1. Introduction
Taxpayer attitudes are important determinants of tax compliance (Torgler, Reference Torgler2007), influencing governments’ abilities to fund public goods and, in turn, enhancing government accountability (Moore et al., Reference Moore, Prichard and Fjeldstad2018). Despite the substantial tax evasion challenges many African countries face, questions remain over what influences citizens’ tax willingness. Existing studies on Africa largely rely on correlational evidence, lacking causal insights into the factors driving people’s willingness to pay taxes. Moreover, questions remain about how individuals’ contributions to non-state community goods or initiatives influence their willingness to pay state taxes. These questions are particularly important in contexts with limited public goods provision, where individuals are often expected to fund local infrastructure or other community goods, supply school materials, contribute to burial funds, or join road crews.
To answer these questions, we employ a factorial experiment.Footnote 1 In doing so, we join a growing literature using experimental approaches to study taxation (Shimeles et al., Reference Shimeles, Gurara and Woldeyes2017; Mascagni and Nell, Reference Mascagni and Nell2022; Antinyan and Asatryan, Reference Antinyan and Asatryan2025). To our knowledge, we are the first to experimentally test how donations to community goods affect tax compliance attitudes. We thus provide a novel contribution to the growing literature examining whether citizens’ donations to community initiatives undermine their willingness to pay state taxes (Ali et al., Reference Ali, Fjeldstad and Sjursen2014; Bodea and LeBas, Reference Bodea and LeBas2016; Lust and Rakner, Reference Lust and Rakner2018; Castañeda et al., Reference Castañeda, Doyle and Schwartz2020).Footnote 2 Additionally, we test key determinants of tax morale: access to public goods and services, government fairness, transparency, and enforcement mechanisms.
The results suggest a substitution effect of donations to community goods on tax willingness, and that this effect is substantially larger for individuals eligible to pay income tax. The results also suggest that better access to public services increases both the likelihood of positive attitudes toward paying taxes and trust in the government’s ability to utilize tax revenue effectively. Interestingly, an increased likelihood of detection is associated with a decreased willingness to pay tax. This result, although unexpected, aligns with previous research indicating that enforcement can negatively affect tax morale by eroding government trust (Wahl et al., Reference Wahl, Kastlunger and Kirchler2010).
2. Drivers of tax compliance attitudes
Our main goal in this paper is to test the effect of community donations on tax compliance attitudes. A growing literature proposes a substitution effect of community contributions on tax willingness, especially where states are weak and public goods provision is limited (Lust and Rakner, Reference Lust and Rakner2018). For instance, Bodea and LeBas (Reference Bodea and LeBas2016) find Nigerians with access to non-state community goods are less likely to support taxation, while Castañeda et al. (Reference Castañeda, Doyle and Schwartz2020) show Mexicans who pay for private healthcare are more inclined to evade taxes. In contrast, López García (Reference Lopez Garcia2025) finds that individuals who contribute more to their communities are willing to pay higher taxes in exchange for municipal goods. However, empirical evidence remains limited and largely observational, making it difficult to disentangle causality. By testing the relationship experimentally, we seek to isolate the causal effect of community contributions on tax compliance attitudes. Consistent with much of the literature, we expect that higher payments for non-state community goods will reduce tax willingness.
The literature on fiscal exchange aligns with the intuition described above. This literature conceptualizes taxation as a social contract in which citizens pay taxes in exchange for governments providing public goods and services (Rousseau et al., Reference Rousseau, Brumfitt and Hall1950). Tax morale and compliance are thus expected to be higher when citizens trust the government to fulfill its promises and believe other taxpayers will not evade payment (Levi, Reference Levi1988).
Africans appear to view taxation in this way. For instance, Ali et al. (Reference Ali, Fjeldstad and Sjursen2014) find a positive correlation between access to public goods and tax-compliant attitudes in Kenya, Tanzania, Uganda, and South Africa. Furthermore, Fjeldstad and Semboja (Reference Fjeldstad and Semboja2001) identify inadequate public goods provision in Tanzania as a significant determinant of tax evasion. Therefore, we expect citizens satisfied with the government’s public service provision to have more positive attitudes toward taxation.
Kirchler et al. (Reference Kirchler, Hoelzl and Wahl2008) and Rothstein (Reference Rothstein2005) further argue that citizens’ trust in the government influences tax compliance. This is supported by both observational (Torgler, Reference Torgler2007) and experimental (Wahl et al., Reference Wahl, Kastlunger and Kirchler2010) evidence. Similarly, perceived government fairness (D’Arcy, Reference D’Arcy2011) and fiscal transparency (Capasso et al., Reference Capasso, Cicatiello, De Simone, Gaeta and Mourão2021) appear to increase tax morale, while corruption erodes it (Jahnke and Weisser, Reference Jahnke and Weisser2019). We thus expect that citizens who perceive the government as fairer or more trustworthy will have more positive attitudes toward taxation.
Taxpayers’ perceptions of the government’s ability to punish tax evasion should also significantly influence tax compliance (Allingham and Sandmo, Reference Allingham and Sandmo1972; Kirchler et al., Reference Kirchler, Hoelzl and Wahl2008). However, empirical evidence on the impact of enforcement is mixed (Mascagni and Nell, Reference Mascagni and Nell2022). Some studies suggest that enforcement erodes trust in the government and reduces voluntary tax compliance (Wahl et al., Reference Wahl, Kastlunger and Kirchler2010), while others find it increases compliance (Castro and Scartascini, Reference Castro and Scartascini2015; Shimeles et al., Reference Shimeles, Gurara and Woldeyes2017). Although the literature is divided, we expect enforcement to enhance tax-compliant attitudes.
3. Taxation in Kenya
We conducted our study in Kenya, a non-rentier, multiparty democracy, where the social contract theory of taxation should apply. Community donations are also prevalent, as they are in much of Africa. Approximately 40% of our survey respondents reported contributing to local initiatives in the past year, much like the Kenyan, Malawian, and Zambian respondents in the 2019 Local Governance Process Indicators (LGPI) survey (Lust et al., Reference Lust, Kao, Landry, Harris, Dulani, Metheney, Nickel, Carlitz, Gatua, Jöst P, Mechkova, Maxim, Tengatenga, Grimes, Ahsan Jansson, Alfonso, Nyasente, Ben Brahim, Jordan, Bauhr and Boräng2020), in which a comparable share of the LGPI respondents reported community contributions, particularly to infrastructure initiatives, such as water provision.
Kenya relies significantly on direct tax revenues, particularly the Pay As You Earn (PAYE) income tax scheme, which applies to individuals earning above 24,000 KSH per month (Kenya Revenue Authority, 2026). Direct taxes on personal and business income constitute about half of tax revenue, but noncompliance remains a major concern, with an estimated tax gap of over 40% (IMF, 2021).
To address these concerns, the Kenyan Revenue Authority (KRA) invests substantial resources in implementing reforms. In 2014, the KRA introduced electronic filing, reducing tax preparation time (World Bank and PriceWaterhouseCoopers, 2018). The KRA has also sought to enhance enforcement and impose stricter penalties, leading some taxpayers to complain that authorities treated unintentional oversights with the same vigor as fraud (ACCA, 2019).
4. Experiment on tax compliance attitudes
To investigate determinants of tax compliance attitudes, we presented participants with a hypothetical scenario in which an individual is required to pay income tax to the government, accompanied by a set of randomized factors (see Table 1 for an overview). Using this factorial experiment allows us to systematically vary multiple factors that we expect influence tax compliance attitudes. It also enables us to isolate the effects of individual factors on tax compliance attitudes, providing a clearer understanding of the relative importance of each factor (Hainmueller et al., Reference Hainmueller, Hopkins and Yamamoto2014).
Experimental factors and levels
Average treatment effects
Note: The table presents the main effects of the experimental analyses. Columns (1)–(3) are estimated on the whole sample, and columns (4) and (5) are estimated on the sample of individuals who report an income high enough to be eligible for the PAYE income tax scheme.
Standard errors in parentheses.
+p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001.
We designed the treatments to test the theoretical determinants of tax willingness. We explored the potential substitution effect between community contributions and tax compliance by varying the share of income donated to local community initiatives. We tested the notion of taxation as a social contract by varying individuals’ access to public goods and services, helping us explore how perceived benefits from tax-funded services influence compliance attitudes. As generalized trust in the government and tax authorities is inherently difficult to manipulate in an experimental setting, we varied factors shown to correlate with trust, such as perceptions of governmental fairness in resource distribution and the transparency of tax authorities. To examine the impact of enforcement, we varied the likelihood of detection and the penalties for tax evasion, and considered how perceived risk of punishment influences individuals’ willingness to comply with tax obligations. Finally, we randomized gender and earnings, although these were not our primary factors of interest in the analysis.
Factor levels are randomly selected with equal probability and inserted into the following text:
Imagine {Gender}, who has a job and earns {Earnings} each month. He/She lives in an area with {Service Provision} access to public services such as electricity, water, and security. Every year, he/she donates a {Community Donations} share of his/her income to schools, churches, and other local development initiatives in his/her community. John/Mary is also required to report his/her monthly earnings and pay government tax accordingly. In his/her county, government resources are {Fairness} distributed across wards, and tax authorities provide {Transparency} information about the use of tax funds. If he/she would decide to evade paying taxes, it is {Detection} that he/she would be caught, and if he/she was caught, he/she would have to pay a {Fine} fine.
Each respondent was exposed to one experimental vignette. Following the vignette presentation, respondents were asked five questions, three of which are the main outcome variables for this paperFootnote 3: (1) Should the person pay income tax to the government? (Pay Tax), (2) Would it be justified if the person did not report all their income to pay less tax? (Justifiable), and (3) Would you trust the government to spend the tax money wisely? (Trust).Footnote 4
The experiment was embedded in a nationally representative survey of 2000 Kenyan adults, conducted by TIFA in June–July 2021. The vignette was conveyed via telephone due to the COVID-19 pandemic. Thus, to ensure the experiment’s design was easy to comprehend, we described the levels of factors like Community Donations, Detection, and Fine using words like Large/Small and Likely/Unlikely rather than providing specific amounts, percentages, or probabilities.
With 8 factors, each with 2 levels, there are 256 experimental conditions, averaging only about 8 respondents per condition. Direct comparisons between conditions would have low power to detect effects. Instead, we estimate main effects by comparing individuals across the two levels of each factor while averaging across the other factors, increasing power by using all subjects for each main effect.Footnote 5 While the power calculations in Table A10 suggest at least moderate statistical power, and the balance checks in Tables A5 and A6 support the validity of the design, the large number of conditions likely introduces noise that may limit the detection of significant effects. A more focused design with fewer factors would likely yield more precise estimates.
5. Results
The descriptive statistics suggest a balanced sample across treatment conditions, sufficient variation in dependent variables, and that the chosen experimental conditions are relevant in the study context.Footnote 6 We estimate the average treatment effect of each experimental factor in a linear probability modelFootnote 7:
\begin{equation}{y_i}\, = \,y\, + \,{\beta _1}C{D_i}\, + \,{\beta _2}S{P_i}\, + \,{\beta _3}{F_i}\, + \,{\beta _4}{T_i} + {\beta _5}{D_i}\, + {\beta _6}F{I_i}\, + \,{\beta _7}{G_i}\, + \,{\beta _8}{E_{i\,}} + \,{\epsilon_i}\end{equation}Here, yi is one of three outcomes for individual i. The independent variables represent the experimental factors: CDi is community donations; SPi is service provision; Fi is government fairness; Ti is tax authority transparency; Di is the likelihood of detection; FIi is the size of the fine for evasion; Ei is earnings; and Gi is gender.Footnote 8
We uncover evidence of a substitution effect between donations to community goods and willingness to pay taxes, particularly for individuals eligible to pay income tax. When restricting our sample to the eligible subgroup, individuals presented with a profile making a large donation are nearly 13 percentage points (p ≤ 0.01) less likely to agree that the individual should pay taxes, compared to a profile making a small contribution. In the full sample, the estimated coefficient is about one-third the magnitude and significant at the 0.10 level. Importantly, the benchmark group sees a profile donating a small share of income; results may have been stronger if the profile donated no income. Moreover, the results suggest that the effect is large and significant for the most relevant group—those eligible to pay taxes. This result aligns with our expectations about non-state community contributions.
Our results also support the idea of taxation as a social contract between citizens and the state (Levi, Reference Levi1988) and align with previous correlational findings on public goods provision and tax morale (Levi, Reference Levi1988; Fjeldstad and Semboja, Reference Fjeldstad and Semboja2001; Ali et al., Reference Ali, Fjeldstad and Sjursen2014). Those shown a profile in which the person has good access to public goods and services were nearly 5 percentage points more likely to agree that they should pay the tax and 4 percentage points more likely to say they would trust the government to spend the tax money wisely. When we restrict the sample to individuals eligible to pay income tax based on their self-reported income, the estimated coefficient more than doubles in size. This aligns with our expectations about fiscal exchange.
Surprisingly, we find little evidence that generalized trust or enforcement drives tax compliance. We observe no statistically significant effect of government fairness or tax authority transparency on tax-compliant attitudes. We also find no evidence supporting the positive impact of enforcement on tax compliance. Individuals exposed to a scenario with a higher likelihood of detection were 6 percentage points less likely to agree with the person paying the tax. This contradicts our initial hypothesis but adds to already ambiguous findings on enforcement and attitudes toward tax compliance (Kirchler et al., Reference Kirchler, Hoelzl and Wahl2008).
Other results are less theoretically interesting but lend credence to the experiment. Most notably, individuals presented with a high-income profile were 11 percentage points more likely to say that they should pay the tax than those who saw a lower-income profile.
It is important to note that none of our factors significantly affect the perceived justifiability of tax evasion, except among those eligible to pay income tax. Surprisingly, in this group, good access to public goods and services increases the likelihood of viewing tax evasion as justifiable by 9 percentage points, though at the 0.10 significance level. Donating a large share of income and a high likelihood of detection influence agreement with the taxpaying individual but have no impact on other outcomes. The service provision factor is therefore the only factor consistently affecting multiple outcomes.
6. Conclusion
In this note, we employ a survey experiment to investigate the causal effect of community donations on taxpayer attitudes in Kenya. Our findings support the existence of a substitution effect between donations to community goods and the willingness to pay taxes. They also align with the notion of taxation as a social contract between the citizen and the state, particularly for tax-eligible citizens.
Surprisingly, however, our results indicate limited support for the influence of government fairness on resource distribution, tax authority transparency, and enforcement on tax compliance attitudes. Importantly, the results were stronger for better-endowed respondents. The effect sizes of access to public goods and services on tax compliance attitudes were nearly three times larger for respondents expected to be tax-eligible compared to those who were not. Similarly, the negative impact of large donations to community goods was substantially larger for the better-endowed subsample. In short, perceptions of a fair exchange with the state may have the greatest impact on tax-eligible individuals.
Our findings are both theoretically and substantively important. Tax authorities invest significant efforts to improve tax compliance, often focusing on easing filing processes and strengthening enforcement. However, our results suggest that improving the efficiency and supply of public goods may be just as important. This poses challenges: taxes required to fund public goods are essential for garnering support and ensuring citizens’ compliance, while citizens’ willingness to pay taxes is required to raise necessary funds. Nevertheless, if taxpayers view taxation as a social contract with the state, as our results indicate, ensuring access to essential goods and services has the potential to not only enhance overall welfare but also boost tax compliance, thereby raising the resources necessary for further development.
Future research could further investigate the relationship between privately funded community goods, public goods provision, and taxpayer attitudes or compliance using a more focused experimental design that isolates these key factors.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/psrm.2026.10098. To obtain replication material for this article, https://doi.org/10.7910/DVN/POT08T.
Acknowledgements
We thank Rose Shaber-Twedt for excellent editing, Barry Maydom, Salih Yasun, and Ana López García for their valuable comments, and four anonymous reviewers for their helpful feedback.
Funding statement
This research was supported by the Swedish Research Council through the International Scholar Recruitment Grant (E0003801, PI: Pamela Fredman) and “Public Goods Provision in the Shadow of Urbanization: Learning from Experiences in Malawi, Zambia, and Kenya” (2018-03967, PI: Erica Ann Metheney).
Competing interests
Dr Alex Oguso was employed by the Kenya Revenue Authority for part of the period during which this study was conducted. The authors declare no other competing interests.
Open access
