Taxability and the Social Contract

Scott Gehlbach∗

University of Wisconsin - Madison and CEFIR

August 4, 2003

∗Prepared for delivery at the 2003 Annual Meeting of the American Political Science Association, August28 - August 31, 2003. Copyright by the American Political Science Association. I am grateful for usefulfeedback from seminar and conference participants at CEFIR, SITE, the October 2002 workshop of theProject on Honesty and Trust in Budapest, Notre Dame, Harvard, Columbia, NYU, Yale, Wisconsin, andChicago. In addition, I received many thoughtful suggestions from George Akerlof, Henry Brady, GeorgeBreslauer, John Earle, Sergei Guriev, Matthew Rabin, David Ralph, Jim Robinson, Konstantin Sonin, andEkaterina Zhuravskaya. Geraint Jones was of great help in providing answers to various questions aboutthe BEEPS data. Any errors are my own. Institutional support was graciously provided by CEFIR, andfinancial support by the Fulbright-Hays DDRA, the IREX IARO, and the SSRC IDRF programs. Commentswelcome.

Abstract

A series of formal models is developed to explore the possibility of an efficient social

contract between a revenue-maximizing state and economic actors who benefit from

some good or service the state can provide. Central to the analysis is the interaction

of two factors: the ability of the state to commit to leaving economic actors with a

portion of their production, and the ability of those actors to hide production from

the state. Economic actors have an incentive to hide revenues from the state when

the state is unable to commit to leaving behind a portion of unhidden production, and

need to be compensated for forgoing that option when the state can commit. Given

that some actors find it easier to hide revenues than others, revenue-maximizing states

will typically favor economic activity which is more “taxable.” This result — that an

efficient social contract will be impossible, even when the state is able to commit —

fails to hold only under a pair of conditions probably not reflective of empirical reality.

Application of these models to the analysis of survey data finds patterns of revenue

hiding and state support strongly suggestive of commitment failures in postcommunist

states, and weakly so of a state which has the ability to commit, albeit only to an

inefficient tax.

2

“Corruption and overregulation are killing my business.”

“Good entrepreneurs know how to avoid paying taxes.”

Two statements often heard in conversation with entrepreneurs in postcommunist countries.

1. Introduction

Modern political economy has revisited and revised the theory of the state developed by the

contract theorists of the seventeenth and eighteenth centuries. As in that earlier era, the

state or sovereign is seen as providing security or some other public good that individuals

cannot provide for themselves, and for which individuals are willing to enter into a social

contract and surrender some portion of their property or liberties. North (1981, p. 23), for

example, suggests in the modern theory’s paradigmatic formulation that “the state trades a

group of services, which we shall call protection and justice, for revenue.”

However, today’s theorists - influenced by the vast literature on “transaction costs” in-

spired by Coase (1960) - are decidedly more skeptical than their forerunners of the scope for

efficient bargaining between the state on the one hand, and individuals on the other.1 The

inherent problem with a contract to which the state is a party, of course, is that there is

typically no third party to enforce the contract. “A state with sufficient coercive power to

[enforce contract and property rights and provide public goods] also has the power to with-

hold protection or confiscate private wealth” (Greif et al 1994). This potential for ex-post

1For a review of the “New Institutional Economics of the State,” see Furubotn and Richter (2000, ch. 9).

3

opportunism on the part of the state implies that a “political Coase theorem,” in which the

state and private entities bargain to an efficient outcome, will often not apply (Acemoglu

2002; see also Robinson 1998).

Compounding this commitment problem is the fact that information asymmetries be-

tween state and society are likely to be large, especially in the context of revenue collection,

where the cost to the state of fully ascertaining taxpayers’ ability to pay and of monitoring

compliance is prohibitive (Levi 1988).2 In particular, if the state is unable to commit to

leaving taxpayers with a portion of their income, then taxpayers may have an incentive to

underreport revenues, even when hiding revenues from the state is costly. In what follows,

we refer to the costliness of hiding revenues, i.e. the ease with which the state may extract

revenues from economic agents, as the taxability of economic activity.3

One immediate implication is that a Pareto-improving trade of revenues for some publicly

provided good might never be consummated. Another is that the extent to which state and

society fall short of the Pareto frontier may depend on both the degree to which the state

can commit and the nature of the information asymmetry between state and society. In

particular, when the public good provided by the state enhances economic productivity, the

provision of that good may be influenced by the taxability of economic activity, as well as

by the nature of commitment power.

2On “Information and the Coase Theorem,” see Farrell (1987).3Problems of collective action also reduce the scope for efficient bargaining between state and society.

See Dixit and Olson (2000). In abstracting from this issue, this paper demonstrates the limits to efficientcontracting even when society can organize to negotiate a social contract.

4

This paper explores these issues formally, presenting four variations on a model featuring a

ruler and a firm. (Reducing “society” to a single economic actor is an analytical convenience:

we are interested in what the model says about the incentive to provide differential support

to firms or sectors which differ in their taxability, which we parameterize in the model.) The

former has the ability to provide support to the latter, but will do so only to the extent that

she expects a return on her “investment.” Since support enhances economic productivity,

that return may come in the form of revenues collected following production. However, only

revenues unhidden may be collected, so both the taxability of economic activity and the

extent to which the ruler is able to commit to leaving the firm with a portion of unhidden

production will affect the provision of support.

As will be seen, in all but the most limiting scenario the ruler will have an incentive to

provide more support to firms which are more taxable. Economic actors have an incentive

to hide production from the state when the ruler is unable to commit to leaving behind a

portion of unhidden production, and need to be compensated for forgoing that option when

the ruler can commit. Given that some actors will find it easier to hide revenues than others,

revenue-maximizing rulers will typically favor economic activity which is more taxable, i.e.

favor those firms which find it harder to hide production from the state. When the ruler

cannot commit, she anticipates that the amount of revenues she can seize after production

takes place will be greater with more taxable firms. Consequently, she will provide more

support to firms which are more taxable. When the ruler can commit, firms may avoid

5

revenue hiding in part or full in return for a promise to leave behind a portion of unhidden

production. However, more must be left behind for firms which find it easier to hide,

potentially reducing the return to the ruler of providing support. This will not be the case

only under a pair of conditions that leave the ruler the residual claimant on any production

resulting from state support: the ruler must be able to commit to a particular lump-sum

tax that provides the firm with an incentive to not hide, and the productivity of hidden

production in the “informal sector” must be unaffected by state support. It is likely that

these two conditions are not simultaneously met in real-world institutional settings.

With the exception of this limiting case, the models presented below suggest that social-

contract failures will be present even in institutional environments in which states have

invested in their credibility. Beyond this basic result, however, the models vary widely in

their predictions about the nature of revenue hiding and provision of state support. In

particular, ruler and firm will fall farthest short of the Pareto frontier when the ruler is

unable to commit to leaving the firm with a portion of its production. Empirically, we

are therefore interested in knowing not only whether state support is indeed related to the

taxability of economic activity, as the models presented below generally predict, but also

whether that bias arises in a context of state commitment or noncommitment.

These questions are explored in the analysis of data from a survey of firms in 25 post-

communist states. As discussed below, postcommunist countries are a particularly apt

environment in which to explore the impact of taxability on state support of economic activ-

6

ity, as the information asymmetries that plague revenue collection in any corner of the world

are particularly acute for postsocialist states, which may also suffer from serious commit-

ment problems. As will be seen, firms across the postcommunist world exhibit patterns of

revenue hiding and state support broadly consistent with a model in which the state cannot

commit to leaving firms with a portion of their production, and moderately consistent with

a model of commitment power but inefficiencies in taxation. In particular, firms which an

outside observer might expect to be less taxable — small firms, firms in sectors dealing in

cash, etc. — do indeed report higher levels of revenue hiding. Further, firms which hide are

less likely to receive state support: they are more burdened by corruption, are less able to

appeal administrative violations to higher authorities, are less likely to have their contracts

and property rights enforced, and are less likely to say that local governments are supportive

in general.

In providing theoretical underpinnings to the argument that the state will tend to favor

economic activity which it finds easier to tax, this paper builds on a substantial literature

that places the state’s desire for revenues at the center of analysis. North (1981), for

example, bases his analysis of economic history on the premise that states are interested in

maximizing revenues; Tilly (1990), Skocpol (1979), and the various essays in Evans et al

(1985) motivate the assumption of state autonomy by emphasizing the revenue needs of the

state; and Brennan and Buchanan (1990) focus, as does this paper, on the credibility of any

promise by the state to leave economic actors with a portion of their production. Further,

7

as in this paper, many studies of the politics of taxation emphasize that economic sectors

differ in their taxability (see, e.g., Levi 1988 and Lieberman 2001), a consideration which

plays a role in a number of analyses of business-state relations, including the literatures

on fiscal federalism, hybrid ownership forms in China, and the “resource curse.”4 The

models presented below expand upon these insights by examining the precise conditions

under which state support of economic activity will depend on the taxability of that activity.

To a substantial degree, then, this paper can be read as a formalization of North’s statement

that a ruler will “attempt to act like a discriminating monopolist, separating each group of

constituents and devising property rights for each so as to maximize state revenue” (1981,

p. 23).

Beyond any theoretical contribution, this paper complements the existing literature in

that the evidence presented is quantitative rather than qualitative in nature. While both

methods of empirical inquiry are important, there has been little statistical evidence to

date of social-contract failures and their consequences. That evidence which does exist is

largely based on data at the level of political units. Jin and Qian (1998), for example,

provide evidence that the presence of township-village enterprises in China (which are more

4On fiscal-federalist arrangements, see Oi 1992, Qian and Weingast 1996, and Zhuravskaya 2000. Ontownship-village enterprises and the impact of local-government retention of revenues, see Che and Qian(1998) and Gordon and Li (1997). With respect to the resource curse, whose multidimensional features andvast literature are beyond the scope of this paper to summarize, Shafer (1994) argues that countries withlarge natural-resource sectors or similar “inflexible leading sectors” will develop “specialized tax authoritiesto tap the huge, concentrated revenue streams such sectors produce, and specialized agencies to monitor,regulate, and promote the activities of these few critical firms” (p. 13), a logic which speaks to both thetheoretical work below and the experience of many postcommunist countries.

8

“taxable” than private firms since they are municipally owned) helps governments to satisfy

their objectives of increasing government revenues, employment, and rural income, while

Acemoglu, Johnson, and Robinson (2001, 2002) demonstrate that the choice of institutions

in former European colonies is driven by historical circumstances, rather than by what would

maximize income today.

The paper proceeds as follows. Section 2 further defines the two key terms used in this

paper: “taxability” and “support.” Section 3 develops the model. Section 4 presents the

results of the empirical analysis. Section 5 concludes.

2. “Taxability” and “Support”

Before proceeding, it may be useful to provide further clarification of two key terms in this

paper: the “taxability” of economic activity, and state “support” of that activity. The body

of this section presents a classification of phenomena associated with these two concepts. For

reasons of space, motivating examples are relegated to the footnotes.

By “taxability,” what is meant is the ability of the state to extract revenues from economic

actors, or conversely, the costliness to economic actors of hiding revenues from the state.

Such revenues might take one of two forms:

• Taxes. Most obviously, the state extracts revenues from economic actors as taxes.

In any economic system, it will be easier for some actors to hide revenues from tax

authorities than for others. For example, the scale and nature of production, as well as

9

the ability of the state to control key bottlenecks (such as pipelines, ports, and border

crossings), makes it especially difficult for natural-resource firms to hide production

from tax inspectors. In contrast, small businesses, especially those operating in retail

or personal service and thus dealing substantially in cash, find it relatively easy to

avoid the attention of tax authorities.5

• Non-tax revenues. The state or its agents may be able to claim a share of economic

actors’ production through non-tax means, including dividends on state ownership,

payment in kind for government services, or off-budget payments. While the first

of these three revenue sources obviously depends on the state’s having an ownership

stake in a business, the control and information provided by ownership may facilitate

the latter two as well. In addition, as with tax revenues, it may be easier to extract

non-tax revenues from large or especially visible firms.6

5Russia is an excellent case in point. As discussed by Easter (2002), the postcommunist tax system inRussia developed around the relatively simple taxation of natural resource companies. In 1997, the fuel andpipeline sectors in Russia provided a quarter of total tax revenues while accounting for only 12 percent ofGDP (MacFarquhar 1997). In contrast, the small-business sector generally stayed below the radar screen oftax authorities. Irina Khakamada, at the time head of the Russian State Committee for the Support andDevelopment of Small Enterprises, told an interviewer that she had been told by entrepreneurs that therewas no need to change the tax system, as “We evade all taxes ... The government won’t get our moneyanyway” (interview on Radio Mayak, December 19, 1997, quoted in Gustafson 1999).

6The non-tax return to ownership of real estate and other assets by the city of Moscow seems to take thelatter of two forms, rather than the payment of formal dividends (personal communication with securitiesanalysts, January 2002; personal communication with hotel industry representative, April 2002). (In thisand subsequent references to personal conversations, names are withheld in agreement with the Committeefor the Protection of Human Subjects at UC Berkeley.) My own analysis of publicly available budget datashows that despite substantial city holdings in real estate and other businesses, property income (dokhodyot imushchestva) in Moscow made up only 3.9 percent of total revenues in 1999. Non-tax revenues fromlarge private businesses have also been an important source of funding for the city, most visibly in the formof donations for the reconstruction of the Cathedral of Christ the Savior (see, e.g., Jensen 2000). Similarly,a large number of the contributors to a fund to finance the reconstruction of a palace in St. Petersburg for

10

Typically, tax revenues will be larger in scale than non-tax revenues, but non-tax revenues

(in particular, off-budget payments) may be valued more highly by state officials. At

the risk of potential confusion, the term “taxability” is chosen over the more cumbersome

“extractability.”

“Support” in this paper means any costly action taken by the state which contributes to

the profitability of economic activity. Such actions can be divided into two broad categories:

• “Helping hand.”7 The state may provide positive assistance to firms and other eco-

nomic actors by enforcing contracts, creating the necessary legal infrastructure for

markets to exist, and otherwise proactively encouraging economic activity. Given

constraints on time and political capital, some sectors or firms may receive more at-

tention than others.8 Alternatively, though less fashionably, the state may provide

direct assistance in the form of investment funds or goods.9

government use are enterprises in which the federal or regional governments have ownership stakes. See,e.g., “Kozhin prosit eshche $50 mln na dvorets,” Vedomosti, August 5, 2002; “Renessans Konstantinovskogodvortsa,” Rosbalt, February 28, 2003.

7Shleifer and Vishny (1998) provide the helping-hand/grabbing-hand metaphor.8Lane (2001, p. 117) notes that the tax importance of the fuel and power sector in Russia contributes

to the large number of government ministries with an interest in the sector. This in turn implies that it isvery difficult for other sectors to get the attention of state officials. As Weinthal and Jones Luong (2002)note, quoting a Russian financial analyst, “In Russia, nothing happens in legislation unless it is driven by theoil and gas companies.” (Obviously, other factors also contribute to the importance of the fuel and powersector, and more generally to sectors which may be more taxable. The empirical analysis below attemptsto isolate the revenue impact through multivariate regression. Weinthal and Jones Luong’s focus on therole of the energy sector in pushing for general tax reform suggests that the interests of a leading sector mayat times be aligned with those of other sectors of the economy. In contrast, Slinko et al (2003) find thatconcentration of legislative favors for large businesses at the regional level in Russia is negatively associatedwith small-business growth.)

9In the context of this paper, it is interesting to note that the large-scale provision of agricultural ma-chinery in the Soviet Union took place only after collectivization increased the ability of the state to extractproduction from the agricultural sector. While such machinery may have been difficult to utilize with small

11

• Restraint of the state’s “grabbing hand.” In many political economic environments,

the status quo is characterized by corruption, overregulation, and other forms of rent

seeking. At some cost (in time, money, or political capital), state officials can reduce

the impact of this behavior on economic activity. For example, the senior officials

that make up the state may engage in costly monitoring of lower-level bureaucrats to

reduce corruption. Alternatively, if senior officials are themselves on the take, the cost

of reducing corruption may be the opportunity cost of rents foregone. In either case,

intervention will often benefit favored sectors or firms, rather than the economy as a

whole.10

Again at the risk of some confusion, the term “support” as used in this paper would

not normally apply to government subsidies to loss-making firms or tax breaks, as the net

revenue impact of such budget outlays would typically be negative. That said, if it is non-

tax revenues that are important to state officials, then it is possible that those officials will

funnel subsidies through firms from which they can more easily extract revenues.

The theoretical and empirical question of this paper is whether the first concept (taxa-

bility) is associated with the second (support). That exploration begins with the following

peasant plots, the state may also have seen it worth its while to invest in the sector only after guaranteeinga return on its investment. On the timing of state investment in the agricultural sector and collectivization,see Davies (1998, pp. 57-58).10In Russia, this seems to be very much the case for businesses which are more “taxable” in the sense of non-

tax revenues. For example, bureaucratic protection is one of the benefits associated with ownership of realestate by the Russian Foreign Ministry’s sizeable property company (GlavUpDK ) (personal communicationwith Moscow property manager, April 2002, who stressed that there are also costs associated with suchownership).

12

section.

3. Theory

In this section, we consider variations on a general model of state support of economic

activity, analyzing in turn four scenarios which differ according to the state’s ability to

commit to not taking all observable revenues from a firm following production; to whether

or not state support is useful for production in the “informal,” i.e. untaxable sector; and

to whether or not taxation takes the form of a lump-sum transfer or a proportional tax

on observable revenues. As we will see, state support will typically be increasing in the

taxability of economic activity for all but the limiting case where the state can commit to

extracting no more than a lump-sum tax and where state support does not augment hidden

production. The models differ substantially, however, in their predictions about the division

of production between taxable and untaxable economic activity, a fact which will be useful

in the empirical work which follows in the next section.

3.1. Model With No Commitment

Consider a model with two players: a ruler and a firm. The firm benefits from state support

of its economic activity, which may come either in the form of a “helping hand” or restraint

of the state’s “grabbing hand.” In either case, provision of state support is costly to the

ruler, so that the ruler will provide support only to the extent that she can expect a return

13

on her investment. In this model, that return comes in the form of revenues.

Formally, the ruler chooses a level of state support e ∈ [0,∞), incurring a cost from

that support of c (e), where ce, cee > 0. Simultaneously, the firm chooses the division of

production between the “formal” (taxable) and “informal” (untaxable) sector, with only

revenues in the former observable and hence potentially expropriable by the state. Let H

be the proportion of production “hidden” in the untaxable sector. Further, let production

in the taxable sector be equal to e, and in the untaxable sector equal to k. In Sections

2.2-2.4 we consider both the possibility that state support has no effect on production in the

untaxable sector (i.e. we consider a version of the model where k equals some exogenous

value b), and the possibility that state support augments hidden production in the same way

as it does unhidden production (i.e. k = e). However, the distinction is unimportant when

the state cannot commit.

In practice, any firm will bear some cost to hiding production, even if it is merely paying

an accountant to maintain a second set of books. Further, this cost will vary according

to the nature of economic activity. For example, it may be substantially more difficult for

pipeline operators to hide revenues from the state than it is for restaurateurs. We capture

these considerations by letting the cost to the firm of diverting production to the informal

sector be equal to k ·g(H,α), where the exogenous parameter α ∈ (α0,∞) reflects the degree

to which hiding revenues is costly to the firm. We will refer to α as the “taxability” of

economic activity. Scaling the cost of hiding by k says that for a given α and H, the

14

per-unit cost of hiding production is independent of the scale of economic activity. This

assumption will considerably ease the analysis to follow. (It should be noted that firms

of different size may still differ in their ability to hide revenues, but that this difference is

captured by the parameter α rather than k.) As discussed above, we are interested in what

the model says about the incentive of the state to provide support to actors which differ in

their taxability, all of whom might potentially be party to a social contract.

In particular, let g(H,α) satisfy the following assumptions, where subscripts denote deriv-

atives:

gH , gHH > 0 (3.1)

limH→0

gH = 0

g(0,α) = 0

gHα > 0

Assuming gHα > 0 is a form of single-crossing condition, equivalent to saying that at any

level of revenue hiding H, the marginal cost to the firm of hiding additional revenues will be

greater, the higher is α. In all that follows below, we will focus on interior solutions to the

model, which will exist so long as the following assumption is satisfied:

α > α0, where α0 satisfies gH (1,α0) = 1 (3.2)

15

Example 1. All assumptions placed on g (H,α) and α are satisfied by:

g (H,α) =α

2H2, with α > 1

Following choice of e and H, production takes place in both the formal and informal

sector, with only production in the former potentially taxable by the state. Given our

assumptions above, total production in the taxable sector is equal to (1−H) e, while in the

untaxable sector it is equal to Hk.

After production, the ruler may extract any revenues in the formal sector as taxes. When

the state is unable to commit to taking less than the total production of the formal sector,

the firm knows that it will keep only revenues hidden in the informal sector. Thus, the

firm’s choice of H solves:

maxHk [H − g (H,α)] (3.3)

For future reference, we will use the superscript NC to refer to the equilibrium in the no-

commitment case. The solution to (3.3) is gH¡HNC ,α

¢= 1, which is independent of k (and

in particular, independent of e for the version of the model when k = e, implying that HNC

is a dominant strategy). Our assumptions about the shape of g (H,α) imply that ∂HNC

∂α< 0,

i.e. the more taxable is economic activity, the less the firm will try to hide production from

the state. For example, if small firms deal more in cash, and thus find it easier to hide

revenues from tax collectors, revenue hiding will be inversely correlated with firm size.

16

In solving for the optimal level of state support, the ruler anticipates her desire to fully

take all unhidden revenues, implying the following maximization problem:

maxe(1−H) e− c(e) (3.4)

Given the firm’s dominant strategy, the optimal level of state support is given by ce(eNC) =

1−HNC. Thus, the less the the firm hides, i.e. the more taxable is economic activity, the

more support the state will provide. Following our earlier example, the model thus implies

that the state may have an incentive to provide more support to large enterprises if they

cannot hide revenues so easily as small firms.

These observations are formally stated in the following proposition.

Proposition 1. In the model with no commitment, the unique equilibrium has HNC and

eNC defined by:

gH¡HNC ,α

¢= 1

ce(eNC) = 1−HNC

Proof. Omitted.

Example 2. If g (H,α) = α2H2, with α > 1, and c(e) = 1

2e2, thenHNC = 1

αand eNC = 1− 1

α.

It is useful to compare the outcome in Proposition 1 to the first-best outcome. Assume

17

in the version of the model where k = b that b < e0, where e0 is defined by ce (e0) = 1. (No

further assumption is needed for the version of the model with k = e.) Since revenue hiding

is costly, and production in the informal sector is less productive than in the formal sector

when the ruler provides the “full” level of support, the first-best outcome is given byHFB = 0

and ce(eFB) = 1. Thus, two sources of inefficiency are present in the no-commitment case:

the firm engages in costly revenue hiding, and the state provides less than the efficient level

of support. In the following subsections, we examine the conditions under which one or both

of these inefficiencies can be eliminated.

3.2. Model With Commitment — “de Soto” Case

When the ruler is unable to commit to leaving the firm with a share of its production in the

formal sector, the firm hides revenues from the state, despite the destruction of wealth this

causes. That in turn influences the degree to which the state supports economic activity,

as the previous section demonstrated. Given these inefficiencies, it may be in the interest

of the ruler to commit to leaving behind a portion of unhidden revenues and thus induce

complete disclosure of economic activity by the firm.

To explore this possibility, we must be more precise about what constitutes hidden eco-

nomic activity. One conceptualization, popularized by de Soto (1990) and subsequently

adopted in much formal analysis (e.g. Johnson et al 1998, Roland and Verdier 1999), is that

the “informal” sector exists in a world without such state services as contract enforcement

18

and protection of property rights: rather than operating out of a store front, a trader sells

his goods on the black market, invisible to the state and its agents. Despite its popularity,

however, this assumption is strong: If the division of production between the “formal” and

“informal” sectors merely reflects tax evasion within a single sphere of activity rather than

the diversion of production to a different form of productive activity, then it is implausible

to think that such state actions as contract enforcement have no impact on the profitability

of “informal” economic activity. For example, for a given proportion of total revenues hid-

den, a restaurant which cooks its books will have more to hide if the state cracks down on

bribe-seeking fire inspectors, as will an oil firm engaging in transfer pricing schemes if the

state provides assistance in establishing foreign markets.

We begin, however, by assuming that the informal sector is truly informal, i.e. we assume

k = b using the notation above, with b exogenous. Assume as before that b < eFB. Further,

to establish a benchmark, assume that the state can commit to a nondistortionary lump-sum

tax conditional on state support e.

Examining Proposition 1, we observe that for the firm to want to fully report its revenues,

it must be given at least as much as when it engages in optimal revenue hiding, i.e. the state

must commit to leaving the firm with b¡HNC − g ¡HNC ,α

¢¢. Consider a commitment by

the ruler to extract no more than T (e), where T (e) satisfies:

T (e) = max£0, e− b ¡HNC − g ¡HNC ,α

¢¢¤(3.5)

19

Note that this tax is conditional on the level of state support, not on the level of unhidden

revenues. One example of such a tax is a user fee on government services. Another is

a real-estate tax tied to the overall value of property in a community. Further below we

consider the impact of a distortionary proportional tax on unhidden revenues.

As constructed, T (e) provides the firmwith as much as it would earn from hiding revenues

so long as e is high enough, and leaves the ruler the residual claimant on unhidden revenues.

Thus, it will be an equilibrium for the firm to choose H∗ = 0, while the state provides

the first-best level of support e∗ = eFB. Clearly, if ruler and firm can coordinate on this

equilibrium, then the ruler will want to commit to T (e), as relative to the equilibrium with

no commitment total surplus is strictly greater and the ruler gets all of the increase. The

following proposition establishes these arguments formally.

Proposition 2. When state support does not augment hidden production (i.e. k = b) and

the ruler can commit to taking no more than a lump-sum tax T (e) =

max£0, e− b ¡HNC − g ¡HNC ,α

¢¢¤, then it is an equilibrium for the firm to play H∗ = 0

and the ruler to choose e∗ = eFB.

Proof. See appendix.

Example 3. If g (H,α) = α2H2, with α > 1, and c(e) = 1

2e2, then T (e) = max

£0, e− b

2α

¤and there will be an equilibrium with no revenue hiding and state support e∗ = 1 = eFB.

Proposition 2 should be considered a limit result: under the assumptions of nondis-

tortionary taxes and no contribution of state support to informal economic activity, the

20

first-best outcome can be achieved. As we will see, this result does not hold when these

assumptions are relaxed.

3.3. Model With Commitment — Tax Evasion Case

As discussed above, it is perhaps more plausible to think that in certain institutional en-

vironments hidden revenues represent simple tax evasion rather than production in a com-

pletely different sector unaffected by state support. To consider the role of commitment

under this alternative assumption, let k = e so that total production in the hidden sec-

tor is eH and total costs of hiding are e · g (H,α). (The results in this section will be

muted but not eliminated if k = βe, for some exogenous β ∈ (0, 1).) Further, as in

Section 2.2.2, assume that the state can commit to taking no more than a lump-sum tax

conditional on e, T (e). For the firm to want to operate without hiding revenues, it must

be guaranteed at least its utility from optimal revenue hiding, which in the present en-

vironment is e¡HNC − g ¡HNC ,α

¢¢. Thus, a commitment by the ruler to take no more

than T (e) = max£0, e− e ¡HNC − g ¡HNC ,α

¢¢¤= e

£1− ¡HNC − g ¡HNC ,α

¢¢¤can sup-

port H = 0, thus avoiding one of the inefficiencies identified in Section 2.2.1. Nonetheless,

a commitment to T (e) cannot provide the first-best outcome. To see why, observe that if

21

the firm is playing H = 0, the ruler’s maximization problem is:

maxemin[(1−H) e, T (e)]− c (e)⇔ (3.6)

maxee£1− ¡HNC − g ¡HNC ,α

¢¢¤− c (e) (3.7)

Thus, the optimal level of state support satisfies ce(e) = 1−¡HNC − g ¡HNC ,α

¢¢< 1. This

failure of commitment to achieve the first-best is the result of the role of state support in

the “informal” sector. In essence, the opportunity to hide can be thought of as an outside

option for the firm. In this version of the model, the size of this option is influenced by

the level of state support e. When the ruler provides more support, the value of hidden (as

well as unhidden) production increases. Thus, the firm must be compensated more to avoid

revenue hiding, implying that the marginal return to the ruler of state support is less than

in the “de Soto” case.

The extent to which this is the case will depend on the taxability of economic activity.

In particular, the level of state support e in this equilibrium will be increasing in taxability

α. In essence, when hiding production is very costly, then the firm will be disinclined in any

event to hide revenues, implying that an increase in state support will have little effect on the

attractiveness of the “informal” sector. In contrast, when hiding production is easy, a small

increase in state support results in a large increase in the amount the firmmust be guaranteed

to avoid revenue hiding. (To see this formally, apply the envelope theorem to the firm’s

22

maximization problem in the no-commitment case (3.3), observing that HNC − g ¡HNC ,α¢

is decreasing in α.) Summarizing:

Proposition 3. When state support augments hidden production (i.e. k = e) and the ruler

can commit to taking no more than a lump-sum tax T (e)=max£0, e− e ¡HNC − g ¡HNC ,α

¢¢¤,

then there exists an equilibrium with H∗ = 0 and e∗ s.t. ce(e∗) = 1−¡HNC − g ¡HNC ,α

¢¢.

In this equilibrium, the level of state support is increasing in taxability α.

Proof. Omitted.

Example 4. If g (H,α) = α2H2, with α > 1, and c(e) = 1

2e2, then T (e) = max

£0, e

¡1− 1

2α

¢¤,

with an equilibrium involving no hiding and support e∗ = 1− 12α.

In understanding Proposition 3, it may be useful to reflect on why taxability has no

effect on state support in the “de Soto” case (Proposition 2). There, as here, the ruler must

promise to leave the firm with a larger share of its production, the less costly it is for the

firm to hide. However, in the “de Soto” case the ruler remains the residual claimant on any

revenues above the guaranteed amount, meaning that she will provide the first-best level of

support. In contrast, when state support augments hidden as well as unhidden production,

the firm captures a portion of the marginal increase in production through the additional

compensation it receives for not hiding revenues.

23

3.4. Model With Commitment — Tax Evasion Case (Proportional Tax)

Up to now we have been giving efficiency as much of a chance as possible by assuming that

the ruler can commit to a nondistortionary tax. As we have seen, even with that assumption

state support is increasing in taxability if support boosts hidden as well as unhidden revenues.

Nonetheless, even Proposition 3 predicts no revenue hiding by the firm, a result that seems to

be at odds with the reality of some degree of revenue hiding in all countries, including those

where states have some ability to commit. In this section, we assume that for exogenous

reasons the ruler may be unable to commit to a nondistortionary tax, but may commit to

a proportional tax on unhidden revenues. Such taxes are widely observed in practice, of

course, so the results of this section may be of empirical relevance.

In particular, assume that the ruler can commit to a proportional tax t on unhidden

revenues. Further, assume as in the previous subsection that k = e, so that state support is

useful in the “informal” sector. We will proceed in steps, first deriving the equilibrium for

a given t, and then solving for the optimal t from the ruler’s point of view.

Given a commitment to t, the firm solves:

maxH(1− t) (1−H) e+ e [H − g (H,α)] (3.8)

Clearly, it will be a dominant strategy for the firm to choose H such that gH (H,α) = t,

which implies an inverse function H = H(t,α). Given the assumptions in (3.1) on the

24

function g (H,α), H(t,α) has the properties Ht > 0 and Hα < 0. Meanwhile, the ruler

chooses e according to:

maxet (1−H) e− c (e) (3.9)

which together with the firm’s dominant strategy H(t,α) implies that the optimal level of

state support e∗ is given by ce (e∗) = t (1−H(t,α)).

Thus, in contrast to the previous two cases with commitment to a nondistortionary tax,

commitment by the ruler to a proportional tax on unhidden revenues implies some level of

revenue hiding, as when the ruler is unable to commit. Further, the level of state support

is decreasing in the level of revenue hiding. The following proposition recapitulates these

results.

Proposition 4. When state support augments hidden production (i.e. k = e) and the ruler

can commit to a proportional tax on unhidden revenues t, the equilibrium level of hiding

and support is given by the function H = H(t,α) (with Ht > 0 and Hα < 0) and e∗ s.t.

ce (e∗) = t (1−H(t,α)). Thus, there will be revenue hiding in equilibrium, with such hiding

negatively associated with state support.

Proof. Omitted.

What Proposition 4 does not state is how the level of revenue hiding depends on the

level of taxability if the ruler chooses t optimally. Examination of (??) shows that there

are two results, one obvious and one ambiguous. As taxability α increases, the direct cost

25

to the firm of hiding revenues increases, discouraging hiding. However, it seems plausible

that if the ruler chooses t to maximize her revenues, then the tax rate may be higher, the

more taxable is economic activity. If so, then higher taxability could indirectly encourage

revenue hiding.

To further explore these considerations, we proceed with the following extended example

(unambiguous analytical results do not seem possible without further assumptions on the

shape of g (H,α)), which demonstrates that under fairly general conditions these two effects

cancel each other out, but only for relatively low taxability α. Above a certain α, revenue

hiding is so costly that the optimal tax rate from the ruler’s point of view is t = 1, i.e.

the ruler’s behavior is identical to that in the no-commitment case. Consequently, for high

α an increase in taxability α unambiguously results in less revenue hiding, as the tax rate

remains unchanged while the cost of hiding revenues increases. In either case, the level of

state support is increasing in taxability α: when taxability is low an increase in taxability

raises the optimal tax rate without affecting revenue hiding, while when taxability is high

an increase in taxability reduces revenue hiding without changing the optimal tax rate.

Example 5. Assume g (H,α) = αxHx, with α, x > 1, and note that this function satisfies

all assumptions in (3.1) and (3.2). (Assume as in the general model that c = c(e), with

ce, cee > 0.) Then (dropping superscripts and arguments for the sake of presentation),

gH = αHx−1 = t, implying optimal hiding H =¡tα

¢ 1x−1 and support e s.t. ce(e) = t [1−H].

26

If the ruler chooses t to maximize her revenues, the optimal t solves:

maxt∈[0,1]

t [1−H] e− c(e) (3.10)

This is a quasiconcave problem, with t∗ satisfying the following first-order condition:

(1−H − tHt) e+ (et + eHHt) t [1−H]− (et + eHHt) ce ≥ 0 (3.11)

Recalling that ce = t [1−H], we observe that for t∗ < 1 (3.11) reduces to:

(1−H − tHt) = 0 (3.12)

Plugging in,

1−µt

α

¶ 1x−1− t

·µ1

x− 1¶(α)

− 1x−1(t)

1x−1−1

¸= 0 (3.13)

⇔ t = α

µx− 1x

¶x−1(3.14)

Thus, the optimal t∗ is given by:

t∗ = max

"α

µx− 1x

¶x−1, 1

#

In other words, for α sufficiently low, t∗ is increasing in α, so taxability will have an indirect

27

positive effect on hiding in addition to its direct negative effect. For α sufficiently high,

t∗ = 1. Solving for H∗(t∗(α),α), we see:

H∗(t∗(α),α) =µt∗ (α)α

¶ 1x−1

=

maxhα¡x−1x

¢x−1, 1i

α

1

x−1

=x− 1x

for t∗ (α) < 1

µ1

α

¶ 1x−1

for t∗ (α) = 1

Thus, the level of revenue hiding is locally independent of the degree of taxability α for α

such that t∗ (α) < 1, and locally decreasing in α for α such that t∗ (α) = 1. Further note that

in either case state support is increasing in α: for t∗ (α) < 1, ce (e∗) = t [1−H∗(t∗(α),α)] =

αx

¡x−1x

¢x−1, while for t∗ (α) = 1, ce (e∗) = 1−

¡1α

¢ 1x−1 .

In practice, of course, there is no point in the ruler’s committing to t∗ = 1, as this is

equivalent to not committing at all. In what follows, when referring to the model of this

section we will restrict attention to the case where the optimal tax t∗is less than 1.

3.5. Summary

There is a recurring theme to the models presented above: rulers interested in tax revenues

will generally provide greater support for economic activity when that activity is easier to

tax. For all three models in which this is the case, greater taxability implies a greater share

for the ruler of the marginal benefit from state support. The nature of that share, however,

differs according to the ability to the ruler to commit to leaving producers with a portion of

28

unhidden revenues:

• When the ruler is unable to commit at all, greater taxability increases the ruler’s share

of production at the margin by increasing the proportion of revenues unhidden.

• When the ruler is able to commit to a lump-sum tax and state support augments

hidden production, greater taxability increases the ruler’s share of production at the

margin by reducing the amount that the firm must be guaranteed to avoid revenue

hiding.

• When the ruler is able to commit to a proportional tax and state support augments

hidden production, greater taxability increases the ruler’s share of production at the

margin by increasing the optimal tax on unhidden revenues.

Thus, while the ruler and firm fall farthest short of the Pareto frontier when the ruler is

unable to commit at all,11 commitment power is not sufficient to achieve an efficient outcome

when state support makes hiding production more profitable. In essence, some portion of

the increase in production from state support must be spent to keep the firm from deviating

to the now-more-profitable informal sector. This effect is exacerbated when the ruler can

11To see this, compare Propositions 1-4 and recall that there are two possible sources of inefficiency inthe model: revenue hiding and underprovision of state support. Revenue hiding is greatest in the no-commitment case, as there is no hiding with a lump sum tax, while with the proportional tax gH = t ≤ 1vs. gH = 1 when the ruler cannot commit. Further, state support is the smallest in the no-commitmentcase: ce = 1 −HNC vs. ce = 1 in the “de Soto” case and ce = 1 −HNC + g

¡HNC

¢in the “tax evasion”

case. With respect to the proportional tax, recall from (3.9) that the ruler could always choose t = 1 as inthe no-commitment case; when she does not do so, it must be the case that the expected share of unhiddenrevenues is greater than in the no-commitment case, implying greater support.

29

commit only to a proportional tax on unhidden revenues.

Only when the ruler is able to commit to a lump-sum tax and state support plays no role

in production in the “informal” sector is taxability unrelated to state support. In that case,

the ruler is the residual claimant on unhidden revenues, so the ruler always provides the

first-best level of support. However, as discussed above, this pair of conditions is unlikely

to be observed often in practice.

Despite their similarities, the models differ substantially in their predictions about the

relationship between revenue hiding on the one hand, and state support and taxability on

the other. Thus, beyond any theoretical interest, these models provide a framework in which

to analyze the data discussed below. Firms in 25 postcommunist countries were queried on

the degree of support received from state officials, as well as the extent to which firms like

theirs hide revenues from tax authorities. Further, firms provided information on various

characteristics which might be associated with the taxability of their operations. We are

interested in what the data say about both the nature of support and state commitment in

these countries. Before proceeding it may be useful to examine Table 1, which highlights

the key empirical predictions of the models presented in this section.

TABLE 1

30

4. Empirical Evidence from Postcommunist States

Postcommunist states are a particularly appropriate environment to examine the issues dis-

cussed in this paper. While information asymmetries between state and society hinder rev-

enue collection in any state, the challenge of collecting revenues has been particularly severe

for many postcommunist countries. Socialist states raised revenues primarily through profit,

turnover, and payroll taxes collected from state-owned enterprises and funneled through the

state banking system (see, e.g., Martinez-Vazquez and McNab 2000 or Tanzi and Tsibouris

2000). Privatization, liberalization, and the growth of the traditionally underdeveloped

service sector following the collapse of communism necessitated the wholesale reform of tax

policy (Hemming, Cheasty, and Lahiri 1995)) and recreation of tax-administration systems

(Ebrill and Havrylyshyn 1999), a task still incomplete in much of the postsocialist world. In

the absence of such reform, entire economies have developed around tax evasion (Yakovlev

2000).

Further, the rulers of postcommunist states may have commitment problems to a greater

degree than their counterparts in many other parts of the world. The now-familiar “ratchet

effect” was first identified in the study of socialist economies (Berliner 1952), with firms

only just fulfilling the plan for fear of what they might be asked to do in the future if

they demonstrated their true capacity. The legacy of the failure of the state to commit to

future planning targets may be generalized mistrust of any policy pronouncement, including

promises to not raise tax rates once firms “come out of the shadows.” In more mature

31

democracies, similar commitment problems may be mitigated by the presence of political

parties, which by encouraging responsible behavior among their members enhance the party’s

reputation (Cox and McCubbins 1994.) However, reputations take time to develop, and most

parties in the postcommunist world are new. Problems of commitment may also be related

to the difficulty of collecting taxes for the state. With revenues falling short of political

demands for spending, state officials may find it difficult to credibly commit to not taking

as much as they can from taxpayers.12

4.1. Data

The empirical work in this paper is based on the Business Environment and Enterprise

Performance Survey (BEEPS) carried out in 1999 by theWorld Bank and the European Bank

for Reconstruction and Development. Through the BEEPS project, firms were surveyed on

various aspects of business-state relations, including the topics covered in this paper. In all,

4104 small and medium-sized enterprises were surveyed in 26 countries. This paper restricts

attention to firms located in the 25 postcommunist countries listed in Table 2. Firms

in Turkey were not included since Turkey is not a postcommunist country.13 As Table

12In Russia, these commitment problems plauge the collection of both tax and non-tax revenues. Taxcollection has typically resembled a “free negotation between tax inspectors and taxpayers” (Aslund 2001,p. 360), with the former under incredible pressure to increase collections (personal communication with taxaccountant, May 2002). In Moscow, non-tax revenues often take the form of side payments to individualsclose to the city administration, with the city’s ownership of real estate used as leverage to guarantee a shareof the profits from real estate development. It is not unheard of for this share to be renegotiated ex-post,after real-estate development has taken place (personal communication with investment analyst, June 2003).13Details on the survey and its implementation can be found in Hellman et al (2000) or at

http://www.worldbank.org/wbi/governance/beeps/htm. Note that while Hellman et al (2000) refers toa survey of firms in twenty countries interviewed in 1999, six countries (Albania, Turkey, Latvia, Bosnia, the

32

2 demonstrates, the revenue challenge imposed by difficulties in tax collection has been

especially difficult for post-Soviet states, where reform along many dimensions has generally

lagged that in Eastern Europe. In the regression results reported below, we control for

variation in institutional environment across states by including country dummies.

TABLE 2

Table 3 summarizes various characteristics of the firms in the sample. We are primarily

interested in the relationship of these characteristics to the taxability of the firm’s operations,

though for various reasons each variable may have an independent impact on the incentives

of the state to support economic activity of that type. For both information and control

reasons, the state may find it easier to extract revenues (including non-tax revenues, as

discussed above) from state enterprises than from private firms (Roland and Verdier 1999),

a fact emphasized in the literature on municipal support for locally owned township-village

enterprises in China (e.g., Oi 1992, Che and Qian 1998). Large enterprises are typically

more taxable than small firms — small firms deal more in cash (and thus find it easier to

hide revenues), and tax inspectors avoid smaller firms to the extent that there are economies

of scale in tax collection — and so may be favored by the state (Gordon and Wilson 1999).

Foreign firms may be constrained for reasons of reputation or home-country legislation to

more fully honor their tax obligations. Firms operating in a more competitive environment

may feel greater pressure to cut costs by avoiding tax payments, while the state may gain

Serb Republic in Bosnia, and Macedonia) were added to the original project late in that year. The data setavailable on the World Bank website is the full data set, with all 26 countries represented.

33

valuable information through its regulatory activities about the profitability of a monopolist’s

operations. Export-oriented firms may be subject to additional checks by tax inspectors,

though they may also find it easier to divert profits offshore. Finally, as discussed in Section

2, technological considerations imply that some sectors will be especially easy to tax.

TABLE 3

The enterprises in the BEEPS data set are overwhelmingly private, although there is a

non-trivial percentage of state-owned enterprises, and most are wholly domestically owned.

Further, they are not large, with a mean employment of 141, maximum employment of 1000,

and employment of 500 or less for 97 percent of firms in the sample. Firms in the sample

generally face at least moderate competition, with 77 percent of respondents reporting that

they have more than three competitors, and only 10 percent stating that they have none.

Approximately one quarter of the firms surveyed export some good directly. A plurality

of firms (30 percent) operate in the manufacturing sector, though other sectors are well

represented.

Table 3 also reports the distribution of firms across city size. Since there is no coding for

the exact location of the firm in the data set, and since it is possible that both taxability and

state support are influenced by institutional environments finer-grained than mere country of

residence, five dummy variables are included for the six town-size categories in all regressions

to control for as much intra-country variation as possible. As discussed below, all regressions

were run on the subsample of firms in capital cities as an additional robustness check, since

34

we know that each country has only one capital.

Table 4 presents our measure of revenue hiding. Firms were asked, “What percentage

of sales of a typical firm in your area of activity would you estimate is reported to the tax

authorities, bearing in mind difficulties in complying with taxes and other regulations?”14

Wording such as “typical firm in your area of activity” is standard in survey research when

questions touch on sensitive matters, and it is typically assumed that respondents answer

based on their personal experience. In our case, we can check this assumption by comparing

the covariation of this variable and various firm characteristics with publicly available data

on tax compliance, as we do in the following subsection.

TABLE 4

As can be seen, variation in responses is quite large, with a mean of 79 percent and

standard deviation of 26 percent, and only 34 percent of firms saying that they report 100

percent of revenues to tax authorities. Thus, despite possible a priori concern that firms

would be afraid to admit to any revenue hiding (by firms like theirs), the modal response

indicates some degree of tax evasion.15 In the regressions below, we invert this variable

14For this question, and the bribe and time-tax questions used as dependent variables in the OLS re-gressions, respondents were allowed to choose from a number of percentage ranges (e.g. “2-9.99 percent”)rather than asked to name a number between 0 and 100. These questions were reconstructed as continuousvariables due to the large number of possible responses, the inherently continuous (and linear) nature of theunderlying variables, and the need to construct a variable that excluded bribe payments to tax and customsofficials. Responses were coded as the midpoint of the range for the category chosen (e.g. “2-9.99 percent”is recorded as 6 percent), except for responses such as “more than 50 percent,” which were coded as the lowend of the range (i.e. 50 percent).15While this question only refers to revenues, it is likely that firms which find it easier to hide revenues

can also more easily hide expenses if they choose to do so, especially if it is the proportion of transactionsin cash that primarily determines firms’ responses. In particular, firms that operate largely in cash may beable to pay a substantial percentage of employee compensation in cash. Thus, this question likely captures

35

for consistency with the model above, with “percent of revenues hidden” calculated as 100

minus “percent of revenues reported.”

Table 4 also provides summary statistics and frequency distributions for the various mea-

sures of state support used in the following section. The first three variables are quantitative

measures of the burden of the state on business, which we recode in the regressions below so

that a higher response for the dependent variable consistently reflects more state support of

economic activity. Thus, for example, “bribe payments as a percent of revenues” is recoded

as the percent of revenues not paid as bribes.

Substantial variation is evident in all three quantitative measures. On average, entrepre-

neurs in all three postcommunist countries report paying 2.6 percent of their revenues (not

profits!) as “unofficial payments to government officials,” with the mean response in Eastern

Europe and the Baltics (1.8 percent) substantially less than that in the former Soviet Union

less the Baltics (3.4 percent). A similar pattern is noted for the variable “non-tax bribe

payments as a percent of revenues,” in which bribe payments to tax and customs officials

are subtracted out to control for the fact that firms may bribe tax officials to allow them to

underreport revenues. (Recall that we are interested in the correlation between tax report-

ing and various measures of state support, including the degree of corruption.) Finally, an

average of 9.8 percent of management time is spent dealing with government officials, with

the mean percentage in Eastern Europe and the Baltics (7.1 percent) again substantially

the degree to which firms are able to collude with employees to avoid social taxes and withholding of incometaxes, as well as their ability to evade profit taxes, VAT, and other taxes related to revenues.

36

lower than that in the former Soviet Union less the Baltics (12.6 percent).16

Among our qualitative measures of state support, the first variable — “opportunity to

appeal administrative violations” — captures in a fairly direct way the degree to which firms

expect that higher-level state officials will protect them against arbitrary behavior by lower-

level bureaucrats. The expectation that “contracts and property rights [will be] enforced” is

obviously critically important for private investment to take place, and many scholars assume

this to be one of the primary services the state can provide to the private sector. Finally,

the variable “local government helpful” proxies for the general evaluation by entrepreneurs

of the degree to which the state provides a supportive business environment. All three

qualitative measures exhibit considerable dispersion among the possible responses.

4.2. Results

In this section we empirically explore the empirical predictions of the models presented in

Section 3. As Table 1 demonstrated, the models provide similar predictions with respect

to the relationship between state support and taxability, but sharply different predictions

about the nature and effect of revenue hiding.

A first rough test of the models presented above (albeit one with little power to distin-

guish among models, as discussed below) is to regress our various measures of state support

on the firm characteristics summarized in Table 3, since each of these characteristics can be

16Much other empirical work has noted the sharp divide in the business environment between the twohalves of the postcommunist world. See, e.g., Frye and Shleifer (1997), Johnson et al (2000), and EBRD(1999). The last report uses data from the BEEPS project.

37

thought of as influencing taxability α for one theoretical reason or another. As Tables 5 and

6 show, state support is positively associated with three firm characteristics in particular

that an outside observer might assume would imply greater taxability. State ownership is

positively correlated with our various measures of support in every case but for the regres-

sion on “percent of management time not spent with government officials” (presumably the

managers must spend time with the owners of a firm whoever they are, so that managers

of state-owned enterprise will spend more time with state officials), and four of the five

positive coefficients are precisely estimated. Employment follows a similar pattern, with

the estimated coefficients significant and positive for all measures of support but the time

variable. Estimated effects for these two variables are large. For example, the probability

that a state-owned firm will respond that local government is mildly helpful is six percentage

points higher than it is for a private firm (relative to an average probability of 18 percent),

while an increase in employment from 1 to 1000 (an increase in log employment from 0 to

6.9) is estimated to reduce the percent of revenues paid as bribes by 3.4.17 Finally, the

estimated coefficient on the monopoly variable is positive for all but the regression on “time

not spent,” and is significant in four of five cases. Among the remaining variables, esti-

mated coefficients are generally imprecisely estimated and of inconsistent sign. Overall, the

primary impression given by Tables 5 and 6 is that state-owned enterprises, large firms, and

17The marginal effects in Table 6 are calculated at the means of the independent variables, and are givenas the discrete change for dummy variables. Further, due to limitations of space, effects are given for thesecond-right-most category of the dependent variable only (as Table 4 suggests, the probability of fallinginto the right-most category is small).

38

monopolies are likely to receive more support.

TABLE 5

TABLE 6

That said, the results in Tables 5 and 6 must be treated with caution, as there may be

other reasons why firms which are more taxable receive more state support. For example,

collective-action theory suggests that large enterprises may find it easier to deal with state

officials than do small firms. Further, while roughly consistent with all but the “de Soto”

model in the previous section, the regressions in Tables 5 and 6 do little to identify why more

taxable firms might receive more support. To identify the role of commitment problems and

efficiency of taxation, we need to look at patterns of revenue hiding across firms.

The fact that two thirds of firms report some level of revenue hiding, as indicated in

Table 4, suggests that business-state relations in postcommunist countries may suffer either

from the inability of the state to commit to an efficient tax schedule, or from inefficiencies

in taxation. Tables 7 and 8 further explore this possibility by reporting the results of OLS

and ordered-probit regressions of our various measures of state support on the proportion of

revenues hidden from tax authorities.

A note on identification: The theory outlined in the previous section assumes that the

ruler observes the taxability of economic activity. More practically, state officials observe

characteristics of firms and sectors which are correlated with taxability, and thus with the

degree of revenue hiding in the case when the state has no commitment power. The dis-

39

cussion above identified various characteristics observed by the interviewer in the BEEPS

survey which may be correlated with taxability. However, all of these variables may be

thought of as independently influencing the degree of support provided by the state, beyond

their influence on revenue hiding. Large firms, for example, may be able to influence pol-

icy outcomes by overcoming the collective-action problems which plague small firms, while

state-owned firms may be more taxable in the sense of non-tax revenues (our revenue-hiding

variable explicitly refers to tax revenues). Thus, the regressions in Tables 7 and 8 include

these characteristics as control variables. Note that this specification implicitly assumes

that there are additional firm- or sector-specific characteristics which:

1. are observable to the state officials making decisions about the level of support to

provide

2. influence the taxability of economic activity, and thus the level of revenue hiding when

the state is unable to commit

3. do not directly influence the optimal level of support

Put differently, these regressions assume that our measure of revenue hiding can instru-

ment itself. An alternative approach is to find a set of variables which are correlated with

taxability but have no independent impact on state support. This would be especially ap-

propriate if there were about some missing variable correlated with both revenue hiding and

state support, or if endogeneity of revenue hiding with respect to support were suspected.

40

Unfortunately, no pure instrument seems to be available in the BEEPS questionnaire. In

the discussion below on robustness checks, various subsets of the firm characteristics listed in

Table 3 are considered as instruments, with each subset addressing the tradeoff between effi-

ciency and exogeneity of potential instruments differently. As will be seen, with one notable

exception the basic qualitative results of this section continue to hold in these alternative

specifications. In any event, the theory presented in the previous section provides no strong

reason to suspect simultaneity between revenue hiding and support, while as discussed below

the most obvious potential omitted variables suggest results different from those obtained.

As can be seen, revenue hiding is a consistently positive, and very precisely estimated,

predictor of state support. For all six variables the estimated coefficient on “proportion of

revenues hidden” is statistically significant at the 1 percent level, and marginal effects are

large. Thus, for example, a firm which hides nothing pays almost one percent less of its

revenues as non-tax bribes than does a firm which hides half of its revenues. (Recall that

we subtract out bribes paid to tax and customs officials for the “non-tax bribes” variable,

since firms may pay bribes to reduce their tax burden.) A similar decrease in revenue

hiding increases the probability that the firm says it always has the opportunity to appeal

administrative violations by three percentage points (relative to an average probability of

10 percent) and that it mostly has that opportunity by two and one half percentage points

(relative to an average probability of 16 percent).

Overall, the results reported on Tables 7 and 8 are consistent with both the no-commitment

41

model and the model of commitment to a proportional tax, as both models predict state

support to be decreasing in revenue hiding. In the specific case of the proportional-tax

model presented in Example 5, however, revenue hiding is unaffected by taxability α, im-

plying that any variation in revenue hiding observed in practice (perhaps because we have

not completely captured institutional variation with our country and town-size dummies)

should be uncorrelated with the taxability of economic activity at the firm level. To some

extent, then, we may be able to differentiate between commitment and distortionary-tax

effects by checking whether revenue reporting is correlated with firm characteristics in a way

that suggests a positive correlation with taxability α.

Table 9 presents the results of a regression of the percent of revenues hidden from tax

authorities on the same firm characteristics used in the previous regressions. As can be seen,

firms reporting higher degrees of tax compliance are generally those that we would expect

to find it harder to hide revenues from tax authorities. Large enterprises hide substantially

less of their revenues from tax authorities, as do firms with some level of foreign ownership.

The percentage of revenues reported by monopolies is 11 points higher than it is for firms

with more than three competitors.

Further, the relationship between business sector and tax hiding follows an intuitive

pattern: firms in sectors dealing primarily in cash hide more from tax authorities than

do manufacturing firms, while natural-resource and finance firms (whose operations may be

relatively easy to track) hide less. As can be seen in Table 10, these results are consistent

42

with publicly available information on tax receipts by sector in Russia. The one significant

discrepancy — the transportation sector in Russia is a disproportionately large contributor of

tax receipts — may be explained in part by the exclusion from the BEEPS survey of pipeline

and other large transportation firms.

Only the relationship between state ownership and tax reporting does not follow the

expected pattern. State-owned firms do not report significantly more of their revenues to

tax authorities than do private firms, despite the advantages of control and information

presumably conferred by ownership. Nonetheless, as we saw in Tables 5 through 8, state-

owned firms do generally receive more support from state officials. One explanation for this

pattern which is consistent with the models presented above is that state-owned firms are

more “taxable” in the sense of non-tax revenues. An alternative possibility is that managers

of state-owned enterprises have access through their social networks to state officials that

their private-sector counterparts do not.18

In sum, the empirical data are strongly suggestive of the no-commitment model, and

somewhat less so of the model in which the state has commitment power but cannot use

non-distortionary lump-sum taxes. The models of Sections 3.2 and 3.3 in which the state

could commit to a nondistortionary lump-sum tax can be rejected on multiple grounds: such

taxes are relatively unusual, revenue hiding is observed in practice, such hiding is strongly

associated with state support along a number of dimensions, and the degree of revenue

18On “Social Networks and Corruption,” see Gehlbach (2001).

43

reporting is positively correlated with firm characteristics which suggest that the firm may

find it difficult to hide revenues.

4.3. Robustness Checks

One possible concern with the regressions reported in Tables 7 and 8 is that the BEEPS

questionnaire contains no question that can be used to directly compare profitability of firms.

Thus, one could question whether revenue hiding and state support of economic activity are

both correlated with profitability, in which case the estimated coefficients on revenue hiding

would be biased. In particular, it is conceivable that more profitable firms are both more

likely to be asked for bribes (if bureaucrats are at least partially able to observe profitability)

and more likely to hide their revenues from tax inspectors. However, if one looks at the

estimated coefficients on the “monopoly” variable, it appears that more profitable firms

(those with less competition) pay less in bribes, not more. Further, it is far from obvious

why more profitable firms would receive less support along other dimensions. For example,

in postcommunist countries there are often various means of contract enforcement available

— bribing judges, hiring “debt-collection agencies,” etc. — and more profitable firms may have

better access to these alternative mechanisms. Finally, all regressions control for sector at

a fairly fine level, so much variation in profitability is likely accounted for.

An additional issue concerns choice of strategy to lower tax payments. One can think

of firms as having two options available: they may hide their tax obligations, or they may

44

report their liability but then refuse to pay. The latter strategy is likely to be employed by

politically powerful firms, which may also find it easier to acquire state support. Thus, if

the variable “political power” is not completely captured by observable characteristics, the

positive correlation between revenue hiding and state support could be spurious.

To check for this possibility, all regressions in Tables 7 and 8 were rerun controlling

for whether or not the firm receives “subsidies (including tolerance of tax arrears) from

local or national government.” In all, 11 percent of firms report that they received such

subsidies. While this variable is generally positively correlated with our various measures of

state support, its inclusion has virtually no effect on the point estimates for the coefficient

on “proportion of revenues hidden.”

This discussion points to another consideration. As mentioned above, the regressions in

Tables 7 and 8 assume that there is no omitted variable correlated with both revenue hiding

and state support, and no simultaneity between the two variables. Two possible omitted

variables have been discussed; others are possible. Further, while the models above suggest

no simultaneity, it is possible that for psychological reasons firm managers who feel they are

being poorly served by the government hide more than is optimal given their taxability.

Ideally, we would deal with these concerns by finding a set of variables correlated with

taxability (and hence revenue hiding in the no-commitment case) but uncorrelated with

support beyond their impact on revenue reporting, and instrument revenue hiding on those

variables. In practice, no pure instrument seems to exist on the BEEPS questionnaire.

45

All of the firm characteristics listed as determinants of revenue hiding in Table 9 can be

thought of as independently affecting state support in some way. However, their “degree”

of endogeneity likely differs, as do their correlation with revenue hiding. Unfortunately, the

firm characteristic most correlated with revenue hiding — log employment — is also one of the

variables most likely to exert an independent effect on state support. Meanwhile, while the

sector dummies may contribute to state support primarily through their impact on revenue

hiding, their contribution to explaining variation in revenue hiding is relatively small.19

Despite these concerns, all the regressions in Tables 7 and 8 were rerun as two-stage least-

squares regressions, instrumenting “proportion of revenues hidden” successively on three

different sets of instruments:

1. All firm characteristics listed in Table 9.

2. All firm characteristics less state ownership and log employment.

3. Only sector dummies.

In comparing the results of these specifications with those in Tables 7 and 8, three

patterns are evident. First, the estimated coefficient on “proportions of revenues hidden”

is almost always larger in magnitude in the instrumental-variables regressions (in the case

of the regressions in Tables 8, relative to the coefficients when the original regressions are

rerun as OLS rather than ordered-probit). While possibly due to an omitted variable which

19On the tradeoff between efficiency and exogeneity in instrumental-variables techniques, see Bartels(1991).

46

resulted in an underestimation of the effect of revenue hiding in the original regressions, the

more likely explanation is that these instruments are not in fact exogenous. Second, as

we move from specification 1 to 2, and especially 2 to 3, some of the estimated coefficients

lose their significance, as is to be expected if the instruments are only weakly correlated

with the endogenous variable. Third, and most interesting, the estimated coefficient on

“percent of management time not spent with government officials” switches sign. In all

three specifications, firms which report more revenues to tax authorities are estimated to

spend more, not less, time with government officials. While it is not obvious what omitted

variable might have resulted in the sign being reversed in the original regression, it should

be noted that this result is consistent with the estimated effect of state ownership: state-

owned firms, presumably more taxable in some sense, also spend more time with government

officials. Thus, while the estimates produced by the instrumental-variables regressions must

be treated with caution, taken as a whole they do not seem to seriously challenge the results

reported above. The one exception concerns the time variable, where the relationship

suggested by OLS estimation should be strongly questioned.

All the regressions reported in this section include country and town-size dummies to

try to control for as much institutional variation as possible. Nonetheless, it is possible

that there is unobserved institutional variation, as might be the case if firms were surveyed

in cities with very different political-economic environments but which fell into the same

town-size category. As an additional robustness check, all regressions were rerun on the

47

subsample of firms in capital cities, since we know that each country has only one capital.

While doing so substantially shrinks the sample size, qualitative results for these regressions

are very similar. Estimated coefficients on our key variables occasionally lose significance,

but are still sizeable in magnitude and of the same sign.

Finally, given that the hypotheses in this paper were motivated to some extent by obser-

vation of Russian political economy, there might be some concern that Russian firms (which

account for one seventh of the sample) are driving the results. To check for this possibility,

all regressions were rerun on the subsample of non-Russian firms. Among our key variables,

the only substantial difference is that the estimated coefficients on “state-owned enterprises”

and “monopoly” in the regression on “opportunity to appeal administrative violations” in

Table 6 acquire statistical significance (at the 10 percent level), and are slightly larger in

magnitude. Further, given the importance of the natural-resource sector to tax revenues

in Russia, it is interesting to note that natural-resource firms in the non-Russian subsample

seem to be less taxable relative to their counterparts in other industries, as the coefficient

on “resource sector” in the regression of revenue hiding on firm characteristics changes from

-1.47 to -0.04 when firms in Russia are dropped.

5. Conclusion

This paper has demonstrated theoretically that an efficient social contract between state

and society, in which the former provides support for economic activity in return for a share

48

of the revenues thus generated, will generally be impossible. In all but a limiting case

probably not reflective of empirical reality, states will provide more support to economic

activity from which they find it easier to extract revenues. Further, empirical evidence from

postcommunist states suggests that the more insidious form of contract failure — when the

state is unable to commit to leaving producers with a share of their unhidden production

— contributes to the large observed variation in the degree to which firms face a supportive

business environment.

In so doing, the theoretical and empirical work here has suggested a number of avenues for

future research. Most obviously, it would be interesting to explore the extent to which the

findings here hold in other political-economic contexts. In particular, one might imagine a

different distribution of outcomes in countries where for historical reasons the state is better

able to commit and has more experience in taxing private economic activity.

That in turn suggests that it would be fruitful to track changes over time in postcommu-

nist countries. It is certainly within the spirit of the transactions-cost literature to expect

that institutions will evolve to minimize the contract failures identified in this paper.

Finally, it is worth noting that this paper has abstracted from what are in reality sub-

stantial differences across countries in the nature of tax systems and revenue sharing among

different levels of government. The latter point is likely to be especially important to the

degree that it is local officials who are largely responsible for creating a more or less positive

business environment, and for whom the tax return to support of economic activity is conse-

49

quently important. For example, while local governments in the Czech Republic, Hungary,

and Poland all retain some tax revenues collected locally, the nature of their revenue base

differs substantially, with personal income taxes relatively more important in the first two

countries and real estate taxes more important in the third.20 Similarly, empirical research

has found substantial variation within Russia in the degree to which local governments retain

any marginal increase in tax revenues (Makrushin et al 2002). The implication is that the

effects identified here may interact in interesting ways with local fiscal incentives. In other

words, there appear to be unrealized gains from trade between the literature on state-society

contracts to which this paper speaks, and that on fiscal federalism, which emphasizes instead

contracts between levels of government.

20See, e.g., the country reports in Horvath (2000). As discussed above, the measure of revenue reportingused in this paper likely captures to a considerable extent the evasion of many taxes not directly calculatedfrom firm revenues.

50

6. Appendix - Proof of Proposition 2

Proof. Focus first on the firm. If e = eFB, then T¡eFB

¢= eFB − b ¡HNC − g ¡HNC ,α

¢¢.

Thus, the firm faces the following problem:

maxHmax

£0, (1−H) eFB − T ¡eFB¢¤+ b (H − g (H,α)) (6.1)

We will partition the firm’s strategy space H ∈ [0, 1] into two intervals, and look for the

optimum over each interval. For H > H 0, where H 0 solves:

(1−H 0) eFB − T ¡eFB¢ = 0 (6.2)

eFB −H 0eFB − eFB + b ¡HNC − g ¡HNC ,α¢¢

= 0

H 0 =b

eFB¡HNC − g ¡HNC ,α

¢¢

the optimum H solves:

maxH∈(H0,1]

0 + b (H − g (H,α)) (6.3)

This is identical to the firm’s maximization problem (3.3) in the no-commitment case, giving

utility of b¡HNC − g ¡HNC ,α

¢¢. (Observe that H 0 < HNC, since from (6.2) eFBH 0 =

b¡HNC − g ¡HNC ,α

¢¢: production in the taxable sector is more efficient with first-best

support than it is in the untaxable sector.)

51

In contrast, for H ≤ HNC, the optimum H solves:

maxH∈[0,H0]

(1−H) eFB − T ¡eFB¢+ b (H − g (H,α)) (6.4)

maxH∈[0,H0]

−HeFB + b ¡HNC − g ¡HNC ,α¢¢+ b (H − g (H,α))

Since by assumption eFB > b, the solution to this problem is H = 0, giving utility of

b¡HNC − g ¡HNC ,α

¢¢. Thus, the firm is indifferent between H = 0 and H = HNC.

Now examine the ruler’s choice. The ruler solves:

maxemin[(1−H) e, T (e)]− c (e) (6.5)

which if the firm is playing H = 0 is:

maxemin

£e,max

£0, e− b ¡HNC − g ¡HNC ,α

¢¢¤¤− c (e) (6.6)

⇔ maxemax

£0, e− b ¡HNC − g ¡HNC ,α

¢¢¤− c (e)Since the ruler is residual claimant on any revenues over e− b ¡HNC − g ¡HNC ,α

¢¢, she will

choose e = eFB so long as her utility from doing so is greater than her utility from playing

e = 0, i.e. so long as eFB − c ¡eFB¢− b ¡HNC − g ¡HNC ,α¢¢ ≥ 0. Clearly that is the case,

52

since total surplus in this equilibrium:

eFB − c ¡eFB¢

is strictly greater than total surplus in the equilibrium with no commitment:

¡1−HNC

¢eNC − c ¡eNC¢+ b ¡HNC − g ¡HNC ,α

¢¢

53

References

Acemoglu, Daron. 2002. “Why Not a Political Coase Theorem? Social Conflict, Commitment

and Politics.” Unpublished manuscript.

Acemoglu, Daron, Simon Johnson & James A. Robinson. 2001. “The Colonial Origins of

Comparative Development: An Empirical Investigation.” American Economic Review

91(5):1369—1401.

Acemoglu, Daron, Simon Johnson & James A. Robinson. 2003. “Reversal of Fortune: Ge-

ography and Institutions in the Making of the Modern World Income Distribution.”

Quarterly Journal of Economics Forthcoming.

Aslund, Anders. 2001. The Development of Small Enterprises. In Russia’s Post-Communist

Economy, ed. Brigitte Granville & Peter Oppenheimer. Oxford: Oxford University

Press pp. 347—366.

Bartels, Larry M. 1991. “Instrumental and ’Quasi-Instrumental’ Variables.” American Jour-

nal of Political Science 35(3):777—800.

Berliner, Joseph. 1952. “The Informal Organization of the Soviet Firm.” Quarterly Journal

of Economics 66:342—365.

Brennan, Geoffrey & James M. Buchanan. 1980. The Power to Tax: Analytical Foundations

of a Fiscal Constitution. Cambridge: Cambridge University Press.

54

Che, Jiahua & Yingyi Qian. 1998. “Insecure Property Rights and Government Ownership

of Firms.” Quarterly Journal of Economics 113(2):467—496.

Coase, Ronald. 1960. “The Problem of Social Cost.” Journal of Law and Economics 3:1—31.

Cox, Gary W. & Mathew D. McCubbins. 1994. Legislative Leviathan: Party Government in

the House. Los Angeles: University of California Press.

Davies, R. W. 1998. Soviet Economic Development from Lenin to Khrushchev. Cambridge:

Cambridge University Press.

de Soto, Hernando. 1990. The Other Path. New York: Harper and Row.

Dixit, Avinash & Mancur Olson. 2000. “Does Voluntary Participation Undermine the Coase

Theorem?” Journal of Public Economics 76:309—335.

Easter, Gerald M. 2002. “Politics of Revenue Extraction in Post-Communist States: Poland

and Russia Compared.” Politics and Society 30(4):599—627.

EBRD. 1999. Transition Report 1999: Ten Years of Transition. London: European Bank

for Reconstruction and Development.

Ebrill, Liam & Oleh Havrylyshyn. 1999. “Tax Reform in the Baltics, Russia, and Other

Countries of the Former Soviet Union.” IMF Occasional Paper No. 182.

Evans, Peter B., Dietrich Rueschemeyer & Theda Skocpol, eds. 1985. Bringing the State

Back In. Cambridge: Cambridge University Press.

55

Farrell, Joseph. 1987. “Information and the Coase Theorem.” Journal of Economic Perspec-

tives 1(2):113—129.

Frye, Timothy & Andrei Shleifer. 1997. “The Invisible Hand and the Grabbing Hand.”

American Economic Review Papers and Proceedings 87:354—358.

Furubotn, Eirik G. & Rudolf Richter. 2000. Institutions and Economic Theory: The Con-

tribution of the New Institutional Economics. Ann Arbor, MI: University of Michigan

Press.

Gehlbach, Scott. 2001. “Social Networks and Corruption.” Paper presented at the 2001

Annual Meeting of the American Political Science Association.

Gordon, Roger H. & David Li. 1997. “Taxes and Government Incentives: Eastern Europe Vs.

China.” University of Michigan Department of Economics Working Paper No. 97-05.

Gordon, Roger H. & John D. Wilson. 1999. “Tax Structure and Government Behavior:

Implications for Tax Policy.” Unpublished manuscript.

Greif, Avner, Paul Milgrom & Barry R. Weingast. 1994. “Coordination, Commitment,

and Enforcement: The Case of the Merchant Guild.” Journal of Political Economy

102(4):745—776.

Gustafson, Thane. 1999. Capitalism Russian-Style. Cambridge: Cambridge University Press.

56

Hellman, Joel S., Geraint Jones, Daniel Kaufmann & Mark Schankerman. 2000. “Measuring

Governance, Corruption, and State Capture: How Firms and Bureaucrats Shape the

Business Environment in Transition Economies.” World Bank Policy Research Working

Paper No. 2312.

Hemming, Richard, Adrienne Cheasty & Ashok K. Lahiri. 1995. The Revenue Decline. In

Policy Experiences and Issues in the Baltics, Russia, and Other Countries of the Former

Soviet Union, ed. Daniel A. Citrin & Ashok K. Lahiri. IMF Occasional Paper No. 133.

Horvath, Tamas M., ed. 2000. Decentralization: Experiments and Reforms. Budapest: Local

Government and Public Service Reform Initiative, Open Society Institute.

Jensen, Donald N. 2000. “The Boss: How Yuri Luzhkov Runs Moscow.” Demokratizatsiya

8(1).

Jin, Hehui & Yingyi Qian. 1998. “Public Vs. Private Ownership of Firms: Evidence from

Rural China.” Quarterly Journal of Economics 113(3):773—808.

Johnson, Simon, Daniel Kaufmann & Andrei Shleifer. 1998. “The Unofficial Economy in

Transition.” Brookings Papers on Economic Activity 2:159—239.

Johnson, Simon, John McMillan & Christopher Woodruff. 2000. “Entrepreneurs and the Or-

dering of Institutional Reform: Poland, Slovakia, Romania, Russia and Ukraine Com-

pared.” Economics of Transition 8(1):1—36.

57

Lane, David. 2001. The Political Economy of Russian Oil. In Business and the State in

Contemporary Russia, ed. Peter Rutland. Boulder, CO: Westview Press pp. 101—128.

Levi, Margaret. 1988. Of Rule and Revenue. Berkeley: University of California Press.

Lieberman, Evan S. 2001. “National Political Community and the Politics of Income

Taxation in Brazil and South Africa in the Twentieth Century.” Politics and Society

29(4):515—555.

Makrushin, Aleksei, Ekaterina Zhuravskaya, Irina Slin’ko & Ruben Enikolopov. 2002.

“Prichiny i Sledstviia Slabosti Byudzhetnykh Stimulov v Rossii na Mestnom Urovne.”

CEFIR manuscript.

Martinez-Vazquez, Jorge & Robert McNab. 2000. “The Tax Reform Experiment in Transi-

tional Countries.” Andrew Young School of Policy Studies Working Paper 00-1.

North, Douglass C. 1981. Structure and Change in Economic History. New York: W.W.

Norton and Company.

Oi, Jean C. 1992. “Fiscal Reform and the Economic Foundations of Local State Corporatism

in China.” World Politics 45(1):99—126.

Qian, Yingyi & Barry Weingast. 1996. “China’s Transition to Markets: Market-Preserving

Federalism, Chinese Style.” Journal of Policy Reform 1(2):149—185.

Robinson, James A. 1998. “Theories of ’Bad’ Policy.” Journal of Policy Reform 3:1—46.

58

Roland, Gerard &Thierry Verdier. 1999. “Law Enforcement and Transition.”WilliamDavid-

son Institute Working Paper No. 262.

Shafer, D. Michael. 1994. Winners and Losers: How Sectors Shape the Developmental

Prospects of States. Ithaca: Cornell University Press.

Shleifer, Andrei & Robert W. Vishny. 1998. The Grabbing Hand: Government Pathologies

and Their Cures. Cambridge, MA: Harvard University Press.

Skocpol, Theda. 1979. States and Social Revolutions: A Comparative Analysis of France,

Russia, and China. Cambridge: Cambridge University Press.

Slinko, Irina, Evgeny Yakovlev & Ekaterina Zhuravskaya. 2003. “Institutional Subversion:

Evidence from Russian Regions.” Paper presented at the 2003 CEPR/WDI Annual

International Conference on Transition Economics.

Tanzi, Vito & George Tsibouris. 2000. “Fiscal Reform Over Ten Years of Transition.” IMF

Working Paper 00/113.

Tilly, Charles. 1990a. Coercion, Capital, and European States. Cambridge, MA: Blackwell.

Tilly, Charles. 1990b. Coercion, Capital, and European States, AD 990-1990. Cambridge,

MA: Blackwell.

59

Weinthal, Erika & Pauline Jones Luong. 2002. “Resource Wealth and Institutional Change:

The Political Economy of Tax Reform in Russia.” Paper presented at the 2002 Annual

Meeting of the American Political Science Association.

Yakovlev, Andrei. 2000. “Pochemu v Rossii vozmozhen bezriskovyi ukhod ot nalogov?”

Voprosy ekonomiki 11:134—152.

Zhuravskaya, Ekaterina V. 2000. “Incentives to Provide Local Public Goods: Fiscal Feder-

alism, Russian Style.” Journal of Public Economics 76:337—368.

60

Table 1: Comparison of Model Predictions

Model Support (with respect to α)

Support (with respect to H)

Hiding (with respect to α)

No Commitment Increasing Decreasing Decreasing Commitment – de Soto Constant N/A Constant (H = 0) Commitment – tax evasion Increasing N/A Constant (H = 0) Commitment – tax evasion (proportional tax, t < 1) Increasing? Decreasing / N/A? Constant (H >0)?

Table 2: Countries Represented in Sample

Firms in BEEPS

Sample

1999 General Government Revenue

as Percent of GDP Albania 163 21.3 Armenia 125 20.3 Azerbaijan 137 18.9 Belarus 132 45.7 Bosnia-Herzegovina 127 Bulgaria 130 39.8 Croatia 127 42.8 Czech Republic 149 38.7 Estonia 132 36.4 Georgia 129 15.4 Hungary 147 39.1 Kazakhstan 147 17.4 Kyrgyzstan 132 24.0 Latvia 166 40.1 Lithuania 112 31.7 Macedonia 136 38.0 Moldova 139 27.4 Poland 246 40.2 Romania 125 33.3 Russia 552 35.1 Serb Republic in Bosnia 65 Slovakia 138 39.7 Slovenia 125 43.6 Ukraine 247 33.7 Uzbekistan 126 30.4 Note: Government revenue figures are imputed from expenditure and balance data in EBRD (2001).

Table 3: Firm Characteristics

Variable Mean/proportion State 0.14 Private 0.86 Employment 141 (173) Foreign ownership 0.13 No foreign ownership 0.87 Number of competitors - None 0.10 - 1-3 0.13 - More than 3 0.77 Exporter 0.24 Non-exporter 0.76 Transportation sector 0.06 Personal-service sector 0.05 Construction sector 0.09 Wholesale sector 0.14 Business-service sector 0.06 Retail sector 0.14 Manufacturing sector 0.30 Resource sector 0.12 Other sector 0.02 Finance sector 0.02 Capital city 0.31 Other, over 1 million 0.05 Other, 250,000 – 1 million 0.13 Other, 50,000 – 250,000 0.19 Other, under 50,000 0.24 Rural 0.09 Note: Standard deviation of employment reported in parentheses.

Table 4: Revenue Reporting and State Support

Variable Name Question on Survey Responses A. Revenue Reporting

Mean Standard deviation

< 100%

100%

Percent of revenues reported

“What percentage of sales of a typical firm in your area of activity would you estimate is reported to the tax authorities, bearing in mind difficulties in complying with taxes and other regulations?”

78.5

25.9

2430

66.4%

1232

33.6%

B. State-Support Variables in OLS Regressions Mean

Standard deviation

Bribe payments as percent of revenues

“On average, what percent of revenues do firms like yours typically pay per annum in unofficial payments to public officials?”

2.6

4.7

Non-tax bribe payments as percent of revenues

Previous question, subtracting out that proportion of unofficial payments spent “to deal with taxes and tax collection” and “to deal with customs/imports.”

1.6

3.2

Percent of management time spent with government officials

“What percentage of senior management’s time per year is spent dealing with government officials about the application and interpretation of laws and regulations?”

9.8

12.6

C. State-Support Variables in Ordered-Probit Regressions Never Seldom Sometimes Frequently Mostly Always Opportunity to appeal administrative violations

“If a government agent acts against the rules I can usually go to another official or to his superior and get the correct treatment without recourse to unofficial payments.”

629

18.9%

738

22.2%

797

24.0%

277

8.3%

515

15.5%

365

11.0% Strongly

disagree Disagree in most cases

Tend to disagree

Tend to agree

Agree in most cases

Fully agree

Contracts and property rights enforced

“To what degree do you agree with this statement? ‘I am confident the legal system will uphold my contract and property rights in business disputes.’”

373

9.5%

591

15.1%

962

24.5%

1120

28.5%

619

15.8%

261

6.7% Very

unhelpful Mildly

unhelpful Neutral Mildly

helpful Very

helpful

Local government helpful “How helpful do you find local/regional governments towards businesses like yours?”

1151 29.9%

661 17.2%

1132 29.4%

718 18.7%

186 4.8%

Notes: In regressions reported in Tables 5 and 7, the dependent variables are recoded so that a higher response consistently reflects more government support of business activity, e.g. “Percent of revenues not paid as bribes” is equal to 100 minus “Bribe payments as percent of revenues. Further, revenue reporting is inverted in the regressions in Tables 7 to 9, i.e. “percent of revenues hidden” is equal to 100 minus “percent of revenues reported.”

Table 5: Effect of Firm Characteristics on State Support – OLS Regressions

(Significance levels: 10% - *; 5% - **; 1% - ***)

Percent of Revenues Percent of Revenues Percent of Management Time Not Paid as Bribes Not Paid as Non-Tax Bribes Not Spent with Government Officials Estimated Robust Estimated Robust Estimated Robust coefficient std. error coefficient std. error coefficient std. error State-owned enterprise 0.49** 0.22 0.46*** 0.14 -3.26*** 0.74 Log employment 0.42*** 0.07 0.31*** 0.05 -0.26 0.16 Foreign ownership -0.04 0.25 0.26* 0.16 -1.02 0.65 Monopoly 1.01*** 0.25 0.62*** 0.17 -0.78 0.80 1-3 competitors 0.08 0.27 0.03 0.19 -0.87 0.66 Exporter -0.04 0.22 0.13 0.16 -0.19 0.56 Personal-service sector 0.06 0.39 -0.07 0.33 0.80 0.96 Transportation sector 0.31 0.31 0.24 0.24 -0.49 0.97 Wholesale sector -0.38 0.30 0.21 0.22 0.03 0.67 Retail sector -0.29 0.33 0.00 0.26 -1.12 0.79 Construction sector -0.66* 0.36 -0.42 0.27 -1.93** 0.90 Other sector 0.22 0.78 0.57* 0.32 2.60* 1.54 Resource sector 0.06 0.32 0.04 0.23 0.11 0.78 Business-service sector -0.21 0.43 0.42 0.26 -1.24 0.96 Finance sector 0.26 0.73 1.02*** 0.23 -2.88 1.84 N 2973 2667 3483 R2 .109 .095 .100 Notes: Wording of dependent variables reflects coding so that a higher response consistently reflects more government support of business activity; see Table 2.2 for details. Private, no foreign ownership, more than three competitors, non-exporters, and manufacturing firms omitted categories. Constant and country and town-size dummies included in all regressions.

Table 6: Effect of Firm Characteristics on State Support – Ordered-Probit Regressions

(Significance levels: 10% - *; 5% - **; 1% - ***)

Opportunity to Appeal Contracts and Property Local Government Administrative Violations Rights Enforced Helpful

Estimated Robust

std. Marginal

effects

Estimated Robust

std. Marginal

effects

Estimated Robust

std. Marginal

effects coefficient error (Pr = .16) coefficient error (Pr = .15) coefficient error (Pr = .18) State-owned enterprise 0.10 0.07 0.01 0.30*** 0.06 0.06 0.26*** 0.06 0.06 Log employment 0.05*** 0.02 0.01 0.08*** 0.01 0.01 0.13*** 0.01 0.03 Foreign ownership 0.07 0.06 0.01 0.00 0.05 0.00 -0.04 0.05 -0.01 Monopoly 0.10 0.08 0.01 0.12* 0.06 0.02 0.19*** 0.07 0.04 1-3 competitors -0.06 0.06 -0.01 -0.03 0.05 -0.01 0.05 0.06 0.01 Exporter 0.03 0.05 0.00 -0.03 0.05 -0.01 -0.01 0.05 0.00 Personal-service sector 0.15 0.10 0.02 0.08 0.09 0.01 0.16* 0.09 0.04 Transportation sector 0.11 0.09 0.02 -0.06 0.08 -0.01 0.09 0.09 0.02 Wholesale sector -0.03 0.06 -0.01 0.06 0.06 0.01 -0.05 0.06 -0.01 Retail sector -0.04 0.07 -0.01 0.01 0.06 0.00 0.05 0.06 0.01 Construction sector -0.06 0.07 -0.01 -0.03 0.07 -0.01 0.00 0.07 0.00 Other sector 0.13 0.19 0.02 0.26 0.16 0.05 0.32** 0.16 0.07 Resource sector -0.01 0.08 -0.00 0.04 0.07 0.01 0.13* 0.07 0.03 Business-service sector 0.09 0.08 0.01 0.08 0.08 0.01 0.10 0.08 0.02 Finance sector 0.17 0.14 0.02 0.36*** 0.12 0.07 0.01 0.11 0.00 N 3226 3818 3742 Maximized log likelihood -5495.4 -6051.0 -5236.9 Notes: Less than 100% of revenues reported, private, no foreign ownership, more than three competitors, non-exporters, and manufacturing firms omitted categories. Constant and country and town-size dummies included in all regressions. Marginal effects are given for second-right-most category (see Table 2.2), are calculated at means of independent variables, and are given as discrete change for dummy variables.

Table 7: Effect of Revenue Hiding and Firm Characteristics on State Support – OLS Regressions

(Significance levels: 10% - *; 5% - **; 1% - ***)

Percent of Revenues Percent of Revenues Percent of Management Time Not Paid as Bribes Not Paid as Non-Tax Bribes Not Spent with Government Officials Estimated Robust Estimated Robust Estimated Robust coefficient std. error coefficient std. error coefficient std. error Proportion of revenues hidden -2.52*** 0.41 -1.82*** 0.30 -3.39*** 0.97 State-owned enterprise 0.51** 0.22 0.46*** 0.14 -3.18*** 0.77 Log employment 0.37*** 0.07 0.27*** 0.06 -0.16 0.17 Foreign ownership -0.12 0.25 0.19 0.16 -1.23* 0.67 Monopoly 0.84*** 0.26 0.50*** 0.18 -1.11 0.85 1-3 competitors -0.01 0.28 -0.05 0.20 -1.03 0.68 Exporter -0.03 0.22 0.12 0.17 -0.48 0.59 Personal-service sector 0.16 0.40 0.02 0.35 1.09 1.01 Transportation sector 0.30 0.32 0.26 0.25 -0.09 1.04 Wholesale sector -0.26 0.31 0.34 0.22 0.50 0.69 Retail sector -0.35 0.35 -0.03 0.27 -0.68 0.82 Construction sector -0.67* 0.37 -0.40 0.28 -2.04** 0.93 Other sector 0.93** 0.43 0.72** 0.29 2.09 1.72 Resource sector 0.04 0.33 0.06 0.24 -0.26 0.82 Business-service sector -0.27 0.45 0.46* 0.27 -1.41 1.02 Finance sector -0.04 0.77 0.87*** 0.24 -3.86** 1.96 N 2806 2522 3248 R2 .127 .111 .104 Notes: Wording of dependent variables reflects coding so that a higher response consistently reflects more government support of business activity; see Table 2.2 for details. Private, no foreign ownership, more than three competitors, non-exporters, and manufacturing firms omitted categories. Constant and country and town-size dummies included in all regressions.

Table 8: Effect of Revenue Hiding and Firm Characteristics on State Support – Ordered-Probit Regressions

(Significance levels: 10% - *; 5% - **; 1% - ***)

Opportunity to Appeal Contracts and Property Local Government Administrative Violations Rights Enforced Helpful Estimated Robust Estimated Robust Estimated Robust coefficient std. error coefficient std. error coefficient std. error Proportion of revenues hidden -0.33*** 0.08 -0.23*** 0.08 -0.21*** 0.08 State-owned enterprise 0.09 0.07 0.29*** 0.06 0.27*** 0.06 Log employment 0.04*** 0.02 0.08*** 0.01 0.12*** 0.01 Foreign ownership 0.06 0.06 -0.01 0.06 -0.05 0.06 Monopoly 0.07 0.08 0.10 0.07 0.15** 0.08 1-3 competitors -0.07 0.06 -0.01 0.05 0.07 0.06 Exporter 0.02 0.06 -0.06 0.05 -0.03 0.05 Personal-service sector 0.19* 0.11 0.11 0.10 0.14 0.09 Transportation sector 0.08 0.09 -0.09 0.08 0.11 0.09 Wholesale sector -0.01 0.06 0.08 0.06 -0.03 0.06 Retail sector -0.04 0.07 0.03 0.06 0.06 0.07 Construction sector -0.10 0.07 -0.04 0.07 -0.01 0.07 Other sector 0.18 0.19 0.25 0.17 0.35** 0.17 Resource sector -0.01 0.08 0.02 0.07 0.15** 0.08 Business-service sector 0.09 0.08 0.07 0.08 0.11 0.08 Finance sector 0.11 0.14 0.34*** 0.13 -0.03 0.12 N 3021 3559 3484 Maximized log likelihood -5136.4 -5626.9 -4872.7 Marginal effects Always

(Pr = 0.10) Mostly

(Pr = 0.16) Fully Agree (Pr = 0.05)

Agree in Most Cases (Pr = 0.15)

Very Helpful (Pr = 0.03)

Mildly Helpful (Pr = 0.18)

- Proportion of revenues hidden -0.06 -0.05 -0.02 -0.04 -0.02 -0.05 - State-owned enterprises 0.02 0.01 0.03 0.05 0.02 0.06 Notes: Private, no foreign ownership, more than three competitors, non-exporters, and manufacturing firms omitted categories. Constant and country and town-size dummies included in all regressions. Marginal effects calculated at means of independent variables and are given as discrete change for state-owned enterprise.

Table 9: Determinants of Revenue Hiding - OLS Regression (Significance levels: 10% - *; 5% - **; 1% - ***)

Percent of Revenues Hidden Estimated Robust coefficient std. error State-owned enterprise -0.46 1.30 Log employment -2.30*** 0.32 Foreign ownership -4.63*** 1.14 Monopoly -8.80*** 1.44 1-3 competitors -2.25** 1.15 Exporter -0.30 1.09 Personal-service sector 3.61* 2.11 Transportation sector 2.71 2.02 Wholesale sector 1.75 1.43 Retail sector 0.60 1.42 Construction sector 0.31 1.57 Other sector -0.79 3.29 Resource sector -1.47 1.53 Business-service sector -1.55 1.93 Finance sector -8.07*** 2.29 N 3573 R2 .187 Notes: Private, no foreign ownership, more than three competitors, non-exporters, and manufacturing firms omitted categories. Constant and country and town-size dummies included.

Table 10: Tax Receipts by Sector in Russia, 1997

Percent of GDP (1) Percent of Tax

Receipts (2) Column (2) / Column (1)

Agriculture 7.2 1.0 0.1 Services 16.1 8.6 0.5 Construction 8.7 6.5 0.7 Other 28.9 26.8 0.9 Transport 10.8 15.1 1.4 Industry 28.1 39.7 1.4 Banking 0.3 2.4 8.0 Source: M.P. Afanas’ev and P.V. Kuznetsov, Nalogi v Rossii i v mire [Taxes in Russia and the World], 1997, Rabochii Tsentr Ekonomicheskikh Reform pri Pravitel’stvo Rossii [Working Center for Economic Reforms – Russian Government], p. 24