اختلافات قومی، اعتماد، و اندازه بخش غیررسمی

This paper investigates the relationship between ethnic fractionalization and the size of the informal, or hidden, economy. Recent experimental and empirical research links, in turn, ethnicity and trust, and trust and tax compliance. In addition, recent empirical studies have identified an unwillingness to contribute to public goods benefiting other ethnic groups. Combining these insights, we argue that increasing ethnic fractionalization decreases voluntary tax compliance, and we present empirical evidence at the macro level in a cross-section of more than 50 countries that more ethnically fractionalized societies have significantly larger informal sectors.

Tax compliance and administration are central issues in public finance. For developing economies, difficulties in tax collection remain a major obstacle for economic development. Tax evasion and large informal sectors can force governments and tax agencies to increase tax rates on activities less prone to evasion, resulting in increased distortions that can have adverse effects on investment and growth and, additionally, non-compliance can result in effective tax systems being less equitable than those legislated. These problems of non-compliance are present also in the developed world; for example, European countries are currently considering how to finance extensive welfare state programs in the face of adverse demographic changes, and the presence of large scale tax evasion will make this even harder.1 This paper provides the first cross-country empirical investigation of two explanations for the size of the informal sector that have received much attention in the literature on tax compliance and administration, particularly regarding developing countries:2(i) the degree of voluntary tax compliance, and (ii) the size of the rural, or agricultural, economy. While the latter is straightforwardly tested using available data, I draw on recent experimental and empirical research in economics, political science and social psychology to formulate the hypothesis that the degree of voluntary tax compliance (and the size of the informal sector) depends on the degree of ethnic fractionalization in society: Ethnic fractionalization (i) decreases the level of trust, which decreases tax compliance, and (ii) increases the unwillingness to contribute to financing public goods to the extent that these (primarily) benefit other ethnic groups. This approach might give some insight into why developing nations, some of which (especially in Africa) are characterized by a high degree of ethnic heterogeneity, have larger informal sectors. While ethnic heterogeneity could be one of many reasons many developing countries have problems securing (voluntary) support for taxation, it is also very relevant for the discussion concerning the financing of Europe’s extensive welfare state programs, given that the historically very homogenous European countries are currently experiencing increasing ethnic heterogeneity due to immigration and, further, that an integrated European Union itself will be an area with a high degree of ethnic, linguistic and religious fractionalization. While there is by now a substantial theoretical literature on tax compliance and the size of the informal sector, comparative empirical evidence (for obvious reasons) is more scarce. In the first cross-country study of the informal sector, Friedman et al. (2000) find that the size of the informal sector increases in the level of corruption, measured in a variety of ways. They do not, however, consider the effect of societal heterogeneity on the informal sector and generally control only for two explanatory variables at a time. A small, recent group of papers address issues related to the analysis presented here. Slemrod, 1998 and Slemrod, 2003 suggests, without testing it empirically, that voluntary tax compliance is linked to social capital and trust.3Scholz and Lubell (1998b) find, linking survey data on attitudes with micro data on tax compliance for the U.S., that trust in others increases tax compliance. La Porta et al. (1997) consider trust in large organizations, instrumenting a survey measure of tax compliance as a proxy for trust by the degree of ethnolinguistic fractionalization, without discussing their finding.4 This paper provides a framework for why ethnic fractionalization, through trust, matters for tax compliance. The resulting hypothesis is tested using as the dependent variable actual estimates of the size of the informal sector, rather than the qualitative survey measure employed by La Porta et al. (1997). Finally, the empirical analysis employs a wider set of control variables in a larger sample. The next section presents the theory and derives the testable hypothesis, and section three describes the empirical model and the data. Section four presents econometric evidence that confirms the main hypothesis at the macro level, and section five concludes. 2. Voluntary tax compliance and ethnic divisions: A Theoretical Framework The economic literature on tax evasion has largely been built on Becker (1968) economics-of-crime approach. Pioneered by Allingham and Sandmo (1972), the possibility of tax evasion is modelled as essentially adding to the tax payer’s optimization problem a risky asset, the return of which is determined by tax rates, fines and the risks of detection. In accordance with the theory, it has been established empirically that tax payers respond to incentives such as changes in the audit probabilities. However, two (related) puzzles for the theory remain. First, tax collection based strictly on enforcement in environments where the return on taxes is very low is remarkably ineffective.5 Second, given the existing degree of enforcement in industrial countries, characterized by low audit probabilities, tax compliance is remarkably high.6 A possible solution to the puzzles is the existence of voluntary tax compliance. Observed already by Machiavelli, “no prince can govern long unless most citizens willingly obey the laws of the land.”7 Tax payers who voluntarily comply choose to pay taxes; that is, they do not (actively) engage in non-compliance. Goode (p. 213), summarizing his experience as advisor to the IMF, made the case for voluntary tax compliance half a century ago, arguing that “a large degree of voluntary compliance on the part of tax payers [is a] requirement for satisfactory income taxation.” Though empirical evidence on the prevalence of voluntary tax compliance is non-existent, and experimental evidence is scarce, many, such as Slemrod, 1998 and Slemrod, 2003, remain convinced of its importance. It is important, therefore, to identify the causes of voluntary tax compliance or the lack thereof. Based on her historical and comparative work on tax evasion, (Levi, 1988, p. 53) argues that voluntary compliance will occur “only when taxpayers have confidence that (1) rulers will keep their bargain and (2) the other constituents will keep theirs. Taxpayers are strategic actors who will cooperate only when they can expect others to cooperate as well. The compliance of each depends on the compliance of the others. No one prefers to be a “sucker.” Following Levi’s classification, voluntary tax compliance is influenced by two (implicit) social contracts. The horizontal contract concerns the perceived fairness of the tax payment: If people expect others to evade or avoid taxes, they will try to do so themselves, whereas people expecting others to comply will comply themselves. The vertical contract concerns what has been called the quid pro quo of taxation: do tax payers get “sufficient” public goods in exchange for taxes paid?8 And do they get the public goods mix they prefer? In the Allingham-Sandmo model, taxes finance an exogenous revenue requirement, and tax payers are not concerned with what they get in exchange for taxes paid; hence, there is no quid pro quo. In the following, I argue that ethnic fractionalization can affect the size of the informal sector through both the horizontal and the vertical contracts. 2.1. Ethnic fractionalization and the horizontal contract Experimental and observational evidence from social and political psychology has established that (a) the level of trust and the degree of trustworthiness decreases with ethnic diversity Zucker, 1986, Tyler, 1998, Glaeser et al., 2000 and Alesina and La Ferrara, 2002, and that (b) tax compliance increases with trust Scholz and Pinney, 1995, Scholz and Lubell, 1998a, Scholz and Lubell, 1998b and Scholz, 1998. The argument given is that people are willing to comply if they know that everyone else complies. If people do not trust others to comply, they choose to evade taxes themselves. Tax payer surveys and experimental evidence support this. For example, Laurin (1986) reports survey evidence from Sweden that people who think that other tax payers underreport are more likely to do so themselves.9 What is the theoretical basis for such a behavior? Tax compliance is typically modelled as a prisoner’s dilemma (PD) game. In PDs, non-compliance is a strictly dominant strategy; the role of tax enforcement, then, is to promote compliance by decreasing the gains from non-compliance. However, as noted by Andreoni et al., the expected penalties found in actual tax enforcement systems are not large enough to make compliance a dominant strategy. Therefore, accounting for tax enforcement still leaves players in a PD. But people pay taxes anyway. The fact that people pay taxes, even though the analysis framed as a PD says they should not, fits a general conclusion of experimental game theory: that people cooperate more than predicted by the theory (Ledyard, 1995). Two recent papers, Fehr and Schmidt (1999) and Bolton and Ockenfels (2000), attempt to solve this inconsistency by modeling individuals as having preferences not only over absolute monetary outcomes, but also over equity in outcomes, and they show that such preferences can support observed behavior in a wide range of games, including public goods games. When people have equity concerns, they compare their resulting payoff with that of others in a similar situation, and incur disutility from over- or undercontributing, a feature confirmed experimentally (Loewenstein et al., 1989). In the public goods game characterized by Bolton and Ockenfels (p. 182), players’ contributions are increasing in their belief that others contribute, which is exactly the behavior postulated by Levi, as noted above. If they encounter other ethnic groups with a greater probability, and they trust such groups less, the expected contribution by others is lower.10 Another avenue through which tax compliance could be affected by ethnic fractionalization is social sanctions or norms. Olson (1965) argues that social sanctions can provide ‘selective incentives,’ facilitating public goods provision and, therefore, as noted by Roth et al., social stigma associated with non-compliance could improve compliance. For social sanctions to be effective, however, it is important that individuals are affected by such sanctions. To the extent that individuals are affected mainly by social sanctions exercised by their own ethnic group, such sanctions will not be as effective in ethnically fractionalized communities as in more homogenous communities; see Miguel and Gugerty (2005) for an application to school financing in Kenya. 2.2. Ethnic fractionalization and the vertical contract According to the quid pro quo argument, tax compliance depends in part on tax payers receiving “sufficient” public goods in return for taxes paid. Thus, if the (perceived) rate of transformation from revenue to (favored) public goods is low, tax payers will feel that the state has not kept its bargain, with voluntary tax compliance deteriorating as a result. There can be two reasons for this: (a) preference divergence/favoritism; and (b) corruption. It has been established empirically that public goods provision and participation is lower in ethnically fractionalized societies; see Alesina et al. (1999) and Alesina and La Ferrara (2000) for US municipal level evidence, and Miguel and Gugerty for micro evidence from school financing in Kenya. The possible reasons suggested are that different ethnic groups may prefer different public goods mixes and, further, that people may not want to contribute to public goods benefiting other ethnic groups.11Alesina et al. (1999) argue that the lower provision is determined through the political process by tax payers exercising their voice option. However, another way of expressing discontent is the exit option: If tax payers feel that the public goods mix provided is very different from what they would have preferred themselves, or that it benefits people with whom the taxpayer does not identify, or that the rate of transformation is low due to pervasive corruption with much rents being appropriated by public officials and politicians, they will feel the attractiveness of the quid pro quo contract diminished, which could lead to lower voluntary compliance.12 This argument is closely related to political legitimacy. As noted by (Alt, 1983, p. 185,) “the legitimacy of a tax is that of the state that levies and collects it.” Therefore, if the state is perceived to have low legitimacy, for example due to non-representativeness, this can lead into a downward spiral of non-compliance. Rabushka and Shepsle (1972), in their analysis of multi-ethnic societies, argue that ethnic salience can result in “ethnicization of collectively provided goods,” (p. 84) in the sense that the political process allocates excludable public goods and transfers based on ethnic characteristics (favoritism). The failure of the state to insure nonexcludeability make individuals turn to their ethnic communities, as a sort of alternative statehoods, for the provision of public goods and this process can initiate a vicious cycle in which ethnic communalism “breeds attitudes of illegitimacy, which in turn reduce the effectiveness of the state, and further intensify attitudes of illegitimacy” (p. 85).

The analysis of this paper underscores the importance of social structure and trust on economic outcomes, in this case the size of the informal sector and the degree of tax compliance. The empirical results presented above suggest a statistically significant and economic sizeable effect of ethnic heterogeneity: In the preferred specification, (4) above, an increase of one-standard deviation in ethnic fractionalization increases the size of the informal sector by 4.7 percentage points, comparable in magnitude to the effects of a one-standard deviation increase in corruption (7.1 percentage points) and in the urban population (−5.4 percentage points). Further, the estimated effect of a one-standard deviation increase in ethnic fractionalization was reasonably stable across specifications, ranging from 4.0 to 5.7 percentage points. Future work should aim at extending the sample of countries to include countries with more variation. For example, we conjecture that including estimates from more African countries would support the hypothesis put forward here: African countries are ethnically very heterogenous and, at the anecdotal level, characterized by large informal sectors. Another possibility would be to use as a dependent variable the share of the informal labor force in percent of total labor force, as this measure is available for a different set of countries. A crucial problem that more data will not solve, however, follows from the nature of the informal sector data; this data is (perhaps very) imprecisely estimated, and therefore one should be very cautious in making results such as these the basis for policy intervention or policy recommendation. Rather, the goal of the present analysis is to emphasize, once again, the importance of social and ethnic heterogeneity on public sector outcomes. As for economic development, the results of this paper contribute to the bad news associated with ethnic fractionalization. While Collier, in a careful review of the evidence, concludes that the difficulties facing ethnically diverse societies have been greatly exaggerated, he concedes that regarding “the public sector there is evidence that ethnically differentiated organizations encounter problems” (p. 154). In particular, the results of Easterly and Levine (1997) and Alesina et al. (1999) suggest that public expenditures are lower, and less efficiently allocated, in fractionalized communities. The results of this paper confirm the problems of the public sector in ethnically fractionalized societies, but suggest that such problems exist also on the revenue side of the public budget. This suggests that trying to improve the public sector in fractionalized societies by improving allocative efficiency on the expenditure side (for example, as suggested by Collier, by reducing ethnic employment patronage) could, while securing a better vertical contract, be insufficient to secure a better public sector if problems of the horizontal contract persist. In a broader context, Collier suggests that one should simply accept the fact that the public sector is relatively less effective in diverse societies than in more homogenous societies, so that the role of the public sector relative to the private sector should be redefined in such societies. This does not necessarily mean that public goods will not be provided, as it could to some extent be provided through ethnic groups, but perhaps that public goods will be provided less effectively and in smaller scale. The results also provide another, stronger case for decentralization in ethnically divided societies. Decentralization into more ethnically homogenous regions is often offered as a way of reducing tensions and avoiding political gridlock in fractionalized societies, and the results provided in this paper suggests that decentralization should include also tax collection and, possibly, tax policy if centralized taxation is not deemed fair by citizens. Finally, the fact that ethnically fractionalized societies have larger informal sectors could also be part of an explanation of the fact that more fractionalized societies have less public goods provision in percent of GDP. If true income is more underestimated in fractionalized communities, then public goods provision relative to true income (rather than reported income) may turn out not to be very different between more and less heterogenous communities. The method adopted in this paper has been to apply results obtained through experiments in the laboratory to real world macro level data, bridging results from political and social psychology with public finance. Interesting avenues for future work could be to use such methods for a broader range of public policy problems.