Corruption Perception and Attitude Towards Taxation in Africa

Abstract

1. Introduction

African countries have stepped up efforts to increase government revenues in recent years, particularly through taxation.¹ But there is a structural limit to these efforts due to the prevalence of ‘hard-to-tax’ sectors (e.g., small/informal businesses or subsistence farmers) in many countries in Africa. This situation poses substantial enforcement problems and provides ample opportunities for non-compliance. In such a situation, a better understanding of attitude towards taxation becomes very important.

The main objective of this paper is to study the relationship between corruption and attitude towards taxation. The latter can be defined broadly as non-pecuniary factors (intrinsic motivation, guilt, shame or reciprocity) that encourage voluntary tax compliance (Luttmer and Singhal, 2014). Specifically, we investigate how the public’s perception of corruption at the highest level of the government (president/head of government office) may affect individuals’ attitude towards taxation, using a large sample that combines Afrobarometer surveys collected in thirty-six African countries between 2011 and 2015.² We find that a low level of perceived corruption in the president’ office is significantly associated with a higher likelihood of tax compliance and tax morale.

A potential problem in studying the effect of governance on attitude towards taxation relates to possible reverse causality. On the one hand, the psychological tax contract literature posits that tax compliance is influenced by government policy, tax authorities’ behaviour and state institutions (Feld and Frey, 2007). On the other hand, the ‘revenue bargaining’ theory argues that the taxpaying process can play a crucial role in the emergence of responsive and effective governments (see Moore, 2008). Our paper attempts to address this possible endogeneity between individuals’ perceptions of corruption and their attitude towards taxation, using an instrumental variables (IV) approach.

We use the ethnicity of the president or head of the government of each country as an instrument for perceived level of corruption in the executive’s office. Our assumption is that individuals from the same ethnic group as the president or head of the government are more likely to have a favourable perception of the executive’s governance, as they may derive ‘psychic benefits’ from seeing him/her in office (Chandra, 2004; Franck and Rainer, 2012; De Luca et al., 2018). In contrast, an individual would prefer to avoid paying their taxes (see e.g., Slemrod, 2007), whether or not the president or head of government is from the same ethnic group.³ The IV results confirm that an individual’s positive perception of governance positively impacts one’s willingness to pay taxes.

The rest of the paper is organised as follows. Section 2 presents a review of the literature on tax compliance, institutions and governance and highlights the contributions of our paper relative to the previous literature. Section 3 describes the data, while Section 4 discusses the empirical approach. We present the results in Section 5 and make concluding remarks in Section 6.

2. Literature review

The deterrence approach to taxation suggests that tax compliance is positively associated with the probability of detection and the severity of punishment, à la Allingham and Sandmo (for a review, see Sandmo, 2005). In contrast, the psychological tax contract view sees the act of tax paying as a quasi-voluntary one. It portrays the existence of the state as a contractual relationship between the elected and the electorate, wherein the latter becomes tax compliant as long as the political process is perceived to be fair and legitimate and public goods are provided. In this case, being tax compliant is influenced by government policy, tax authorities’ behaviour and state institutions (Feld and Frey, 2007).

The psychological tax contract argument has been corroborated in a series of cross-country studies comprising both developing and developed countries (see, e.g., Torgler, 2006; Frey and Torgler, 2007; Richardson, 2008; Anderson, 2017) or single country analyses (see, e.g., Torgler and Schneider, 2005; Alm and Torgler, 2006; Timmons and Garfias, 2015). Most of these studies use the World Value Surveys (WVSs) to measure attitude towards taxation combined with country-level measures of governance perception, potentially introducing a discrepancy between the data sources.⁴ In this paper, both tax morale and governance perception data were obtained from the same individual, permitting more precise individual-level analysis. Specifically, we use over 70,000 survey-level responses identified by individual, country and year of survey (see Table 1).

Kirchler et al. (2008) combined the deterrence and psychological tax contract models into the ‘slippery slope’ framework, arguing that trust in authorities increases voluntary compliance, whereas the power of tax authorities to enforce tax payments determines involuntary compliance. The power of authorities refers to taxpayers’ perceptions of tax officers’ capacity to detect tax evasion, while trust in authorities stems from citizens’ general belief that the tax authorities are benevolent and work beneficially for the public welfare. The power of the government and taxpayers’ trust in government (or tax authorities) both, independently and jointly, determine tax compliance levels.

The review of the theoretical literature above suggests that, while enforcement remains a key driver of compliance, attitude towards taxation, specifically tax morale, can play a significant role in tax compliance decisions. A few studies provide empirical evidence on the importance of tax morale. For example, Kleven et al. (2011) distinguish between third-party and self-reported income in a tax audit study in Denmark. They found that the compliance rate for self-reported income was about 83% for total positive income, 95% for capital income, 86% for stock income and 82% for self-employment income.

Dwenger et al. (2014) conducted a field experiment on local church tax payments in Germany. This tax (which can be combined with donations through overpayment) is legally binding, but the church does not exercise its auditing rights, giving rise to a zero-deterrence situation. They found that 20% of individuals pay at least their true taxes owed at the zero-deterrence baseline, suggesting intrinsically motivated compliance is substantial, although a majority behave as rational, self-interested taxpayers. They also find evidence suggesting that deterrence has strong compliance effects on extrinsically motivated payers but insignificant effects on the intrinsically motivated.

Kogler et al. (2015) provide empirical support for the ‘slippery slope’ framework in four European countries by presenting participants with different scenarios of trust and power. Likewise, Kastlunger et al. (2013) surveyed 389 self-employed Italian taxpayers and entrepreneurs and found that trust is positively related to voluntary tax compliance and that coercive power and legitimate power are correlated with enforced compliance, with the latter leading to increased evasion. A laboratory experiment and an online experiment also show that trust and power, modelled by describing fictive situations, positively influence tax payments (Wahl et al., 2010). Again, these studies point to the importance of tax morale (here trust) in compliance decisions.

In an African context, using a field experiment approach, Shimeles et al. (2017) find that appealing to tax morale promotes compliance but slightly less than does the threat of an audit. Ali et al. (2014) find that tax compliance attitude is positively correlated with the provision of public services (a proxy for good governance) in Kenya, Tanzania, Uganda and South Africa, using the 2011–12 Afrobarometer survey data. Likewise, using the Afrobarometer surveys (2011/2013), Jahnke (2017) finds that bribe payment is negatively correlated to tax morale. The previous two studies focus on correlation and do not attempt to address the reverse causality bias between an individual’s perception of governance quality and his/her attitude towards tax payment. In contrast, our paper attempts to address the causality issue explicitly (see details in the results section below).

3. Data

This section presents the data used to quantify attitudes towards tax payment and perception of governance quality, including the control variables in our regressions. We use data from Afrobarometer surveys, a collection of nationally representative surveys that provide a series of information on African citizens’ opinions on economic, social and political aspects in thirty-six African countries. We rely on two questions related to people’s attitudes towards tax payment in Round 5 (collected between 2011 and 2013) and Round 6 (collected between 2014 and 2015) of the surveys. The total number of observations is over 100,000 for the dependent variables and about 88,000 for the corruption perception variable (see Table 1).

3.1 Attitude towards tax payment

The literature has proposed different measures of tax compliance, ranging from measures that capture the actual act of paying taxes to measures that capture individuals’ attitudes towards taxation—the so-called tax morale—which relates to voluntary tax compliance, an aspect deterrence models have typically pushed to the residual (see Feld and Frey, 2007). In this paper, we focus on both reported tax payment behaviour and tax morale.

To measure attitude towards tax payment, we use the question that asks interviewees whether they have refused to pay taxes or fees to the government in the last year. The possible responses are ‘yes’, ‘no, would never do this’ and ‘no, but would do if had the chance’.⁵ We define ‘never refused to pay taxes’ as a dummy variable that takes the value 1 if the respondent never refused to pay taxes or fees to the government and 0 otherwise.⁶ Responses to this type of questions are likely to be biased in tax evasion surveys in Western countries due to strong social disapproval (e.g., Kinsey, 1992). However, these response biases could be much lower in an African context (Ali et al., 2014) due to lower social stigma associated with tax evasion and lower enforcement power of authorities.

Given the response bias issue, we also use an indirect question to capture attitude towards taxation while avoiding the direct implication of ‘wrongdoing’ by the respondent. Indirect survey questions have been used to reduce response biases (see, e.g., Kaufmann, 1997; Reinikka and Svensson, 2006). By taking the focus of the question away from the respondent’s own actions, indirect questions are believed to be less threatening and less prone to social desirability bias. In addition, these answers are assumed to be indicative of the respondent’s own actions, given that they will use their own behaviour or attitude as a reference point (McNeeley, 2012). Specifically, a survey question asks citizens whether they agree or disagree that the tax authorities always have the right to make people pay taxes. We create a variable ‘Right to make people pay taxes’, which takes the value 1 for respondents who agree with the statement and 0 otherwise. For discussion purposes only, we will refer to cases where the ‘right to make people pay taxes’ equals 1, as positive tax attitude; and cases where the ‘right to make people pay taxes’ equals 0, as negative tax attitude.

Finally, it is worth mentioning that an implicit assumption in studies like ours is that self-reported attitudes towards taxation would translate into actual compliance behaviour. However, such an assumption is difficult to verify, given the complexity of linking survey questions with actual behaviour. Nevertheless, some existing studies report a strong negative correlation between the level of tax morale and the extent of tax non-compliance (Williams and Martinez, 2014).

Table 2 below provides some summary statistics. The first column, ‘has or would refuse to pay taxes’, shows for each country the percentage of citizens who have refused to pay taxes in the past 12 months or would do so if they had the opportunity. The overall average is 26.97% of the respondents. Mauritius records the lowest value at 8.67%, while Sao Tome and Principe have the highest value at 44.31%, followed by Togo at 42%. Notably, there are significant differences across countries, and we can observe a high standard error of around 8.23%.

The second column, called ‘disagree that citizens must pay taxes’, contains the percentage of people who disagree that the tax authorities have the right to make people pay taxes. On average, around 26% of citizens disagree that tax authorities have the right to make people pay taxes. The minimum value reported is 12.39% for Sierra Leone, while Togo has the highest percentage, at 45.98%.

3.2 Measuring perception of corruption

To assess the effect of perceived governance quality on tax compliance, we consider the question from the Afrobarometer surveys on citizens’ perceptions of corruption in the country’s executive branch. The question asks about the level of corruption in the president/prime minister and officials in his office: ‘How many of the following people do you think are involved in corruption, or haven’t you heard enough about them to say: The president and the officials in his office?’. The possible answers are the following: ‘none of them’, ‘some of them’, ‘most of them’ and ‘all of them’.

For the empirical analysis, we create a dummy variable: ‘perceived corruption of president’s office’. The dummy variable equals 0 if a respondent’s answer is ‘none’ and 1 if the respondent thinks that ‘some’, ‘most’ or ‘all’ of the officials in the president’s office are corrupt. Those who refused to respond or answered ‘I don’t know’ are coded as missing.

Table 2 also presents summary statistics of the perceived corruption in the president’s office. The numbers indicate a high level of perceived corruption in the president’s office. On average, 80.67% of the people interviewed think there is corruption in the president’s office. The minimum is 63.42 in Algeria, rising to a maximum of 96.94 in Nigeria.

3.3 Individual control variables

In addition to the measures of perceived corruption, we include a series of individual socioeconomic characteristics that may affect people’s willingness to pay taxes. Table 3 shows the list of the individual-level variables with summary statistics. Among them, bribe payments made to officials measure people’s experience of corruption. We use two questions from the surveys. One asks whether, over the past 12 months, the respondent paid a bribe to government officials to get an official document or permit. The second question asks whether the respondent paid a bribe to avoid a problem with the police. There is a high incidence of bribe payments in our sample. At least 87% of the people reported paying a bribe for a document or permit, and roughly 90% paid a bribe to the police to avoid problems. This high incidence of ‘actual’ bribery is in line with the high level of perceived corruption among public officials, as shown in Table 2.

We also take into account the gender, education, age, geographical location and employment status of the respondents. To measure access to information, we refer to the survey questions that ask interviewees how often they get news from sources such as the radio, TV, newspapers and the Internet. Access to information makes it more likely citizens will be informed about abuses of power and other illegal activities so that governments can be held accountable and possibly be changed through voting.

Access to basic social services and infrastructure could also influence citizens’ perceptions of tax compliance. To capture this, we created a community infrastructure variable to measure the quality of infrastructure in the primary sampling unit (PSU) where people live. In the surveys, the interviewers reported whether the following items were present in the PSU where the interviews took place: an electricity grid, piped water, sewage systems, paved roads and cell phone services. They also indicated if there is a post office, school, police station, health clinic and market in the PSU or within walking distance. Using all these pieces of information and factor analysis method, we construct the variable ‘availability of infrastructure in PSU’ to control for infrastructure provision in the place of residence.

4. Empirical approach

The variable |${corruption}_{ic}$| measures perceived level of corruption of the president’s office. The corruption variable equals 1 if the respondent thinks that none of the president or prime minister is involved in corruption and 0 otherwise. The vector X denotes a set of respondent’s individual socioeconomic characteristics, which are gender, education, age bracket, employment status, geographical location (rural or urban) and access to information. The variables country, time and country×time control for country and time fixed-effects as well as their interacts, while |${\varepsilon}_{irc}\sim N(0,1)$| is the error term.

In addition to personal experience with bribery, Canache et al. (2019) hypothesise that variations in corruption perceptions may be at least partly systematic due to predictable psychological foundations. In this paper, we argue that co-ethnicity with the president or head of government can lead to such a systematic variation.⁸ Specifically, individuals from the same ethnic group as the president or head of government would tend to have a more favourable perception of the quality of governance because they may derive ‘psychic benefits’ from having a co-ethnic leader in office (Chandra, 2004; Franck and Rainer, 2012; De Luca et al., 2018). These ‘psychic benefits’ imply that members of the leader’s ethnic groups will tend to support him/her unconditionally (Franck and Rainer, 2012), including through a positive appreciation of governance quality. As a result, co-ethnicity with the president or head of government is likely to be correlated with an individual’s perception of corruption in the president’s office.

However, co-ethnicity as an IV should not affect an individual’s attitude towards taxation for several reasons. First, individuals would prefer to avoid paying their taxes (Slemrod, 2007), regardless of co-ethnicity with the president or head of government because an individual would bear the full financial burden of tax payment without being able to exclude free-riders from benefiting from public goods funded by his/her taxes. Second, Padro and Miquel (2007) and Burgess et al. (2015) argue that ethnic groups cannot be directly discriminated using taxes in developing countries, given African governments’ limited capacity to effectively discriminate individuals with regard to taxation (Kasara, 2007). As a result, there should not be any systematic variation in attitude towards taxation due directly to co-ethnicity with the president or head of government. Third, it is difficult for citizens to gather accurate and direct information about the pervasiveness of corruption (Olken, 2006), particularly at the highest level of government authority. Consequently, individuals are likely to turn to external sources such as news of major corruption scandals from media or corruption-relevant information pertaining to individuals’ surrounding environment, including the same ethnic group. In contrast to corruption perception, attitude towards taxation is likely to be indicative of the respondent’s actions or beliefs, given that all citizens are taxpayers in principle.

5. Empirical results

In this section, we start by providing the results from the simple probit regression before turning to the IV approach.

5.1 Simple probit

We start our empirical analysis by investigating the effects of perceived corruption in the president’s office on the probability that a citizen would indicate that he has not refused to pay taxes or fees to the government, using a simple probit regression. To keep as many observations as possible, in column (1) of Table 4, we only control for the perceived corruption variables and exclude individual variables. In column (2), we add a series of individual-level variables discussed in Section 3.3.

The estimated coefficients on the perceived level of corruption of the president’s office are all negative and statistically significant at the 1% significance level, suggesting that an individual who thinks that at least some of the officials in the president’s office are involved in corruption has a lower probability of responding that he/she has not refused to pay a fee or tax to the government in the past 12 months, than does an individual who perceives that none of the officials in the president’s office are corrupt. These findings are not affected by the inclusion of control variables (column (2)).

We also find that individuals’ experiences of bribe payment in exchange for official documents or for avoiding problems with the police is significant and negatively associated with tax morale. We find a positive and significant relationship between positive tax attitude and the following variables: being a woman, having completed secondary education or a higher level of education, and being able to access information. Older respondents have higher tax morale than people younger than 25 years. As expected, the presence of necessary infrastructures in the primary sample unit where the individuals are located positively affects their tax morale.

Columns (3) and (4) of Table 3 show results relating to whether citizens agree or disagree that the tax authorities always have the right to make people pay taxes. The estimated coefficients on the perceived level of corruption of the president’s office are also all negative and statistically significant at the 1% significance level, suggesting that an individual who thinks that at least some of the officials in the president’s office are involved in corruption has a lower probability of agreeing that the tax authorities always have the right to make people pay taxes, compared to an individual who thinks that none of the officials in the president’s office are corrupt.

5.2 IV approach and robust check

As noted earlier, there can be reverse causality between attitudes towards tax payment and perception of corruption. To address this potential endogeneity, we propose an IV approach based on the ethnicity of the president or head of government as an instrument for perception of corruption (see Section 4 for details).

The IV estimation results are reported in Table 5. These include both the results for the first-stage corruption equation (see columns (1) and (3)) and the second-stage tax compliance attitude equation (see columns (2) and (4)). The results for the perceived corruption equation show a negative and statistically significant coefficient on the dummy related to ethnic group of the president. This indicates that, on average, people from the same ethnic group as the country leader have a lower probability of indicating that the president’s office is corrupt, in line with the ‘psychic benefits’ hypothesis.

For individual-level control variables, paying a bribe increases the probability of agreeing with the statement that officials in the president’s office are corrupt. Education, living in an urban area and having access to information also positively affect the probability of perceiving the president’s office as corrupt.

Columns (2) and (4) of Table 5 show the second-stage estimation results for the tax attitude regressions. The estimated coefficients of the perceived level of corruption of the president’s office remain all negative and statistically significant at the 1% level, thereby suggesting that a perceived corruption in the president’s office has a negative effect on the probability that a citizen would indicate he has not refused to pay a fee or tax or would agree with the statement that tax authorities always have the right to make people pay taxes, in line with the findings of the simple probit regression model.

Regarding the variable ‘never refused to pay taxes’, the coefficient of perceived corruption is almost five times larger when using the IV approach in Table 5 (−0.166) compared to simple Probit in Table 4 (0.036). Similarly, the coefficient of perceived corruption is almost six times larger when using the IV approach in Table 5 (−0.221) compared to simple probit in Table 4 (0.038), for the variable ‘Right to make people pay taxes’. Such results suggest that controlling for the favourable bias in the corruption perception of the president’s co-ethnics amplifies the negative effects of corruption on tax-compliant attitude.

6. Conclusion

How to increase domestic resource mobilisation through taxation has been a key policy concern in African countries, given formidable development financing needs. However, according to the Afrobarometer surveys conducted over the period 2011–2015, on average, 27% of citizens in Africa have or would refuse to pay taxes, and around 26% of them agree that the tax authorities do not have the right to make people pay taxes. These figures vary significantly across countries.

With the prevalence of ‘hard-to-tax’ sectors, positive attitudes towards taxation and voluntary compliance can be instrumental for increasing revenue mobilisation in African countries. Using survey answers from Afrobarometer, this paper provides evidence of a negative and significant relationship between perceived corruption of the president’s office and tax compliance attitudes, even after controlling for people’s experience of corruption (that is bribe payments made to government officials in the past 12 months). Specifically, a citizen who thinks that at least some of the officials in the president’s office are involved in corruption has a significantly lower probability of having a positive attitude towards paying taxes, compared to an individual who perceives that none of the officials in the president’s office is corrupt.

In contrast to previous papers, we also attempt to establish a causal relationship between perceived corruption and tax attitude, using co-ethnicity with the president or head of government as an instrument for the perceived level of corruption in the executive’ office. As hypothesised, we find that co-ethnics of a country leader have a lower probability of indicating that the president’s office is corrupt. The IV results also confirm a negative and significant relationship between perceived corruption and attitude towards taxation.

In summary, our result suggests that besides the enforcement power of tax authorities, other factors, such as the quality of governance, can influence attitude towards taxation. Better governance, by showing that the executive branch is acting beneficially for the public welfare, can provide incentives for voluntary compliance and more positive attitudes towards taxation, thereby resulting in higher tax revenues. This can be achieved by fighting corruption or promoting transparency and accountability mechanisms (including international initiatives such as the Extractive Industries Transparency Initiative) in government, in addition to providing quality public goods and services.

Acknowledgments

We thank participants in the AERC Biannual Research Workshop (December 2017) in Arusha (Tanzania) as well as Stephen D. O'Connell, for constructive comments. We would also like to thank Yaye B. Camara, Gideon Ndubuisi, and Zackary Seogo for excellent research assistance. The views expressed in this article are those of the authors and do not necessarily express the views of the African Development Bank Group, African Economic Research Consortium, Columbia University, or UNU-MERIT.

Footnotes

References

Appendix