## wpiea2024234-print-pdf

## Source details

**Canonical URL:** [wpiea2024234-print-pdf](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024234-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024234-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024234-print-pdf.pdf.json)

---

### Research question and motivation
- Examines the relationship between citizens’ perceptions of trust in the tax authority and governments’ efficiency in collecting value-added tax (VAT) and corporate income tax (CIT) in Africa.
- Focus region chosen because it has the lowest average VAT efficiency and CIT Productivity according to IMF Fiscal Affairs Department (FAD) data for the period 2000-2021.
- Hypothesis: higher perceived trust in tax authorities increases voluntary tax compliance and thus tax efficiency; effect expected to be stronger in fragile states.

### Data and sample
- Panel dataset for the period 2014-2019.
- VAT analysis: panel of 32 countries.
- CIT analysis: subset of 24 countries.
- Perceptions of trust sourced from Afrobarometer survey (rounds 6, 7, and 8); typical country sample size in each round: 1,200-2,400 adult citizens.
- VAT efficiency and CIT productivity sourced from FAD at IMF.
- Additional controls sourced from World Bank, Medina and Schneider (2018), and UN e-government survey.
- Because the “trust in tax department” question did not appear in round 7, observations associated with years 2016-2018 were dropped.

### Key definitions and measures
- VAT Efficiency: ratio of actual VAT revenues to the product of the standard VAT rate and final consumption.
- CIT Productivity: ratio of CIT revenues (as percent of GDP) to the CIT rate.
- Lack of trust in tax department: share of respondents who answered “not at all” or “just a little” to the Afrobarometer question about trust in the tax department.
- Online Service Index (OSI): captures scope and quality of public sector online services; values between 0 and 1.
- Human Capital Index (HCI): proxy for digital literacy; values between 0 and 1.
- Government Effectiveness Index (GovEff): ranges from -2.5 to 2.5 (transformed in regressions by multiplying by 20 and adding 51 so the least possible value is 1).

### Stylized facts motivating the study
- Taxes on goods and services accounted for 52 percent of tax revenues in 2021 in African countries (twenty percent higher than the share for OECD countries).
- Corporate taxes accounted for 19 percent of tax revenues in Africa in 2021, compared to 9 percent for OECD countries.
- Informality data from Medina and Schneider (2018) ends in 2017; a two-year lag is used given high persistence of informality (correlation between 2012 and 2014 is 0.990 and between 2013 and 2015 is 0.983).

### Econometric specification (log–log)
- Estimated model for VAT efficiency:
  - VATEff_it = β0 + β1 lack_trust_it + β2 GDP_it + β3 Agr_it + β4 informality_it + β5 trade_it + β6 OSI_it + β7 HCI_it + β8 GovEff_it + ε_it
- Dependent variable: VATEff_it = log of VAT efficiency.
- Same regressors used for CIT Productivity regressions.
- Variable transformations:
  - lack_trust = log of share with little or no trust.
  - GDP = log of GDP per capita.
  - Agr = log of share of agriculture in GDP.
  - informality = log of share of informal sector in GDP.
  - trade = log of trade as share of GDP.
  - OSI = log of online service index.
  - HCI = log of human capital index.
  - GovEff = log of transformed government effectiveness index (multiplied by 20 and added 51).

### A priori expectations
- lack_trust: negative.
- GDP: positive.
- Agr: negative.
- informality: negative.
- trade: positive.
- OSI, HCI, GovEff: positive.

### Sample and summary statistics (selected)
- Countries covered (VAT sample includes): Benin, Botswana, Cape Verde, Cote d’Ivoire, Egypt, Eswatini, Gabon, Ghana, Guinea, Kenya, Lesotho, Madagascar, Malawi, Mauritius, Morocco, Namibia, Senegal, Sierra Leone, South Africa, Tanzania, The Gambia, Togo, Tunisia, Uganda, Zambia, Burkina Faso, Burundi, Cameroon, Ethiopia, Mali, Mozambique, Nigeria (last seven are fragile per World Bank classification).
- Observations for summary statistics:
  - VAT Efficiency and Lack of Trust: 53 (full), 11 (fragile), 42 (non-fragile).
  - CIT Productivity: 38 (full), 8 (fragile), 30 (non-fragile).
- Selected means (original scale, not log-transformed):
  - VAT Efficiency (Full sample mean): 0.366
  - VAT Efficiency (Fragile states mean): 0.337
  - VAT Efficiency (Non-fragile states mean): 0.374
  - CIT Productivity (Full sample mean): 0.723
  - CIT Productivity (Fragile states mean): 0.882
  - CIT Productivity (Non-fragile states mean): 0.680
  - Lack of Trust in tax department (Full sample mean): 0.517
  - Lack of Trust (Fragile states mean): 0.547
  - Lack of Trust (Non-fragile states mean): 0.509
  - GDP per capita (mean): 5,953
  - Share of agriculture in GDP (mean): 17.29
  - Share of informal sector in GDP (mean): 33.07
  - Share of trade in GDP (mean): 68.14
  - Online Service Index (mean): 0.36
  - Human Capital Index (mean): 0.48
  - Government Effectiveness Index (mean): -0.49
- Notable descriptive points:
  - Mean share of citizens with little or no trust ≈ 0.52 (52 percent); higher in fragile (0.55) vs non-fragile (0.51). Maximum = 0.72 (Nigeria).
  - Mean trade/GDP = 68 percent; agriculture and informal sectors large at 17.29 percent and 33.07 percent respectively.
  - OSI and HCI means below 0.5 but with high maximum values.

### Main empirical findings — VAT efficiency
- Baseline elasticities and significance:
  - A one percent increase in the share of citizens perceiving little or no trust is associated with a 0.22 percent decrease in VAT efficiency, ceteris paribus.
  - Table 3 (selected): Sample Obs. = 50; R2 = 0.627 (Column I), R2 = 0.665 (Column II).
    - Column I: lack_trust coefficient = -0.222* (standard error 0.120).
    - Column II (interaction with fragile): lack_trust coefficient = -0.237*** (standard error 0.076).
- Controls (signs and selected coefficients):
  - Agr: negative (e.g., -0.395 (0.242) in Column II context).
  - informality: negative but not always statistically significant in baseline.
  - trade: positive and statistically significant.
  - OSI and HCI: positive effects.
  - GDP: unexpected negative coefficient in baseline (possible multicollinearity).

### Corruption and VAT efficiency
- tax_corrupt = log of share reporting most or all tax officials are involved in corruption (mean ≈ 0.39; fragile mean 0.47, non-fragile mean 0.36).
- Table 4 (selected): Obs. = 50; R2 reported 0.627, 0.671, 0.630, 0.674 across columns I–IV.
  - Column I: tax_corrupt = -0.199 (standard error 0.118), p-value = 0.101 (not conventionally significant).
  - Separation by fragile vs non-fragile shows stronger negative effects in fragile states; interaction corrupt_fragile indicates larger negative magnitude in fragile contexts.
- Correlation between lack_trust and tax_corrupt = 0.62 (high correlation; identification difficulty noted).

### CIT Productivity results (selected)
- Dependent variable: log of (CIT revenues as percent of GDP divided by CIT rate).
- Table 5 (lack_trust on CIT Prod): Obs. = 36; R2 = 0.303 (Column I), R2 = 0.484 (Column II).
  - Column I: lack_trust coefficient = 0.153 (0.313) — not significant.
  - Column II (interaction with fragile): trust_fragile shows a significant negative effect (evidence that lack of trust matters primarily in fragile states).
- Table 6 (tax_corrupt on CIT Prod): Obs. = 36; R2 values 0.330, 0.464, 0.350, 0.552.
  - Corruption effects significant only in fragile states in Columns I–II.
  - When both lack_trust and tax_corrupt are included, lack_trust tends to dominate for CIT productivity in fragile states.

### Interpretation and mechanisms
- Lack of trust and perceived corruption impair tax collection:
  - VAT efficiency: lack_trust has a negative and statistically significant effect (baseline elasticity ~ -0.22).
  - Effects stronger in fragile states (interaction results).
- Proposed mechanisms:
  - In informal economies VAT payments are not automatic; lack of trust can facilitate buyer–seller collusion or businesses not remitting VAT receipts.
  - Digitalization (OSI) and human capital (HCI) support improved VAT efficiency by simplifying procedures and enhancing compliance.
  - Large agricultural share associated with lower VAT efficiency due to lower taxes/subsidies in agriculture.
  - For CIT productivity, prevalent bribery in fragile states may explain stronger negative effects of lack_trust and perceived corruption on corporate tax collection.

### Robustness analysis (summary)
- Simplified macro specification (dropping Agr and HCI) due to high correlation with GDP per capita:
  - Informality becomes significantly negative when Agr dropped.
  - OSI and HCI lose power when omitted.
  - Negative effect of lack of trust on tax efficiency strengthens; distinction between fragile and non-fragile states disappears.
  - Adding corruption to simplified specification renders survey variables generally insignificant, except when effects separated by fragility (high corruption in fragile states significantly impairs VAT efficiency).
- For CIT:
  - Lack of trust and corruption significantly negative in fragile states under robustness checks; lack_trust dominates when both survey variables included.

### Selected regression coefficients reported in Conclusion (robustness and alternative specs)
- Effect of Lack of Trust on VAT Efficiency (Dependent Variable: VATEff)
  - Specification I: lack_trust -0.337*** (0.123); Obs 50; R squared 0.491
  - Specification II: lack_trust -0.325** (0.121); trust_fragile -0.287 (0.236); Obs 50; R squared 0.509
  - Informality: -0.502* (0.298) in I; -0.433 (0.307) in II
  - trade: 0.577*** (0.172) in I; 0.587*** (0.170) in II
- Effect of Corruption on VAT Efficiency (Dependent Variable: VATEff)
  - Specification I: tax_corrupt -0.338** (0.143); Obs 50; R squared 0.501
  - Specification II: tax_corrupt -0.337** (0.133); Obs 50; R squared 0.526
  - Specification III: tax_corrupt -0.255 (0.212); lack_trust -0.145 (0.200); Obs 50; R squared 0.504
  - Specification IV: tax_corrupt -0.217 (0.195); trust_fragile 0.557 (0.475); Obs 50; R squared 0.533
- Effect of Trust on CIT Productivity (Dependent Variable: CITProd)
  - Specification I: lack_trust 0.018 (0.321); Obs 36; R squared 0.254
  - Specification II: lack_trust -0.066 (0.291); trust_fragile -1.130*** (0.276); Obs 36; R squared 0.444
  - trade: 0.531* (0.277) in I; 0.493** (0.210) in II
- Effect of Corruption on CIT Productivity (Dependent Variable: CITProd)
  - Specification I: tax_corrupt 0.256 (0.255); Obs 36; R squared 0.274
  - Specification II: tax_corrupt 0.216 (0.254); corrupt_fragile -0.722*** (0.256); Obs 36; R squared 0.411
  - Specification III: tax_corrupt 0.782 (0.760); lack_trust -0.772 (0.804); Obs 36; R squared 0.312
  - Specification IV: tax_corrupt 0.648 (0.779); trust_fragile -2.782* (1.448); Obs 36; R squared 0.523

### Policy implications and recommendations
- Strengthening citizens’ perceptions of trust in tax authorities is important for revenue mobilization and broader trust in government.
- Fiscal capacity building should prioritize ensuring citizens believe resources will be properly used, especially in fragile states where:
  - Distrust in public institutions is higher on average.
  - Negative effects of lack of trust on VAT and CIT efficiency are generally stronger.
- Recommended actions:
  - Enhance fiscal transparency, including publishing details about the use of government expenditures.
  - Provide technical assistance to revenue administrations to improve effectiveness.
  - Support digitalized tax submissions to help limit and better track non-payment.

### Limitations and interpretation cautions
- Small sample sizes and data constraints:
  - VAT analyses use 50 observations in many regressions; CIT productivity regressions use 36 observations.
  - Perceptions data missing for Afrobarometer round 7 reduced sample.
- Data limitations prevented separating VAT policy gap versus compliance gap.
- High multicollinearity among macro controls (e.g., GDP, Agr, HCI) affects some coefficient signs (GDP negative in some specs).
- Causal inference limited by linear regression framework; results demonstrate importance of trust but are not definitive causal proof.
- Behavioral inference: evidence consistent with the view that visible, effective use of tax revenues increases taxpayers’ willingness to pay.

*Source: IMF working paper (Introduction, Econometric Specification, Results, and Conclusion sections, wpiea2024234-print-pdf).*

### 1. Introduction ........................................................................................................

### 1. Introduction

### Research question and motivation
- Examines the relationship between citizens’ perceptions of trust in the tax authority and governments’ efficiency in collecting value-added tax (VAT) and corporate income tax (CIT) in Africa.
- Focus region chosen because it has the lowest average VAT efficiency and CIT Productivity according to IMF Fiscal Affairs Department (FAD) data for the period 2000-2021.
- Hypothesis: higher perceived trust in tax authorities increases voluntary tax compliance and thus tax efficiency; effect is expected to be stronger in fragile states.

### Data and sample
- Panel dataset for the period 2014-2019.
- VAT analysis: panel of 32 countries.
- CIT analysis: subset of 24 countries.
- Perceptions of trust sourced from Afrobarometer survey (rounds 6, 7, and 8); typical country sample size in each round: 1,200-2,400 adult citizens.
- VAT efficiency and CIT productivity sourced from FAD at IMF.
- Additional controls sourced from World Bank, Medina and Schneider (2018), and UN e-government survey.
- Because the “trust in tax department” question did not appear in round 7, observations associated with years 2016-2018 were dropped.

### Key definitions and measures
- VAT Efficiency: ratio of actual VAT revenues to the product of the standard VAT rate and final consumption.
- CIT Productivity: ratio of CIT revenues (as percent of GDP) to the CIT rate.
- Lack of trust in tax department: share of respondents who answered “not at all” or “just a little” to the Afrobarometer question about trust in the tax department.
- Online Service Index: captures scope and quality of public sector online services; values between 0 and 1.
- Human Capital Index: proxy for digital literacy; values between 0 and 1.
- Government Effectiveness Index: ranges from -2.5 to 2.5.

### Stylized facts motivating the study
- Taxes on goods and services accounted for 52 percent of tax revenues in 2021 in African countries (twenty percent higher than the share for OECD countries).
- Corporate taxes accounted for 19 percent of tax revenues in Africa in 2021, compared to 9 percent for OECD countries.
- Informality data from Medina and Schneider (2018) ends in 2017; a two-year lag is used given high persistence of informality (correlation between 2012 and 2014 is 0.990 and between 2013 and 2015 is 0.983).

### Main empirical findings (overview)
- Negative and significant association between lack of trust in the tax department and VAT C-efficiency:
  - A 1 percent increase in the share of citizens’ perception of little or no trust in the tax department is associated with a 0.22 percent decrease in VAT efficiency, ceteris paribus.
  - The negative effect is significantly greater in fragile than non-fragile states.
- For CIT efficiency (productivity), a significant negative association is found for fragile state countries:
  - A 1 percent increase in the share of citizens’ perception of little or no trust in the tax department is associated with a 1 percent decrease in CIT efficiency for fragile state countries.

### Literature context
- Prior literature emphasizes technical tax reforms (reforming tax laws, simplifying tax structures, widening the tax base, increasing rates) and administrative capacity as drivers of revenue mobilization.
- Related behavioral literature finds:
  - Trust in public institutions is positively associated with tax morale and willingness to pay taxes.
  - Perceived corruption is negatively associated with tax compliance and tax revenue outcomes.
  - Digitalization of government services is associated with improved VAT efficiency and CIT productivity in other settings.
- This study differs by directly linking perceptions of trust in tax authorities (Afrobarometer) to VAT efficiency and CIT productivity measures, thereby focusing on behavioral channels and avoiding multicollinearity concerns from using tax ratios and GDP per capita simultaneously.

### Contribution and scope
- Provides evidence that perceptions of trust in the tax authority are a significant determinant of VAT efficiency and CIT productivity in African countries, after controlling for macroeconomic indicators used in the literature (GDP per capita, share of agriculture in GDP, informality, trade openness, government effectiveness, online service index, human capital).
- Tests heterogeneity between fragile and non-fragile states.

*Source: IMF working paper (Introduction and Data sections, wpiea2024234-print-pdf).*

### 3. Econometric Specification

### 3. Econometric Specification

### Model specification
- Estimated linear regression (log–log specification) for VAT efficiency:
  - VATEff_it = β0 + β1 lack_trust_it + β2 GDP_it + β3 Agr_it + β4 informality_it + β5 trade_it + β6 OSI_it + β7 HCI_it + β8 GovEff_it + ε_it
- Dependent variable:
  - VATEff_it = log of VAT efficiency for country i at time t
- Same regressors used for CIT efficiency (CIT Productivity) regressions.

### Variable definitions and transformations
- lack_trust = log of the share of citizens with little or no trust in the tax department.
- GDP = log of GDP per capita.
- Agr = log of the share of agriculture in GDP.
- informality = log of the share of the informal sector in GDP.
- trade = log of trade as a share of GDP (trade openness).
- OSI = log of the online service index.
- HCI = log of the human capital index.
- GovEff = log of government effectiveness after transforming the index by multiplying by 20 and adding 51 so the least possible value is 1 (index originally between -2.5 and 2.5).
- Error term ε_it assumed i.i.d.
- Log specification chosen to interpret coefficients as elasticities.

### A priori expectations
- lack_trust: expected negative.
- GDP: expected positive (higher development → larger formal sector and VAT base).
- Agr: expected negative (agriculture tends to benefit from lower taxes).
- informality: expected negative (larger informal sector reduces actual revenues relative to potential).
- trade: expected positive (trade openness improves VAT on imports and formality/competitiveness).
- OSI: expected positive (better online public services improve compliance and administration).
- HCI and GovEff: expected positive.

### Sample and summary statistics (selected)
- Countries covered: Benin, Botswana, Cape Verde, Cote d’Ivoire, Egypt, Eswatini, Gabon, Ghana, Guinea, Kenya, Lesotho, Madagascar, Malawi, Mauritius, Morocco, Namibia, Senegal, Sierra Leone, South Africa, Tanzania, The Gambia, Togo, Tunisia, Uganda, Zambia, Burkina Faso, Burundi, Cameroon, Ethiopia, Mali, Mozambique, Nigeria (last seven are fragile per World Bank classification).
- Observations reported for summary statistics: 53 (full sample), 11 (fragile states), 42 (non-fragile states) for VAT Efficiency and Lack of Trust; 38 (full), 8 (fragile), 30 (non-fragile) for CIT Productivity.
- Selected means (original scale, not log-transformed):
  - VAT Efficiency (Full sample mean): 0.366
  - VAT Efficiency (Fragile states mean): 0.337
  - VAT Efficiency (Non-fragile states mean): 0.374
  - CIT Productivity (Full sample mean): 0.723
  - CIT Productivity (Fragile states mean): 0.882
  - CIT Productivity (Non-fragile states mean): 0.680
  - Lack of Trust in tax department (Full sample mean): 0.517
  - Lack of Trust (Fragile states mean): 0.547
  - Lack of Trust (Non-fragile states mean): 0.509
  - GDP per capita (mean): 5,953
  - Share of agriculture in GDP (mean): 17.29
  - Share of informal sector in GDP (mean): 33.07
  - Share of trade in GDP (mean): 68.14
  - Online Service Index (mean): 0.36
  - Human Capital Index (mean): 0.48
  - Government Effectiveness Index (mean): -0.49
- Notable descriptive points:
  - Mean share of citizens with little or no trust ≈ 0.52 (52 percent); higher in fragile (0.55) vs non-fragile (0.51). Maximum for this variable = 0.72 (Nigeria).
  - Governments appear highly open to trade (mean trade/GDP = 68 percent); agriculture and informal sectors are large (17.29 percent and 33.07 percent of GDP respectively).
  - Online service index and human capital index means below 0.5 but with high maximum values.

### Main regression results — VAT efficiency (Table 3, selected)
- Sample: Obs. = 50; R2 = 0.627 (Column I), R2 = 0.665 (Column II).
- Effect of lack_trust:
  - Column I: lack_trust coefficient = -0.222* (standard error 0.120).
  - Column II (interaction with fragile states): lack_trust coefficient = -0.237*** (standard error 0.076) and trust_fragile included (interaction results reported in table; full interaction interpretation in text).
- Interpretation presented in text: A one percent increase in the proportion of citizens with little or no trust in the tax department is associated with a 0.22 percent decrease in governments’ efficiency in generating VAT revenues, ceteris paribus.
- Controls (signs and key coefficients from Table 3):
  - Agr: negative (e.g., -0.395 (0.242) in Column II context; negative effect interpreted as larger agricultural share reduces VAT efficiency).
  - informality: negative but not always statistically significant in the baseline specification (high correlation with Agr noted).
  - trade: positive and statistically significant (trade openness associated with higher VAT efficiency).
  - OSI and HCI: positive effects on VAT efficiency (digitalization and human capital improve collection efficiency).
  - GDP: unexpected negative coefficient in baseline (significant negative; authors note potential multicollinearity).

### Corruption and VAT efficiency (Table 4, selected)
- tax_corrupt = log of share reporting most or all tax officials are involved in corruption (mean ≈ 0.39; fragile mean 0.47, non-fragile mean 0.36).
- Table 4 sample: Obs. = 50; R2 values reported 0.627, 0.671, 0.630, 0.674 across columns I–IV.
- Selected coefficients:
  - Column I: tax_corrupt = -0.199 (standard error 0.118), p-value = 0.101 (not significant at conventional levels).
  - Columns separating fragile vs non-fragile show stronger negative effects in fragile states; interaction term corrupt_fragile indicates larger negative magnitude in fragile contexts.
- Correlation: lack_trust and tax_corrupt correlation = 0.62 (high correlation noted; difficulty separating effects).

### CIT Productivity results (Tables 5 and 6, selected)
- CIT Productivity dependent variable = log of (CIT revenues as percent of GDP divided by CIT rate).
- Table 5 (lack_trust on CIT Prod): Obs. = 36; R2 = 0.303 (Column I), R2 = 0.484 (Column II).
  - Column I: lack_trust coefficient = 0.153 (0.313) — not significant.
  - Column II (interaction with fragile): trust_fragile shows a significant negative effect (evidence that lack of trust matters primarily in fragile states).
- Table 6 (tax_corrupt on CIT Prod): Obs. = 36; R2 values 0.330, 0.464, 0.350, 0.552 across columns I–IV.
  - Columns I–II: corruption effects significant only in fragile states.
  - When both lack_trust and tax_corrupt included, lack_trust tends to dominate for CIT productivity in fragile states (text interpretation).

### Interpretation and substantive analysis
- Lack of trust in the tax department and perceived corruption among tax officials both impair tax collection:
  - VAT efficiency: lack_trust has a negative and statistically significant effect; magnitude about -0.22 elasticity in baseline.
  - Effect of lack_trust and tax_corrupt is stronger in fragile states (interaction results).
- Mechanisms discussed:
  - In largely informal economies VAT payments are not automatic; lack of trust can facilitate buyer–seller implicit collusion or businesses not remitting VAT receipts.
  - Digitalization (OSI) and human capital (HCI) support improved VAT efficiency by simplifying procedures and enhancing compliance.
  - Large agricultural share is associated with lower VAT efficiency due to lower taxes/subsidies in agriculture.
  - For CIT productivity, bribe prevalence in fragile states may explain stronger negative effects of lack_trust and perceived corruption on corporate tax collection.
- Data limitations noted:
  - Unable to separate VAT policy gap vs compliance gap due to limited observations when both measures are included (resulted in only three observations).
  - Including tax expenditure data reduces sample by about half and the tax expenditure variable was insignificant (results available on request).
  - High multicollinearity among macro controls (e.g., GDP, Agr, HCI) affects some coefficient signs (GDP negative in baseline).

### Robustness analysis (section 4.3)
- Simplified macro specification by dropping Agr and HCI due to high correlation with real GDP per capita:
  - Informality becomes significantly negative when Agr dropped.
  - OSI and HCI lose economic power when those variables are omitted.
  - Negative effect of lack of trust on tax efficiency strengthens under simplified specification, but distinction between fragile and non-fragile states disappears.
  - Adding corruption to simplified specification renders survey variables generally insignificant, except when effects are separated between fragile and non-fragile states (in which case high corruption in fragile states significantly impairs VAT efficiency).
- For CIT:
  - Lack of trust and corruption significantly negative in fragile states under robustness checks; lack_trust dominates when both survey variables included.

*Source: IMF working paper — section "3. Econometric Specification" and accompanying Results (sections 4.1–4.3) as provided in the content unit.*

### 5. Conclusion

### 5. Conclusion

### Main empirical findings
- Negative and significant association between citizens’ perception of lack of trust in the tax department and VAT efficiency.
  - As the share of citizens with perceptions of little or no trust in the tax department increases by 1 percent, VAT efficiency decreases by 0.22 percent, ceteris paribus.
- The magnitude of the negative effect of lack of trust on VAT efficiency is significantly greater in fragile relative to non-fragile states.
- Perceptions about corruption in the national tax authority have a similar negative impact on VAT efficiency, with the effect being significantly higher for fragile states.
- For corporate taxation (CIT productivity):
  - The effect of perceptions of lack of trust/corruption is significant only in fragile states.
  - Lack of trust plays a stronger role in explaining CIT productivity than corruption in fragile states.
- Other determinants of VAT efficiency:
  - Higher share of trade in GDP is linked to higher VAT efficiency.
  - More digitalized governments and a higher degree of digital literacy are linked to higher VAT efficiency.
  - Larger informal and agricultural sectors are linked to lower VAT efficiency (informal sector effect is only statistically significant in the regression without the agriculture output share).

### Robustness and selected regression results (as reported)
- Effect of Lack of Trust on VAT Efficiency (Dependent Variable: VATEff)
  - Specification I: lack_trust -0.337*** (0.123); Obs 50; R squared 0.491
  - Specification II: lack_trust -0.325** (0.121); trust_fragile -0.287 (0.236); Obs 50; R squared 0.509
  - Informality coefficients: -0.502* (0.298) in I; -0.433 (0.307) in II
  - trade: 0.577*** (0.172) in I; 0.587*** (0.170) in II
- Effect of Corruption on VAT Efficiency (Dependent Variable: VATEff)
  - Specification I: tax_corrupt -0.338** (0.143); Obs 50; R squared 0.501
  - Specification II: tax_corrupt -0.337** (0.133); Obs 50; R squared 0.526
  - Specification III: tax_corrupt -0.255 (0.212); lack_trust -0.145 (0.200); Obs 50; R squared 0.504
  - Specification IV: tax_corrupt -0.217 (0.195); trust_fragile 0.557 (0.475); Obs 50; R squared 0.533
- Effect of Trust on CIT Productivity (Dependent Variable: CITProd)
  - Specification I: lack_trust 0.018 (0.321); Obs 36; R squared 0.254
  - Specification II: lack_trust -0.066 (0.291); trust_fragile -1.130*** (0.276); Obs 36; R squared 0.444
  - trade: 0.531* (0.277) in I; 0.493** (0.210) in II
- Effect of Corruption on CIT Productivity (Dependent Variable: CITProd)
  - Specification I: tax_corrupt 0.256 (0.255); Obs 36; R squared 0.274
  - Specification II: tax_corrupt 0.216 (0.254); corrupt_fragile -0.722*** (0.256); Obs 36; R squared 0.411
  - Specification III: tax_corrupt 0.782 (0.760); lack_trust -0.772 (0.804); Obs 36; R squared 0.312
  - Specification IV: tax_corrupt 0.648 (0.779); trust_fragile -2.782* (1.448); Obs 36; R squared 0.523

### Policy implications and recommendations
- Increasing citizens’ perceptions of trust in tax authorities is important for revenue mobilization in Africa and for fostering trust in government generally.
- Policies aimed at building fiscal capacity should prioritize ensuring that citizens believe resources will be properly used, especially in fragile states where:
  - Distrust in public institutions appears higher on average.
  - The negative effect on VAT and CIT efficiency of lack of trust in the tax authority is generally stronger.
- Specific recommended actions:
  - Enhance fiscal transparency, including publishing details about the use of government expenditures.
  - Provide technical assistance to revenue administrations to improve effectiveness.
  - Support digitalized tax submissions to help limit and better track non-payment.

### Limitations and interpretation cautions
- Sample size and data constraints:
  - Analysis performed on fifty observations since data on perceptions of trust/corruption is not available in one of the survey rounds.
  - For CIT productivity regressions, sample size is 36 observations.
- Given the small sample size, results should be seen as a starting point for discussions about the role of trust and perceptions of trust in tax authorities on tax efficiency, especially in fragile economies.
- Data limitations prevented separating the effect of perceptions of lack of trust on the VAT policy gap versus the compliance gap.
- Causal inference:
  - Use of linear regression models makes it hard to make causal statements.
  - Results should be interpreted as demonstrating the importance of trust in the tax authority in improving tax compliance in Africa, not definitive causal proof.
- Behavioral inference:
  - Once the public sees that tax revenues are being effectively used for their assigned purposes, it can reasonably be inferred that they are more likely to accept paying taxes.

*IMF Working Paper — Conclusion and appendices as provided in the source content.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024234-print-pdf.pdf_