - Adjunct: The table reports model-level diagnostic and fixed-effects inclusion consistently across columns; the raw numeric reporting in the table is preserved as presented. *Italic attribution line: Chapter 3 of “Well Spent: How Strong Infrastructure Governance Can End Waste in Public Investment”, International Monetary Fund (contents excerpted as provided).* ### Annex V. GovTech Adoption Status in Other ### Annex V. GovTech Adoption Status in Other Regions ### GovTech Adoption Timing by Region: European, African, Western Hemisphere, Middle East & Central Asia - Source: GovTech Maturity Index, 2020 and 2022. - Note from figures: "The dark blues indicate the period after the digitalization. The countries that are yet to digitalize respective government function are categorized as a control group." - Four GovTech functions displayed for each region (year of adoption): - Establishment of a Dedicated GovTech institution - Launch of TSA IT System (linked with the FMIS) - Digitalized e-Procurement Portal - Digitalized PIMS - Regions covered with year-of-adoption visualizations: - European Region (Annex Figure AV.1) - African Region (Annex Figure AV.2) - Western Hemisphere Region (Annex Figure AV.3) - Middle East & Central Asia Region (Annex Figure AV.4) ### Annex VI. Specification Test ### Annex Table AVI.1. Average Treatment Effect of GovTech Adoption on the Pre-tax Income Share of the Bottom 50th Percentile: Comparison across Alternative Specifications - Model specifications: (1) OLS (TWFE), (2) Synthetic control (country FE), (3) Synthetic control (year FE), (4) (implicit fourth specification shown in table) - Reported estimates by relative year (rows labeled -4 through 10). Values reproduced exactly as in source: - -4: (1) 0.082 (2) 0.035 (3) 0.034 (4) 0.034 - -3: (1) 0.057 (2) -0.020 (3) -0.026 (4) -0.003 - -2: (1) 0.078 (2) -0.048 ** (3) -0.043 * (4) -0.041 - -1: (1) 0.000 (2) -0.009 (3) -0.020 (4) -0.008 - 0: (1) 0.057 (2) 0.006 (3) 0.003 (4) 0.007 - 1: (1) 0.112 (2) -0.025 (3) 0.002 (4) 0.014 - 2: (1) 0.122 (2) -0.060 (3) -0.027 (4) -0.028 - 3: (1) 0.092 (2) -0.077 * (3) -0.067 (4) -0.014 - 4: (1) 0.226 (2) -0.116 ** (3) -0.024 (4) 0.030 - 5: (1) 0.219 (2) -0.038 (3) 0.106 ** (4) 0.112 - 6: (1) 0.300 (2) -0.031 (3) 0.188 *** (4) 0.145 - 7: (1) 0.264 (2) 0.095 * (3) 0.373 *** (4) 0.277 * - 8: (1) 0.282 (2) 0.163 *** (3) 0.460 *** (4) 0.281 * - 9: (1) 0.314 (2) 0.209 *** (3) 0.535 *** (4) 0.371 ** - 10: (1) 0.610 * (2) 0.258 *** (3) 0.654 *** (4) 0.318 * - Sample: (1) ALL (2) ALL (3) EMDEs (4) ALL - MSPE: ... 0.000085 0.000082 0.000290 - Significance legend: * p < 0.10, ** p < 0.05, *** p < 0.010 - Note: "The blue shaded cells indicate the estimate under the baseline synthetic control model with country fixed effect (reported in Figure 11, Panel A)." ### Annex Table AVI.2. Average Treatment Effect of GovTech Adoption on the Pre-tax Female Income Share: Comparison across Alternative Specifications - Model specifications: (1) OLS (TWFE), (2) Synthetic control (country FE), (3) Synthetic control (year FE), (4) (implicit fourth specification shown in table) - Reported estimates by relative year (rows labeled -4 through 10). Values reproduced exactly as in source: - -4: (1) -0.531 (2) -0.143 * (3) -0.147 * (4) -0.141 - -3: (1) -0.372 (2) -0.120 *** (3) -0.107 *** (4) -0.103 - -2: (1) -0.340 (2) 0.043 *** (3) 0.048 ** (4) 0.073 - -1: (1) 0.000 (2) 0.065 (3) 0.040 (4) 0.091 - 0: (1) -0.138 (2) 0.177 (3) 0.163 (4) 0.225 - 1: (1) 0.072 (2) 0.321 *** (3) 0.284 ** (4) 0.386 * - 2: (1) 0.128 (2) 0.582 *** (3) 0.500 *** (4) 0.607 ** - 3: (1) -0.090 (2) 0.624 *** (3) 0.578 *** (4) 0.509 * - 4: (1) -0.146 (2) 0.715 *** (3) 0.590 *** (4) 0.503 * - 5: (1) 0.090 (2) 0.956 *** (3) 0.819 *** (4) 0.726 ** - 6: (1) 0.161 (2) 1.131 *** (3) 0.960 *** (4) 0.645 * - 7: (1) 0.422 (2) 1.238 *** (3) 1.094 *** (4) 0.598 - 8: (1) 0.405 (2) 1.405 *** (3) 1.232 *** (4) 0.646 - 9: (1) 0.569 (2) 1.555 *** (3) 1.358 *** (4) 0.687 - 10: (1) 0.765 (2) 1.836 *** (3) 1.554 *** (4) 0.790 - Sample: (1) ALL (2) ALL (3) EMDEs (4) ALL - MSPE: ... 0.000077 0.000080 0.000726 - Significance legend: * p < 0.10, ** p < 0.05, *** p < 0.010 - Note: "The blue shaded cells indicate the estimate under the baseline synthetic control model with country fixed effect (reported in Figure 11, Panel B)." *Inclusive GovTech: Enhancing Equity and Efficiency Through Public Service Digitalization — Working Paper No. WP/2023/226* --- _Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023226-print-pdf.pdf_