The Origins of Voluntary Compliance: Attitudes Toward Taxation in Urban Nigeria

Abstract

Voluntary compliance is an important aspect of strong tax regimes, but there is limited understanding of how social norms favoring compliance emerge. Using novel data from urban Nigeria, where tax enforcement is weak, this article shows that individuals with a positive experience of state services delivery are more likely to express belief in an unconditioned citizen obligation to pay tax. In addition to support for this fiscal exchange mechanism, social context is consequential. Where individuals have access to community-provided goods, which may substitute for effective state services provision, they are less likely to adopt pro-compliance norms. Finally, the article shows that norm adoption increases tax payment. These findings have broad implications for literatures on state formation, taxation and public goods provision.

Full text

The Origins of Voluntary Compliance:
Attitudes toward Taxation in Urban Nigeria
Forthcoming at British Journal of Political Science
Cristina Bodea
Department of Political Science
Michigan State University
Email: bodeaana@msu.edu
Adrienne LeBas
Department of Government
American University
Email: adrienne.lebas@gmail.com
** This project was funded by a grant from the Department for International Development (DfID)
of the United Kingdom. The authors thank Mike Bratton, Melani Cammett, Jeff Colgan, Jan
Leighley, John Miller, Emmanuel Teitelbaum, Matt Winters, the comparative politics group at
Michigan State University, and audiences at American University, George Washington University,
McGill University, and Northwestern University for comments on earlier drafts. The authors also
extend thanks to Oxford University’s Centre for the Study of African Economies, the Nigerian
Governors Forum, CLEEN, and participants at our October 2012 workshop in Abuja, Nigeria.

1
Abstract
Voluntary compliance is an important aspect of strong tax regimes, but there remains
limited understanding of how social norms favoring compliance emerge. Using novel
data from urban Nigeria, where tax enforcement is weak, we show that individuals
with positive experience of state services delivery are more likely to express belief in
an unconditioned citizen obligation to pay tax. In addition to support for this fiscal
exchange mechanism, we find that social context matters. Where individuals have
access to community-provided goods, which may substitute for effective state
services provision, they are less likely to adopt pro-compliance norms. Finally, we
show that norm adoption actually increases tax payment. The paper has broad
implications for literatures on state formation, taxation, and public goods provision.

2
How do states convince citizens to pay tax? Rather than focusing on enforcement, most
accounts emphasize voluntary or “quasi-voluntary” compliance as an essential element of
successful tax regimes.1 In these settings, individuals evade taxes far less commonly than would be
expected if decisions were driven by fear of state sanctioning, and they express an intrinsic
motivation to pay tax.2 This intrinsic motivation, which scholars term tax morale, is often framed as
citizen obligation or as civic duty. Where tax morale is weak, the cost of enforcing tax compliance
is much higher; widespread evasion and organized popular resistance to taxation are more likely to
occur.3 Norms favoring voluntary compliance are therefore important, but there remains limited
understanding of how these norms come into being. The bulk of the literature on state-building and
taxation presumes that social contracts emerge slowly out of multiple rounds of conditional
exchange between states and citizens.4 From this perspective, it is repeated rounds of bargaining –
not the expansion of enforcement – that produces quasi-voluntary compliance. But how do pro-
compliance norms become general across a population? Does the provision of public goods simply
produce a quid pro quo exchange with taxpayers, or can it result in deeper and less contingent
norms of citizen obligation fairly quickly? And are there features of social context that either
facilitate or impede this process?
In explaining how conditional compliance becomes normative commitment, the emphasis
has long been on fiscal exchange, which is the relationship between receipt of public goods and
increased willingness to pay tax. But the broader context surrounding exchange is important.
Research has shown that higher levels of institutional trust, involvement in participatory decision-
1 Levi 1988.
2 Alm, McClelland, and Schulze 1992; Andreaoni et al. 1998; Feld and Frey 2002.
3 Levi 1988; Fjeldstad 2001; Bernstein and Lu 2003.
4 Bates and Lien 1985; Levi 1988; Tilly 1985; Tilly 1990; Lieberman 2003.

3
making, and national identity all strengthen societal norms against tax evasion.5 In this paper, we
broaden this attention to context to include the social landscape surrounding individuals. We argue
that two factors give definite predictions about individuals’ attitudes toward taxation: experience of
state services provision (public goods) and the availability of community-provided substitutes for
state services (club goods). Thus, our theoretical starting point is the standard exchange-driven
narrative of social contract emergence: individuals more readily adopt pro-compliance norms when
they have a reasonable expectation that the state will deliver services in exchange for taxes.6 More
novel is this paper’s argument that tax morale is affected by the existence of substitutes for state
services. Where communities are able to effectively engage in “self-help” provision of club goods,
individuals with access to these community-provided goods are less likely to adopt norms favoring
tax compliance, as they are less likely to turn to the state to provide security, contract enforcement,
or other basic services. These two arguments might be seen as consistent with capturing both supply
and demand dimensions of state-society bargaining. Where either the state supply of services or
societal demand for such services is weak, we would be less likely to find norms supporting
voluntary compliance with state tax demands.
To test our hypotheses, we use new public opinion data from urban Nigeria to investigate
why individuals adopt attitudes about taxation that we view as consistent with higher levels of tax
morale. Ordered logit and multilevel estimations provide empirical support for our posited drivers
of individual tax morale. First, individuals who report receipt of public goods and those who are
more satisfied with the government’s use of tax revenue are more likely to express support for an
unconditioned citizen obligation to pay tax. Second, we test our innovative and more controversial
intuition regarding the effect of club goods using several proxies, and we account for possible
5 Feld and Frey 2002; Alm et al 2006; Torgler 2007; d’Arcy 2011.
6 Alm, McClelland, and Schulze1992; Timmons 2005.

4
reverse causality. Overall, we find support for the notion that greater access to community-provided
or club goods has a depressive effect on tax morale. Individuals who view group relations in their
communities as harmonious, which would presumably make collective action easier and more
effective, express lower degrees of tax morale. Similarly, individuals with access to club goods,
such as those who participate in savings clubs or rely on vigilante groups for security, express lower
degrees of tax morale. It is not the involvement with these groups in particular that produces a
weaker orientation toward the state; instead, we argue that reported reliance on these groups serves
as an indicator of a local landscape that favors self-help collective action by communities or groups
of individuals.
Do norms of citizen obligation to pay tax translate into actual tax compliance? The
literature commonly assumes that individual tax morale generates actual tax compliance, but
empirical support for this relationship is sparse.7 Our data allow us to explore the effects of pro-
compliance norms, state services delivery, and the availability of club goods on individual tax
payment. Bivariate logit models show that unconditioned belief in a citizen obligation to pay tax has
a significant, positive effect on actual tax payment, and this norm mediates the effect of public
goods provision on tax payment. In other words, the delivery of state services does not
independently produce tax compliance; instead, receipt of public goods generates higher degrees of
tax morale, which in turn boosts the likelihood that an individual will pay tax.
Our paper makes several contributions to the literature. First, we suggest a new avenue of
inquiry for scholars interested in attitudes toward taxation and state authority. The literature on
state-society relations has often relied on a binary opposition between clientelistic and
7 Torgler 2007.

5
programmatic linkage, commonly seen as rivals.8 In an attempt to move beyond the debate over
clientelistic versus programmatic linkage, this paper focuses on a contextual factor often neglected
in the literature: the presence of community-level goods that sometimes serve as substitutes for state
services. Non-state provision of needed services is particularly important where state penetration is
historically very weak, as is the case across the developing world and, certainly, in urban Nigeria.
Because we find that individuals with access to club goods are less likely to adopt pro-compliance
norms, our findings suggest reasons to be skeptical that social capital necessarily improves
government performance.9 In Putnam’s Italy, rotating credit associations and other trust-based
community institutions can provide information to government institutions and make them more
efficient. These same types of communities, however, when faced with a weak or predatory state,
may serve as effective bulwarks against state penetration.
Secondly, our findings expand the evidentiary base for the fiscal exchange hypothesis to a
region of the world for which evidence remains sparse. This paper increases our confidence that
government delivery boosts tax morale across a range of environments, including those in which
distrust of state institutions and a past history of predatory state rule might be expected to
undermine the strength of this mechanism. Thirdly, our tax payment findings suggest that public
goods provision does not generate compliance on its own, as a purely exchange-based account of
tax contracts would suggest. Instead, state services provision generates higher levels of tax morale,
which in turn makes individuals more likely to actually pay tax. This is presumably good news for
states, as we would expect a normative pathway to compliance to be more durable than one based
on simple assessments of government performance.
8 For example, Wantchekon 2003, Kitschelt 2000.
9 Putnam 1993.

6
In the next section, we lay out how the literature currently addresses the emergence of tax
regimes and the place of individual attitudes in that literature. The next section explores the
analytical distinction between public, private, and club goods, from which we derive our working
hypotheses. The fourth section describes the data, measures and research design. This section also
explains why urban Nigeria is a good context in which to examine the question of tax norm
emergence. Finally, we turn to the discussion of results and conclude.
STATE FORMATION, TAXATION, & INDIVIDUAL ATTITUDES
In studies of the formation of modern Western European states, the development of tax
collection capacity proceeds in tandem with the increasing societal penetration of states. The
literature stresses the importance of iterated bargaining between states and societal actors.10 The
outcome of this bargaining is the establishment of a mutually beneficial social contract, in which
citizens defer to the authority of the state, pay taxes, and receive public goods. Studies of tax
regimes in other settings have often adopted this framework, focusing on the evolution of
consensual relations between tax-seeking states and tax-paying dominant classes.11 At base, much
of the historical institutionalist literature posits a simple exchange: individuals agree to pay taxes,
and states agree to provide public goods and, possibly, to include citizens in state decision-making.
One problem with this approach is its applicability to contexts where states are not reliant on
tax revenues. For instance, in sub-Saharan Africa, few governments are dependent on direct
taxation to finance their budgets, and negotiation around taxation has not been central to state-
society relations. In order to explain weak state accountability, scholars have tended to emphasize
macro-structural factors, such as ethnic diversity, group inequality, or the availability of alternative
10 Bates and Lien 1985; Levi 1988; Tilly 1990.
11 Lieberman 2003; Jones Luong and Weinthal 2004.

7
revenue sources.12 For instance, the resource curse literature suggests that state-society bargaining
does not occur if states rely on natural resource rents rather than tax revenues. In these contexts,
states have few incentives to deliver social services or develop ties of accountability with their
populations. 13 Structural features at the national level, such as resource profiles,affect states’
motivations to invest in direct taxation, but state capacity to collect tax – once that strategy is
chosen – can be powerfully influenced by pre-existing social institutions. Particularly in the early
stages of state-building, these institutions mediate the effects of state strategies, possibly leading to
significant geographic variation in state capacity within a single state territory.14 For instance,
according to Boone, where pre-existing local authority structures are centralized and have their own
revenue sources, states must build stronger bureaucratic capacity and deliver more public goods
than they do in areas where they do not face these kinds of rivals. Drawing on this central insight,
we investigate the effects of meso-level organizations – i.e., the social institutions that coordinate
club goods provision – on individual tax morale.
The paper argues that neither exchange nor coercion are sufficient for explaining tax regime
emergence, but we depart from the above literature in stressing the importance of individual
attitudes. Tax morale scholars have established that attitudes play a primary role in boosting
compliance with taxation, but these scholars are largely reliant on evidence from the developed
world, where compliance and state capacity are already high. Like the literature discussed above,
the tax morale literature has relied heavily on exchange mechanisms to explain the emergence of
norms favoring voluntary compliance. It has also acknowledged the effect of contextual factors on
12 Alesina et al 1999; Baldwin and Huber 2010; Ross 2012.
13 Ross 2012.
14 Boone 2003.

8
tax morale.15 For instance, Alm et al. show that longitudinal changes in levels of trust in the Russian
state partly explains shifts in tax morale.16 Both Torgler and Feld and Frey find that political
participation in decision-making improves tax morale and the treatment of citizens by tax
authorities.17 In a rare study of tax compliance in sub-Saharan Africa, Fjeldstad and Semboja find
that social influence exerts at least as strong an effect as the perceived likelihood of prosecution.18
Our paper fills an important hole by examining how context and public goods delivery affect norms
of citizen obligation in a context of low state capacity and high levels of tax evasion.
INDIVIDUAL ATTITUDES TOWARD TAXATION
To develop hypotheses about individual tax morale, we consider research on risk-pooling,
public goods investment, and clientelism in the developing world. These distinct literatures suggest
the importance of three different kinds of goods that individuals may receive from their
environments: public, club, and private goods. In deriving testable hypotheses, we focus on the first
two types of goods (club and public), since the literature on what we term private goods does not
generate clear expectations regarding effects on our dependent variable, as we will discuss.
Public goods. Our starting hypothesis is the straight-forward fiscal exchange thesis that
underlies much of the work on taxation.19 According to this mechanism, individuals who directly
benefit from state services are more likely to pay taxes and to support the state’s right to tax. This
may be because individuals expect tax contributions to be returned to them via government
spending. Or public goods provision may create greater trust in government institutions and greater
belief in the procedural fairness of government decisions, both of which scholars have found to be
15 Alm, Jackson, and McKee 1992; Scholz and Lubell 1998; Frey and Torgler 2007.
16 Alm et al. 2006.
17 Torgler 2007; Feld and Frey 2002.
18 Fjeldstad and Semboja 2001.
19 Timmons 2005; for Africa, Fjeldstad 2004; d’Arcy 2011.

9
positively associated with tax morale20 and with payment.21 Evaluations of government
performance can be affected by individuals’ political allegiances. Those who support the sitting
government express more positive evaluations of government performance, and their evaluations
are less likely to be affected by evidence of corruption or mismanagement.22 Despite these potential
complications, the logic of fiscal exchange is straight-forward: Individuals with positive evaluations
of government performance are more likely to support the state’s right to tax.(H1)
Club goods. Our second mechanism focuses on the link between tax morale and internal
community attributes, specifically the availability of group-delimited or club goods. In Nigeria and
in other low-capacity states, public goods are provided at suboptimal levels, and communities and
individuals often engage in “self-help” provision of such goods. These efforts range from the
forging of somewhat amorphous trade and trust networks to the collective provision of schools and
roads to the creation of concrete non-state agents of order, such as ethnic militias and vigilante
groups. Access to these goods is almost always limited to those who contribute to their creation and
maintenance.
Why are club goods provided in some communities to a greater extent than in others? The
literature on risk-pooling in the developing world, reaching back to James Scott, suggests that
individuals are often embedded in networks based on norms of reciprocity and more tangible
enforcement mechanisms.23 Co-ethnicity is a strong basis for the creation and maintenance of these
kinds of networks,24 as are religious co-affiliation and other kinds of social ties.25 Via income
20 Alm et al 2006.
21 Feld and Frey 2002; Fjeldstad 2004.
22 Anderson and Tverdova 2003; Chang and Kerr 2009.
23 Scott 1976; Fafchamps 1992.
24 Grimard 1997.
25 Grief 1994; Cox and Jimenez 1998; Cassar and Wydick 2010.

10
transfers within the group, these networks serve as valuable protection against income shocks for
the poor in both urban and rural settings. But the benefits of socially-embedded networks extend
beyond income insurance for the poor. They can be powerful resources for organizing and policing
collective action. Thus, it is easier to enforce participation and contributions to community goods
where there exist community cohesion and norms of reciprocity.26 Where communities are instead
sharply divided, individuals may shy away from investing in community goods from which rivals
cannot be excluded.27 Stated differently, individuals are more likely to construct closed networks of
exchange where information is poor, as we might presume it to be in communities characterized by
diversity, conflict, or distrust.28
Given the above, there are two mechanisms through which club or community-provided
goods may shape tax morale. First, effective club goods provision may crowd out demand for state-
provided services. Community-provided insurance or saving would reduce demand on the state to
provide a social safety net; similarly, vigilante and community policing may reduce citizen reliance
on state police. Secondly, negative experiences with the state and positive experiences with
investment in club goods can reinforce one another. Where public officials prioritize “primordial
publics” over the civic realm, as scholars suggest is the case in Nigeria, state corruption and
predation become commonplace.29 This would in turn reinforce individuals’ decisions to invest in
alternative authority structures, such as ethnic groups, market associations, or even militias. Our
second hypothesis flows from the posited inverse relationship between a community’s success at
club goods provision and its residents’ willingness to invest in the state: Individuals are less likely
26 Habayarimana et al 2007; 2009.
27 Miguel 2004.
28 Geertz 1978; Grief 2004.
29 Ekeh 1975; Lewis 1996.

11
to support the state’s right to tax if they live in communities that effectively engage in “self-help”
provision of club goods.( H2)
Private goods. In electoral regimes, core or electorally crucial constituencies are often
granted greater access to state services and state spending.30 Though this phenomenon occurs in
much of the world, targeted redistribution of state resources and clientelism is even more central in
Africa. Here, differential public investment is often organized on ethnic lines, and the effects of
favoritism toward ruling coalition members can be seen in the distribution of roads, education and
health, and electrification.31 Clientelism is a winning strategy for politicians, and voters often expect
– and even demand -- targeted redistribution over more general delivery of public goods.32 Thus, for
many African voters and politicians, politics involves the transformation of public finances into
private goods. Membership in ethnic or partisan winning coalitions is often treated as an important
determinant of individuals’ access to resources, voting preferences, and political participation.
Can we link membership in winning coalitions to attitudes toward taxation? Would likely
“winners” in the game of clientelistic redistribution be more likely to support a citizen obligation to
pay taxes? Clientelistic redistribution can explain some group-level differences in government
services, but these effects are rarely uniform across types of services even within a single country.33
Further, as Wantchekon points out, even where clientelistic benefits exist, they rarely extend to all
individuals within a coalition.34 Looking at taxation explicitly, there is a great deal of ambiguity
about the effect of membership in ruling coalitions. On the one hand, members of ruling coalitions
30 Dixit and Londegran 1996; Golden and Picci 2008.
31 Kramon and Posner 2013.
32 Wantchekon 2003; Lindberg 2003.
33 Kramon and Posner (2013) point out that members of presidential coalitions may receive benefits in terms
of one public good, such as electrification, even as they are disadvantaged in another.
34 Wantchekon 2003.

12
might be more likely to support an expansion of the state and tax revenue, as they are likely to
benefit from heavier state coffers. But these same individuals might think twice if the tax burden
falls disproportionately on their shoulders, as Kasara suggests it does.35 She finds that African states
more heavily tax their core constituencies – in this case, presidents’ co-ethnics – because they have
greater extractive capacity when dealing with these groups. Greater tax burdens, however, need not
be inconsistent with willingness to pay. In his study of tax policy in South Africa and Brazil,
Lieberman shows that individuals willingly pay taxes where they see state policy as benefiting a
group to which they belong, even if they themselves will not be beneficiaries.36 Clientelism’s effect
on tax morale therefore seems indeterminate. The beneficiaries of clientelistic redistribution may
wish others to pay taxes, as expanded tax revenue may benefit them, but they may reject obligation
when it comes to their own payment. If members of ruling coalitions are disproportionately taxed,
as Kasara suggests, the perceived unfairness of this tax burden could conceivably lower tax morale.
Yet Lieberman’s suggestion of voluntary payment by ruling coalition members would also be
consistent with Kasara’s findings of higher tax payment within the ruling coalition. More broadly,
clientelistic linkage has an ambiguous effect on individuals’ attitudes toward state authority. On the
one hand, clientelistic linkage can be a basis for state legitimacy and trust in state institutions,37
which is often found to increase tax morale. On the other hand, clientelistic linkage may undermine
civic attitudes and reinforce the notion of governance as conditional exchange between particular
governments and particular constituencies.38 This runs counter to the idea of non-conditioned
citizen obligation to pay tax, which is the focus of this paper. Though the empirical sections below
35 Kasara 2007.
36 Lieberman 2003.
37 Kitschelt 2003.
38 Letki 2006.

13
examine this dimension more carefully, we do not view ruling coalition membership as a
convincing explanatory variable due to the inconsistency of its hypothesized effects.
DATA AND EMPIRICAL DESIGN
We use data from an independent academic survey on taxation and political attitudes
conducted in eleven Nigerian cities in December 2010 (N=2750), more details of which can be
found on the author’s website. City-representative samples were selected using a stratified,
clustered sampling procedure; individual informants were selected using a random-walk protocol
from GIS-selected starting points. To ensure adequate female respondent representation, especially
in the Muslim north, gender parity was imposed. Cities are drawn from each of the three zones into
which Nigeria is often divided, and they vary in terms of demographics, politics, and riot-proneness.
There are three northern cities (Bauchi, Kano, Sokoto), three middle belt cities (Jos, Kaduna, Lafia),
and five cities in southeastern and southwestern Nigeria (Ibadan, Lagos, Aba, Enugu, and Onitsha).
Nigeria is a valuable context in which to examine the initial drivers of tax morale. The
federal and state governments’ dependence on oil revenues has generated a classic resource curse
dynamic, in which both government accountability and tax collection infrastructure are low.39 The
oil boom of the late 1970s generated an increasingly corrupt and predatory state in Nigeria, and
successive authoritarian governments did not invest in rule of law or services delivery. State neglect
fostered a large informal economy and a do-it-yourself attitude toward infrastructure and contract
enforcement on the part of communities and individuals. In some cities, informal vigilante groups
have often been the most visible agents of order; in others, communal riots over indigeneity and
religious differences are common. Despite the return to multiparty elections in 1999, distrust of
39 Lewis 1996; Sala-i-Martin and Subramanian 2003.

14
state institutions remains widespread.40 This is an environment in which tax morale would be
expected to be low.
Though this disconnection between state and society is general in Nigeria, the federal
system allows for a degree of variation across states. In terms of formal taxing powers, the federal
government retains control over the taxes that are least costly to collect, such as customs and excise.
State governments are formally empowered to collect personal income tax and other levies, and
some states collect council rates and taxes on behalf of local governments.41 In terms of services,
there is a great deal of devolution from the federal level, and the bulk of schools, roads, and other
public goods are provided at the state and local level. Overall, tax collection infrastructure at both
federal and state levels is weak in Nigeria, even by African standards. Many state governments
depend on federal oil transfers, and they consequently underinvest in tax collection capacity. Over
the past decade, however, some Nigerian states have undertaken reforms aimed at increasing
internally-generated revenue and creating a bureaucratic presence in their urban centers. This
creates interesting variation across the cities sampled. For instance, in a surprising test of the
durability of the resource curse, federal transfers to Lagos State were suspended in 2003 due to a
bureaucratic dispute, and the state government responded by stating that it would fund itself via tax
revenues. The state government increased its tax revenue more than ten fold between 2003 and
2007. Dramatic improvements in revenue collection coincided with visible expansion of social
services, even in slums. In areas where states have started this process of reform and delivery, we
would expect higher levels of tax morale, even if the overall national context remains deleterious.
Tax Morale. Our measure of tax morale captures individuals’ attitudes toward a citizen obligation
to pay tax. Individuals are asked to grade their agreement with the following alternative statements:
40 Lewis 2006.
41 E.g., “Why state collects rates on behalf of LGs,” Daily Independent (Lagos), 9/9/2013.

15
“citizens should always pay their taxes, even if they disagree with the government” (statement A);
or “citizens should only pay taxes if they believe in the government” (statement B). Individuals can
respond by strongly agreeing with A, agreeing with A, agreeing with B, or strongly agreeing with
B. They can also choose to agree with neither statement (4.4 percent of the sample). Based on this
question we create a single scale ranging from 1 to 4, with larger numbers showing more agreement
with the notion of an unconditioned citizen obligation to pay tax (statement A).
Is this a valid measure of an individual’s intrinsic motivation to pay tax? The tax morale
literature uses a range of questions to capture the same concept.42 Many of these questions ask
respondents to express the strength of their agreement with a single normative statement, or they
ask respondents to rate the justifiability of tax evasion on particular grounds. Importantly, there is
no objective measure of tax morale. All existing measures imperfectly capture the concept of
intrinsic motivation to pay tax, but our measure is both consistent with other authors’ use and also
avoids some potential pitfalls. For instance, one widely used measure of tax morale is a World
Values Survey (WVS) question that asks respondents to rate the justifiability of “cheating on
taxes.”43 This question style, very common in the literature, both runs the risk of normative priming
and also may be unsuitable in low-enforcement settings. In these settings, tax evasion rarely
requires active “cheating” so much as it does avoidance, and a cheating-based question may
produce implausibly high level of tax morale. For instance, according to the WVS, a high
percentage of Nigerians – 63.3 percent – say that cheating on taxes is never justified (with a further
14.71 percent responding with a 2 out of 10, with 10 being always justifiable). This is a higher level
42 Torgler 2007.
43 The World Values Survey asks how often the following can be justified: ‘Cheating on taxes if you have the
chance’. Respondents are required to rate justifiability on a 10-point scale, from 10 being “always justifiable”
to 1 “never justifiable.”

16
of tax morale than in most Latin American countries and is comparable to or exceeds levels in high-
income industrialized countries.44 The WVS seems to significantly over-report levels of tax morale
in Nigeria given known societal attitudes toward taxation and actual evasion behavior, which is well
over 50 percent of the population. We follow the practice of other scholars in using a question that
probes for the acceptability of particular grounds for non-payment (here, a citizen’s disagreement or
non-belief in the government).
One concern with our choice of dependent variable may be the extent to which it taps into
the overall support for the sitting government versus a deeper long-term orientation toward the state.
As we explained earlier, the literature on clientelism and targeted redistribution does not suggest a
single mechanism linking taxation and ruling coalition membership. But it is worthwhile to examine
the correlation between tax morale and these measures. We examine several measures of individual
support for or membership in ruling coalitions: (i) stated co-partisanship with the sitting president or
expressed intention to vote for him in the then-upcoming 2011 elections; (ii) shared partisanship
with the sitting state governor; (iii) shared ethnicity with the state governor and the president45; (iv)
expressed trust in one’s own governor or in the president. Univariate correlations between our
dependent variable and these measures are very low (under 0.1). Low correlations increase our
confidence that tax morale, as measured here, taps into a deeper orientation toward state authority
independent of support for or expected benefit from any particular administration.
In the analysis below, we retain the full 4-point scale of responses, ranging from “strongly
agree with statement B” (value 1) to “strongly agree with statement A” (value 4). Roughly 42
44 Torgler 2007: Chapter 6.
45 In the eleven states where our cities are located, there are three Hausa, two Yoruba, three Igbos, one Jukun,
and two governors from much smaller ethnic minorities. We have a small number of the President’s
coethnics in the dataset, given the small size of his ethnic group. We therefore include as co-ethnics Igbos,
another southeastern tribe, who are typically perceived as beneficiaries of his rule.

17
percent of the sample agrees with statement A to some extent. Patterns of responses, however, vary
across cities. Figure 1 shows city-level means for the variable, with bars representing 95%
confidence intervals around the means. The solid line represents the global sample mean.
Differences in means are significant across several cities, and particular cities (Lagos, Ibadan,
Bauchi, Jos) serve as interesting outliers. This variation does not seem associated with the
effectiveness of the tax collection apparatus, the capacity of which is low across Nigeria. There also
seems to be little correlation between a state’s level of budgetary tax dependence and either pro-tax
attitudes or tax compliance. For instance, both Lagos and Oyo States rely on internally-generated
tax revenue to fund state budgets to a greater extent than other states; however, they lie on opposite
ends of the spectrum in terms of both tax morale and residents’ reporting of tax payment. Among
states with intermediate levels of tax dependence, there is similar variation. If budgetary
dependence on taxation is correlated with state investment in tax collection, it does not seem likely
that variation in our measures is driven by differences in state collection capacity.
Nor is variation directly due to differences in state government performance. Lagosians have
higher tax morale than their counterparts in other cities, which is unsurprising given expansion of
services. Other outliers are more difficult to explain. Jos, for instance, is the most riot-prone of the
cities surveyed. During our survey, deadly conflict broke out in a village near Jos and residents
feared that riots would erupt within the city. Ibadan, on the other hand, is a fairly homogenous city,
with Yorubas composing 88 percent of the population, and it has never experienced ethnic or
religious riots. As Figure 1 shows, these two cities are strong outliers with regard to attitudes toward
taxation. The battle-scarred residents of Jos express much stronger approval of citizens’
unconditioned obligation to pay tax, while the residents of Ibadan have much lower mean tax
morale. Jos and Ibadan suggest that conflict and trust may work as suggested above: Less
conflictual communities may be able to solve collective action problems and engage in “self-help”

18
provision of club goods without the assistance of the state, while those who cannot are more likely
to look to the state as a potential solution.
Figure 1: Unconditioned Support for Citizen Obligation to Pay Tax across Nigerian Cities
22.2 2.4 2.6 2.8
Agreement with 'Citizens should always pay tax'
Aba
Bauchi
Enugu
Kaduna
Kano
Lagos
Lafia
Onitsha
Ibadan
Jos
Sokoto
Independent Variables
For our fiscal exchange hypothesis, we use two measures of tax-related government
performance. The first of these indicators is expressed satisfaction with how the current state
administration has spent tax revenue. Approximately 38 percent of our sample reported that they
were very or somewhat satisfied with the way that their state governments had used tax revenue.
This question directly prompts state government use of tax revenue, but this does not undermine its
fitness as a measure of satisfaction with overall government public goods delivery or revenue use.
Fees and levies are also sometimes termed “taxes” by ordinary Nigerians, and state governments are
the primary referent for most urban residents. There is a strong association between state governors
and elected local officials, as the former typically serve as political “godfathers” within their

19
territories. Further, as noted above, state governments are often involved in collection of local
government levies and provide most large public goods. Our second indicator of public goods
delivery is an additive measure of concrete public goods provision. Respondents are asked about
seven different public goods that may have been built in their communities by the current
government (hospitals, schools, police posts, water infrastructure, etc.). Using these responses, an
index of public goods provision was constructed with a maximum of seven and a minimum of zero
public goods provided. About 73 percent of the sample reported the construction or provision of at
least one new public good by the sitting government; 23 percent of the sample reported the
provision of four or more public goods.
To test our second hypothesis, we use several proxies for the ability of an individual’s
community to provide club goods to its own members. The first is an individual’s perceived level of
conflict within her own community, since we assume conflictual communities are less effective at
club goods provision. Individuals are asked to characterize the relationship between ethnic and
religious groups in their city. Responses are on a 5-point scale ranging from “very hostile” (coded
1) to “very cordial” (coded 5). In addition, we also code respondents’ answers to the question of
how often ethnic and religious conflict in their community had resulted in loss of life or destruction
of property. There is a high rate of refusal for this question (1000 observations lost), but the
correlation between perception of community relations and individual’s estimate of rioting is high at
0.56 for remaining observations.
Both of these remain imperfect measures of club goods provision, and we include more
direct measures of individual access to community goods. Thus, a third measure is a dummy
variable that takes the value 1 if the individual relies on informal savings clubs to safeguard the bulk
of their savings. States are rarely expected to provide banking services to their citizens, but we see
use of savings clubs as a rough proxy for community-level social capital. In poor countries,

20
individuals weather risks on their own without a state-provided security net, and savings clubs are
one of a variety of community solutions to contract enforcement and other services usually provided
by states. About 8% of our sample report using savings clubs, though that constitutes 22 percent of
those who actively save money. Another direct proxy for club goods availability is reported reliance
on non-state security forces. In Nigeria, as elsewhere in the developing world, communities often
rely on vigilante groups and informal militia for protection.46 We therefore include a measure of the
reliance of individuals on non-state actors for security provision: the variable takes on a value of 1
for individuals who report that they would find vigilante groups most helpful in solving a theft.
About 7 percent of our sample report that they would find vigilante groups most helpful (before
police, for instance). We expect weaker tax morale from individuals living in harmonious
communities, those who invest their savings with members of their social networks, and those who
see vigilante groups as potentially effective security enforcers.
Besides our key variables, we use standard controls, including respondent’s age, gender,
education, ethnicity, religion, religiosity, the respondent’s status as indigenous to state of residence,
a measure of food deprivation, and household asset ownership. Occupation was recoded to produce
usable “class-like” categories, such as informal sector employment, unskilled formal sector
employment, and white-collar professional status. We also include variables that capture different
aspects of an individual’s interaction with the state: a measure of an individual’s interest in politics
(a four-point range from “not at all interested” to “very interested”); a measure of whether
individuals reported meeting with a government official or representative in the prior year, ranging
from never (1) to often (4); and a binary measure of whether respondents reported paying a bribe to
or being asked for a bribe by a government official in the prior year.
46 Smith 2007; LeBas 2013.

21
Due to the ordinal nature of our dependent variable, the statistical analysis relies on ordered
logistic models.47 City fixed effects are included to control for unobservable city level
characteristics, but the results are similar when they are excluded. As a robustness check, we also
incorporate city-level factors via multi-level estimation techniques that explore nested data.48 This
latter approach recognizes the hierarchical structure of that data, correcting potentially
underestimated standard errors of regression coefficients. Overall, our results are robust across a
variety of model specifications.
RESULTS
In discussing the statistical estimations, we first assess the direction of effects and statistical
significance. We then compare the substantive effects of the variables identified as statistically
significant and explore robustness. Table 1 shows the results from ordered logit estimations, testing
our hypotheses separately and then combining the key independent variables in a full model (Model
3). Model 4 uses the alternative operationalization of community conflict. The dependent variable is
individual tax morale (with higher values indicating greater agreement with the statement that
“citizens should always pay their taxes, even if they disagree with the government”).
Several findings stand out. There is solid support for fiscal exchange. Both variables
operationalizing our public goods hypothesis have strong, significant effects on tax morale.
Increased satisfaction with the use of tax revenue and receipt of concrete public works (hospitals,
roads, etc) increases the likelihood that individuals express unconditioned support for a citizen
obligation to pay tax. Both indicators are statistically significant in Model 1 and in the full model
(Model 3). Secondly, we find robust support for the depressive effective of club goods supply on
individual tax morale. All of the measures used to capture community-level social capital receive
47 Results are robust when we collapse tax morale into a binary variable and use binary logit models.
48 Rabe-Hesketh and Skrondal 2008.

22
support from the estimations. Individuals who see communal relations in their local area as cordial
or very cordial are more likely to condition their support for a citizen’s obligation to pay tax. The
negative effect of harmonious community relations on tax morale is robust both in Model 2 and in
the full model. As an additional check on this measure, in Model 4 we use an individual’s
aggregated report of ethnic and religious clashes in her community. Individuals who report rioting
in their communities are more likely to express higher levels of tax morale. Being a member of a
savings club also has a negative effect on tax morale, as posited. The variable is just shy of the
conventional 10% confidence level in Model 2 (p=0.118) but is statistically significant in Model 3
and in a joint F test of all variables operationalizing H2. Respondents who express a willingness to
turn to vigilante groups in the event of a theft are also more likely to condition a citizen’s obligation
to pay tax (negative and statistically significant coefficients in Models 2 & 3).49 Overall, there is
strong support for the idea that individual access to club goods reduces tax morale.
Other measures of individual contact with the state also have statistically significant effects
on tax morale. Having paid or been solicited for a bribe is found to be strongly and positively
associated with tax morale, as is being a victim of state-directed evictions (reported in Table 2).
This corroborates the idea that any contact with the state -- even if that contact is predatory – might
increase an individual’s orientation toward the state as a locus for claims-making.50. Far less robust
is the effect of meeting with state representatives, which is signed in the expected direction
(meeting with a representative increases tax morale) but is not consistently statistically significant
across models.
49 The variable loses statistical significance in Model 4 (p=0.156). This is due to smaller sample size, since a
similar loss of significance occurs when we limit the Model 3 sample to match the availability of perceptions
of riot pervasiveness. All variables operationalizing H2 are statistically significant in a joint F test.
50 Tilly 1985; 1990.

23
Table 1. Ordered logit model of individual level determinants of attitudes toward taxation
Club Goods
Public Goods
Full model
Alternative
Specification
Model 1
Model 2
Model 3
Model 4
H1
Satisfaction with revenue spending
0.192
0.218
0.202
(0.049)***
(0.052)***
(0.062)***
Specific spending on public goods
in individual’s community
0.102
0.094
0.097
(0.024)***
(0.026)***
(0.032)***
H2
Community relations
-0.214
-0.216
(0.036)*** ,♦
(0.037)***
Member of saving club
-0.206
-0.242
-0.330
(0.132) ,♦
(0.137)*
(0.168)** ,♦
Agree with vigilante group
protection
-0.386
-0.438
-0.341
(0.146)*** ,♦
(0.154)***
(0.240) ,♦
Individual level assessment of riot
prevalence in community
0.062
(0.013)*** ,♦
Met with government official or
representative
0.067
0.138
0.105
0.011
(0.062)
(0.060)**
(0.066)
(0.077)
Direct contact with corruption
0.372
0.258
0.237
0.337
(0.094)***
(0.096)***
(0.098)**
(0.121)***
Interest in politics
-0.112
-0.053
-0.047
-0.029
(0.042)***
(0.042)
(0.044)
(0.056)
Adjusted R2
0.028
0.033
0.037
0.05
Observations
2,115
2,155
1,982
1,302
Substantive effects of key independent variables – Use Model 3 above
Strongly agree with statement
A: Citizens should always pay
their taxes, even if they disagree
with the government
Strongly agree with
statement B: Citizens should
only pay taxes if they believe
in the government
Predicted Probability (p)
Satisfaction with revenue spending
1 (not at all satisfied)
0.156
0.287
(range 1 to 4)
3 (somewhat satisfied)
0.218
0.209
% Change in p (X: 1 to 3)
35.85%
-25.17%
Specific spending on public goods
0 (none)
0.165
0.272
in individual’s community
3 (3 distinct projects)
0.207
0.222
(range 0-7)
% Change in p (X: 0 to 3)
24.24%
-18.18%
Community relations
2 (hostile)
0.236
0.191
(range 1-5)
4 (cordial)
0.169
0.264
% Change in p (X: 2 to 4)
-29.11%
40.3%
Member of saving club
0 (no)
0.199
0.234
(range 0/1)
1 (yes)
0.165
0.278
% Change in p (X: 0 to 1)
-18.09%
19.41%
Agree with vigilante group protection
0 (no)
0.20
0.231
(range 0/1)
1 (yes)
0.142
0.312
% Change in p (X: 0 to 1)
-28.5%
34.1%
Note: Robust standard errors in parentheses; *** p<0.01, **p<0.05, *p=<0.10. Models include city fixed effects and all
control variables. The dependent variable takes higher values for higher tax morale (unconditioned support for a citizen
obligation to pay tax). ♦: Jointly statistically significant in an F test.
Many demographic controls do not have an effect on tax attitudes. Education, age,
socioeconomic status, religion, and religiosity do not achieve statistical significance, though female
respondents do have higher tax morale. Hausa ethnic identity is also positively associated with tax

24
morale, which may be due to a pre-colonial history of stronger, more centralized states in the areas
of Northern Nigeria where Hausa reside or to different governance patterns during colonialism.
Apart from the Hausa, other ethnic effects are not significant. Igbo identity is negatively associated
with social contract attitudes, but it is not statistically significant, suggesting that being part of the
president’s ethnic coalition has no effect on tax morale.
With logit models, the size of effects cannot be directly inferred from model coefficients.
Table 1 therefore includes predicted probabilities and average marginal effects for statistically
significant variables. The probabilities and effects are computed for meaningful values of each of
the variables, keeping the remaining covariates at their observed values and then averaging across
respondents. The variables that test our core theoretical hypotheses have sizeable marginal effects.
More satisfaction with the use of tax revenue (a move from 1 to 3) and recent public good projects
in a community (a move from 0 to 3) increase the likelihood of expressions of tax morale by 36 and
24 percent, respectively. Residents of harmonious communities are about 29% less likely to agree
with an unconditioned citizen obligation to pay tax. Being a member of a savings club and
supporting vigilante group protection reduce the expression of tax morale by 18% and 28.5%,
respectively.
Our main findings are robust to a battery of additional empirical specifications. As a first
step, we include several variables that code for individual membership in ruling coalition, which
allows us to test whether potential access to clientelistic benefits shapes tax morale. These code
whether the respondent was: a co-partisan of the president or expressed the intention to vote for him
in the then-upcoming 2011 elections; a co-partisan of the sitting governor in her state; or co-ethnic
with the state governor. Co-ethnicity with the president is already included in Igbo ethnicity as a
control, since Igbos are typically presumed to be the beneficiaries of Goodluck Jonathan’s
administration. We also add alternative evaluations of confidence in government (perception of the

25
fairness of elections; trust in the governor; and trust in the president) and other measures of social
and political engagement (attendance of community meetings; likelihood of reporting theft to the
police; and reported voting in the last election). Finally, we include variables capturing personal
security: perception of the prevalence of electoral intimidation; direct experience of crime in the last
year; and direct experience with state-sponsored clearance of illegal structures.
Most importantly, Table 2 shows that our key variables retain statistical significance after
adding additional variables to our preferred Model 3. In addition, membership in winning coalitions
does not affect individual attitudes toward taxation (Model 5). Some of the additional measures of
attitudes toward the state attain statistical significance (Model 6). Perception of the fairness of
elections increases support for an unconditioned citizen obligation to pay tax. The trust variables
also achieve statistical significance, but the effect of trust is not uniform. Trust in the governor
increases tax morale, while trust in the president reduces it. In Model 7, reported attendance of
community meetings, which we might see as involvement in neighbourhood civic life, reduces tax
morale. Since individuals who participate in neighbourhood civic life presumably have greater
access to networks or institutions that provide club goods, this result supports our club goods
hypothesis. In model 8, two measures of personal insecurity – perceptions of the prevalence of
electoral intimidation and direct experience with state clearances – have significant, positive effects
on support for citizen tax obligation. As with contact with bribery, these measures might be seen as
a proxy for enforcement, or they may reinforce the notion that any contact with the state increases
an individual’s familiarity with and therefore orientation toward the state.
One concern with our findings on the fiscal exchange mechanism may be that political
allegiances affect how individuals respond to government performance. However, similar to our
theoretical discussion of potential direct effects of clientelistic coalition membership, we believe
that partisanship and co-ethnicity could mediate the public goods-morale relationship in different

26
Table 2. Ordered logit model of individual level determinants of attitudes toward taxation
Robustness checks
Partisan support
Trust
Community and
national
engagement
Insecurity
Model 5
Model 6
Model 7
Model 8
Satisfaction with revenue spending
0.201
0.197
0.238
0.239
(0.053)***
(0.054)***
(0.055)***
(0.052)***
Specific spending on public goods in
0.094
0.100
0.099
0.082
individual’s community
(0.026)***
(0.027)***
(0.028)***
(0.026)***
Community relations
-0.227
-0.190
-0.193
-0.130
(0.038)***
(0.039)***
(0.039)***
(0.040)***
Member of saving club
-0.283
-0.223
-0.314
-0.271
(0.137)**
(0.141)*
(0.151)**
(0.136)**
Agree with vigilante group protection
-0.428
-0.376
-0.468
-0.311
(0.155)***
(0.157)**
(0.172)***
(0.156)**
Co-partisan with state governor
0.137
(0.100)
Co-ethnic with state governor
0.153
(0.141)
Co-partisan with the president
0.048
(0.096)
Perception of elections (fair=1; unfair=0)
0.084
(0.044)*
Trust in governor
0.223
(0.052)***
Trust in president
-0.187
(0.053)***
Attendance of community
meetings
-0.155
(0.046)***
Would report theft to police
-0.038
(0.093)
Voted in last election
-0.023
(0.098)
Evicted
0.359
(0.110)***
Subject to electoral intimidation
0.408
(0.106)***
Subject to crime
0.190
(0.109)*
Adjusted R2
0.04
0.044
0.04
0.046
Observations
1,946
1,921
1,783
1,969
Note: Robust standard errors in parentheses; *** p<0.01, **p<0.05, *p=<0.10. All control variables are included in the
estimations. The dependent variable takes higher values for higher tax morale.
directions. Those who share a partisan or ethnic identity with office-holders may react more
positively to received services than non-group members, such that the effect of the fiscal exchange
mechanism is magnified. On the other hand, coethnics and copartisans may view receipt of services
as mere clientelistic spoils, which then has an ambiguous effect on tax morale. As suggested above,
clientelistic linkage may foster a view of state-citizen relations as contingent or conditional, making
individuals more likely to agree with the statement that views tax compliance as only contingently

27
necessary. We test these conditional hypotheses by adding interaction terms to Model 5 We find no
significant interaction effect between the number of public goods in a community and indicators of
either co-partisanship or co-ethnicity. This suggests that our findings regarding fiscal exchange are
not mediated by partisanship or ethnicity. This is important because it shows that government
services delivery can promote tax morale – and, possibly, a deeper orientation toward the state –
across ethnic and partisan lines, even in divided societies.
As a further robustness check, we run multi-level ordered logit regressions (Table A2,
Supplementary material). In these models, we include the individual level variables from Model 3
and city/state level variables. Two relevant measures at the state level are available: the percentage
of the state budget funded from internally generated tax revenue (as opposed to transfers from the
federal government)51 and statewide personal victimization rates52. State budget reliance on
internally generated revenue can serve as a proxy for varying levels of tax enforcement across
states, while crime victimization may capture the difference in safety and security across states. We
also include city-level aggregations of individual level survey responses, including trust attitudes
vis-à-vis the governor and the president, co-partisanship with the governor/president, and co-
ethnicity with the state governor. The intuition here is that living in an environment characterized
by lower levels of trust may independently affect individual attitudes. Similarly, an individual’s
attitudes may be affected by the partisan or ethnic character of her city, regardless of her own
identity. We find little evidence of effects above and beyond individual level effects. Only living in
a city with a large proportion of co-partisans of the president independently increases the likelihood
of expressions of higher tax morale. At the same time, our hypotheses regarding public and club
goods continue to find robust support in these multi-level estimations.
51 Central Bank of Nigeria 2010: 239.
52 CLEEN 2010: 5-6.

28
Despite the robustness of our findings so far, there remains some concern about a recursive
relationship between our dependent variable and some of the variables we use to operationalize our
club goods hypothesis. In other words, those individuals with low tax morale may be more likely to
seek out and invest in community-level substitutes for the state. Though we acknowledge this
theoretical possibility, we believe it does not affect the substance of our inferences for several
reasons. First, the causal direction from tax morale to club goods seems less compelling in a context
like urban Nigeria. Forms of community self-help predate government presence in the informal
slum communities where the bulk of urban Nigerians live; indeed, state governments have even
attempted to piggy-back atop and expand their reach via pre-existing associations and vigilante
groups.53 In the rest of Africa, the pattern is similar: burial societies, rotating credit associations,
and other informal institutions were already established in the colonial period.
Secondly, our results are robust when we rerun our models with community-level proxies
that are less vulnerable to this endogeneity concern. For instance, it is implausible that a
respondent’s orientation toward tax obligation would substantially affect his count of the number of
riots that had occurred in his town. As we noted earlier, this less qualitative measure of community
conflict is highly statistically significant in Model 4 (Table 1). Finally, to further bolster our
confidence in the findings, we employ a filtering technique described in Frey and Torgler54 to
estimate and correct the potential bias in some of our proxies for the supply of community-provided
or club goods. For instance, assume that those who express belief in an unconditioned citizen
obligation to pay tax always have a favorable orientation toward the state. These state-favoring
53 Smith 2007; LeBas 2013.
54 Frey and Torgler (2007) study tax morale as a function of the perceived tax evasion of others, and they
show similar results from a two stage least square procedure with both instrumental variables and the
filtering technique. In fact, filtering results in more conservative estimates.

29
individuals may see then their communities as less harmonious, may be less likely to turn to
vigilante groups for help with a crime, or may be less likely to participate in savings circles.
To deal with this potential bias, we calculate a bias correction by computing two city-level
averages: (1) the average value of all residents’ responses for each of our main club goods proxies
(quality of community relations; reliance on vigilantes; and participation in savings clubs); and (2)
the average value of responses for those individuals with the highest levels of tax morale (the
response of 4 or “strongly believe” that citizens should always pay tax). If there exists a positive
difference between the city average of all respondents and the city average of high-morale
respondents, then we add this city-specific bias correction value to the assessments of all high-
morale individuals in that city. By following the same method for all three of our proxies, we bring
the response values of high-morale individuals closer to the average responses of those who live in
the same city. When our preferred model is rerun with these bias-corrected measures, both the
qualitative assessment of community relations and the reliance on vigilante groups remain
statistically significant (Table A3, Supplementary material). This fairly conservative estimation
method roughly halves the coefficients on these two variables, but our marginal effects remain
large, increasing our confidence in a causal relationship between the supply of community goods
and tax morale.
ATTITUDES AND BEHAVIOR: WHO PAYS TAXES?
We turn now to the important question of whether higher tax morale has concrete effects on
actual tax compliance. Our data allows the construction of a behavioral measure that is not
susceptible to the problems of self-reported tax payment. Direct questioning regarding individuals’
own tax payment will generate some degree of over-reporting, given incentives to conceal tax
evasion. It may also generate respondent hostility toward enumerators, which could corrupt other
responses as well. We therefore pursued an indirect means of ascertaining individual tax payment.

30
In the course of several questions about taxation and awareness of reforms, each respondent was
asked if she currently paid more taxes and levies than she had three years previously. To this
question, 44 percent of respondents volunteered that they did not pay taxes. We use this un-
prompted response to construct a binary measure of tax payment. Those who reported that their
taxes increased or decreased were presumed to have paid taxes, while those that volunteered non-
payment were coded as non-payers. Our findings are robust to an alternative coding of this variable.
Because the answer “No, I do not pay more taxes and levies than three years ago” can also be
construed as “no I do not pay taxes, but I also did not pay in the past”, we also code our variable as
a 1 only if the respondent volunteered that he/she pays more now than in the past (Model 11). This
variable codes 69 percent of respondents as not paying tax.
Are our rates of tax evasion plausible in the Nigerian context? State governments do not
calculate or publicly release estimates of tax evasion rates. In our own discussions with state and
federal government officials, there has been a general acknowledgement that the size of the tax base
and the overall level of evasion in Nigeria are simply unknown. Lagos State has, however, started a
process of tax auditing, and their estimates are instructive. Despite better tax enforcement than
elsewhere in Nigeria, less than 35% of Lagosians paid tax in 2011.55 Nor should this be seen as
primarily a problem of the informal economy. Recently, Nigeria’s Minister for Economy & Finance
estimated that 75% of officially registered firms do not pay federal corporate taxes.56 Given these
numbers, our self-reported measures seem plausible.
We can only view volunteered reporting of non-payment as evidence of evasion if the
respondent is liable for tax. In Nigeria, income tax rates are set federally, and all workers –
regardless of whether employed in the formal or informal sector – are assessed personal income tax
55 “Lagos to prosecute 5.3 million tax evaders,” This Day (Lagos), March 21, 2012.
56 “Government loses N80b monthly to tax evasion,” The Guardian (Nigeria), August 14, 2013.

31
liability starting at 5 percent of income per year.57 Those with annual incomes below 30,000 naira
(UK £258) are not required to file or pay tax. This is a very high bar for exemption, far less than a
pound a day, since income earners typically support several dependents. Though the majority of
Nigerians are poor, very few workers would be eligible for formal exemption. Further, even those
exempted from income tax are still likely to be required to pay a property tax known as tenement
rate. Legislation on tenement rates varies from state to state, but individual exemptions are not
granted based on income, property value, or whether the inhabitant is a tenant or owner.58 The vast
majority of our respondents would be legally required to pay both personal income tax and
tenement rate.
In order to further explore the reliability of our non-payment measure, we run additional
tests that assume some subset of non-evading non-payers among our respondents. Individuals may,
for instance, earn less than 30,000 naira and also live in a low-income area with habitations exempt
from tenement rate. These individuals would not owe tax. We use our indicators on individual
income, food deprivation, and assets to create tax payment variables that do not consider extremely
poor respondents as liable for tax. We consider as non-evaders individuals that have high levels of
food deprivation (highest two or three levels of the variable), possess few assets (lowest two or
three levels of the variable), and reported earning less than 5000 Naira (£19) per month.
Importantly, all the findings reported below on determinants on tax payment remain robust to using
such alternative measures of tax payment (Table A4, Supplementary material). It is also important
to underline that, overall, neither wealth nor employment in the formal economy are statistically
significant predictors of tax payment.
57 Federal Government of Nigeria 2004. In rural areas, tax is often assessed communally, but this is not the
case for our urban respondents.
58 Ogunba et al. 2006.

32
So who pays tax? Findings in Table 4 are derived from logistic models with city-fixed
effects, but results hold when city fixed effects are excluded or when we use the same number of
observations across Models 9-11. Our key finding, reported in Models 10 & 11 is that attitudes
towards taxation have a significant effect on tax payment. Those who express higher levels of tax
morale are actually more likely to pay tax. The variable has an important substantive effect too: A
change from strongly supporting statement B (“citizens should only pay taxes if they believe in the
government”) to strongly supporting statement A (“citizens should always pay their taxes, even if
they disagree with the government”) increases the predicted probability of tax payment by 26.4%.
Since satisfaction with revenue spending and public goods receipt do not directly boost tax
payment, services provision seems to produce tax compliance through its effect on individual
attitudes. In other words, public goods receipt produces higher degrees of tax morale, which in turn
is associated with tax payment. This finding underlines the importance of examining shifts in social
norms alongside more concrete improvements in state delivery of services and state enforcement
capacity.
Of the other factors that achieve significance in our earlier model of tax morale, only one
has consistent and robust independent effects on actual tax payment. In all specifications, those who
report harmonious community relations are significantly less likely to pay tax. We posit that this is
due to the same substitution effect earlier hypothesized to explain attitudes: residents of areas where
there exists effective community provision of club goods are less likely to pay taxes, as they simply
need the state less than those who do not have access to effective community goods provision.
Neither savings club membership nor vigilante group reliance, however, is statistically significant.
Measures of public goods provision also perform weakly in explaining actual tax payment.
Satisfaction with revenue spending does not boost tax payment. Concrete public goods provision
has a positive effect on tax payment, but is not significant across all models. In Model 11, which

33
Table 3. Logit model of individual level determinants of tax payment
Individual level
variables
Attitude toward
tax payment
Alternative coding
of tax payment
Model 9
Model 10
Model 11
Satisfaction with revenue spending
0.032
0.004
-0.037
(0.058)
(0.061)
(0.060)
Specific spending on public goods in
individual’s community
0.065
0.049
-0.051
(0.028)**
(0.030)*
(0.031)
Community relations
-0.205
-0.183
-0.096
(0.048)***
(0.049)***
(0.047)**
Member of saving club
-0.005
0.004
-0.064
(0.182)
(0.188)
(0.204)
Agree with vigilante group protection
-0.256
-0.145
0.085
(0.200)
(0.210)
(0.214)
Met with government official or representative
0.169
0.217
0.345
(0.078)**
(0.084)***
(0.077)***
Food deprivation
-0.036
0.004
0.086
(0.054)
(0.057)
(0.057)
Religiosity
-0.020
-0.016
-0.014
(0.036)
(0.037)
(0.040)
Direct contact with corruption
0.463
0.442
0.375
(0.113)***
(0.117)***
(0.116)***
Assets
0.027
0.011
0.154
(0.059)
(0.062)
(0.062)**
Interest in politics
0.022
0.017
0.128
(0.051)
(0.053)
(0.052)**
Tax morale
0.233
0.118
(0.051)***
(0.053)**
Adjusted R2
0.12
0.13
0.09
2,015
1,919
1,919
Substantive effects for the statistically significant variables – Use Model 2 above
Value of independent
variable
Predicted
probability of tax
payment (p)
% Change in p
Specific spending on public goods
0 (none)
0.57
5%
in individual’s community
3
0.6
Community relations (range 1 - 5)
2 (hostile)
0.64
-10.9%
4 (cordial)
0.57
Tax morale (range 1 - 4)
1
0.53
26.4%
4
0.67
Met with government official or representative
0 (no)
0.58
8.6%
1 (yes)
0.63
Direct contact with corruption
0 (no)
0.57
14%
1 (yes)
0.65
Note: Robust standard errors in parentheses; *** p<0.01, **p<0.05, *p=<0.10. Models include city fixed effects and all
control variables. The dependent variable takes the value of 1 if individual pays taxes and 0 otherwise.
uses our alternative measure of tax payment, it is insignificant and signed in an unexpected
direction. Other proxies for direct contact with the state have strong, positive effects on tax
payment. Direct contact with official corruption is significantly and positively associated with tax
payment, as it is for tax morale. Another measure of contact with the state – meeting with a

34
government official - is also significantly and positively associated with tax payment. The results
for bribery and meetings with officials may be consistent with either of the mechanisms posited
above. That is, state contact may be increasing individuals’ orientation to the state and individual
tax payment, or bribe solicitation and meeting with government representatives may serve as
proxies for state enforcement capacity.
CONCLUSION
Our findings have several implications for those who are interested in the development of
state capacity and accountability in weak states in the developing world. Most simply, we show that
tax morale exists even in contexts like Nigeria, where resource curse and predatory state dynamics
are presumed to stunt the adoption of norms favoring state authority and voluntary compliance with
state demands. Belief in an unconditioned citizen obligation to pay tax is not general: Only about 40
percent of our urban sample express a degree of agreement with this view, and these attitudes are
likely to be more rare in rural contexts. Secondly, our findings lend further evidentiary support to
established findings on fiscal exchange. Individuals in urban Nigeria are more likely to support a
citizen obligation to pay tax if they have received concrete benefits from the state and view the
state’s use of tax revenue favorably. Support from Sub-Saharan Africa for this mechanism is
particularly important, as the existing literature is heavily reliant on evidence from the developed
world. Finally, a more complicated story is suggested by our findings about the effect of social
context on tax morale. We find that individuals with access to club goods express lower levels of
tax morale or belief in an unconditioned citizen obligation to pay tax. Individuals who live in
conflict-prone areas, where effective club goods provision is less likely, express higher degrees of
tax morale, even if they lack past positive experience of state delivery. These findings suggest that

35
conflict and ethnic diversity – factors often statistically associated with weaker public goods
provision59 – may exist alongside popular support for ideas of citizen obligation to the state.
We suggest two directions for future research on the creation of state authority and effective
taxation. First, both scholars and policy-makers have recently -- and rightly -- focused on public
goods delivery as a route to tax revenues and accountability. There has been a concomitant
downplaying of coercion’s role in state-building. But we show that state contact need not be
benevolent in order to change individual orientations toward the state. Urban Nigerians who have
been solicited for bribes – and those who have suffered due to state-sponsored destruction of illegal
structures – are more likely to express unconditioned support for a citizen obligation to pay taxes
than those who lack these negative experiences of state contact. This is controversial evidence that
warrants more research, but it is also direct evidence for the link Tilly poses between state extortion
and societal acceptance of state authority. Secondly, future work would profit from greater attention
to pre-existing institutions that serve as rivals to some state functions. Communities often construct
substitutes for state services provision in the absence of the state. These substitutes can serve as
impediments to greater reliance on and responsibility to the state. Overall, we suggest that the
emergence of tax morale – or norms supporting the quasi-voluntary compliance long presumed to
be instrumental in the construction of state authority -- is a reciprocal process of extending state
services while prompting citizens to lessen reliance on community-provided goods.
59 Alesina et al 1999.

36
REFERENCES
Alesina, Alberto, Reza Baqir, and William Easterly. 1999. Public Goods and Ethnic Divisions.
Quarterly Journal of Economics 114: 1243-84.
Alm, James, Betty Jackson, and Michael McKee. 1992. Institutional Uncertainty and Taxpayer
Compliance. American Economic Review 82(1):1018-1026.
Alm, James, Gary McClelland, and William Schulze. 1992. Why Do People Pay Taxes? Journal
of Public Economics 48: 21-38.
Alm, James, Jorge Martinez-Vazquez, and Benno Torgler. 2006. Russian attitudes toward paying
taxes – before, during and after the transition. International Journal of Social Economics
33(12): 832-860.
Anderson, Christopher, and Yuliya Tverdova. 2003. Corruption, Political Allegiances, and Attitudes
toward Government in Contemporary Democracies. American Journal of Political Science
47(1): 91-109.
Andreoni, James, Brian Erard, and Jonathan Feinstein. 1998. Tax Compliance. Journal of Economic
Literature 36(2): 818-860.
Baldwin, Kate, and John Huber. 2010. Economic versus Cultural Differences: Forms of Ethnic
Diversity and Public Goods Provision. American Political Science Review 104(4): 644-662.
Bates, Robert, and Da-Hsiang Donald Lien. 1985. A Note on Taxation, Development and
Representative Government. Politics & Society 14(53): 53-70.
Bernstein, Thomas, and Xiaobo Lu. 2008. Taxation without Representation in Contemporary Rural
China. Cambridge: Cambridge University Press.
Boone, Catherine. 2003. Political Topographies of the African State: Territorial Authority &
Institutional Choice. Cambridge: University of Cambridge Press.

37
Cassar, Allesandra, and Bruce Wydick. 2010. Does social capital matter? Evidence from a five-
country group lending experiment. Oxford Economic Papers 62: 715-739.
Central Bank of Nigeria. 2010. Annual Report. Abuja, Nigeria.
Chang, Eric, and Nicholas Kerr. 2009. Do Voters have Different Attitudes toward Corruption? The
Sources and Implications of Popular Perceptions and Tolerance of Political Corruption.
Afrobarometer Working Paper No. 116.
CLEEN Foundation. 2010. National Criminal Victimization and Safety Survey, 2010: Summary of
Findings. Lagos, Nigeria: CLEEN.
Cox, Donald, and Emmanuel Jimenez. 1998. Risk Sharing and Private Transfers: What about Urban
Households? Economic Development and Cultural Change 46(1): 621-637.
D’Arcy, Michelle. 2011. Why Do Citizens Assent to Pay Tax? Legitimacy, Taxation, and the
African State. Afrobarometer Working Paper No. 126.
Dixit, Avinash, and John Londregan. 1996. The Determinants of Success of Special Interests in
Redistributive Politics. Journal of Politics 58(4): 1132-1155.
Ekeh, Peter P. 1975. Colonialism and the Two Publics in Africa: A Theoretical Statement.
Comparative Studies in Society and History 17(1): 91-112.
Fafchamps, Marcel. 1992. Solidarity Networks in Preindustrial Societies: Rational Peasants with a
Moral Economy. Economic Development and Cultural Change 41(1): 147-174.
Feld, Lars, and Bruno Frey. 2002. Trust breeds trust: how taxpayers are treated. Economics of
Governance 3:2, 87-99.
Fjeldstad, Odd-Helge. 2001. Taxation, Coercion and Donors: Local Government Tax Enforcement
in Tanzania. Journal of Modern African Studies 39(2): 389-206.
Fjeldstad, Odd-Helge. 2004. What’s trust got to do with it? Non-payment of service charges in local
authorities in South Africa. Journal of Modern African Studies 42(4): 539-562.

38
Fjeldstad, Odd-Helge, and Joseph Semboja. 2001. Why People Pay Taxes: the Case of the
Development Levy in Tanzania. World Development 29:12, 2059-2074.
Frey, Bruno, and Benno Torgler. 2007. Tax morale and conditional cooperation. Journal of
Comparative Economics 35: 136-159.
Geertz, Clifford. 1978. The Bazaar Economy: Information & Search in Peasant Marketing.
American Economic Review 68(2): 28-32.
Golden, Miriam, and Lucio Picci. 2008. Pork-Barrel Politics in Postwar Italy, 1953-1994. American
Journal of Political Science 52(2): 268-289.
Grief, Avner. 1994. Cultural beliefs and the organization of society: a historical and theoretical
reflection on collectivist and individualist societies. Journal of Political Economy 102(5): 912-
950.
Grimard, Franque. 1997. Household consumption smoothing through ethnic ties: evidence from
Cote d'Ivoire. Journal of Development Economics 53: 391-422.
Habyarimana, James, Macartan Humphreys, Daniel Posner, and Jeremy Weinstein. 2007. Does
Ethnic Diversity Undermine Public Goods Provision? American Political Science Review
101(1): 709-725.
Habyarimana, James, Macartan Humphreys, Daniel Posner, and Jeremy Weinstein.
2009. Coethnicity: Diversity and the Dilemmas of Collective Action. New York, NY:
Russell Sage.
Jones Luong, Pauline, and Erika Weinthal. 2004. Contra Coercion: Russian Tax Reform,
Exogenous Shocks, and Negotiated Institutional Change. American Political Science Review
98: 139-152.
Kasara, Kimuli. 2007. Tax me if you can: Ethnic Geography, Democracy, and the Taxation of
Agriculture in Africa. American Political Science Review 101(1): 159-172.

39
Kitschelt, Herbert. 2000. Linkages between Citizens and Politicians in Democratic Polities.
Comparative Political Studies 33 (6): 845-879.
Kramon, Eric, and Daniel Posner. 2013. Who Benefits from Distributive Politics? How the
Outcome One Studies Affects the Answer One Gets. Perspectives on Politics 11(2): 461-474.
LeBas, Adrienne. 2013. Violence and Urban Order in Nairobi, Kenya, and Lagos, Nigeria. Studies
in Comparative International Development 48(3): 240-262.
Letki, Natalia. 2006. Investigating the Roots of Civic Morality: Trust, Social Capital, and
Institutional Performance. Political Behavior 28: 305-325.
Levi, Margaret. 1988. Of Rule and Revenue. Berkeley: University of California Press.
Lewis, Peter. 1996. From Prebendalism to Predation: the Political Economy of Decline in Nigeria.
Journal of Modern African Studies 34(1): 79-103.
______. 2006. Performance and Legitimacy in Nigeria’s New Democracy. Working Paper.
Afrobarometer Briefing Paper No. 46.
Lieberman, Evan. 2003. Race and Regionalism in the Politics of Taxation in Brazil and South
Africa. New York: Cambridge University Press.
Lindberg, Staffan. 2003. It’s Our Time to “Chop”: Do Elections in Africa Feed Neo-Patrimonialism
Rather than Counteract It? Democratization 10(2): 121-140.
Miguel, Edward. 2004. Tribe or Nation? Nation Building and Public Goods in Kenya versus
Tanzania. World Politics 56(3): 327-362.
Nigeria, Federal Government of. 2004. Personal Income Tax Act (PITA), Cap P8 LFN 2004;
amendment to Personal Income Tax Decree 104 of 1994. Abuja, Nigeria.
Ogunba, Olusegun, Mashood Tomori, and Olatoye Ojo. 2006. An Appraisal of the Efficiency of
Local Property Tax Administration in Nigeria. Journal of Land Use & Development Studies
2(1).

40
Putnam, Robert. 1994. Making Democracy Work: Civic Traditions in Modern Italy. Princeton:
Princeton University Press.
Rabe-Hesketh, Sophia and Anders Skrondal. 2008. Multilevel and Longitudinal Modeling using
Stata (2nd Edition). College Station, TX: Stata Press.
Ross, Michael. 2012. The Oil Curse: How Petroleum Wealth Shapes the Development of Nations.
Princeton University Press.
Sala-i-Martin, Xavier, and Arvind Subramanian. 2003. Addressing the Natural Resource Curse: an
Illustration from Nigeria. NBER Working Paper No. 9804 (June 2003).
Scholz, John, and Mark Lubell. 1998. Trust and Taxpaying: Testing the Heuristic Approach to
Collective Action. American Journal of Political Science 42(2): 398-417.
Scott, James. 1976. The Moral Economy of the Peasant: Rebellion & Subsistence in Southeast Asia.
New Haven: Yale University Press.
Smith, Daniel. 2007. A Culture of Corruption: Everyday Deception and Popular Discontent in
Nigeria. Princeton: Princeton University Press.
Tilly, Charles. 1985. War Making and State Making as Organized Crime. In Evans et al, Bringing
the State Back In. Cambridge University Press.
Tilly, Charles. 1990. Coercion, Capital, and European States, AD990-1992. Oxford: Blackwell
Publishing.
Timmons, Jeffrey. 2005. The Fiscal Contract: States, Taxes, and Public Services. World Politics 57:
530-567.
Torgler, Benno. 2007. Tax Compliance and Tax Morale: a Theoretical and Empirical Analysis.
Cheltenham, UK: Edward Elgar Publishing.
Wantchekon, Leonard. 2003. Clientalism and Voting Behavior: Evidence from a Field Experiment
in Benin. World Politics 55: 399–422.