Fact 1: Rich countries have made successive investments in their fiscal capacities over time.

Fact 2: Rich countries collect a much larger share of their income in taxes than do poor countries.

Fact 3: Rich countries rely to a much larger extent on income taxes as opposed to trade taxes than do poor countries.

Fact 4: High-tax countries rely to a much larger extent on income taxes as opposed to trade taxes than do low-tax countries.

Fact 5: Rich countries collect much higher tax revenue than poor countries despite comparable statutory rates.

Fact 6: Countries with strong executive constraints collect higher tax revenues, when income per capita is held constant, than do countries with weak executive constraints.

Fact 7: Countries with strong executive constraints rely on a higher share of income taxes in total taxes, when income per capita is held constant, than do countries with weak executive constraints.

5.1 Economic Development

In this subsection, we discuss how economic change affects choices of fiscal capacity and the implications for observed taxation. Against the background of the stark time-series and cross-sectional facts in Section 3, we focus on the role of economic development for the introduction and expansion of the income tax. As discussed at the outset, this has also been the standard focus in the taxation and development literature. We begin by discussing exogenous differences in the economy across countries or time, turning then to changes that are endogenous to the government’s investment in fiscal capacity.

Exogenous economic differences: We noted in Section 3 (recall Figure 4) the typical path of change involves the two discrete steps of introducing the income tax and upgrading its reach via direct withholding. In a contemporary cross-section, we also saw (recall Figures 7 and 8) that rich and high-taxing states rely much more on the income tax than poor and low-taxing states. Through which channels does our framework explain such patterns in the data?

To answer this question, we specialize the model to include only one consumption good, in addition to the numeraire good and labor—i.e., we set . Moreover, there is no fixed cost in building fiscal capacity for the taxable consumption good, whereas a fixed cost may exist for the income tax—i.e., we have and . Of course, this stark difference is for illustrative purposes only. To keep things simple and pin down the value of public funds, we specialize the utility function to be linear in public goods, i.e., , and the value of public goods to be equal across groups exceeding the value of transfers, i.e., . These assumptions are relaxed in later subsections on politics and the value of public spending. For now, they allow us to focus on a government that spends only on public goods with a constant marginal value of funds.

We start by assuming that wages are given by the simple expression

i.e., every group has the same wage. Different values of could represent natural exogenous income differences across countries, or across time, due to, say, geography or total factor productivity.

In this specialized framework, the marginal first-order conditions(8) associated with the two tax bases are

(9)

If there were no fixed costs, this expression would tell us that the government invests more in the tax base that raises more revenue on the margin at the future value of public funds (the first term), induces a lower utility cost for consumers via the cost of tax evasion (the second term), or has a lower marginal cost of investing at the current value of public funds (the third term). Provided the positive first term outweighs the negative second term, for , we observe positive investments in both types of fiscal capacity since .

We now revisit the question of when an income tax is worth levying at all and why economic growth might typically induce the introduction of an income tax, as we have seen historically. Suppose fiscal-capacity building for the income tax has a fixed cost and the period-1 level of this capacity is zero, . Recall that the government raises no revenue at zero fiscal capacity. In order for the income tax to be introduced, the perceived welfare gains from doing so, by bringing fiscal capacity up to locally optimal level given by (9), have to be large enough to outweigh the effective fixed cost associated with setting up a compliance and monitoring system. Using the definitions and additive separability of the government payoff (6), the net tax bases (4), and the indirect utility function (2), and recalling that when we have (private evasion cost and taxes are zero), we can write the formal condition as follows:

(10)

where solves (9).

There are three main considerations. The term on the first line reflects the value of transferring funds from private incomes to public spending, recognizing that lower labor supply induces a deadweight loss. This expression is positive only if is sufficiently high (above one).¹⁹ Also, the first-line term is proportional to exogenous productivity , as this determines how lucrative the income tax base is. The second term on the left-hand side reflects the possibility of non-compliance. It has two parts, the first reflecting the gain from a new source of profits from tax avoidance. However, this is offset by the fact that greater avoidance reduces valuable public spending. If there was full compliance at , then this expression would be zero. Finally, the term on the right-hand side reflects the costs of introducing a new tax base—fixed costs and the cost of the investment in fiscal capacity up to .

Notice that the tax base in the first term of (10) is increasing in the productivity factor . Moreover, the optimal income tax rate associated with a given level of fiscal capacity will generally be higher if income is higher. To see this, recall the Ramsey-tax formula (5), where is decreasing in , and hence increasing in (since is decreasing in ).

If captures income growth over time, this can naturally explain the eventual introduction of an income tax, as in Figure 4, by reference to (10). If instead captures differences across countries, at a given point in time, this can explain the higher reliance of the income tax in rich and high-tax countries, as in Figures 7 and 8. To explicitly link up with the data on income taxes vs. trade taxes discussed in Section 3, the argument would have to be recast in a setting where trade rather than consumption is the alternative tax base (see Besley & Persson, 2011c, chap. 2 for such a model).

Endogenous economic differences: In this section, we make the level of fiscal capacity endogenous to other government decisions. The general modeling follows the analysis in Besley and Persson (2011c).

Let wages be given by , where scalar represents endogenous government investment to increase productivity and where is an increasing concave function. As Besley and Persson (2011c, chap. 3) show, one can microfound such a formulation if represents the capacity to carry out legal support to the private sector concerning contract enforcement or, alternatively, protection of property rights. In this interpretation, which we will maintain in this subsection, captures the legal capacity of the government: its courts, its supply of educated judges, or its registers for credit or property. An alternative interpretation would be to think of as government infrastructure that raises the productivity in the private sector. We could also let be a vector of productive capacities, rather than a scalar.

Analogous to fiscal capacity, we assume that legal capacity in period 2 can be augmented by investment in period 1 at cost . We assume that legal-capacity investments have no fixed costs, for simplicity, and that is a convex function with . As a consequence, the total investment costs for the period-1 government are now given by

What happens to the investment in fiscal capacity in the specialized model that we just studied, when we replace exogenous wages with endogenous wages The marginal investment conditions in (9) are not affected, because neither nor depend on legal-capacity investments . However, the condition for incurring the fixed costs of the income tax now becomes:

(11)

Only the first term from (10) is affected with higher legal capacity increasing wages. There are good reasons to expect that this key expression is increasing in and . For example, in the case of a constant elasticity of labor supply, , the first expression in (11) becomes:

which is clearly increasing in . Thus, a country with higher legal capacity and endogenously higher income is more likely to have an income tax than one with low legal capacity.

Of course, this raises the question what drives investments in legal capacity. Maximizing the investment objective (7) with regard to , under the assumptions of the specialized model and using Roy’s identity, we obtain the first-order condition

(12)

Since the two terms in the first bracket, the net benefit of legal capacity, are both non-negative and since , there are always positive investments in legal capacity. Moreover, a higher level of fiscal capacity in the income tax raises the equilibrium tax rate . This way, a higher value of raises the net benefit of investing in legal capacity, by raising the private marginal surplus from higher wages as well as boosting the fiscal benefits of the income tax through a higher tax base.

This result and the earlier result, that a higher makes (11) more likely to hold, make the investment in legal capacity and the investment in fiscal capacity necessary to introduce the income tax complementary decisions. This is a close relative to the complementarity discussed in Besley and Persson (2009a,b, 2011a,b,c). Thus, the endogenous growth of income triggered by investments in the productive side of the state makes it more likely that a country at some point in time will incur the fixed costs necessary to put an income tax in place.

As discussed at length in this earlier work, measures of fiscal capacity—like a high share of total tax income collected by the income tax—and measures of legal capacity are strongly positively correlated across countries in the data, and both of these capacities indeed have a strong positive correlation with income.

This point is illustrated in Figure 12 which plots the share of income tax in total tax revenue in 1999 against the ICRG measure of property-rights protection. Countries that raise more in income tax (have more fiscal capacity) also tend to enforce property rights in a better way (have more legal capacity).

Structural change: Development is about a lot more than raising income per capita. The process of rising incomes typically goes hand in hand with structural change toward a more urbanized and non-agriculturally based economy. As a consequence, more economic activity operates in the open, particularly in the formal sector where transactions and employment relations are recorded. To some extent, informality in production is just the flip side of tax avoidance. But it is more than that. Firms also choose not to become part of the formal sector in order to avoid an array of regulations. But this has a cost: such firms are not able to take advantage of formal legal protection and contract disputes have to be resolved informally, often placing trust between parties at a premium. This limits the scope of business, which often becomes restricted to social networks.

The move toward formality tends to facilitate tax compliance. More employment takes place in legally registered firms rather than self-employment, as stressed by Kleven et al. (2009), and more financial transactions take place via formal intermediaries (such as banks), as stressed by Gordon and Li (2009). Both of these make transactions more visible to tax authorities and enable tax authorities to obtain corroborating evidence from cross-reported transactions. Falsifying these requires collusion rather than unilateral secrecy. Such changes result from transformations in the nature of economic activity whereby larger firms take advantage of scale economies in production. To the extent that this is reflected in higher wages, the arguments from the last section apply and we expect investments in fiscal capacity to occur.

The typical discussion of development and taxation couches structural change as an exogenous feature of economic development with causality running from economic development to fiscal capacity. This can be captured in our model either by allowing the function to depend on the sector of the economy in which an individual is operating. Suppose we exogenously assign individuals to the formal and informal sectors denoted by where stands for “formal” and for “informal” with evasion functions . We may then reasonably suppose that

i.e., the marginal impact of an investment in fiscal capacity is more effective in deterring evasion for those operating in the formal sector. In this event, more formality would boost the revenues that can be generated from fiscal-capacity investments, all else equal. This is consistent with the observation that countries with smaller informal sectors also raise more taxes. This is illustrated in Figure 13 which plots a measure of the size of the informal economy in 1999/2000 from Schneider (2002) against the share of income taxes in total tax revenue in 1999 from Baunsgaard and Keen (2005). The downward sloping relationship is extremely clear.

The literature has paid less attention to the possibility that the size of the informal sector and the structural development of the economy evolve endogenously with the development of fiscal capacity, as in our discussion of legal capacity above. However, we may also take a further step and think of legal capacity as affecting the returns to being formal. It is very hard for an individual to simultaneously be largely invisible to the tax system and take full advantage of the formal legal system. This creates a further complementarity between the legal and fiscal capacities of the state. A state which invests in the infrastructure to support formal financial intermediation will overcome some of the barriers to formality and enhance the ability to raise more taxes. A good example are efforts to build credit and land registries in the process of development, to increase property rights and contract enforcement. Such registries bring the patterns of ownership and credit contracts into the daylight for tax authorities. To study these issues explicitly, we would have to extend the model with an endogenous decision to choose the sector based on costs and benefits. While a higher cost of tax evasion is a cost of choosing the formal sector, there may be benefits in the form of a better trading environment.²⁰

5.2 Politics

No account of the development process can be complete without considering the political forces that shape policy selection. It is widely held that the failure of states to build strong institutions might reflect weak motives embedded in political institutions. In this section, we explore the implications of introducing a government which operates under institutional constraints and faces the possibility of political turnover. The specific framework that we use is based on Besley and Persson (2010a,b, 2011a,b,c). This belongs to a wider body of work and thinking in dynamic political economics which is reviewed in Acemoglu (2006). As we shall see, this adds new issues to the analysis of fiscal-capacity building and allows us to uncover additional forces which can explain high or low investments.

Cohesive institutions: Suppose the government in power acts on behalf of a specific group in the spirit of the citizen-candidate approach to politics—see Besley and Coate (1997) and Osborne and Slivinsky (1996). There is no agency problem within groups: whoever holds power on behalf of a group cares only about the average welfare of its members.

We model how political institutions constrain the incumbent’s allocation of transfers in a very simple way. Specifically, the incumbent group in period , called , must give (at least) a fixed share to all non-incumbent groups for any unit of transfers awarded to its own group. That is to say, we impose the restriction

The parameter represents the “cohesiveness” of institutions with closer to representing greater cohesiveness.

This is an extremely simple and tractable, but reduced-form, way of looking at politics and is used extensively in Besley and Persson (2011a,b,c). We can interpret a higher value of in one of two broad ways. One real-world counterpart might be minority protection by constraints on the executive, due to some constitutional separation of powers. In practice, we expect democracies to impose greater constraints on the executive than autocracies. An alternative real-world counterpart might be stronger political representation of the interests of political losers in policy decisions through proportional representation elections or parliamentary democracy. The literature on the policy effects of constitutional rules suggests that both of these institutional arrangements make policymakers to internalize the preferences of a larger share of the population—see, e.g., Persson, Roland, and Tabellini (2004), Aghion, Alesina, and Trebbi (2000), and Persson and Tabellini (2000).

In this representation of political institutions, we can solve for transfers allocated to the incumbent group and all the groups in opposition . In the model of Section 4, these are

where

(13)

For , any residual tax revenue is equally divided in transfers to all groups. Otherwise, the incumbent group receives a higher per-capita share of transfer spending.

We maintain the simplifying assumption of a linear utility function for public goods, but allow the valuation of public goods to differ across groups. The shadow value of public revenue now compares the incumbent’s value of transfers to spending on public goods . As in the general model, we have two cases. If , all spending is allocated to public goods, i.e., , else the marginal value of public funds is .²¹ Suppose now that a single group is in power in period 1 as well as period 2, i.e., there is a natural political elite and no political turnover for sure. In this case, the preferences of the elite determine policy and investment in fiscal capacity. For simplicity, we assume away any fixed costs in investment (or alternatively for all so that the fixed costs have already been incurred). Then, we get the following first-order conditions for investment in fiscal capacity:

(14)

The analysis requires only a modest modification of the benchmark model, where we recognize that the driving force behind the decision to build fiscal capacity is now the preference of the ruling elite for tax revenue, rather than society as a whole. Clearly, an elite that greatly values public goods is more likely to spend on public goods compared to one that does not. Spending on public goods rather than transfers is more likely as institutions become more cohesive, and the ability of the incumbent group to extract transfers diminishes. However, an elite can also be motivated to build capacity to collect tax revenue as a means of increasing transfers for itself when , because the elite faces few constraints on its power to pursue group interests (i.e., is low which makes high).

Political turnover: The model becomes more interesting when we introduce the possibility of political turnover, i.e., the identity of the incumbent group may shift over time. To home in on this issue, we specialize the model to the case of only two groups each comprising half the population, . Let be the probability that the incumbent group is replaced between the two time periods. Clearly, is a natural measure of political (in) stability. This new feature adds new and important dimensions to the analysis of policy and investments in fiscal capacity.

Let the period- payoff of being either the incumbent or the opposition, , be:

where

is the total budget available for transfers, and and are the shares of transfers going to the incumbent and opposition groups. Now the level of fiscal capacity will be chosen to maximize

(15)

The effect of political turnover follows from the fact that enters this expected payoff. The optimization of the incumbent over the vector of fiscal capacity yields:

(16)

which, of course, just says that marginal costs and benefits are equated (at an interior solution). The third marginal cost term in (16) is by now familiar. However, some additional considerations go into computing the marginal benefit represented by the first and second terms.

After some simple algebra, we can rewrite (16) as:

(17)

where

(18)

and

The third and fourth terms in (17) are the same as in earlier cases, capturing the utility costs of greater compliance and the marginal costs of investment in fiscal capacity. As before, the first term represents the value of extra revenue. However, the weight on this is now more complicated since the value of future public revenue to the current incumbent is different when a marginal future dollar is spent by a future incumbent different than herself, especially when the spending is on transfers rather than public goods. Unless there is agreement on the valuation of public goods and/or institutions are fully cohesive , we would expect , so this effect will tend to diminish the incentive to invest in fiscal capacity, and more so the higher is the probability of turnover .

The second term is entirely new. It represents an effect familiar from the work on strategic policy making in dynamic models of politics, which began with Alesina and Tabellini (1990) and Persson and Svensson (1989). The fact that the current incumbent and opposition may differ in their views about optimal period-2 taxes, means that the period-1 incumbent should structure investments in fiscal capacity to influence those decisions. For example, she may overinvest (underinvest) in the income tax if she likes the income tax more (less) than the opposition, so as to encourage (discourage) the opposition in using the income tax in the future, and the more so the higher the likelihood that the opposition takes over.

The size of this effect and whether it is positive or negative cannot be determined without going into details. A specific example that may lead to underinvestments is the case of a period-1 high-wage incumbent, who might be unlikely to invest heavily in income tax compliance if she anticipates being replaced by a period-2 low-wage incumbent (see Section 5.3 for more details) who would like to engage in more redistribution.

On balance, we may therefore expect higher political turnover to diminish investments in fiscal capacity, especially if there are few executive constraints so that is low and transfers are unequally shared.

Three types of states: Following Besley and Persson (2011a,b,c), the political model of the previous section allows us to think about three types of fiscal states that can emerge, depending on the combination of political cohesiveness and turnover. For simplicity, and to focus on a specific set of issues, we will work through the case where and , so the valuations of public goods as well as earnings opportunities are identical across the two groups.

A common-interest state: As long as is high enough relative to the value of transfers, we have:

(19)

In this case, all incremental tax revenue is spent on public goods and there is agreement about the future value of public funds. We will refer to this as a case of common interests, as both groups agree that the state should be for a common purpose, either because public goods are valuable (so that is high), or political institutions are very cohesive (so that is low). In this case, we have a common-interest state, where the level of investment is driven entirely by the motive to invest in tax revenue to provide public goods. Moreover, both groups agree on the level and structure of taxation. The Euler equations for investing in fiscal capacity become identical to the benchmark model in Section 3, namely:

Political institutions do not affect these decisions since the two groups agree on policy, and the state is run with a common purpose, no matter who is in charge. Although somewhat stylized, the nearest real-world example might be what happens in a state of war, or a common external threat where common interests are paramount. We return to this theme in Section 5.3 on the value of public spending.

A redistributive state: Now consider what happens when

(20)

In this case, the marginal dollar is spent on transfers, i.e., . Moreover, the value of public funds to the opposition is . Now each group values public revenues differently and the period-1 incumbent cares about whether his group will remain in power to reap the rewards from investing in fiscal capacity which will accrue to the incumbent. The expected value of public revenues in period-2 to the period-1 incumbent is now:

which is decreasing in for all . Indeed, this value is maximized at , when . This is the case, when an incumbent faces no threat of removal and no executive constraints. The desire to build a revenue base is then based on the desire to redistribute resources toward the incumbent group.

Besley and Persson (2011a,b,c) refer to the case where a strong group-based motive to redistribute is the driving force for state building as a redistributive state. Such states thrive on low turnover and low cohesion. In the limiting case of , the Euler equations are:

Since the incumbent is guaranteed to remain in power, the strategic effect disappears.

A weak state: A weak state combines non-cohesive institutions so that (20) holds with high political instability. To illustrate this, consider what happens if an incumbent expects to lose power for sure and his successor faces no meaningful executive constraints, i.e., and . Then, the expected value of public revenues created by investments in fiscal capacity is zero! The future incumbent, i.e., the current opposition is the residual claimant on all revenue created by fiscal-capacity investments. In this special case, the fiscal-capacity Euler equations are:

Since the second and the third terms are both negative, the only potential argument for building fiscal capacity would be to influence strategically the decisions over taxation of a future incumbent, according to the first term. However, this term is negative too: because , the future incumbent (the current opposition) wants (much) higher taxation than the current incumbent. Hence, the strategic motive makes the current incumbent not want to invest at all, perhaps even destroy fiscal capacity if that is a feasible option.

While we have illustrated this mechanism for an extreme case, the logic is much more general. Political instability and little political cohesion (weak executive constraints) generally mean that the incentives to invest in fiscal capacity are very weak, so we expect tax compliance and hence tax revenues to stay poorly developed under these conditions.

One bottom line of this discussion is that we should expect countries that have operated on more cohesive institutions in the past to have a higher stock of fiscal capacity today. Besley and Persson (2011c, chaps. 2 & 3) show that this is indeed the case, when fiscal capacity is measured in different ways and cohesive political institutions are measured by executive constraints. Political instability is harder to measure in a convincing way, but there seems to be some evidence that more stability is correlated with higher fiscal capacity.²²

Figure 14 illustrates the relationship between current fiscal capacity and past cohesive political institutions using a partial correlation plot. As a measure of cohesiveness, we use the history of the strength of a country’s executive constraints from 1800, or its year of creation, up to 2000. As in Section 3, the data come from the Polity IV database, specifically the variable executive constraints measuring various checks and balances on the executive. Following the theory outlined above, the underlying regression controls for the value of public spending, through measures of ethnic fractionalization and (past) external wars, and the degree of political instability, through measures of openness and competition in the selection of the executive. We see a very clear upward slope in the regression line, consistent with the argument in this section—countries with a history of more cohesive institutions appear to have built more fiscal capacity.²³

Social Structure: The two-group model with common valuations of public goods and identical wages has served well to illustrate some key points. But clearly, it misses a lot in terms of social structure which may affect the struggle for power. Indeed political struggle often has different values concerning public spending and/or an unequal distribution of resources at its heart. We now briefly explore these issues and their implications for investing in fiscal capacity.

Group size and elite rule: Suppose the two groups in the specialized model have different size but face the same institutions, as represented by , when in office. Observe that , as defined in (13), is larger for the minority group than for the majority group, meaning that transfer behavior becomes more cohesive when larger groups are in office, as their value of extracting a dollar in transfers is much lower than for a small group. A lower value of means that the condition

for all spending to be on public goods is more easily fulfilled. Thus, there is a greater chance the state pursues a policy in the common interest when large groups hold power. In view of this, majority rule is more likely to stimulate the build-up of fiscal capacity than minority rule. Indeed, if a country is governed by a small elite, it seems rather unlikely that a common-interest state will emerge—instead the state will become redistributive or weak.

The same basic effect emerges when a leader rules on behalf of a small elite rather than on behalf of her group as a whole. When such narrow elites alternate in power, it is difficult to create common interests in the use of public resources. Therefore, the value of political reform that raises toward 1 can be particularly strong in such countries. Similarly, measures which reduce the agency problem between elites and rank and file group members could also push a country toward a common-interest state.

Income inequality: The discussion in Section 5.2, abstracted from heterogeneity in earnings. To focus on this, we now look exclusively at income taxation and assume away all other forms of heterogeneity, e.g., in group size or preferences for public spending. Using the standard logic from Romer (1975), Roberts (1977), and Meltzer and Richards (1981), we would expect a low-income group to prefer a higher rate of income taxation than the high-income group, due to the redistributive effect of income taxation whether it is spent on public goods or transfers. We might also expect these policy preferences to translate into different incentives to invest in fiscal capacity to increase income taxation.²⁴

These mechanisms are most simply illustrated in the case of a common-interest state where (19) holds. Suppose we specialize the model to two groups, , where stands for “poor” and for “rich” with . By dropping the time subscripts, we are assuming that the wages of the rich and the poor stay constant over periods 1 and 2. In the case with constant elasticity of labor supply, the income tax preferred by group (assuming an interior solution) becomes:

(21)

where is the ratio of group ’s labor income to average labor income. For the rich to want a positive income tax, the value of public spending as represented by has to be great enough. Clearly, we have

in general. We can now say something concrete about the strategic effect in Eq. (18) provided we assume, as the discussion following Eq. (5) suggests, that

This says that an incumbent of any type would wish to implement a higher rate of income taxation if there is greater fiscal capacity in the income tax. Recall that this effect arises because greater fiscal capacity in the form of income tax enforcement raises the marginal yield from any given statutory income tax rate. The strategic effects become:

To see why, take the case of a rich incumbent. If the poor group takes over in the future, then it will tax too much from the viewpoint of the rich, so a rich group contemplating being in future opposition would gain a strategic advantage by lowering fiscal capacity. In the same way, a poor incumbent contemplating being in future opposition would gain a strategic advantage by pushing investment in income tax capacity further. These incentives are larger the higher is income inequality and the larger is political instability. The logic is the same as the one that makes a right-wing group want to impose a larger debt on a left-wing successor, and a left-wing group to impose a smaller debt on a right-wing successor, in Persson and Svensson (1989).

It is perhaps unsurprising that inequality creates a conflict of interest over investing in fiscal capacity, which mirrors the conflict of interest over the tax base itself. However, the patterns of political control also matter to whether income tax capacity gets built. If the rich have a secure hold on power, they will invest in fiscal capacity to support public spending. However, if they fear losing power, they will invest less as their investment would encourage the poor to use income taxation more intensively in the future making the rich pay for an even larger share of public spending. If the poor are securely in power, this should encourage investment in income tax capacity. To the extent that transitions to more democratic rule lead to lower-income citizens being in political ascendancy, we should observe a tendency to build income tax capacity. This would be spurred on even further if the poor are more fearful of a reversion to elite rule. Generally, the poor have a strategic incentive to overinvest.

While we have applied this argument to income tax capacity, the same argument applies to any tax base which generates a strong conflict of interest between groups that could hold power. We have made the argument in the case of a common-interest state, where there is agreement over the disbursement of public resources. But the basic logic could equally well be applied to a redistributive or weak state.

The bottom line from this discussion is that we may expect income inequality to play an important role in the development of fiscal capacity. Given that a high level of income inequality particularly curtails the investment incentives for a rich incumbent, this conclusion is strengthened if we are willing to assume that economic power and political power tend to go hand in hand. Cárdenas (2010) considers the question empirically, using cross-sectional data for 100 + countries, and finds that political and (especially) economic inequality appears to be associated with lower incentives to invest in state capacity. In fact, he uses income inequality to explain Latin America’s generally underdeveloped fiscal capacity.

Polarization: In the political models above, we have assumed that there is a common way of valuing public goods across the two groups. However, this need not be the case. Alesina, Baqir, and Easterly (1999), e.g., have forcefully argued that ethnic conflicts may lead to polarized preferences that diminish society’s spending on public goods.²⁵ Differences in valuation may reflect, e.g., ethnic, linguistic, or religious cleavages in society. We now briefly consider the implications of such divergent views, which we think about in two different ways. First, we consider what happens when groups inherently differ in their value of public goods in a way unrelated to whether they are incumbents. Second, we consider the possibility that differences arise according to whether a group is an incumbent, since an important dimension of policy choice may be the type of public goods that are chosen.

To illustrate the first case, we suppose that with . For simplicity, we focus on the common-interest case where all public spending is allocated to public goods. Now it is clear that the marginal value of public spending depends on which group is in office. Any group for whom has a higher value of future public funds than the low-valuation group, and will therefore invest more in fiscal capacity of all types, everything else equal. For such groups, securely holding power will encourage investing. One interpretation of such heterogeneity in values may be that certain groups have stronger social capital and hence can provide public goods on their own, e.g., through ethnic or family networks. Then, arranging the public-goods provision through the state will be of lesser interest. For example, authors such as Esping-Andersen (1999) and, more recently, Alesina and Guiliano (2010) have argued that countries with strong family ties invest less in the welfare state.

To illustrate the second case, where the decision rights of being in power affects the mix of public goods when in office, we suppose that .²⁶ In this case, becomes a natural measure of the polarization in preferences. The expected value of future public revenues to an incumbent becomes

It follows from the expressions above that more polarization and higher political instability both reduce the incentive to invest in fiscal capacity of all types. Figure 15 illustrates the partial correlation (controlling for political stability, executive constraints, and external wars) between ethnic fractionalization and the share of taxes in GDP.²⁷ There is a clear negative relationship between the two.²⁸

5.3 Value of Public Spending

Our approach gives the value of collective goods a central place among the motives to build fiscal capacity. Formally, parameter affects the value of public revenues in the eyes of group . In this section, we discuss some factors that go into determining this value. Of course, in the standard interpretation, these are just fixed preference parameters. But there are strong reasons to think that they depend on factors which can be shaped by history as well as policy.

Common-interest spending and war finance: As discussed in the introduction, war has played a central role in the history of public finance. In terms of the model, external threats can help determine the structure of preferences . The threat of war may also act like a common-interest shock that moves a society close to a common-interest state, or from the status of a weak to a redistributive state (at least during a period where the threat is felt). In our approach, the mechanism is to raise the value of public revenues and make it incentive compatible to spend these revenues on public goods rather than redistribution. This allows our framework to capture the arguments made by Hintze (1906), Tilly (1985, 1990), and others. Dincecco and Prado (2010) use pre-modern war causalities to explain fiscal capacity today (measured as direct taxes as a share of total taxes), and relate GDP per capita to fiscal capacity. Gennaioli and Voth (2011) build a two-country model, where endogenous external conflict interacts with the fragmentation of political institutions and the cost of war to shape state-building motives. They then apply the insights of the model to explain the divergent paths of taxation of European states in the years between 1500 and 1800. Feldman and Slemrod (2009) link tax compliance to episodes of war. Figure 16 illustrates the partial correlation between the share of years in war from 1800 (or year of independence) and the share of taxes in GDP in 1999.

War may have other effects which are more non-standard to the extent that war actually shapes social preferences. One interpretation may be that it diminishes polarization, as citizens forge a clearer sense of national identity—see Shayo (2009) on the endogenous formation of national identity. This might translate a transitory shock to a permanent effect. Thus, war may have lasting effects in a dynamic model where fiscal capacity investments are long-lived. The fact that a country built a strong tax system during a past war may raise its long-term tax take to the extent that such investments are permanent. This could be true, for example, in countries that introduced collection of income taxes at source as a means to help finance their war expenditures.²⁹

An important aspect of income tax compliance is direct withholding of taxes from wage packets. So its introduction is an interesting discrete investment in income tax capacity. Figure 17 illustrates the introduction of withholding over time, for a sample of 76 countries for which we have been able to find data. We compare the 19 countries in this sample that participated in the second world war (WWII) with the 57 that did not. The significant increase in the proportion of countries with direct withholding among the war participants is striking, especially when compared to the non-perceptible effect among the non-participants. Although this figure represents no more than casual empiricism, it is consistent with the arguments in this section.

Identifying public projects: We could also see as reflecting the ability of governments to identify good projects. An important line of development research in recent years has been instrumental in using Randomized Controlled Trials (RCTs) to identify the value of public interventions. These can be thought of as trying to find ways of better allocating resources to public goods by identifying high-benefit interventions. (See Duflo, Glennerster, & Kremer, 2007 and Banerjee & Duflo (2009) for a discussion of the methodology.)

In our framework, we can represent an RCT as a particular form of experiment to evaluate project effectiveness. To model this, suppose there is a continuum of possible public projects indexed by where some have high returns and other low returns. Preferences for public goods are now:

where is spending on projects of type . In the absence of discriminating information, we assume that the expected return on each project is the same, such that . In this case, spending will be identical on all projects. For the sake of illustration, let us suppose that utility from public goods is quadratic, i.e., .

Suppose now that RCTs have been conducted on a subset of projects, which we assign to the interval , to establish which have high and low returns. For simplicity, suppose all projects are equally likely to be high return, or low return and that

Given the outcomes of the trials and a given level of public spending, , the government chooses three numbers—, and —to maximize:

subject to

This will lead to governments spending more on projects that have value , and less on those with value . Denote the solution by . Solving this for the quadratic case, the marginal value of public goods spending is

an expression which is increasing in , the fraction of spending in which the government is informed about returns. In words, better information about worthwhile projects raises the value of public spending. Moreover, this information effect is larger the greater is the difference between high and low returns, .

This illustrates how public interventions found through randomized-controlled trials—provided they could be scaled up to achieve large aggregate returns—might assist the creation of common-interest states. Arguably, the argument may also illustrate why Western welfare states have gradually become the engine of state development during times of peace. Creating effective public health-care systems seems like an especially important example. Such systems persist essentially because the returns are perceived as common-interest spending with high returns.

One could develop a related argument regarding improvements in cost efficiency in the delivery of public spending. In that case, there could be a role for using knowledge about best practice to enhance the value of public spending. This could include innovations in the mode of delivery or lower cost forms of delivery, such as making better use of information and communication technologies. Our modeling approach links such efficiency enhancement to the scale of demand for public goods at the expense of transfer payments.

The key point here is to point to a complementarity between creating fiscal capacity and finding better and more efficient ways of spending public resources.

Corruption: Our model assumes that all resources that are spent on public projects find their way into actual spending on public goods. But in many countries, this is a poor assumption due to high levels of corruption. Many studies, following the pioneering work by Reinikka and Svensson (2008), have shown the value of interventions which reduce corruption and increase the effective flow of spending benefiting the end users.

This argument is especially poignant when fiscal capacity is endogenous. Suppose that only a fraction of the intended spending on public goods actually finds its way into actual spending on the ground. If so, the value of public goods is

In terms of accounting, a share of the spending, , ends up in the hands of citizens who earn corruption rents. Indeed, if is a pure transfer, then corruption is also a pure transfer. In practice, corruption in this or other forms creates constituencies in favor of maintaining the status quo. In terms of our approach, if the corruption rents flow disproportionately to ruling groups, this can affect the decisions to build fiscal capacity.

To understand the implications for public finance, we ask how the parameter affects incentives to build fiscal capacity. Two broad effects need to be understood. First, we have an effect on the marginal value of spending on public goods. This depends on how

depends on . As long as the elasticity is less than unity, greater corruption reduces the marginal value of spending on public goods.

The second effect comes from the distribution of rents from corruption . If these accrue exclusively to the incumbent group, this will enhance the value of holding power—in effect, there is a blur between spending on transfers and on public goods since:

With low political turnover , corruption tends to enhance motives for building fiscal capacity, as in the case of a redistributive state above. But this effect is weakened by turnover, as in the case of the redistributive state. Moreover, even as , a redistributive motive for building fiscal capacity remains due to extra-budget transfers accruing to incumbents through corruption. To the extent that corruption rents are widely held, i.e., are not distributed toward incumbent status, these motives will be weakened.

In summary, the first effect, via the marginal value of public goods, likely cuts the value of public funds and thus reduces the motives for investing in fiscal capacity. The second effect, via corruption rents, may go the other way—at least when incumbent groups capture a large share of the rents from corruption.

Summary: The discussion in this section ties together the taxation and spending sides of the state. A requirement for building a state run on common-interest grounds is that public revenues are spent on goods valued by a wide group of citizens. In history, war has arguably been an important source of such common interests and provides a key motive for creating fiscal capacity. Our framework suggests that states which lack common interests will have fiscally weak states, all else equal. One way to foster such interests might be to improve project evaluation and to identify which public interventions work in practice. This may not only improve the use of a given budget, but may also foster endogenous increases in fiscal capacity. As we have seen, combatting corruption in public spending is also linked to the motives for building an effective tax system.

5.4 Non-Tax Revenues

Our model framework permits states to have a source of non-tax revenues, denoted , in the form of aid or natural resources. How this affects incentives to invest is clear in the first-order conditions for fiscal capacity (8). The conditions show that non-tax income matters for investments in the state through changing the marginal value of tax revenue, as represented by and .

Aid and development finance: Anticipated period-2 aid, embodied in , reduces the incentive to invest, whenever marginal spending is allocated to public goods. However, current non-tax income, , reduces costs of investing in the short term, when marginal spending is on public goods, thus boosting the incentives to invest. When the transfer motive for investing in the state is dominant, we would expect aid and resources to go into transfers having no effect on the incentive to build fiscal capacity. As a result, political institutions matter—in the way we have already discussed—by governing the likelihood that a common-interest, rather than special-interest, motives dominate politics.

This discussion justifies the standard focus of development finance on lending to government rather than handing out cash grants. Lending promotes the incentives to build an effective tax system. When public goods are valuable, a period-1 grant or loan should increase investment in fiscal capacity. Forcing repayment of the loan, thereby increasing , further reinforces the investment effect. But the incentives would be reversed in a Samaritan’s dilemma, where a period-1 failure to invest in fiscal capacity elicits more aid to be paid in period 2. This dilemma seems relevant to some debates about the situation in aid-dependent countries where part of the gain from building indigenous fiscal capacity would be taxed away in the form of lower aid.

Figure 18 looks at the relationship between fiscal capacity, measured by the total tax take, and aid receipts as a share of gross national income from the World Development Indicators database. The graph shows that the partial correlation (controlling for political stability, executive constraints, external war, and ethnic fractionalization) is negative, in line with what we would expect within the framework presented here. Of course, the direction of causality implied by a picture like this is far from clear. One justification of aid is often the difficulty that poor countries have in raising revenue domestically. Therefore, aid is unlikely to be exogenous to the process of fiscal-capacity investment.

Resource revenues: The model also gives insights into why natural resource discoveries can stifle the efforts to build fiscal capacity. A government that discovers oil in period 1 with anticipated revenues in period 2 will reduce their investment in fiscal capacity. Of course, such resource revenues may be beneficial but may necessitate a catch-up period of fiscal-capacity building and leave country vulnerable to negative commodity-price shocks.

Some data support the prediction that fiscal-capacity building is related to resource dependence. Jensen (2011) presents econometric evidence, using panel data with country-specific price indexes constructed for natural gas and oil and weighted by respective shares in total national energy production. He finds that a 1% increase in the share of natural resource rents in total government income is associated with a 1.4% decrease in the fiscal capacity of a country.

Informal taxation: The previous subsection discussed the role of corruption on the spending side of the state and touched upon the revenues generated by corruption. But corruption may also work as a direct, non-tax revenue-raising device for governments or government bureaucrats. Like explicit taxation, such informal means of extracting revenue through corruption imposes static and dynamic distortions on the business of the private sector. Here, we briefly discuss how such considerations can be brought into the approach.

Suppose that there are now two kinds of taxation on activity in period , the formal tax rate studied above and informal taxation at rate . Unlike formal taxation, we suppose that returns to corruption accrue directly as transfers to the ruling group, rather than being funneled through the public budget and subject to any checks and balances in place to constrain government spending. Moreover, we suppose that the governing group has some “informal fiscal capacity” and that it is impossible to avoid corruption. This may be extreme, but will serve us well to make a few important points. It is clear that we could extend the treatment and make informal and formal fiscal capacity more alike.

The individual budget constraint is now:

and the earnings from informal taxation are

The existence of such informal taxation affects optimal formal tax rates, as formal and informal tax rates interact for each tax base.

An increase in has a static effect in that it cuts available formal tax revenue by reducing goods demand or labor supply. This is a negative externality for formal taxation. If the motive for informal taxation is purely redistributive, as here, it also reduces the resources available to spend on public goods. There are no constraints on raising such revenues except any informal controls that may exist within a group. Incentives for informal taxation are particularly high when the revenues accrue to a small subset, an “elite,” within the ruling group. In addition to this static effect, however, informal taxation through corruption can also have a dynamic effect. Specifically, it may undermine the incentive to invest in the formal tax base, since the latter shrinks in response to informal taxation. The lower tax base therefore diminishes investments in formal fiscal capacity, paralleling the effect of a lower level of development in Section 5.1.

Unlike corruption on the spending side, we would thus expect higher corruption on the revenue side to be associated with less tax collection, everything else equal. This is confirmed in Figure 19, which plots the partial correlation between fiscal capacity (measured by total tax take) and corruption, measured by a perceptions index from Transparency International (a higher score denotes less corruption).³⁰ Countries with a higher share of taxes in GDP are also the least corrupt. Of course, as with our other correlations, this should not be interpreted as a causal relationship.

5.5 Compliance Technologies

So far, we have basically left the technology for evading taxes and for increasing compliance as a black box. In this section, we will open this black box a little to see how this can enrich the analysis. We begin with a simple model of the forces that may shape the costs of non-compliance, and then extend it in a few ways to motivate interventions to increase compliance.

A simple micro-foundation for the costs of non-compliance: The simplest micro-foundation for the evasion cost function , which plays a crucial role for tax compliance, is a variant of the classic analysis of detection and punishment. Let be a non-pecuniary punishment for non-compliance with the tax code, which is increasing and convex in the amount of evasion and let be the probability of detection, which is increasing in .³¹ Then

This is the classic Allingham and Sandmo (1972) model of evasion, except that we have supposed that punishments are non-pecuniary. To the extent that is pecuniary, it adds directly to tax revenue and would have to be added to the government budget constraint. However, this would be a fairly minor difference with little effect on the main insights and we therefore stick to the non-pecuniary punishment case.

The other important part of the compliance technology is —factors shaping the probability of being caught and facing a sanction. A raft of measures based on technological improvements in record keeping and the competence of tax authorities belongs here. It is questionable whether low-income countries generally use best-practice procedures, so there might be scope for technology transfer. At least, this seems to be a presumption motivating extensive technical development assistance, in the form of capacity building in the area of taxation.

The function also depends on the production structure, as we discussed in Section 5.1, with some kinds of economic activities intrinsically easier to monitor than others depending on the degree of formality, the need for transparent record keeping, and the use of the formal financial system.

Social norms and tax morale: The model can be used to consider the role of social norms in affecting tax compliance. Suppose that shame or stigma from non-compliance in a particular tax base depends on the average amount of non-compliance in the population as a whole, which we denote by . Thus

with , i.e., an increasing amount of evasion in the population as a whole lowers the stigma/shame from cheating. In this simple case, evasion decisions, corresponding to (1) in Section 4, will form a Nash equilibrium where:

With , we get the possibility of multiple Pareto-ranked, tax-evasion equilibria, since the reaction functions for evasion slope upwards.

This opens the door for tax culture to affect compliance. Countries with a strong culture of compliance may find it much cheaper to achieve a similar level of fiscal capacity compared to one where the norm is unfavorable. Such issues have been discussed by political scientists, e.g., Levi (1998), Rothstein (2000), and Torgler (2007).

Obviously, the simple model considered here could be modified in different directions. For instance, there could be spillover effects between different tax bases, so that common cheating on some tax base spreads by contagion and erodes compliance with other taxes. Also, the relevant reference group for the social norm espoused by some particular individual may be more local than the entire set of tax payers. Local reference groups of this sort might help explain local pockets with widespread tax evasion, like the favelas at the outskirts of large Brazilian cities in which whole communities function largely outside the formal sector.

If tax morale is important, then interventions that increase the stigma from non-compliance may be an important form of intervention to improve compliance. It may even make sense to increase the visibility of compliers and to associate compliance with social approval—see Chetty, Mobarak, and Singhal (in press). But the real and fundamental question here, about which we know preciously little, is how legal and administrative interventions interact with social norms—see, however, Benabou and Tirole (2011) for a recent interesting theoretical analysis.

Our discussion of tax morale has been speculative and sketchy. But the issue is certainly important and it is plausible that different tax cultures in, say, Sweden and Greece contribute to the large differences in their tax take. The idea of tax morale also goes to the heart of debates about state legitimacy, a concept we have not dealt with at all. However, the interactions between social norms of compliance, state legitimacy, fiscal capacity, and institutions are an interesting and important topic for further research. There is, in particular, scope for more experimental interventions which can change behavior along the lines studied in Chetty et al. (in press).

Incentives for tax inspectors: In many countries, a major problem in collecting tax revenues is the weak motivation of tax inspectors. These could reflect either low incentives to detect tax evasion or a willingness to take bribes from non-compliers if caught. Our simple model allows us to think about both issues.

Suppose that detection of evasion requires that inspectors put in effort . Such effort increases the chances of catching a non-complier, but is privately costly to the tax inspector. Denote the probability that an evader is caught by with . For any given tax base and level of fiscal capacity, let equilibrium non-compliance be

It is easy to see that is decreasing in . Let now be the private profit per capita from non-compliance when tax inspectors put in effort .

An important question, on which much tax administration has tripped up, is what motivates inspectors to put in such costly monitoring. A traditional view is that this is taken care of by some kind of pro-social motivation, i.e., inspectors are intrinsically honest. But as governments have learned to their cost, this cannot be taken for granted.

Assuming that inspectors have to be compensated for their disutility of labor in a competitive labor market, the socially optimal level of tax-raising effort is:

(22)

where , as above, is the marginal value of public revenues. The maximand includes two terms—the private non-compliance profits and the value of tax revenues net of effort cost. Higher effort will reduce the first term and increase the second term and the balance between the two will define the optimum.

The main question is how the government can implement such an optimal effort level. If is not observed, inspectors face a potential moral-hazard problem—see Mookherjee and Png (1995) and Besley and McLaren (1993) for studies along these lines. If the tax inspector were offered a fixed wage and is not strongly intrinsically motivated, he would set . In this case, there would be no point in employing inspectors at all. In this framework, we can think of changes in fiscal capacity as corresponding to alternative ways of organizing the tax-collection service to avoid this outcome.

One regime would be to focus on recruiting tax inspectors who set by establishing some kind of rigorous recruiting and training regime. Such merit-based professionalization of the bureaucracy is certainly a feature of fiscal history.

Another possibility would be to contemplate tax farming, a popular solution in historic times where tax inspectors are sold a franchise to collect taxes on a particular tax base, in exchange for becoming a residual claimant. In this case, we would expect:

Comparing this to the expression in (22), we see that tax farming would never be optimal in our framework, for reasons that make sense given its somewhat checkered history. Specifically, tax farming would lead to too much effort in extracting taxes, as tax farmers would fail to internalize the utility costs they impose on the public. In practice, tax farmers would have tended to use brutal methods of collecting taxes, ignoring most of the costs to the populations from whom they were collecting.

Another option would be to pay tax inspectors efficiency wages, as discussed in Besley and McLaren (1993). To see how this might work, assume inspectors are themselves subject to inspection in a hierarchical structure. Suppose that inspectors are asked to put in effort and that the probability a tax inspector is monitored and caught is . Finally, assume that if inspectors are caught, they are fired without being paid. Now, an inspector will put in effort at a wage of if:

Solving this inequality says that the wage needed to elicit effort is an increasing function:

Compared to a benchmark model with observable and contractible effort, getting effort is more expensive, meaning that the level of non-compliance will be higher for any . However, compared to a world which relies entirely on public spiritedness, or a world where , this could be a worthwhile proposition.

There is little work to date that has explored empirically how changing incentives for tax inspectors can change revenue collection. However, recent ongoing work by Khan, Khawaja, and Olken (2013) is exploring such issues using experimental interventions for wage and incentive schemes for property tax collectors in Pakistan. This is an area where there is future innovative work to be done.

Corrupt tax inspectors: Now consider how the possibility of corruption affects these arguments. Suppose for the moment that the level of effort put into detection is fixed. After detection, however, a bribe of can be paid by an evader to the inspector, which exempts the evader from suffering the punishment . Assume that the inspector and the evader engage in Nash bargaining, so that the bribe paid is:

In this case, higher penalties for non-compliance are partly transferred to tax inspectors since they grant the ability to tax inspectors to extort money from non-complying taxpayers.

Somewhat paradoxically, bribery can motivate inspectors to put in greater detection effort since their payoff is:

Moreover, this effort is sensitive to tax rates, with greater taxes actually motivating tax inspectors to put in greater effort. This suggests the possibility that efforts to reduce bribery in a world with a great deal of unobserved effort need not necessarily increase tax compliance. This is not to say that bribery should be condoned, but that we need to consider the full set of incentives in a second-best world. Note also that if some component of fiscal capacity is independent of the incentive scheme for the inspectors’ own efforts (such as income tax withholding), then higher independent capacity raises equilibrium effort by complementing the inspector’s own efforts.

Exploiting local information: How far should local information be harnessed in improving tax compliance? Focusing on the formal inspection process underestimates the scope for schemes, which fall under the heading of “cross reporting”. This has become very important in the development literature on peer monitoring in micro-finance but has received less attention in the taxation literature. It has been brought to the fore in a recent paper by Kleven et al. (2009). Moreover, Kleven, Knudsen, Kreiner, Pedersen, and Saez (2011) show that, even in a country like Denmark with large fiscal capacity, income that is not reported using third-party enforcement is susceptible to underreporting.

The main idea exploits something well known in mechanism design, namely that once more than one person is informed about something then a variety of means can be used to illicit that information (see Maskin, 1999 and Moore & Repullo, 1988, among others). This has an obvious counterpart in taxation.

The following canonical model illustrates the idea. Suppose that evasion activity is observed by two parties—whom we can call a purchaser (denoted by a subscript ) and a vendor (denoted by a subscript ). Suppose that the vendor is asked first to declare and then the purchaser either agrees or disagrees. If there is disagreement, the government audits the transaction and the honest party is given a small reward and the dishonest party has to pay .

In the unique subgame perfect equilibrium of this game, there is full compliance as long as . In other words, it is as if there is complete and costless auditing of transactions—the gamble in the traditional Allingham and Sandmo (1972) model goes away. This simple “mechanism” is simply an illustration of the potential power of cross-reporting. However, it only works under two key assumptions. First, there is no scope for collusion between the vendor and purchaser. Second, both parties to a given transaction can be observed and verified. The latter is true when there exist formal contracts of employment or purchase, or where a receipt or record of the transaction is kept.

The evidence on informal taxation in Olken and Singhal (2011) suggests that traditional societies have developed ways of mutually raising revenues using local information and enforcement. In modern economies, firms have taken on this role as argued by Kleven et al. (2009).

Tax remittance by firms: Our model of compliance has focused exclusively on individual decisions. But in reality much of the onus for compliance is on firms and the gains from non-compliance accrue to firm owners.³² Substantively, this makes a difference when the size of the firm affects its non-compliance costs due to the visibility of its operations. It is straightforward to incorporate this feature into our model if we introduce some factor which affects the size distribution of firms. Consumers are assumed to face the same post-tax price but decisions not to comply are now only made at the firm level.

For simplicity, we focus only on commodity taxes and suppose that each industry comprises firms indexed by . At each date, we suppose that firms service demands (expressed in per-capita terms) of from consumers for good at date , where , i.e., the firms exhaust the market. We further suppose that all firms have identical costs of production denoted by . Now let be the firm’s cost of non-compliance which we denote by . We allow this cost to depend on the firm’s sales because this makes evasion more visible.

Now, a firm’s profits are:

This is the sum of profits from sales and profits from evasion. Such pure profits have to be distributed across citizens in proportion to their ownership of firms. If we assume competitive pricing, i.e., , then the only pure profits are from tax non-compliance. However, notice that nothing essential is changed from the basic model if we define:

to be the total profits from evasion in firms. Note, however, that we have put the value of sales to be an upper bound on evasion, so firms cannot evade more tax than they could in principle collect from final consumers. It is clear that the size distribution of firms can now make a difference to compliance.³³ A firm with small sales could find it worthwhile not to comply at all if

This condition would endogenously create informal-sector firms in our framework. If there are some large firms, these would tend to comply more with taxes as they are more visible to tax authorities. And we might even have firms who deliberately reduce their sales to a point they benefit from reduced compliance, in the spirit of Dharmapala, Slemrod, and Wilson (in press).

This extension shows why changes in the economy that make it optimal for firms to grow will also change the structure of compliance. For example, in an oligopoly model with firms producing a good , firm 1 would normally decide to price a shade below the others and take the entire market. However, with endogenous tax compliance, it may prefer to divide the market with other firms in order to remain less visible.

Firm-to-firm transactions provide a further interesting issue and are central to adoption of a VAT. This could be incorporated by supposing that some fraction of the cost is based on purchased inputs that carry a tax but which are deductible against taxation of final goods. This gives the firm an incentive to comply in order to claim back input taxation. It also increases the scope for cross-reporting as we discussed above. We could then suppose that the cost of evasion depends on the extent to which a firm is claiming a rebate on taxes on inputs which raises the marginal cost of non-compliance of taxation on final goods. The evidence from a clever field experiment, in cooperation with the Chilean tax administration reported in Pomeranz (2011), suggests that cross-reporting in the value-added chain indeed helps enforce payments of the VAT except when the majority of transactions do not have firms but consumers (who cannot deduct) on the other side. There is much that could be done using administrative and other kinds of data to further explore compliance with VAT at the firm level in developing countries.³⁴

More generally, and as argued by Gordon and Li (2009), when firms start to use formal financial markets, their costs of non-compliance rise and this increases the feasible array of tax bases. In the simple model above, suppose firms have to publish accounts to inform outside investors. When these accounts declare profits, , this is a useful piece of information to tax authorities who care about estimating for tax purposes. It is a short step from this to having accountants serve as agents of revenue authorities, and to report data needed for VAT and income tax compliance. Of course, there would be difficulties in preventing collusion between firms and accountants. But it is clear that the need to raise external funds will put pressure to limit such collusion. This example further highlights how economic development may support compliance. But it also illustrates the complementarity between legal and fiscal capacity. Building legal structures to protect outside investors and to demand transparent dealings in financial markets creates positive spillover effects on fiscal-capacity building.

Summary: Many factors shape the costs of non-compliance with statutory taxes. A wider range of micro-studies should investigate these issues. Revenue authorities interested in reducing non-compliance have strong incentives to work with researchers to increase knowledge as means of improving policy. The nascent movement toward field experiments and data collection that we have mentioned is likely to become an important research field, which should eventually provide a better understanding of how to more effectively raise broad-based income taxes and the VAT.