CAF - Reciprocity and willingness to pay taxes evidence from a survey experiment in Latin America

CAF

DOCUMENTOS DE TRABAJO

CAF

WORKING PAPERS

RECIPROCITY AND WILLINGNESS TO PAY TAXES: EVIDENCE

FROM A SURVEY EXPERIMENT IN LATIN AMERICA

N° 2012/01

June, 2012

Ortega, Daniel Ronconi, Lucas Sanguinetti, Pablo

CAF - Ave. Luis Roche, Torre CAF, Altamira. Caracas, Venezuela 01060

© CAF, 201 2 por Ortega, Daniel; Ronconi, Lucas y Sanguinetti, Pablo . Todos los derechos reservados. Pequeñas secciones del texto, menores a dos párrafos, pueden ser citadas sin autorización explícita siempre que se cite el presente documento. Los resultados, interpretaciones y conclusiones expresados en esta publicación son de exclusiva responsabilidad de su(s) autor(es), y de ninguna manera pueden ser atribuidos a CAF, a los miembros de su Director io Ejecutivo o a los países que ellos representan. CAF no garantiza la exactitud de los datos incluidos en esta publicación y no se hace responsable en ningún aspecto de las consecuencias que resulten de su utilización.RECIPROCIDAD Y VOLUNTAD DE PAGAR IMPUESTOS: EVIDENCIA DE UN EXPERIMENTO DE ENCUESTAS EN AMERICA LATINA Ortega, Daniel; Ronconi, Lucas y Sanguinetti, Pablo CAF Documento de trabajo N° 2012/01 June, 2012

RESUMEN

Este artículo ofrece el primer intento experimental de medir la reciprocidad en la recolección de impuestos en los países en vías de desarrollo, donde las instituciones

CAF Documento de trabajo N° 2012/01 June, 2012

RESUMEN

Este artículo ofrece el primer intento experimental de medir la reciprocidad en la recolección de impuestos en los países en vías de desarrollo, donde las instituciones de recaudación son débiles, y donde las tasas de impuesto y en general el cumplimiento de las obligaciones tributarias es bajo. En una encuesta de hogares llevada a cabo en 17 ciudades de América Latina, se le entregó aleatoriamente a los encuestados información positiva o negativa sobre las administraciones de los gobiernos locales; esta información alteró significativamente las percepciones sobre la calidad del gobierno local en dos ciudades, y se encuentra que en Rio de Janeiro – donde un grado relativamente alto de autonomía permite a los ciudadanos establecer una clara conexión entre los impuestos y la provisión de servicios públicos - la percepción de los ciudadanos sobre la calidad del gobierno local puede tener efectos importantes sobre su disposición a pagar impuestos.

Palabras clave: evasión, reciprocidad, moral tributaria, América Latina

RECIPROCITY AND WILLINGNESS TO PAY TAXES: EVIDENCE FROM A SURVEY

EXPERIMENT IN LATIN AMERICA

Ortega, Daniel; Ronconi, Lucas y Sanguinetti, Pablo CAF Working paper N° 2012/01 June, 2012

ABSTRACT

We provide the first experimental attempt at measuring reciprocity in tax collection in developing countries, where enforcement institutions are weak, and where tax rates and in general tax observance is lower. In a household survey carried out in 17 Latin American cities, we randomly provide respond ents with positively or negatively information on the local Government's administration; this information significantly altered perceptions on the quality of the local government in two cities, and we find that in Rio de Janeiro -where a relative high deg ree of tax autonomy allows citizens to

American cities, we randomly provide respond ents with positively or negatively information on the local Government's administration; this information significantly altered perceptions on the quality of the local government in two cities, and we find that in Rio de Janeiro -where a relative high deg ree of tax autonomy allows citizens to make a clear link between taxes and the provision of public services - people's perception on the quality of the local government can have sizable effects on their willingness to pay taxes.

Keywords: Evasion, Reciprocity, Tax Morale, Latin America

Daniel Ortega Lucas Ronconi Pablo Sanguinetti

CAF CIAS CAF

Instituto de Estudios Superiores de Administración

(IESA)

Universidad Torcuato Di Tella dortega@caf.com ronconilucas@gmail.com psanguinetti@caf.com1

Reciprocity and Willingness to Pay Taxes: Evidence from a Survey Experiment in Latin America1

Daniel Ortega

CAF and IESA

Lucas Ronconi CIAS Pablo Sanguinetti

CAF and UTDT

June 2012

Abstract We provide the first experimental attempt at measuring reciprocity in tax collection in developing countries, where enforcement institutions are weak, and where tax rates and in general tax observance is lower. In a household survey carried out in 17 Latin American cities, we randomly provide respondents with positively or negatively information on the local Government's administration; this information significantly altered perceptions on the quality of the local government in two cities, and we find that in Rio de Janeiro -where a relative high degree of tax autonomy allows citizens to make a clear link between taxes and the provision of public services - people's perception on the quality of the local government can have sizable effects on their willingness to pay taxes.

Keywords: Evasion, Reciprocity, Tax Morale, Latin America

JEL Codes:

1 We would like to thank Mariana Urbina and Mauricio Stern for excellent research assistance. Seminar

quality of the local government can have sizable effects on their willingness to pay taxes.

Keywords: Evasion, Reciprocity, Tax Morale, Latin America

JEL Codes:

1 We would like to thank Mariana Urbina and Mauricio Stern for excellent research assistance. Seminar participants at CAF's RED 2012 meeting in Santiago de Chile provided useful comments and discussion. Errors are our own.2 1 – Introduction Why do people pay taxes? One obvious answer is that there are enforcement mechanism (fines and penalties) established by the State. This deterrence approach to tax compliance has been formally analyzed in the seminal paper by Allingham and Sandmo (1972) where they apply the canonical rational choice model (expected utility maximization) to study this issue obtaining that tax evasion is negatively associated with the probability of detection and the severity of punishment. However this view has been criticized by various authors (Graetz and Wilde, 1985; Alm et al., 1992; Frey and Feld, 2002) on the account that deterrence only cannot explain the level of tax compliance actually observed in many countries. In other words, as Andreoni, Erard and Feinstein (1998) have indicated, the level of detection probabilities and fines are too low to explain the rather low levels of tax evasion observed in developed economies. This has given rise to a growing literature that analyzes the importance of behavioral and cultural aspects in explaining taxpayer behavior (Scholz and Witte, 1989; Alm et al., 1992; Alm et al., 1993; 1999; Pommerehne et al., 1994; Frey 1997, 2003; Frey and Torgler, 2002). These other non deterrence arguments have been grouped under the broad concept of “tax morale”. This concept encompasses moral rules and sentiments of citizens which make them fulfill with their tax obligations, social norms that makes cheating on taxes an undesirable action when the rest of the population is complying, and a sort of reciprocity response of the individual in its relation to the

grouped under the broad concept of “tax morale”. This concept encompasses moral rules and sentiments of citizens which make them fulfill with their tax obligations, social norms that makes cheating on taxes an undesirable action when the rest of the population is complying, and a sort of reciprocity response of the individual in its relation to the State in the sense that citizens will comply with their tax obligation when they see that the government also performs well its commitments in terms of delivering services and other public goods. Though Torgler (2003) and other authors have provided evidence that some of these different determinants associated with tax morale affect positively tax compliance, the empirical relevance of these non-economic reasons have also been disputed. In particular Fellner, Sausgruber and Traxler (2011) using a field experiment in Austria find that “neither appealing to morals nor imparting information about others‟ behavior enhances compliance on aggregate.” Similarly, Kleven et al. (2011) find that, in Denmark, proxies of social and cultural factors (i.e., gender, age, marital status, church membership and place of residence) have a very modest effect on tax behavior; and3 Blumenthal, Christian and Slemrod (2001) find that normative appeals to social norms and equity have no effect on tax compliance in Minnesota. The purpose of this paper is to provide new evidence on the relationship between tax morale and tax compliance. Within the various channels affecting tax morale we emphasize that related to the interaction between the individual and the State/Government. In particular, we analyze the potential reciprocal relationship that could develop between good government performance (in terms of provision of social services and public goods) and citizens‟ willingness to pay taxes. We analyze the importance of reciprocity in Latin American countries, a region where tax evasion is significantly larger compared to the US or Western Europe. This fact, already signaling a poor government performance, could make the empirical investigation of the reciprocity hypothesis more interesting compared to countries whether taxation and governmental institutions work reasonably well as is the case of developed economies.

significantly larger compared to the US or Western Europe. This fact, already signaling a poor government performance, could make the empirical investigation of the reciprocity hypothesis more interesting compared to countries whether taxation and governmental institutions work reasonably well as is the case of developed economies. We use microdata coming from the CAF survey (CAF 2011) covering 17 cities in Latin America. Our empirical methods include the analysis of conditional questions where the connection between government performance and willingness to pay taxes is made explicit, simple OLS regressions between perceptions about service delivery and attitude toward taxation and finally, and more important, experimental-type exercises where we correlated randomized assignment of information about how the government is doing with individuals‟ intention to pay taxes. This variation in the empirical methods allow us to gauge the robustness of the results and, when we use the experimental data, to evaluate the causal relationship going from changes in perceptions about government performance to willingness to pay taxes. Beyond a purely academic interest, the topic is informative for policymakers. Finding that taxpayers behave in a reciprocal manner suggests that governments interested in increasing tax revenues should not only focus on escalating the coercive power of the state (i.e., more audits and fines) but also on providing better services to society. The paper is organized as follows: The next section offers a short conceptual discussion of the determinants of tax morale and how the concept is related to the4 reciprocity hypothesis. In section three we describe the data and test for the presence of reciprocity using conditional questions wherein respondents express their willingness to pay more (or less) taxes if the government improves its performance. We explore whether Latin Americans are more likely to respond in a reciprocal manner to improvements in institutional aspects (such as more transparency and less corruption) or in the provision of services (such as health and education). Section four presents OLS estimates of the relation between an overall measure of government performance and four different measures of willingness to pay taxes (or pass over easy opportunities of tax evasion). The fifth section exploits an exogenous change in perception of government performance and

services (such as health and education). Section four presents OLS estimates of the relation between an overall measure of government performance and four different measures of willingness to pay taxes (or pass over easy opportunities of tax evasion). The fifth section exploits an exogenous change in perception of government performance and computes instrumental variable estimates of the effect of changes in these perceptions on willingness to pay taxes; finally, section six concludes.

2. Tax morale and reciprocity “Tax morale” usually means the self reported citizen‟s perception that paying taxes is the right thing to do. In this section we distinguish two concepts that are usually bundled under the broad concept of tax morale: reciprocity towards the government and conditional cooperation.

People‟s perception that paying taxes is the right thing to do is related with the idea that the State has some social value , so it is a citizen ‟s obligation to contribute to its financing. In the case of a tyrannical State that exploits its citizens without providing any services whatsoever the notion of tax morale would lose its ethical and political meaning. When seen from this point of view, the concept of tax morale is then closely related to the idea of reciprocal behavior from the individual towards the State. That is, according to the “reciprocity towards the government” hypothes is, people are more likely to pay taxes when they perceive the government is doing a better job. Notice that reciprocity implies a deviation from purely self -interested behavior: it “means that in response to friendly actions, people are frequently much ni cer and much more cooperative than predicted by the self-interested model” (Fehr and Gächter, 2000). In the case of taxes one could say that the individual would be willing to pay more taxes to the exten t that the use of thes e resources by the State is ben eficial for him and for society as a whole. The departure5 from self -interested behavior is clear , since the individual ‟s contribution to the government‟s budget is negligible and thus cannot expect to make a difference in the amount of goods and services r eceived. Thus, reciprocal behavior is altruistic, but

from self -interested behavior is clear , since the individual ‟s contribution to the government‟s budget is negligible and thus cannot expect to make a difference in the amount of goods and services r eceived. Thus, reciprocal behavior is altruistic, but conditional on receiving satisfactory goods and services from the State. Having argued that tax morale is very closely related to the concept of reciprocity towards the government, it is also worth mentioning that reciprocity has also been related to the notion of social norms or “societal institutions” (Bird et al., 2004) whereby the individual‟s willingness to evade/pay taxes in part also depends on whether he perceives other society members comply with their tax obligations or not (Frey and Torgler, 2007). This may also be interpreted as reciprocal behavior towards “the fellow citizen” instead of towards the government. Another way to see this other dimension of tax morale is as a coordination problem, where even though people realize that their individual contribution to the State makes no difference in terms of overall government income and service delivery, if most people don‟t comply with taxes, government performance will be negatively affected, and if everyone complies, performance will improve . In this interpretation a high level of tax morale implicitly means a decision of most individuals to cooperate in the tax/expenditure game of financing the State. Given this other possible interpretation of tax morale, which is complementary to reciproc ity towards the government, in the estimation s we present below we include a social norm variable as a control in all regressions.2 Finally, there may be other reasons beyond reciprocity and condi tional cooperation that explain why people think paying taxes is the right thing to do. This may be related to deep moral judgment or values that are induced by religious beliefs or educational background, and that are unconditional on the behavior of othe r people or institutions. Given that it could be difficult to capture all these other dimensions, in the OLS and experimental exercises we are going to include directly the usual measure of tax morale: whether people think tax evasion is morally never justifiable.

institutions. Given that it could be difficult to capture all these other dimensions, in the OLS and experimental exercises we are going to include directly the usual measure of tax morale: whether people think tax evasion is morally never justifiable.

2 In equation 1 of section 4 we use the individual perception of overall evasion in the country/city as the social norm variable.6 3 – Evidence based on conditional questions The data used in this paper comes from a household survey of 17 cities in Latin America that samples 600 households per city (making it representative at the city level), totaling 10,200 observations (CAF, 2011).3 Survey respondents are individuals between 25 and 65 years of age regardless of their occupational status. The survey has been undertaken yearly since 2008 and in 2011 included a module on taxes and perceptions about government quality. In the Appendix, Table A1 shows descriptive statistics on the sampled population in each city. With some differences between cities, the surveyed population is on average 41.7 years old, has 2.2 children and lives in a household with 4.4 persons, over 65% are married or live with their partner, one third lean to the right ideologically, and overall are not very satisfied with democracy (5.7 out of a possible score of 10). We begin analyzing the relationship between government performance and willingness to pay taxes using conditional questions. We ask people whether they are willing to pay less, the same or more taxes if the government improves its performance among several dimensions. We explore six different measures of performance: Government corruption, transparency, tax collection, quality of public health and education services, crime and social assistance (the exact wording is, for example, “If the government were to improve health and education services, would you be willing to pay more or less taxes than you do presently?”). A plausible measure of reciprocity in this setting is the share of the population that is willing to pay more minus the share that is willing to pay less tax as a reaction to an improvement in government performance.

government were to improve health and education services, would you be willing to pay more or less taxes than you do presently?”). A plausible measure of reciprocity in this setting is the share of the population that is willing to pay more minus the share that is willing to pay less tax as a reaction to an improvement in government performance. The majority of the population (56 percent) reports that they are willing to pay the same amount of tax regardless of improvements in government performance.4 The rest of the population is generally more likely to report a willingness to pay more taxes under hypothetical situations of improved government quality, but there is large heterogeneity across cities and measures of government performance, as is shown in table 1. Taking the region as a whole, there is a statistically significant net share of the population that is

3 The surveyed cities are Buenos Aires, Córdoba, La Paz, Santa Cruz, Sao Paulo, Rio de Janeiro, Bogotá, Medellin, Quito, Guayaquil, Lima, Arequipa, Caracas, Maracaibo, Montevideo, Salto and Panama City. 4 This percentage is an average across cities and measures of government performance.7 willing to pay more taxes in all of the six subjects analyzed (less tax evasion is one aspect where there is some disagreement within several cities).5 The magnitude of the coefficients is often small, ranging between 9 and 12 percent of the population on average, but some coefficients are large, such as the net share willing to increase tax payments in the event of an improvement in health and education in Lima, La Paz, Bogota and Salto which ranges from 30 to 40%. On average Latin Americans declare themselves just as likely to reciprocate when the government improves the quality of services (i.e., social assistance, protection against crime, and particularly health and education – columns 4 through 6) as when there are improvements in procedural or institutional aspects such as transparency, corruption and tax collection effectiveness (columns 1 through 3), although in Sao Pablo and specially in Rio de Janeiro households seem to be willing to contribute more in the case of improvements in institutions while in

education – columns 4 through 6) as when there are improvements in procedural or institutional aspects such as transparency, corruption and tax collection effectiveness (columns 1 through 3), although in Sao Pablo and specially in Rio de Janeiro households seem to be willing to contribute more in the case of improvements in institutions while in the other cities we usually find the opposite. The average across measures of performance is positive in 13 out of the 17 cities. Interestingly, although the average is negative in Medellin, Guayaquil and Maracaibo, it is positive in Bogotá, Quito and Caracas suggesting heterogeneity within countries.6 <<Table 1>> A potential criticism to this evidence is that using conditional questions could generate a social desirability bias.7 People –who have a tendency to want to be accepted by otherscould think that responding “pay more taxes” is the socially desirable answer to an improvement in government performance, leading to an overestimation of reciprocity. The magnitude of the above coefficients suggests this bias may not be very

5 This question merits two comments: First, a large fraction of the respondents presumably evade taxes since the sample is representative of the population and tax compliance is low in Latin America. It is not surprising that tax evaders would not be willing to pay more tax if the government reduces evasion. Second, for those who comply with their tax obligations it is unclear whether they interpret the reduction in evasion as an improvement in government performance or as more cooperation from their fellow citizens. In the next section we distinguish the “reciprocity towards the government” hypothesis from the “conditional cooperation” hypothesis. 6 The coefficients are larger for male and for younger cohorts, although differences are small. People who have a left-leaning ideology (defined as those who consider that the priority of the government is to reduce poverty, inequality and provide free and universal access to health and education) are more likely to have reciprocity in all aspects except in exchange for reductions on crime. (Results are not shown in the text, available upon request). 7 See Demaio (1984) for a review of social desirability bias.8

poverty, inequality and provide free and universal access to health and education) are more likely to have reciprocity in all aspects except in exchange for reductions on crime. (Results are not shown in the text, available upon request). 7 See Demaio (1984) for a review of social desirability bias.8 important, but it is impossible to rule it out. We therefore move to other methods to test for the presence of reciprocity.

4 – Econometric evidence According to the reciprocity hypothesis, people are willing to pay more taxes (or pass over easy opportunities of tax evasion) if they think that the government would use the money to improve the welfare of the population. We test for the presence of reciprocity conditional on tax morale and perceptions of tax evasion using the following model: (1) Yi = βGovt. performancei + πPerception of tax compliancei + θTax moralei + Xi δ + ei where Yi is a measure of the willingness to pay taxes of individual i. We use four alternative measures of willingness to pay taxes. First, we ask people whether they are willing to pay a small increase in two local taxes: housing and vehicle; and create two indicators (Housing tax and Vehicle tax) which take a value equal to 1 if the individual answers positively and 0 if negatively. Only 12 percent of the sample is willing to pay a small increase in housing taxes (ranging from 35 percent in Maracaibo to 4 percent in Sao Pablo) and 8 percent is willing to pay a small increase in automobile taxes (ranging from 23 percent in Maracaibo to 1 percent in Córdoba). Third, we ask whether they consider that the amount they pay for local services (i.e., water, electricity, street lighting and garbage collection) is too low or too high using an ordinal scale from 1 to 5 (Local taxes too high). Finally, we present a hypothetical situation and ask what they would do if the owner of a shop offers them a 10 percent discount in exchange for not providing a receipt. We then create the variable Tax evasion which is an indicator that takes a value

taxes too high). Finally, we present a hypothetical situation and ask what they would do if the owner of a shop offers them a 10 percent discount in exchange for not providing a receipt. We then create the variable Tax evasion which is an indicator that takes a value equal to 1 if the individual responds that he would accept the deal and 0 otherwise. Almost 40 percent of the sample said that they would accept the deal, ranging from 23 percent in Bogotá to 61 percent in La Paz. We create a measure of Government performance asking people to express their level of agreement with the following statement: “The taxes that are collected in your city are used to finance public policies aimed at improving the welfare of the population”. They could choose between a value of 1 (meaning that they totally disagree with the statement) and a value of 10 (meaning that they totally agree with the statement). The9 sample mean is 5.3 with almost half of the sample choosing a value of 5 or 6, but a nonnegligible share took extreme positions (i.e., 9.6 percent expressed a total disagreement and 7.5 percent total agreement with the statement). Contrary to the evidence presented in the previous sections –where we use issue-specific measures of government performance – here we use an overall measure.8 According to the reciprocity hypothesis, people are willing to pay more taxes (or pass over easy opportunities of tax evasion) if they think that the government uses the money to improve the welfare of the population. An individual‟s willingness to pay taxes could also be influenced by her perception of the compliance of others (Frey and Torgler, 2007). That is, according to the conditional cooperation hypothesis, people who think that fewer of their fellow citizens evade taxes are more willing to pay or to pass over easy opportunities of evasion. To test this hypothesis we use a variable (Perception of tax compliance) that ranges between 1 and 10, adopting a value equal to 1 if the individual thinks that none of her fellow citizens fully comply with taxes and 10 if she thinks that all fully comply.

this hypothesis we use a variable (Perception of tax compliance) that ranges between 1 and 10, adopting a value equal to 1 if the individual thinks that none of her fellow citizens fully comply with taxes and 10 if she thinks that all fully comply. We also explore the importance of a narrow definition of tax morale which, contrary to the previous two hypotheses, is an unconditional concept. An individual has tax morale if she thinks it is right to pay taxes regardless of how the government uses the money and regardless of the tax behavior of others. Several studies argue that tax morale affects compliance (Alm, McClelland and Schulze, 1992; Frey 2003; Lewis, 1982; Torgler, 2001), although the term has been used vaguely sometimes. We measure Tax morale as a binary variable based on the following question “Do you think it is morally justifiable to evade taxes?” (1 if totally unjustifiable, 0 otherwise).9

8 In Appendix Table A2 we present OLS estimates of the effects of subject specific measures of government performance on willingness to pay taxes. We analyze four areas: Satisfaction with the garbage collection service, with electricity services, with water services and with public transportation. We find that 6 out of the 16 coefficients (i.e., four measures of performance and four measures of willingness to pay taxes) have the expected sign and are statistically significant; the other 10 coefficients are not significant. 9 As it was discussed in section 2 Frey and Torgler (2007) analyze the determinants of tax morale, and find that it is positively and statistically significantly correlated with both government performance and perceptions of others‟ tax compliance. Based on this finding we may tend to conclude that tax mor ale should not be included as a separate regressor because it is actually caused by the other two factors. But those two variables explain very little of the variation in tax morale (pseudo R2 below 0.05 in Frey and Torgler (2007) and below 0.01 in our sample).10 Finally in X we include a set of controls (i.e., age, sex, educational attainment,

those two variables explain very little of the variation in tax morale (pseudo R2 below 0.05 in Frey and Torgler (2007) and below 0.01 in our sample).10 Finally in X we include a set of controls (i.e., age, sex, educational attainment, employment status, nationality, ideology, and wealth of the individual).10 Table 2 presents the results for several measures of willingness to pay taxes. The first column uses the whole sample and the remaining columns restrict the sample by city. In panel A the dependent variable is willingness to pay more Housing tax, in panel B it is willingness to pay more Vehicle tax, in panel C it is whether the individual thinks that Local taxes are too high, and in panel D it is Tax evasion. All models include the full set of controls