Long-Term Decline in Real Public Wages

Evidence from a number of developing countries suggests that real wage levels for public sector employees have been declining over long periods of time.⁵ Table 1 presents the trend growth in real wage levels in the general government for selected developing countries. Our estimates are for all the countries for which time-series data were available and for as many years as the data permitted. The trend regressions indicate that real wage levels in the general government declined for 13 out of the 21 countries. In the eight countries that registered positive growth rates, Ghana had the highest growth, with an annual rate of 4.4 percent. For the sample as a whole, real wages declined annually by about 1.4 percent.⁶

Table 1.

Selected Developing Countries: Trends in Real Wages in General Government

(Annual percent change)

Source: National authorities.

^aEstimated from a fixed-effects pooled regression of the countries listed. The coefficient is significant at the 5 percent level.

Table 1.

Selected Developing Countries: Trends in Real Wages in General Government

(Annual percent change)

	Time period	Real wage trend in general government
Argentina	1976–89	-3.1
Bolivia	1985–91	2.3
Bulgaria	1989–92	-17.7
Congo	1980–83	-3.8
Costa Rica	1974–93	1.7
Czechoslovakia	1989–92	-11.7
Fiji	1985–93	-1.7
Gabon	1985–91	3.4
Ghana	1986–90	4.4
Hungary	1989–92	-0.8
India	1979–84	3.8
Kenya	1982–92	-2.2
Mauritius	1974–92	1.3
Morocco	1980–89	-1.6
Myanmar	1987–92	-0.5
Panama	1973–91	0.6
Poland	1989–92	-10.3
Romania	1989–92	-8.9
Rwanda	1985–89	-2.0
Solomon Islands	1988–91	1.6
Suriname	1984–86	-8.3
Averagea		-1.4

Source: National authorities.

^aEstimated from a fixed-effects pooled regression of the countries listed. The coefficient is significant at the 5 percent level.

View Table

Table 1.

Selected Developing Countries: Trends in Real Wages in General Government

(Annual percent change)

	Time period	Real wage trend in general government
Argentina	1976–89	-3.1
Bolivia	1985–91	2.3
Bulgaria	1989–92	-17.7
Congo	1980–83	-3.8
Costa Rica	1974–93	1.7
Czechoslovakia	1989–92	-11.7
Fiji	1985–93	-1.7
Gabon	1985–91	3.4
Ghana	1986–90	4.4
Hungary	1989–92	-0.8
India	1979–84	3.8
Kenya	1982–92	-2.2
Mauritius	1974–92	1.3
Morocco	1980–89	-1.6
Myanmar	1987–92	-0.5
Panama	1973–91	0.6
Poland	1989–92	-10.3
Romania	1989–92	-8.9
Rwanda	1985–89	-2.0
Solomon Islands	1988–91	1.6
Suriname	1984–86	-8.3
Averagea		-1.4

Source: National authorities.

^aEstimated from a fixed-effects pooled regression of the countries listed. The coefficient is significant at the 5 percent level.

In transition economies, large-scale liberalization of prices, accompanied by wage controls, led to declines in real wages in the public sector, particularly during the initial stages.⁷ Table 1 also reports declines in real wages in the public sector for five Eastern European countries during 1989–92. A similar pattern, but characterized by even sharper declines, is evident in the former U.S.S.R, countries. Table 2 shows that, with the exception of Latvia and Estonia (the two countries that reformed earlier than others), real wages declined in all of the former U.S.S.R. countries. Thus, during the initial transition years, falls in real wages were common in the transition economies.

Table 2.

Selected Former U.S.S.R. Countries: Real Public Wage Indices

(1992= 100)

Sources: National authorities; and IMF staff estimates.

Table 2.

Selected Former U.S.S.R. Countries: Real Public Wage Indices

(1992= 100)

Country	1992	1993	1994	1995
Armenia	100	22.8	15.2	16.6
Belarus	100	93.5	59.4	63.3
Kyrgyz Republic	100	57.2	44.1	51.5
Russia	100	99.0	90.5	69.4
Ukraine	100	50.7	45.0	50.5
Estonia	100	106.1	111.3	119.9
Latvia	100	107.5	120.2	121.1
Lithuania	100	62.4	72.1	73.2
Average	100	74.9	69.7	70.7

Sources: National authorities; and IMF staff estimates.

View Table

Table 2.

Selected Former U.S.S.R. Countries: Real Public Wage Indices

(1992= 100)

Country	1992	1993	1994	1995
Armenia	100	22.8	15.2	16.6
Belarus	100	93.5	59.4	63.3
Kyrgyz Republic	100	57.2	44.1	51.5
Russia	100	99.0	90.5	69.4
Ukraine	100	50.7	45.0	50.5
Estonia	100	106.1	111.3	119.9
Latvia	100	107.5	120.2	121.1
Lithuania	100	62.4	72.1	73.2
Average	100	74.9	69.7	70.7

Sources: National authorities; and IMF staff estimates.

Real wage declines in the public sector are a frequent consequence of fiscal adjustments in stabilization programs of developing countries (Kraay and Van Rijckeghem, 1995). Such programs appear to protect employment at the expense of reducing real wages (Hewitt and Van Rijckeghem, 1995).

Evidence on Wage Differentials in the Public and Private Sectors

Evidence on government wages relative to the private sector was reported in a fairly comprehensive cross-country study by Heller and Tait (1984). Several measures of wage differentials indicate that, in general, public wages relative to private sector wages were lower in developing countries than in industrial countries during the late 1970s and the early 1980s. Declining trends in real wage differentials over a longer time period in countries for which data were available are presented in Table 3. To control for changes in the structure of wages in the government and in the private sector, Chaudhry, Reid, and Malik (1994) focus on wages at the highest and the lowest grades at comparable levels in the government and the private sector. They find that, in most cases, public wages were lower than private sector wages at both grade levels, particularly at the highest grade levels.

Comparable data on public and private sector wages are not yet available for the transition economies, with the exception of the Czech Republic. In a rare study that compares public sector to private sector wages in the Czech Republic, Flanagan (1995) finds that full-time employees in the private sector earn considerably more than their counterparts in the state sector. After controlling for schooling and potential experience, survey results show that workers (owners) in new private firms earn 18 (43) percent more than those in current or former state enterprises.

Table 3.

Selected Developing Countries: Trends in the Ratio of Government to Private Sector Average Wages

(Annual percent change)

Source: National authorities.

Table 3.

Selected Developing Countries: Trends in the Ratio of Government to Private Sector Average Wages

(Annual percent change)

	Time period	Trend
Bolivia	1985–91	4.0
Costa Rica	1974–93	-0.8
Fiji	1985–93	-27.0
Ghana	1986–90	-8.0
Kenya	1982–92	-3.0
Mauritius	1974–92	-0.4
Panama	1973–91	2.0
Peru	1985–92	-25.0
Poland	1989–92	-0.4
Suriname	1984–86	-1.0
Average		-6.0

Source: National authorities.

View Table

Table 3.

Selected Developing Countries: Trends in the Ratio of Government to Private Sector Average Wages

(Annual percent change)

	Time period	Trend
Bolivia	1985–91	4.0
Costa Rica	1974–93	-0.8
Fiji	1985–93	-27.0
Ghana	1986–90	-8.0
Kenya	1982–92	-3.0
Mauritius	1974–92	-0.4
Panama	1973–91	2.0
Peru	1985–92	-25.0
Poland	1989–92	-0.4
Suriname	1984–86	-1.0
Average		-6.0

Source: National authorities.

Public Wages, Corruption, and Misallocation of Resources

The limited recorded evidence suggests that corruption is quite pervasive in developing countries and transition economies, and that it imposes significant economic costs.⁸ While arguments supporting corruption have been made by Leff (1964) and others, who claim that it cuts red tape and undermines rigid and unjustified regulations, Gould and Amaro-Reyes (1983) find that the practice of giving “speed money” was actually the cause of administrative delays because civil servants developed the habit of slowing paperwork until some kind of payment was made to them. Shortages in the supply of critical government-controlled inputs (such as electricity, transportation services, and imported inputs) have also created rent-seeking opportunities and led firms to create wasteful excess capacity to get higher allocations of inputs (Sahay, 1990 and 1991). Mauro (1995) uses qualitative measures of corruption to test its effects on economic growth and finds a significant negative relationship between corruption and growth for a number of developing countries.⁹

One of the most comprehensive projects on tax evasion and bribes was funded during 1975–78 by the International Development Research Center (IDRC) of Canada (Alfiler, 1986).¹⁰ The IDRC found that under Nepal’s dasturi system monthly average bribes for customs officials ranged from 50 percent to 200 percent of the staff’s monthly pay.¹¹ Officers administrating revenues estimated that 20–30 percent of Nepal’s targeted revenue was “lost” on account of bribes. With respect to other countries covered in the project, an ex-prime minister in Thailand was quoted as stating that the country could collect about 47 percent more revenue if there were no corruption and tax evasion, particularly in the Bangkok region. Also, only 15 percent of the total tax revenue due in the timber industry in Malaysia was collected by the state during 1964–74. At the other end of the spectrum, bribe taking was found to be a high-risk activity in Singapore because penalties were high and anticorruption measures strictly enforced.¹² In Hong Kong, corruption was lucrative in the 1960s, but this situation changed in the 1970s when the Independent Commission Against Corruption was established.

A recent econometric study by Van Rijckeghem and Weder (1996) provides direct evidence of the negative relationship between public wages and corruption. Using a panel data estimation procedure for a sample of 20 countries over the 1982–94 period, they find that the ratio of civil service wages to manufacturing wages was a significant determinant of the corruption index. In the countries studied, the simple correlation between these relative wages and the corruption index was more than 0.8.

Tanzi (1994) argues that less-developed countries may be more prone to corruption than industrial countries because of differences in “cultural factors.” However, as the evidence documented above indicates, corruption appears to be more common in these countries because the administrative system and institutions are poorly developed and government wages are low. While it is true that, with the passage of time, corruption can become institutionalized in any country and may appear to be more characteristic of developing countries, it is not obvious that these are cultural factors per se. In fact, because there is greater impartiality and competence in the public order in industrial countries, migrant societies in industrial countries do not appear to engage in the corrupt behavior that is prevalent in their own countries.

Related Work

Well-known models of tax evasion (Allingham and Sandmo, 1972; and Srinivasan, 1973) are based on asymmetric honesty between taxpayers (who could be dishonest) and tax collectors (who are honest by assumption). Although this scenario is likely in industrial countries, it is not a reasonable assumption in most developing countries.

Our model allows for the possibility that both tax collectors and taxpayers are corrupt. Virmani (1987) also allows for symmetric dishonesty and considers different bureaucratic regimes (honest or dishonest, depending on what he calls a social weakness factor) that are independent of the incentive scheme for tax collectors. Within the dishonest regimes, he determines optimal incentive schemes that ensure the honesty of bureaucrats in “corruption-deterring” societies and “weak” societies. He shows how, if bribes exist, evasion does not increase in a corruption-deterring society when penalties are raised: the larger penalties merely lead to transfers of rent from payer to collector. In a weak society, however, larger penalties lead to increased evasion. The intuition for the latter outcome is that, if penalties are increased, it becomes more profitable for the tax evader and collector to collude.

In contrast to Virmani’s model, our model does not assume a priori that some regimes are more corruptible than others; the emergence of corruption is simply a result of the incentive mechanisms in the administrative system.

Following Becker and Stigler (1974). Rose-Ackerman (1975), Klitgaard (1989 and 1995), Basu, Bhattacharaya, and Mishra (1992), and Mookherjee and Png (1995), our model also focuses on the principal-agent problem of corruption. The model presented below analyzes the interaction between the government (the principal), on the one hand, and the tax collector and the private sector firm (the agents), on the other, to determine how statutory taxes, penalties, and public wage policy determine the level of tax evasion and bribes. In addition to examining ways of motivating the agent to be honest, we derive macroeconomic budgetary implications of alternative wage and tax policies.¹⁴ While Shleifer and Vishny (199.1) also derive the consequences of corruption for resource allocation, they take the principal-agent solution as given.

The Model

The model basically consists of four sets of players: the government, the auditors (or the monitors), the tax collectors, and the private sector firms. The private sector firm produces output, y_i, on which the government imposes a uniform statutory tax rate, t, across all firms such that total taxes in the economy that are due to the government are T= tY= ΣT_i = tΣy_i, where Y is the aggregate output in the economy, T_i are total taxes due by each firm i, and y_i is the output of firm i. It is assumed that the government knows aggregate output, y (and, therefore, T), in the economy perfectly.¹⁵ However, taxes due by each firm, T_i = ty_i, can be observed only imperfectly by the government and the auditors, who monitor the tax collectors and the private firms. The government (the principal) collects taxes through its agents, the tax collectors.

Collusion and Asymmetric Information

In the model, there is perfect information between the government and the auditors but imperfect information between the auditors and the tax collectors. The tax collector observes the firm’s output perfectly (and, hence, the firm’s tax liability), but the government and the auditors do not. Thus, the extent of tax evasion or the presence of side deals between the private sector and the tax collector are not fully observed by the government and the auditors. The assumptions on information flows among the various players are illustrated in Figure 1.

View Full Size

Description of Information Content and Flows

Citation: IMF Staff Papers 1996, 004; 10.5089/9781451930931.024.A004

• Download Figure
• Download figure as PowerPoint slide

Since the government and the auditors cannot fully observe the activities of the tax collector and the firm, the collector and the firm can potentially cheat the government of taxes. We assume that, if tax evasion occurs, the tax collector and the firm play a cooperative game to determine τ_i = μT_i, where 0 ≤ μ ≤ 1 and τ_i is the total tax declared by firm i and collected by the government. The firm pays the lax collector a bribe, b_i, and reduces its tax liability by T_i - τ_i.¹⁶ Because the firm and the bureaucrat share in the gains from collusion, b_i≤T_i-τ_i.

Institutional Setup: Monitoring and Penalties

The government is aware of the potential for collusion between the firms and tax collectors to cheat it of tax revenues. In response, it sets up institutional arrangements to monitor their activity. Specifically, the government hires auditors who discover evasion with a finite probability P, where 0≤P≤ 1 and P is directly proportional to the auditors’ ability to detect cheating. The auditors’ actions are fully observable by the government. An auditor’s ability to detect cheating depends on her skills, which are a function of the wage that she receives. Her skill level is, therefore, assumed to be directly proportional to the public sector wage (w_g) relative to the wage in the private sector (w_p); as both w_g and w_p can be observed, the level of her skill is an observable.¹⁷ The underlying assumption is that a higher relative wage in the public sector, represented by = = w_g/w_p, allows better human capital to flow into the government sector.¹⁸ Thus, the probability of being caught is proportional to α because α determines the auditors’ ability to detect cheating. In addition, we assume that the probability of getting caught increases as tax evasion increases. The probability of getting caught is

$\begin{array}{cc}p=p\left(\text{α,}T,-\text{τ}\right),& \left(1\right)\end{array}$

View Expanded

where T=ΣT_i and τ = Στ_i are aggregate taxes due and collected in the economy and are, therefore, observable to the government and the auditors. The function p is assumed to be differentiable, increasing, and convex in both arguments. It is also assumed that each firm’s decision to cheat or not affects x since τ = Στ_i. Thus, P is, in part, a choice variable for the firm.

If corruption is discovered, both the firm and the tax collector are fined such that they individually pay penalties that are larger than the amount that is cheated. Thus, the firm pays a penalty that is a multiple of the level of tax evasion, that is, δ_p(T_i - τ_i), with δ_p > 1. Similarly, in the event of detection, the collector will pay δ_g(T_i - τ_i), where δ_g > 1.

The Payoffs

We assume that all parties are risk neutral. If no evasion occurs, neither the firm nor the collector gains or loses anything. The payoffs, as well as the structure of the game, are summarized in Figure 2 and described below.

View Full Size

Payoff Functions of Tax Collector and Private Firm

Citation: IMF Staff Papers 1996, 004; 10.5089/9781451930931.024.A004

• Download Figure
• Download figure as PowerPoint slide

The Firm

If evasion occurs and is not discovered by the auditors, the firm’s payoff is

$\begin{array}{cc}{T}_i-b_i-{\text{τ}}_j.& \left(2\right)\end{array}$

View Expanded

If auditors discover the evasion, the firm pays a penalty, δ_p on the tax evaded, as described in the previous section. Thus, when tax evasion occurs and is discovered by the auditors, the firm’s payoff is

$\begin{array}{cc}{T}_i-b_i-{\text{τ}}_i-{\text{δ}}_p\left(T_i-{\text{τ}}_i\right).& \left(3\right)\end{array}$

View Expanded

The total expected payoff to the firm for participating in the evasion game is simply the probability-weighted sum of equations (2) and (3):

$\begin{array}{cc}p\left[\left(T_i-b_i-{\text{τ}}_i\right)-{\text{δ}}_p\left(T_i-{\text{τ}}_i\right)\right]+\left(1-p\right)\left(T_i-b_i-{\text{τ}}_i\right),& \left(4\right)\end{array}$

View Expanded

which simplifies to

$\begin{array}{cc}\left(T_i-{\text{τ}}_i\right)\left(1-p{\text{δ}}_p\right)-b_i.& \left(5\right)\end{array}$

View Expanded

The Tax Collector

The net payoff to the tax collector (net of the wages received) when he is not discovered by the auditor is b_i. As discovery of cheating leads to a penalty assessment of δ_g on the amount of tax that was evaded, the net gain to the collector when he is discovered is b_i - δ_g(T_i-τ_i).¹⁹ The total expected payoff of the collector is, therefore,

$\begin{array}{cc}p\left[b_i-{\text{δ}}_g\left(T_i-{\text{τ}}_i\right)\right]+\left(1-p\right)\left(b_i\right),& \left(6\right)\end{array}$

View Expanded

which simplifies to

$\begin{array}{cc}{b}_i-p{\text{δ}}_g\left(T_i-{\text{τ}}_i\right).& \left(7\right)\end{array}$

View Expanded

The Existence of Bribes

The choice between cheating the government of its tax revenues or not will depend on whether the expected payoffs to the tax collector and the firm are more if they cheat than if they do not. A necessary and sufficient condition for the existence of bribes and tax evasion in equilibrium is that each of the two players gains. This condition is satisfied if expressions (5) and (7) are each greater than zero. Alternatively, the conditions can be collapsed into

$\begin{array}{cc}\left(T_i-{\text{τ}}_i\right)\left(1-p{\text{δ}}_g\right)-b_i\le 0,& \left(8\right)\end{array}$

View Expanded

and

$\begin{array}{cc}{b}_i-p{\text{δ}}_g\left(T_i-{\text{τ}}_i\right)\ge 0,& \left(9\right)\end{array}$

View Expanded

A necessary condition for the nonnegativity conditions in equations (8) and (9) to be satisfied such that T_i - τ_i is strictly positive is

$\begin{array}{cc}p\left({\text{δ}}_p+{\text{δ}}_g\right)\le 1.& \left(10\right)\end{array}$

View Expanded

The economic interpretation of equation (10) is that bribes and tax evasion will exist only if the monitoring arrangements are weak (that is, if p is small) and penalties for such activities (that is, δ_p and δ_g) are small. Alternatively, weak institutional setups raise the payoffs for corrupt and illegal activities.²⁰

Optimal Level of Tax Evasion

If the condition specified in equation (10) is met, the tax collector and the firm bargain to determine the optimal levels of tax evasion and bribes. We solve the problem in two stages. In the first stage, the firm and the collector determine τ_i (the taxes declared) to maximize their joint net payoffs (that is, they maximize the size of the total pie). In the second stage, they bargain to divide the gains from evasion (that is, they determine how to split the pie). This process determines the bribe that the collector receives. We basically follow Mookherjee and Png (1995) in taking this approach. A simple Nash bargaining technique is used to find a solution to this second stage of the problem.

In the first stage, then, the sum of equations (5) and (7) with respect to τ_i is maximized as follows:

View Expanded

The first-order condition is

View Expanded

Given that ∂_p/∂τ_i is negative (from equation (1)), if equation (10) holds, T_i > τ_i (that is, taxes are evaded).²¹ Given equations (10) and (12), the following proposition can be established:

Proposition I: Tax evasion occurs when institutional arrangements are weak, that is. when there is a low probability of detection and punishment in the event of conviction is relatively minor.

We can obtain τ_i (the tax that the government actually collects) as a function of the policy variables of the government. Thus,

View Expanded

Using the optimality conditions, the qualitative effects of changes in the policy variables on revenue collection can be derived.

Under fairly reasonable assumptions,²² the total derivatives of τ_i with respect to α, δ_p, and δ_g are positive. That is, when civil service wages and penalty rates are raised, tax collection increases. The intuition is that, as wages increase, the quality of auditors improves, raising the probability of detection, which, in turn, deters tax evasion. Also, raising penalty rates on either the tax collector or the private sector firm deters tax evasion. However, raising the statutory rate of taxation, T_i, does not guarantee an increase in tax collection. Under certain conditions, it is possible to show that raising the statutory lax rate may actually lead to a decline in collection.²³ This can be summarized in the following proposition:

Proposition 2: Tax collection rises unambiguously when penalty rates on private sector firms or tax collectors are raised, or when government wages are raised. Raising the statutory tax rate may or may not increase revenue collection.

So far. we have not differentiated between the effects of raising the wages of tax collectors and of auditors. Assume that α₁ and α₂ are the relative wages of the tax collectors and the auditors, respectively. Then, in equation (13), we can see that τ* = τ*(T, α₂. δp, δ_g). Consequently, increasing the salaries of tax collectors (α₁) only has no effect on the level of tax evasion. The intuition is fairly clear: as all players, including tax collectors, are assumed risk neutral, higher expected remunerations (either from wages or from bribes) are always preferred to lower, for given probability of detection and penalty rates. Thus, even if their wages are raised, there is no incentive on the part of tax collectors not to cheat unless the government takes complementary measures to improve tax administration (that is, to raise δ_g and p). Proposition 3 follows:

Proposition 3: Ceteris paribus, raising auditors’ wages reduces tax evasion; however, raising only the tax collectors’ wages has no effect on tax evasion.

Of course, proposition 3 holds under the assumption that tax collectors are risk neutral. If tax collectors were risk averse, it can be shown that an increase in their wages would also lead to a reduction in tax evasion. Propositions 1–3 point to the importance of setting the right incentives in government institutions to raise the efficiency of public services. While setting the right wage rates attracts better human capital into tax administration, other supporting measures, such as effective penalty rates, are also required to deter evasion.

Optimal Level of Bribes

We use a simple Nash bargaining solution to determine how the total gains from tax evasion are divided between the tax collector and the private firm. We assume in this simple framework that there is one tax collector for each firm and that the two have equal bargaining power. Equal bargaining power in this Nash bargaining game would imply that the two players agree to equalize their expected payoffs ex ante.²⁴ Setting the net expected gains of the two players equal to each other, we can obtain the optimal level of the bribe:²⁵

View Expanded

The term γ can be interpreted as the “share” parameter that determines the split of the tax evasion pie. Bribes emerge only when γ and T_i - τ_i are strictly positive and γ is strictly less than one.²⁶ These conditions are satisfied when

View Expanded

It can be shown that if equation (10) holds the restrictions imposed by equation (15) are satisfied.

The effect of exogenous changes in the policy variables (δ_p, δ_g, α, and T) on bribes in equation (14) arises from two distinct channels: the direct effect of the policy variables on the amount evaded, T_i - τ_i, and their indirect effect on the share parameter, γ. Since ∂b_i/∂τ_i = − γ in equation (14), a one-unit increase in the revenue collection leads to a decline in b_i by the extent of the share parameter, γ. The effect of increasing penalties on bribes depends on whether δ_p, or δ_g. is raised. We have shown in the previous section that, as penalties rise, revenue collection increases and tax evasion (T_i - τ_i) falls. An increase in δ_p, reduces the share parameter (as the firm has a higher cost of cheating), while an increase in δ_g raises the share parameter (as the collector faces a higher cost of cheating). A one-unit increase in both δ_p and δ_g leaves γ unchanged as the costs borne by both parties rise by equivalent proportions. Thus, when δ_p, is raised, b_i falls unambiguously while, if δ_g is raised, T_i - τ_i falls but γ increases, leaving the net effect on bribes ambiguous. We find then that raising the penalties on civil servants alone may not be sufficient to reduce the emergence of bribes.

A general increase in wage rates in tax administration has an unambiguously negative effect on bribes because of the rise in the auditors’ wages (α₂). As α₂ rises, the level of bribes falls through two channels: directly, because γ declines as the probability of an audit (p) increases; and indirectly, because an increase in p reduces tax evasion (T-τ).

Finally, the effect of an increase in the statutory tax rate on bribes is ambiguous. As the tax rate rises, the level of bribes increases because ∂γ/∂T>0 at a given level of γ. However, the probability of detection,p (see equation (1)), increases as T - τ increases and the share parameter, γ declines. Hence, the level of bribes could fall at higher rates of taxation. The effect of exogenous changes in the policy variables on bribes can be summarized in the following proposition:

Proposition 4: The level of bribes falls unambiguously when wages of auditors or penalty rates on private firms are raised. However, raising the statutory tax rate or increasing the penalty rates on tax collectors has an ambiguous effect on bribes. In addition, a ceteris paribus increase in the wages of the tax collectors only has no effect on bribes.