Lace A 08

Unofficial Economy and Corruption: Complements or Substitutes? Fabio Klein (EAESP/FGV) Abstract Much empirical research has shown that corruption and unofficial economy are somehow related, while only few theoretical models have been developed to explain how precisely they do so. Thorough this paper, I try to approximate theory and practice by developing a simple theoretical model that shows how corruption and unofficial economy relate to each other as either substitutes or complements, depending on variables such as taxes, quality of public goods and services and the institutional context. The main theoretical predictions is supported by a data set comprising of 87 countries observed between 1998, 2000 and 2002. The main conclusion is that the unofficial economy and corruption are complements when (i) taxes are high and quality of public goods are low, and (ii) both taxes and quality of public goods are high, with overall levels of corruption and unofficial economy being higher on (i). Inversely, corruption and unofficial economy are substitutes when (iii) both taxes and quality of public goods are low, and (iv) taxes are low and quality of public goods are high, though the levels of corruption and unofficial economy should be greater in (iii). JEL-class: D73, O17, H26, H41, K42 Key-words: Corruption, Unofficial Economy, Taxes, Public Goods, Rule of Law Introduction The unofficial economy literature has been experiencing a growing institutional focus. The first studies usually adopted a labour market perspective, which then further evolved to a greater concern over public finances and institutional variables, such as effective tax burden, level of bureaucracy, corruption and rule of law. One of the first authors to make this transition was Loyaza (1996), who studied the determinants and effects of the informal sector on economic growth in Latin America. He found that the relative size of the informal sector increases with higher tax burden, higher labour market restrictions and lower quality of government institutions. Moreover, he shows that a higher informal sector is associated with lower economic growth, because congestion over the use of public services by the informal sector reduces their provision for all economy, reducing total factor productivity. Worth mentioning is that he considers quality of bureaucracy, corruption and rule of law as the 3 proxies for the broader category of strength and efficiency of government institutions. Johnson, Kaufmann and Shleifer (1997) – JKS – develop and test a simple theoretical model of 3 possible equilibria between formality and informality (i.e. good, unstable and bad). Their empirical results show that unofficial economy increases with corruption, weak legal system and excessive regulations and tax burden. Additionally, a higher unofficial economy is correlated with lower economic growth. JKS compare the differences in growth rates and size of the unofficial economy between former Soviet Union and those of Eastern Europe and conclude that the main explanation for those differences is the lower quality of the institutions of the former.

1

The study by Johnson, Kaufmann and Zoido-Lobaton (1998) – JKZ – is very much an extension of JKS, where a broader sample of countries (Transition, OECD and Latin America countries) is considered, allowing group comparisons. Their new results point to a less significant direct impact of regulations and taxes on unofficial economy. They suggest that besides and perhaps more importantly then the formal taxes and regulations, is how they are applied, collected and enforced, which ultimately depends on the level of discretionary power (determined by the strength of the rule of law) and consequently on the levels of corruption. Johnson and Kaufmann (2001) summarize these findings affirming that firms go underground not because of high statutory taxes, but because of inadequate institutions (e.g. overregulation, corruption, weak legal system and lack of transparency of laws). If the results in the literature are not conclusive as to how tax burden, tax rate and to a lesser extent administrative bureaucracy affect unofficial economy, the majority of the studies affirm that higher corruption, weaker rule of law and excessive regulations are the main drivers for going underground. And within those, corruption has shown to be the most robust, followed by rule of law (Friedman et al, 2000; Johnson, Kaufmann and Shleifer, 1997; Johnson, Kaufmann and Zoido-Lobaton, 1998, 1999; Johnson and Kaufmann, 2001; Loyaza, 1996). However, some studies differ slightly on these conclusions. Johnson et al (2000) for example, working with firm-level data from more than 1000 companies in five transitional economies do affirm that corruption increases unofficial economy, although they cannot say in what direction the causality runs: if firms go underground to flee from bribes taken place in the official sector or if they pay bribes to safely hide their activities. As for rule of law, their results show that at the macro level (cross country), it is negatively associated with unofficial economy (as predicted), but its statistical significance is weakened at the micro level1. The theoretical model of Choi and Thum (2005) provides a quite different proposition of how corruption, shadow economy and law enforcement interact. If initially corruption at the official sector drives firms underground, it is exactly the possibility of becoming informal that will constrain public officials to exploit official activity, forcing a reduction in the level of bribes, which consequently lowers the costs of operating formally and thus promotes more official economic activity. Therefore, they claim that the unofficial economy works as a complement to the official one, and not as a substitute 2. Moreover, better law enforcement increases the risks of being caught in the unofficial sector, further allowing public officials to increase the bribe levels in both sectors. This could lead to a higher number of illegal activities, since the decision to operate formally or informally would still be made on the basis of the expected bribery payment.

1

They suggest that this happens because most firms in the sample are only partially informal, so that previous access to courts and legal system is not altered if a firm decides to underreport an extra portion of its output. 2 They show that these results hold even when the model is extended to allow for bribes in the informal sector.

2

Dreher and Schneider (2006) present a similar argument on substitutability and complementarity between formal and informal sectors, although their results are sensitive to the level of development. By dividing the countries in low and high income groups, they suggest that in poorer countries, the unofficial economy tends to increase corruption (and vice-versa), thus they act as complements, whereas in richer countries the shadow economy acts as a substitute to corruption, promoting its reduction as firms go underground to avoid bribes, which is aligned with the predictions of Choi and Thum (2005). Worth adding is that Dreher and Schneider (2006) find that the effects of corruption on unofficial economy become insignificant when interacted with institutional quality, which suggests that corruption reflects poor institutions. This paper focuses precisely on the debate over the complementarity and substitutability of corruption and unofficial economy started by Choi and Thum (2005) and Dreher and Schneider (2006) and hopes to contribute to a better understanding of how and when such relationship takes place. While with their empirical findings Dreher and Schneider (2006) have only began to scratch the surface of this concept, and Choi and Thum (2005) did not go beyond their pure theoretical model, I try to fill this gap by putting theory and empirical evidence together in a more systematic way. In the following section, a theoretical model is presented. Data information is presented on the third section, while the estimation strategy is presented on the fourth section. Preliminary results are introduced on the fifth section, followed by the main conclusions. Theoretical Model The model hereby presented is a variation of Loyaza’s (1996), where the capital rate of return depends on the available amount of public services relative to total production, represented by the endogenous production function below: α

G Yi = A  k i Y 

(1)

where Yi and ki are the production of and capital owned by individual firm i; A = exogenous productivity parameter; G = flow of public services; Y = total production in the economy (national income); and α = elasticity of output Yi with respect to G/Y with 0 < α ≤ 1, which measures the productivity of public services relative to private services. The net expected incomes of firms in the official and unofficial economies are given by α

(2)

G y iOE = (1 − [τ + E (C )] ) A  k i → “Corruption Without Theft” Y 

or α

(2’)

y

OE i

G = (1 − E (τ ) ) A  k i → “Corruption With Theft” Y 

and

3

α

 δG  (3) y = (1 − E (π ) ) A  ki  Y  where (2) is the income of firm i in the official economy (OE) net of taxes τ3 and an expected level of corruption E(C), (2’) is the income of firm i in the official economy (OE) net of expected taxes E(τ) and (3) is the income of firm i in the unofficial economy (UE) net of expected penalties E(π). The parameter δ expresses the fraction of G available to informal firms. The higher is δ, the more unofficial firms can illegally use the available public services and infra structure, leading to congestion of their use, which should reduce the level and quality of public services available to official firms, who originally paid for them in the form of taxes (G is mostly financed by τ and only marginally by π, the legal penalty paid by the informal firm if caught operating underground). UE i

Note that at (2), the net expected income from operating officially is given by tax τ to be paid with certainty, plus an uncertain level of corruption C, reflecting something close to the concept of “corruption without theft” as defined by Shleifer and Vishny (1993),meaning that the government does not lose tax revenues due to corruption. This formulation assumes that C enters as an extra and uncertain cost that does not substitute the official tax τ, reflecting situations where a firm is not sure whether it will have to pay extra bribes to “grease the wheel”4 to make its operations viable. But the net expected income of operating officially could be formulated as something closer to the concept of “corruption with theft” defined by these same authors. A suggestive formulation is shown in (2’), where C enters as a substitute of taxes τ, so that the total expected costs would be E(τ) = pτ + (1 – p)C. For the development of the mathematical arguments that follow, the assumption of corruption without theft will be used for the sake of space saving. Such dual formulation does not hold in the unofficial economy. Looking at the net income from operating unofficially (equation 3), there is no tax to be paid with certainty, given the very hidden nature of unofficial operations. Therefore, even the legal penalty π can only occur with some probability p, depending on whether the illegal firm is caught and punished. Even if it is caught, it can still avoid π by adhering to C. In this case, it makes more sense to think of corruption as a substitute to π, since a fiscal official that has uncovered an underground operation may still keep silence about it to his superiors and benefit by entering in collusion with the firm’s manager. It is unrealistic to think of a case where a firm pays the legal penalty π plus a certain amount of C: if it is brought to courts, it pays π and nothing else. If it wants to escape from courts, then entering in a corruptive deal is a possible alternative. Based on the above discussion, the expected costs of corruption at OE and expected penalty at UE are respectively given by (4)

E (C ) = p ⋅ 0 + (1 − p ) ⋅ C = (1 − p)C

and 3

Taxes τ include all monetary taxes required to run a business officially, such as income tax, social security contributions, value added taxes and others. 4 See Huntington

4

E (π ) = p ⋅ π + (1 − p ) ⋅ C (5) Assume that the penalty rate π is a linear function of the unofficial economy UE

π = aUE

(6)

where UE = YUE/Y is the size of the unofficial economy relative to the national income Y, and 0 < a ≤ 1 is a policy parameter determining the level of the penalty rate for a certain size of UE. For simplification, assume a = τ + (1 – τ ) = 1. This means that a firm operating underground, if caught and punished, would pay a fine equivalent to a 100% income tax. The above equation suggests that if the government is active in fighting increasing levels of UE, it would increase the penalty rate for higher levels of informality. OE UE After substituting (4) on (2), (5) on (3), setting the equilibrium condition y i = yi and replacing the result using (6), we get

(7)

δ α − 1 + τ (1 − p ) (1 − δ α ) UE = + C pδ α pδ α

which can be rewriten as (8)

UE =

(1 − p ) (1 − δ α ) C + 1 − pδ

α

p

1 1 + τ α pδ pδ α

Assume p = f(I), where I is a general measure of institutional quality with 0 < I ≤ 1, capturing things such as the rule of law, quality of government bureaucracy and regulations, important institutions that should work against UE. The maximum level of p = 1 would resemble the best level of institutional quality. It can be seen from the above equation that higher levels of p (better institutions) lead to a reduction in UE. Let us go briefly over each of the terms in (8). The first term shows how corruption C affects the unofficial economy UE, where higher levels of C promote higher levels of UE. The second term expresses the impact of institutions p(I) alone over UE, where better institutions lead to a higher probability of detection and punishment, thus reducing UE. But it is worth noting that institutional quality is weighing all other terms, showing that their impacts on UE are influenced by the current level of p(I). The third term can be interpreted as the impact of public service delivery on unofficial economy given by δα, which measures the accessibility (δ) of productive (α) public services by the unofficial sector. For a given level of δ, if public services improve (α increases), UE is reduced (note the negative sign of the term). For a given level of α, if unofficial firms have greater access to public services (δ increases), UE is increased. Note that δα is present in all but one term in (8). Finally, the fourth term expresses the impact of taxes τ on UE for given levels of p and δα, which should be positive (higher taxes lead to higher unofficial economy).

5

The above model differs from Loyaza’s in two main aspects. First, corruption is explicitly inserted in the model and separated from penalty π. In Loyaza, π reflects either the legal sanctions imposed to unofficial firms or bribes paid to fiscal officials as an alternative for escaping from law enforcement, which does not accurately captures the fact that C in the unofficial economy is a substitute to π. In fact, in Loyaza’s theoretical model, corruption occurs only within the unofficial economy. Second, the present model incorporates choice under uncertainty, given that the net incomes of official and unofficial firms are expected returns based on expected levels of corruption C and penalties π. Worth mentioning is that if p = 1, the second term in (7) disappears and the above model becomes exactly the same as the one suggested by Loyaza. This means that even under a perfect institutional setting, there would still be some level of unofficial business operations depending on taxes τ, productivity of public services α and accessibility to these services by unofficial firms given by δ. Corruption and Unofficial Economy: Complements or Substitutes? The present model shows that a firm’s decision to operate in the official (OE) or unofficial (UE) economy depends on the following variables: (i) the rate of official taxes (τ) and the expected rate of penalty (π) for operating unofficially; (ii) the current expected level of corruption (C) at both OE and UE; (iii) the level (G/Y) and productivity (α) of public services and its availability (δ) to unofficial firms. Moreover, this decision is influenced by the level of institutional quality p(I) that ultimately determines the probabilities given in (i) and (ii). The equilibrium condition determines the equilibrium levels of C, τ and π for given levels of p(I), α and δ. Let us now see the possible ways in which C and UE relate to each other. The decision to enter the Unofficial Economy UE Model 1: Corruption Without Theft (equation 2) Imagine a firm deciding to enter a new market or an existing firm deciding how to allocate its investments in a given moment. Assume that initially it wants to operate officially, agreeing to pay the current tax rates τ, which it perceives as being fairly low when there is good provision of public goods and services (α is closer to 1). If the context is of complete institutional quality (p = 1), the firm will not need to resort to corruption C to make its operations viable. But if p < 1, there is a certain chance that it will face additional costs of corruption given by (4), reducing the net income of operating officially. This would be specially true for lower levels of α, since poor provision of public services would force firms to corrupt public officials in order to overcome bureaucratic difficulties and expedite their operations (or public official would condition their delivery on offside extra payments). Given such conditions, the firm could decide to operate underground if yiOE < yiUE. As the current expected level of corruption at OE increases, it also increases the firm’s potential decision to migrate to UE, up to the point where yiOE = yiUE. This is the case where UE is a substitute to C taking place at OE. Model 2: Corruption With Theft (equation 2’)

6

Now assume that the official firm wants to evade from the official tax rate τ, which it perceives as being unfairly high when there is low provision of public goods and services (α is closer to 0). One way it could do so is by resorting to corruption C, “buying the silence” of fiscal officials so that it would not have to pay the high official taxes. In such case, C would work as a substitute to τ. Again, if p = 1, this would not be possible. But if p < 1, it could resort to C and not pay τ, as long as (1 – p)C < pτ. In this case, the government’s loss of tax revenues is directly reflected in higher levels of UE. Such situation shows the case where C is a complement of UE, where C works as the channel through which a firm hides part of its taxable output. Note that the two models differ in relation to how taxes and public services affect a firm’s decision. In the first, taxes are low and firms agree to pay them, but run the risk of paying an additional expected corruption if doing so is necessary to keep its operations running (or running faster). In the second model, taxes are high and firms want to run away from their fiscal duties, even if public goods and services are fairly good. Imagine an extreme situation where τ = 1, α = 1 and p = 1. This resembles a hypothetical country where there is zero corruption and public goods are perfect, but all income is taxable, so that it would not make sense to run a private business, thus making all entrepreneurship unofficial. In such case, there is no complementarity between C and UE, because C = 0. But if the firm is caught hiding, it will surely suffer the penalty π. But if p < 1, then C can work as a protection mechanism for unofficial firms. In this case, C and UE become closely linked as weapons against high τ. A similar, but worse situation would occur if p = 1, τ = 1, but α ≈ 0: even if the firm agreed to pay the high taxes, it could not count on good public services. This resembles a state that keeps all private resources to itself. Such situation forces firm to hide, although with the great risk of being caught and punished due to the government’s efficient detection mechanisms given by p = 1. Therefore, resorting to corruption is one of the few alternatives left. Given high taxes, the fiscal official may well benefit by entering in collusion with the firm, asking for a payment below τ to deliver public services and even protection. Comparing both models, we see that UE emerges from quite different causes. In the first model, UE is caused by corruption C taking place in the official economy OE, so that UE is a substitute to C. As suggested by Choi and Thum(2006), the fact that a firm has the option to go underground should promote a reduction in C: the more firms enter UE in order to avoid C, the lower are the expected bribes to be received by corrupt fiscal officials, forcing them to reduce the level of C. In this case, they show that UE would be a complement to OE. One could argue that fiscal officials could still extract bribes from unofficial firms after detecting them, but that should bring these officials lower payoffs compared to extracting bribes from official firms, due to the difficulties of detecting hidden operations. In the second model, UE is caused by the firm’s attempt to escape from its fiscal duties, so that taxes are the ultimate cause of UE. In such situation, C is the channel through which the firm is able to hide its earnings and go underground. In this case, C is a complement to UE. The decision to remain in the Unofficial Economy Once inside the unofficial economy, the firm runs the risk of being caught. This risk should increase as either p increases (improved institutions leading to greater detection capability), 7

yiUE increases (higher unofficial earnings and/or outputs leading to greater company size and visibility), or a combination of both occurs. If caught, the firm may resort to C as an strategy to avoid π, continuing to do so up to the point where yiUE = yiOE. Such situation describes C working as a complement to UE. If the expected costs of maintaining its operations underground becomes too high, then the firm may decide to become official. Theoretical Predictions The equilibrium matrix below resumes our main theoretical predictions: Table 1 – Equilibrium Matrix Public Goods and Services

High Taxes Low

High

Low

CT > CNT C ↔ UE CNT > CT C → UE

CT > CNT C ↔ UE CNT > CT C → UE

Based on the matrix, we have four possible equilibria5: (a) (Low, Low): In countries with low official tax rates and low quality of public goods and services, corruption without theft should be greater than corruption with theft (CNT > CT) and most of the unofficial economy should be interpreted as a substitute to corruption (C→UE). This happens because the problem faced by firms here are not taxes, but the low quality of public goods, for which they may have to pay an extra amount to get. (b) (Low, High): In countries with low official tax rates and high quality of public goods and services, corruption without theft should be greater than corruption with theft (C NT > CT) and most of the unofficial economy should be interpreted as a substitute to corruption (C→UE). Though this constitutes the best equilibrium of the four, some level of C may occur for reasons other than taxes and public services, like an public official favouring a firm in a procurement process. Compared to (a), (b) should bring lower overall levels of both corruption and unofficial economy (given higher public goods and services, it is less necessary to resort to C “to grease the well”), so that (b) brings greater welfare (w) than (a): 5

Note that in the four possible equilibria, we are assuming a given level of institutional quality p(I) < 1 so that there will always be some level of corruption. Institutional quality is not explicitly inserted in these four equilibria because it unnecessarily complicates the analysis, since it is highly related to C, UE, α and δ. In simplified terms, higher p(I) should produce lower C, UE and δ, and higher α equally in all four cases.

8

C L , H < C L , L → UE L , H < UE L , L ⇒ ( Low, High ) >( Low, Low) w

(c) (High, Low): In countries with high official tax rates and low quality of public goods and services, corruption with theft should be greater than corruption without theft (CT > CNT) and the unofficial economy should be interpreted as a complement to corruption (C↔UE). This is the worse equilibrium of all, because paying the high taxes is no guarantee of getting the public service delivered, so that paying extra offside money may still be required. In this case, both the high taxes and the “grease money” are a problem for the firm. (d) (High, High): In countries with high official tax rates and high quality of public goods and services, corruption with theft should be greater than corruption without theft (CT > CNT) and the unofficial economy should be interpreted as a complement to corruption (C↔UE). In this case, getting the service is not the problem, but paying for it might be too costly for the firm, so that entering the unofficial economy becomes profitable. Compared to (c), (d) should bring lower levels of both corruption and unofficial economy (for the same reason as explained in b) , so that (d) brings greater welfare (w) than (c): C H , H < C H , L → UE H , H < UE H , L ⇒ ( High, High ) >( High, Low) w

It is worth pointing out that the above predictions are in quite contrast with the hypotheses stated in Dreher and Schneider, who simply said that “…corruption and shadow economy are substitutes in high income countries, while they are complements in low income countries” (2006, page 1). The predictions above do not limit the correspondence of the complementarity and substitutability of corruption and unofficial economy to poor and rich countries respectively. This relationship does not depend on wealth, but on how taxes and public services are related. Data Definitions and Sources The data used in the present study comprises an unbalanced panel of 87 countries observed in 3 years (1998, 2000 and 2002)6, containing information on variables related to economic development, public finance and quality of governance, listed below7: 6

The number of countries for each year is: 87(1998), 86(2000) and 87(2002). There are 13 divergent observations between 1998 and 2000, 12 between 1998 and 2002 and only 3 between 2000 and 2002. 7 Most data related to economic development and public finance were obtained from the World Bank Development Indicators 2004 (World Bank, 2005). The exceptions are “mean years of schooling of labour force”, taken from Barro and Lee (2000), and “unofficial economy (%GDP)”, taken from Schneider and Enste (2000) and Schneider (2005). The governance indicators were obtained from Kaufmann, Kraay and Mastruzzi

9

Development indicators: Unofficial economy(%GDP, 1999/2000, 2001/2002 and 2002/2003), GDP per capita (purchasing power parity, $, all 3 years), Mean years of schooling of labour force (age ≥ 25, 2000), Improved sanitation facilities (% of population with access, 2002), Improved water source (% of population with access, 2002), Roads, paved (% of total roads, all 3 years), Freedom of the Press (measures the degree of freedom of the press and media, all 3 years). Public Finance: Public health expenditure (% of GDP, 2002), Public spending on education (% of GDP, 2000 and 2002), Highest marginal corporate tax rate (%, all 3 years), Highest marginal individual tax rate (%, all 3 years), Tax revenue (% of GDP, all 3 years). Quality of Governance: Control of Corruption (measures the exercise of public power for private gain, including both petty and grand corruption and state capture), Government Effectiveness (measures the competence of the bureaucracy and the quality of public service delivery), Rule of Law (measures the quality of contract enforcement, the police, and the courts, as well as the likelihood of crime and violence), Regulatory Burden (measures the incidence of market-unfriendly policies), Voice and Accountability (measures political, civil and human rights), Political Instability and Violence (measures the likelihood of violent threats to, or changes in, government, including terrorism). Data Treatment Unofficial Economy: The definition is borrowed from Schneider and Enste (2000) and Schneider (2005), from which the data was obtained, and is as follows: the unofficial economy “…includes all market-based legal production of goods and services that are deliberately concealed from public authorities…to avoid (i) payment of income, value added or other taxes, and social security contributions, (ii) having to meet certain legal labour market standards, such as minimum wages, maximum working hours, safety standards, etc., and (4) to avoid complying with certain administrative procedures, such as completing statistical questionnaires or other administrative forms” (Schneider and Enste, 2000). The estimates of unofficial economy provided are a combination of three different methodologies8, and are available for the years 1999/2000, 2001/2002 and 2002/2003. Because such periods do not perfectly match the years of the study (1998, 2000 and 2002), it is not possible to estimate equation 8 using the same time t for the dependent and independent variables. Therefore, the dependent variable UE will be measured at t + 1, while the independent variables will all be measured at t. For example, the estimate of the unofficial economy for Argentina is 25.4% for the period 1999/2000. This estimate is plotted in the panel data set for the year 1998. Ideally, it would be desirable to have the dependent and independent variables all at the same t, specially because the theoretical model indicates a possible simultaneity between C and UE, where current expected levels of corruption leads to the current decision to go underground. Although it might be somewhat arbitrary to define an interval of one to two years as the required time lag for the economy to adjust between formality and informality, it is still a reasonable assumption if (2005). 8 The methods are MIMIC (Multiple Indicators, Multiple Causes), Currency Demand and Electricity Differential. Please, refer to the papers for detailed information.

10

we think of fiscal years comprised of one year, which determine private and public agents budget planning. This means that one to two fiscal years provide sufficient time and information for economic agents to decide how much to invest, as well as where to do it: in the official or unofficial economy. In this sense, the size of the unofficial economy in Argentina in 1999/2000 can be thought as reflecting its public finances, services and institutional conditions in 1998. Productivity of Public Goods and Services: In the absence of an objective measure capturing the productivity of public goods and services (α), one was created based on the following formula:

ϕG 2 αG = ⋅ε (I G / Y )

(9)

where φG = performance of public good/service G, IG = total public investment/expenditure on public good/service G, Y = national income, and ε = government effectiveness. This provides a rough measure of the quality and efficacy of public investment that allows country comparison. The squared term is used to reward those countries with overall higher performance. If it were not used, than two very dissimilar countries could be ranked equally in terms of α, provided that one of them had lower levels of both φG and IG/Y. For example, imagine that country A had φG = 5 and IG/Y = 10, whereas country B had φG = 1 and IG/Y = 2, ignoring ε for the moment. Without the squared term, both would have equal productivity of α = 0.5, while applying the quadratic form makes country A 5 times more productive than B. The public goods and services (G) hereby considered are Education, Health and Roads’ infra-structure. Government effectiveness (ε) is one of the six governance indicators present in Kaufmann, Kraay and Mastruzzi (2005). Using this additional measure of overall government effectiveness helps on calibrating the individual measure of productivity based on only one policy dimension. Based on the available information from the data as described above, the productivity of Education and Health policies becomes: (10)

(11)

α Education =

α Health

( Mean Years of Schooling ) 2 Public Spending on Education (% GDP)

( Sanitation *

⋅ε

Water Access) = ⋅ε Public Spending on Health (% GDP) 2

Since the data used does not provide information on the percentage of public spending on roads, the formula for this public good is just given by (12)

α Roads = Roads Paved ⋅ ε 11

The resulting numbers were all normalized so that 0 < α ≤ 1, where higher levels mean higher productivity. Corruption and other Governance Indicators: The governance indicators in Kaufmann, Kraay and Mastruzzi (2005) all range from -2.5 to 2.5. This range was normalized to vary between 0 and 1, where lower rates mean lower institutional quality. To make it coherent with section 2 and avoid misinterpretations, the values in “control of corruption” were subtracted from 1 so that the new variable “corruption” also ranges from 0 to 1, but now having higher values associated with higher corruption. Worth mentioning is that these governance indicators are mainly subjective, perceptions based measures compiled from different surveys and polls applied to selected firms and individuals. The intrinsic subjectivity of these indicators has been subject to critics, suggesting that they are not very reliable proxies for objective indicators. Although plausible, these critics are not strong enough to invalidate the efficacy of such indicators9. But for the purpose of this article, such controversy is not even a problem, since the model hereby developed is based on expected levels of corruption, so that using a perceptions based measure of corruption is in fact preferred to any other objective measure, even if it was available. Descriptive Statistics Table 2 below resembles the equilibrium matrix from section 2, using αHealth. Similar tables for Education and Roads are provided in the Appendix. They were constructed by grouping the countries according to the four possible equilibria discussed in the preceding section. The threshold used to separate countries from high and low values of both taxes τ (corporate tax rate) and public services productivity α was the group average. For example, the average highest income corporate tax for the group below (n = 60) was 29.2%. Countries with τ ≤ 29.2% were classified as having low tax, whereas those with τ ≥ 29.2% were classified as having high tax. The same reasoning was applied for α. For each group, the average corruption C, size of unofficial economy UE and GDP/capita are provided. It can be seen from Table 2 that the evidence provides quite strong support for the theoretical predictions with regards to the level of corruption and unofficial economy in each equilibrium and its welfare consequences in terms of income per capita. Obviously, the resulting groups described in these three tables are far from being perfect for a number or reasons, such as the variables chosen as proxies for τ and α, the formula adopted to capture α, the threshold applied to separate countries from low and high values, 9

For arguments in favour of such indexes, see the authors’ self-defence at pages 27-31. For an argument in favour of their corruption index as opposed to the popular index from Transparency International (TI), refer to Kaufmann, Kraay and Mastruzzi (2003, pgs.32-39). Refer also to the interesting working paper by Fisman and Miguel (2006), a natural experiment on parking tickets violations committed by diplomatic representatives visiting New York during United Nations missions, where they show a high correlation between their infractions ranking and the corruption index present in Kaufmann, Kraay and Mastruzzi (2005).

12

and data availability. These problems may lead to clusters that look strange, having outliers in them. Nevertheless, the results provide us with a powerful picture of how the levels of corruption and unofficial economy change in different equilibria, which in turn strongly depend on the fairness of taxes and the quality of public goods and services. Table A1 in the Appendix presents the statistical summaries of the main variables analyzed in this study. Estimation Strategy Based on equation (8), the following linear regression model is tested: (13) UEi ,t +1 = αt + β1 ln ( Cit ⋅ I it ⋅ PSit ) + γ1 ln I it + γ 2 ln( I it ⋅ PSit ) + γ 3 ln(τ it ⋅ I it ⋅ PSit ) + uit with t = 1,2 and 3 where C stands for Corruption, I for Institutional Quality, PS for Public Services Productivity and τ for Taxes. The first three of these are transformed into their natural logarithmic form so that we can interpret their respective coefficients as the unit change in UE given a percentage change in their related variable. The symbols α and u are the usual constant and error terms. The fact that all independent variables are lagged one year in relation to the dependent variable is explained in subsection “Data Treatment”. Step 1 I begin estimating (13) without the interaction terms (columns 1 through 5) in Tables B-E, while use both interacted and individual terms in a subsequent step (columns 6) and finally check the exact specification given by equation 13 (columns 7). This procedure is applied for each of the three proxies for PS, using αEducation, αHealth and αRoads (see equations 10, 11 and 12). For Education, the only available year is 2000, while for Health is 2002. Information on Roads are available for all three years, but 1998 is be used given its higher number of observations10. For the variable I, Rule of Law is used in all specifications. For the variable τ, highest marginal corporate tax rate is used. These same procedures are applied using the overall indicator of Government Effectiveness as a proxy for PS. Since data on this variable is available for all years, I combine the observations in one pooled cross-section. The results are shown in Tables B through E in the Appendix.

10

For 1998, information on Roads are available for 72 observations. For the years 2000 and 2002, there are 36 and 40 observations respectively.

13

Table 2 - Equilibrium Matrix using Health (2002) Public Goods and Services (α) - Health * Government Effectiveness (2002)

n = 60 High (τ > 29.2%) Avg = 32.8% SD = 2.65% Max = 38% Min = 30% n = 15 Taxes (τ)

High (α > 0.22) Avg = 0.38 SD = 0.10 Max = 0.56 Min = 0.24 n = 15

Low (α ≤ 0.22) Avg = 0.10 SD = 0.07 Max = 0.22 Min = 0.00 n = 25

Australia, Austria, Canada, Costa Rica Jamaica, Japan, Mexico, Netherlands Philippines, Russian Federation Sri Lanka, Thailand, Ukraine United States, Uruguay

Cameroon, China, Colombia, Congo Cote d'Ivoire, Ghana, Guatemala India, Indonesia, Kazakhstan, Kenya Morocco, Mozambique, Namibia Nigeria, Panama, Paraguay, Senegal South Africa, Tanzania, Turkey, Uganda Venezuela, Zambia, Zimbabwe

C = 0.37 Low ( τ ≤ 29.2%) Avg = 21% SD = 6.5% Max = 29% Min = 12% n=8

UE = 31.5%

GDP/cap = $15,248

Bulgaria, Chile, Finland Hungary, Iran, Oman, Slovak Republic, Sweden

C = 0.31 UE = 22.6%

GDP/cap = $14,288

C = 0.62

UE = 43.2%

GDP/cap = $3,372

Azerbaijan, Bolivia, Botswana, Brazil, Cambodia Dominican Republic, Ecuador,Nicaragua, Papua New Guinea, Peru, Romania, Uzbekistan

C = 0.60

UE = 46%

High (τ > 29.2%) Avg = 32.7% SD = 3% Max = 40% Min = 30% n = 25

Low ( τ ≤ 29.2%) Avg = 23.2% SD = 4.2% Max = 27% Min = 15% n = 12

Taxes (τ)

GDP/cap = $4,353

High (α > 0.22) Low (α ≤ 0.22) Avg = 0.37 SD = 0.09 Max = 0.56 Min = 0.31 n = 8 Avg = 0.10 SD = 0.07 Max = 0.21 Min = 0.00 n = 12 Public Goods and Services (α) - Health * Government Effectiveness (2002)

14

Step 2 I run a panel data estimation using fixed-effects (FE) method in order to eliminate timeconstant effects that may be related to the independent variables and that also affects the dependent variable. The results are shown in Table F. Columns 1 through 3 refer the whole period (1998, 2000 and 2002), while in columns 4 and 5 only the years 2000 and 2002 were considered, since this period provides us with an almost complete balanced panel. As in the case of the pooled regression, the unavailability of data forces me to use Government Effectiveness (ε) only as a proxy for PS. Step 3 Motivated by the equilibrium matrix of section 2, I ran a clustered regression for four different clusters of countries for both the cross-sectional and panel models. The thresholds applied to divide countries between low and high taxes and public services productivity are their respective sample average for 2000 and 2002. Step 4 There is some reason to suspect that I and C are closely related to each other (higher/lower institutional quality leading to lower/higher corruption)11, creating endogeneity biases in the results. Additionally, even though the dependent variables are all lagged in our model, there may still be some degree of simultaneity between C and UE when the explanatory variables are measured at t = 2002, because the estimated size of UE for such cases is for the period 2002/200312. In order to address these problems of endogeneity, the most significant specifications obtained in the previous procedures were selected and tested using instrumental variables (IV) estimation. The following variables were used as instruments for C and I: Press Freedom and Voice & Accountability.The assumption is that once all the required controls had been accounted for in the estimation of UE, these instruments should not be correlated with the error term, but only correlated with C and I. Econometric Results Step 1 Looking at columns 1 through 3 in Tables B through D, we see that corruption is significant and presents the expected negative signs. Take column 1 from Table B for example: an increase of 1 percentage point in C reduces UE in 0.119% points. But C is very sensitive to the inclusion of Law (stands for Rule of Law, the proxy used for institutional quality). But this is fairly straightforward to interpret: as Law increases, it is expected that C decreases, since Law captures the government’s capacity to detect and punish illegal activities. It is also notable that Rule of Law is very sensitive to the inclusion of the variable income per capita. This probably happens because there is a quite substantive variation of income per capita among countries with similar levels of institutional quality. This can be seen by observing the countries grouped in the equilibrium matrixes discussed in section 3. But the interesting point is that even though there is variation of income level among countries 11

In fact, the correlation between corruption and rule of law (the measure to be used as proxy for institutional quality) for the whole sample (260 country-years) is -0.972. 12 This simultaneity problem does not happen for t = 1998 and 2000, since for these years, the estimates of UE are for the periods 1999/2000 and 2001/2002 respectively.

15

grouped under similar institutional and fiscal conditions, there is still some relationship between income level, corruption and unofficial economy, where lower income is generally associated with higher corruption and unofficial economy. Tables D and E show interesting results regarding the interaction terms suggested by the theoretical model. Column 6 in Table E suggests that the coefficient of corruption alone is positively correlated with unofficial economy (as expected), but when interacted with a given level of institutional quality (rule of law in our case) and quality of public goods and services, it is negatively correlated with UE. Note that combining both still makes C to be positive on UE. A comment on the effects of taxes on unofficial economy is important. First, in most cases taxes were not statistically significant, and when they were, in all but one case it had negative signs, indicating that taxes alone do not seem to impact much on UE. As our model suggests, its effects can only be understood when public goods and services, as well as institutional quality, are taken into account. It is worth noting that I also tested highest marginal individual tax rate instead of corporate tax rate in order to capture cases where the legal constitution of a company confounds itself with that of the owner, making the individual to be the firm and vice-versa. This is often observed in smaller and informal businesses, which can constitute an important portion of the unofficial economy. But the statistical results were even less significant than those for corporate tax. Step 2 The results obtained from the panel estimation using FE also shows that C and UE are related. Such relationship seems to be even stronger when C is interacted with institutional quality (Rule of Law) and Government Effectiveness. The surprising result is its negative sign. This is probably caused by the fact that the growth rates of C and UE are different, which is emphasized in a panel estimation, since it reports changes in UE from changes in C along time. As can be derived from information on Table A1 in the Appendix, the growth rates of UE and C between 1998 and 2000 are quite similar on average (3.8% and 3.9% respectively), but from 2000 to 2002, UE only grew 1.18% on average, while C grew 3.2%. The results for steps 3 and 4 are not shown here because they ended up being weak in virtually all cases. As mentioned, there is some potential endogeneity on the model, in part confirmed by the results presented on the tables showing that Law many times captures the effects of C on UE . I tried controlling for the endogeneity of C using Freedom of Press as an instrument, based on the intuitive notion that a more active and strong press is able to uncover and publish corruption scandals, but should not impact on a private agent’s decision to operate inside or outside the unofficial economy. Unfortunately, in none of the cases was the IV/2SLS fruitful, as it caused most of the coefficients to be statistically insignificant. I also tried using Voice & Accountability as instrument, but results were still weak, which were further confirmed by statistics of over and under-identifying restrictions. This posits the challenge of controlling for the endogeneity bias in our estimates through the use of adequate instruments, which is not a trivial task, specially because it requires a precise specification of a model for corruption, something beyond the scope of this paper.

16

As for step 3, the results did not show any statistically meaningful correlation between the explanatory variables and the dependent one when countries were grouped in clusters. This is possibly caused by some degree of within-group variation in those variables, so that even though they were grouped as similar according to the threshold specified, they still vary considerably in their taxes and public services productivity. Also, dividing the sample of countries into four groups creates weaker asymptotic properties, since there is a substantial reduction in the number of observations. Conclusions This paper has presented a theoretical and empirical basis to help our understanding of how corruption and unofficial economy are related to each other depending on variables such as taxes, provision of public goods and services, and institutional quality. Based on the theoretical predictions and preliminary empirical findings, it concludes that the unofficial economy and corruption are complements when (i) taxes are high and quality of public goods are low, and (ii) both taxes and quality of public goods are high, with overall levels of corruption and unofficial economy being higher on (i). Inversely, corruption and unofficial economy are substitutes when (iii) both taxes and quality of public goods are low, and (iv) taxes are low and quality of public goods are high, though the levels of corruption and unofficial economy should be greater in (iii). This conclusion adds to the model suggested by Choi and Thum (2005) and differs from Dreher and Schneider (2006), showing that different types of corruption (with or without theft) take place under different fiscal and institutional contexts, which affect the expected costs (and benefits) of corruption and operating in the unofficial economy. These two types of corruption end up determining different links with the unofficial economy, and should not at principle depend on a nations’ wealth, as suggested by Dreher and Schneider.

17

Appendix Table A1 - Statistical Summary of the Main Variables

Variable gdp_capita gov_effe law corruption ue tax_corp edu roads

Obs 87 87 87 87 87 87 0 72

1998 Mean Std. Dev. Min Max 11165.18 9372.772 497.567 32909.81 0.588723 0.2032374 0.233876 1.018764 0.592442 0.20235 0.240468 0.972047 0.417388 0.2259804 0.000878 0.722044 31.61609 14.58275 8.6 67.1 30.28736 7.478509 6 45 57.06625

33.78509

3.5

100

Obs 86 86 86 86 86 86 58 36

Mean 11207.98 0.567861 0.568154 0.43328 32.86395 29.33372 4.461804 57.69472

2000 Std. Dev. 9704.103 0.19259 0.204112 0.223696 14.64801 7.920571 1.524867 36.2149

Min 516.6046 0.1686 0.261424 0.000731 8.7 0 1.244017 5.5

Max 34164.46 0.987857 0.922135 0.711805 68.1 54 8.387465 100

Obs 85 87 87 87 87 87 62 40

Mean 11768.32 0.5694 0.549561 0.447852 33.25402 28.24483 4.930938 59.9465

2002 Std. Dev. 10082.94 0.206076 0.201421 0.220919 15.00524 7.92718 1.56125 34.16588

Min 559.2109 0.142589 0.220793 0.009772 8.4 0 1.060274 6.67

Max 35652.91 0.977318 0.892403 0.763403 68.3 40 8.512175 100

Equilibrium Matrix using Education (2000)

n = 51 High (τ > 30.8%) Avg = 34.4% SD = 2% Max = 38% Min = 31% n = 10

Taxes (τ) Low ( τ ≤ 30.8%) Avg = 26.7% SD = 4.4% Max = 30% Min = 15% n = 16

Public Goods and Services (α) - Education*Government Effectiveness (2000) High (α > 0.38) Low (α ≤ 0.38) Avg = 0.61 SD = 0.22 Max = 1.0 Min = 0.39 n = 10

Argentina, Australia, Canada Czech Republic, France, Greece Israel, Philippines, Russian Federation, Spain

Avg = 0.16 SD = 0.09 Max = 0.34 Min = 0.03 n = 11

High (τ > 30.8%) Cameroon, Colombia, India, Iran, Italy, Jamaica, Malawi, Mexico Portugal, Senegal, Zambia

C = 0.33 UE = 25.7% GDP/cap = $18,018 Chile, Denmark, Ecuador, Finland Germany, Hungary, Indonesia Japan, Korea, Norway, Poland Romania, Slovak Republic, Sweden United Kingdom, Uruguay C = 0.30 UE = 23.7% GDP/cap = $17,141 High (α > 0.38) Avg

UE = 34.1% GDP/cap = $6,790 Bolivia, Brazil, Costa Rica, Dominican Republic, Kenya, Malaysia, Nicaragua, Panama, Paraguay, South Africa, Thailand, Turkey, Uganda, Zimbabwe C = 0.56 UE = 42.7% GDP/cap = $5,376 Low (α ≤ 0.38)

= 0.61 SD = 0.16 Max = 0.94 Min = 0.39 n = 16

Avg = 0.16 SD = 0.08 Max = 0.33 Min = 0.04 n = 14

Avg = 37.6% SD = 5.8% Max = 54% Min = 33% = 11

n

C = 0.53

Low ( τ ≤ 30.8%)

Taxes (τ)

Avg = 27.7% SD = 4.2% Max = 30% Min = 15% n= 14

Public Goods and Services (α) - Education*Government Effectiveness (2000)

18

Equilibrium Matrix using Roads (1998) Public Goods and Services (α) - Roads * Government Effectiveness (1998) High (α > 0.36) Low (α ≤ 0.36) Avg =

n = 72

0.65 SD = 0.18 Max = 0.97 Min = 0.39 n = 20

Austria, Belgium, Czech Republic Denmark, France, Greece, Ireland Israel, Italy, Japan, Kyrgyz Republic Lithuania, Netherlands, New Zealand Poland, Slovak Republic, Spain Thailand, United Kingdom, Uruguay

High (τ > 28.8%) Avg = 34.1% SD = 3.1% Max = 40% Min = 29% n = 20

Taxes (τ)

C = 0.27

UE = 23.5%

GDP/cap = $18,390

Low ( τ ≤ 28.8%) Avg = 23% SD = 7.7% Max = 28% Min = 6% =8

n

Finland, Kuwait, Malaysia, Norway Puerto Rico, Singapore, Sweden United Arab Emirates C = 0.15

UE = 22% GDP/cap = $20,842 High (α > 0.36)

Avg = 0.64 SD = 0.19 Max = 1.0 Min = 0.38 n = 8

Avg = 0.16 SD = 0.09 Max = 0.32 Min = 0.00 n = 28

Argentina, Australia, Azerbaijan, Bulgaria Cameroon, Colombia, Costa Rica Cote d'Ivoire, Guatemala, India, Indonesia Jamaica, Kazakhstan, Kenya, Mexico Morocco, Namibia, Nicaragua, Oman Peru, Philippines, Romania, Russian Federation Senegal, South Africa, Tanzania Ukraine, Zimbabwe C = 0.55 UE = 38.4% GDP/cap = $5,435 Bolivia, Botswana, Brazil, Chile Dominican Republic, Ecuador El Salvador, Estonia, Honduras Hungary, Iran, Korea, Nigeria Panama, Papua New Guinea, Turkey C = 0.53 UE = 38.9% GDP/cap = $6,086 Low (α ≤ 0.36)

High (τ > 28.8%) Avg = 33.4% SD = 3% Max = 40% Min = 30% 28

n=

Taxes (τ) Low ( τ ≤ 28.8%) Avg = 20.4% SD = 5.8% Max = 28% Min = 12% n= 16

Avg = 0.14 SD = 0.11 Max = 0.34 Min = 0.00 n = 16

Public Goods and Services (α) - Roads * Government Effectiveness (1998)

19

Table B - Cross Sectional Regressions on Unofficial Economy using Health Services Productivity (2002) Dependent Variable: Unofficial Economy (% GDP) 1 2 3 4 corruption 11.9464*** 5.2072** 5.3963** 5.5411* 1.47 2.3 2.33 3.23 rule of law -17.8469*** -17.3154*** -14.8973** 4.87 4.98 7.16 tax 0.083 0.0071 0.15 0.26 health productivity -0.9749 1.54 gdp/capita corrup*law*health law*health tax*law*health Constant

5 6 7 3.3956 0 3.2 . -5.2181 -4.6019 -6.4364 7.44 7.89 7.82 -0.1541 -1.9808 0.25 1.5 1.2543 0 1.64 . -7.7824*** -8.1628*** -7.6307*** 2.63 2.63 2.62 3.6337 3.5096 3.19 3.21 -47.4388 0.2084 36.89 7.72 45.15 -2.5116 36.58 6 45.5207*** 26.6319*** 24.8381*** 28.2332*** 109.1371*** 16.5873 111.7409*** 1.94 5.47 6.34 10.11 28.43 80.14 35.33 0.43 0.50 0.50 0.39 0.45 0.45 0.45 87 87 87 60 59 59 59

R-squared N * p<0.10, ** p<0.05, *** p<0.01 standard errors are shown below the coefficients

Table C - Cross Sectional Regressions on Unofficial Economy using Education Services Productivity (2002) Dependent Variable: Unofficial Economy (% GDP) 1 2 3 4 5 corruption 6.4904*** 1.4721 1.4067 1.4152 1.179 1.06 1.29 1.31 1.72 1.67 rule of law -24.0716*** -24.4033*** -13.9619** -7.0375 4.39 4.47 6.91 7.54 tax -0.0638 -0.4237* -0.4055* 0.15 0.25 0.24 edu productivity -4.2104* -1.0492 2.27 2.71 gdp/capita -5.5305* 2.77 corrup*law*edu law*edu tax*law*edu Constant

40.5890*** 19.4199*** 1.83 4.17 R-squared 0.30 0.48 N 86 86 * p<0.10, ** p<0.05, *** p<0.01 standard errors are shown below the coefficients

6 50.2248 31.39 -4.3773 8.16 -2.0227* 1.08

7 0 -6.4507 8.31

-5.3287* -5.6117* 2.74 2.81 -48.7986 1.128 31.19 1.7 0 7.0908 . 7.37 47.131 -8.9968 30.67 6.9 21.0041*** 32.3437*** 89.3232*** -23.0627 108.3577*** 5.57 9.27 29.95 78.86 37.18 0.47 0.40 0.44 0.46 0.43 86 51 51 51 51

20

Table D - Cross Sectional Regressions on Unofficial Economy using Roads Services Productivity (2002) Dependent Variable: Unofficial Economy (% GDP) 1 2 3 4 5 6 7 corruption 5.9239*** 2.1598* 2.0721* 1.3692 1.2669 0 0.89 1.16 1.17 0.94 0.91 . rule of law -21.1166*** -21.1193*** -15.0228*** -5.4665 -1.3957 -2.7794 4.7 4.7 3.74 4.39 4.61 4.6 tax -0.1482 -0.0483 -0.103 -0.8483* 0.16 0.12 0.11 0.44 roads*gov -4.3271*** -2.4815** 0.95 1.05 gdp/capita -6.1232*** -6.1034*** -5.9301*** 1.69 1.68 1.69 corrup*law*roads 1.6455* 1.3283 0.92 0.91 law*roads -19.5619** -1.6878 -9.74 -3.12 taxcorp*law*roads 15.4013 -2.2612 9.48 2.58 Constant 39.5236*** 22.1668*** 26.5377*** 19.1447*** 84.1822*** 55.8236** 86.9801*** 1.75 4.17 6.21 4.54 18.34 25.74 20.28 R-squared 0.33 0.46 0.46 0.54 0.57 0.58 0.57 N 87 87 87 147 146 145 145 * p<0.10, ** p<0.05, *** p<0.01 standard errors are shown below the coefficients

21

Table E - Pooled Cross Section Regressions on Unofficial Economy using Goverment Effectiness (1998, 2000, 2002) Dependent Variable: Unofficial Economy (% GDP) 1 2 3 4 5 corruption 7.0920*** 2.1895*** 2.1443*** 1.9980** 1.4517* 0.63 0.78 0.79 0.81 0.74 rule of law -22.1789*** -22.3376*** -18.6463*** -6.955 2.53 2.55 4.64 4.53 tax -0.0421 -0.0336 -0.0933 0.09 0.09 0.08 gov_effectiveness -4.4279 -1.0902 4.64 4.28 gdp/capita -6.8543*** 0.93 corrup*law*gov

6 25.5079*** 9.18 -0.8431 8.26 -1.0365*** 0.34

7

-2.9985 8.37

-6.8561*** -6.7339*** 0.92 0.93 -23.8050*** 1.5034** 9.13 0.74 law*gov dropped -1.3446 . 4.9 taxcorp*law*gov 20.7101*** -2.7449 7.86 2.08 d98 0.5467 0.958 1.0439 0.8579 -0.338 -0.2739 -0.6506 1.84 1.62 1.63 1.64 1.51 1.46 1.48 d00 0.7689 0.8406 0.8854 0.734 0.0315 0 0 1.84 1.61 1.62 1.63 1.49 . . d02 0 0 0 0 0 -0.0057 -0.0402 . . . . . -1.49 -1.51 Constant 40.5362*** 20.6280*** 21.6652*** 20.9443*** 92.8009*** 52.4734*** 98.4561*** 1.45 2.6 3.35 3.44 10.26 19.16 12.28 R-squared 0.33 0.48 0.48 0.48 0.57 0.58 0.57 N 260 260 260 260 258 254 254 * p<0.10, ** p<0.05, *** p<0.01 standard errors are shown below the coefficients

22

Table F - Panel Data Estimation with Fixed Effects (1998, 2000 e 2002) Dependent Variable: Unofficial Economy (% GDP) 1 2 3 4 5 corruption -0.4617*** -2.2696* -0.3953** 0.13 1.18 0.17 rule of law 0.6648 0.1782 0.24 1.3178 2.2891 1.00 1.28 1.29 1.21 1.72 tax 0.0021 0.0625 -0.034 0.01 0.05 0.03 gov_effectiveness 0.1586 -1.0221 0.62 0.95 gdp/capita -9.2579*** -9.2717*** -9.3347*** -9.6138*** -9.6165*** 0.36 0.37 0.37 0.38 0.38 corrup*law*gov 1.7983 -0.4711*** -0.4024** 1.17 0.14 0.17 law*gov dropped 1.0264 0.2036 . 0.72 1.1 taxcorp*law*gov -1.6663 -0.364 -0.9423* 1.04 0.36 0.54 d00 1.2330*** 1.2490*** 1.2403*** -0.9839*** -0.9791*** 0.13 0.13 0.13 0.12 0.12 d02 2.2529*** 2.2533*** 2.2485*** 0 0 0.15 0.15 0.15 . . Constant 113.3422*** 117.1054*** 115.2710*** 119.4347*** 121.5961*** 3.43 3.93 3.69 3.83 4.16 R-squared 0.73 0.73 0.73 0.78 0.79 N 258 254 254 171 167 * p<0.10, ** p<0.05, *** p<0.01 standard errors are shown below the coefficients

References Bardhan, P. (1997) Corruption and development: A review of issues, Journal of Economic Literature 3: 1320-1346. Barro, R. and Lee, J. (2000) International data on educational attainment updates and implications, NBER Working Paper, no. 7911. Blackburn, K., Bose, N. and Haque, M.E. (2004) Endogenous corruption in economic development, Research Paper 2004/16, GEP, University of Nottingham. Blanchflower, D. and Oswald, A. (1995) The Wage Curve, Cambridge, MIT Press. Bruegger, E. (2005) Endogenous Institutions and the Dynamics of Corruption, Bern University Discussion Papers. Burnside, C. and D, Dollar (2000) Aid, Policies, and Growth, American Economic Review, 90:4, 847-868. Chand, S.K. and Moene, K.O., (1999) Controlling Fiscal Corruption, World Development Vol. 27, No. 7, pp. 1129-1140. 23

Choi, J. and M. Thum, (2005) Corruption and the shadow economy, International Economic Review, Vol.46, n.3, 817-836. Dreher, A., Kotsogiannis, C. and McCorristony, S. (2004) Corruption Around the World: Evidence from a Structural Model, Exeter Working Paper Dreher, A. and Schneider, F. (2006) Corruption and the Shadow Economy: An Empirical Analysis, IZA Discussion Paper 1936. Easterly, W., Kremer, M., Pritchett, L., and Summers, L.H., (1993) Good policy or good luck? Country growth performance and temporary shocks, Journal of Monetary Economics, 32 (3), pp. 459-483. Easterly, W., (2001) The lost decades: Developing countries’ stagnation in spite of policy reform 1980-1998, Journal of Economic Growth, 6 (2), pp. 135-157 Fisman, R. and Miguel, E. (2006) Cultures of Corruption: Evidence from Diplomatic Parking Tickets, NBER Working Paper 12312. Friedman, E., Johnson, S., Kaufmann, D. and Zoido-Labton, P. (2000) Dodging the grabbing hand: The determinants of unofficial activity in 69 countries, Journal of Public Economics, 76/4, pp.459-493. Gupta, M. (1993) Rural-Urban Migration, Informal Sector, and Development Policies: A Theoretical Analysis, The Journal of Development Economics, 41: pp.137-151. Gupta, S., Davoodi, H. and Alonso-Terme (1998) Does Corruption Affect Income Inequality and Poverty?, IMF Working Paper 98/76, International Monetary Fund. Hall, R and Jones, C (1999) Why do some countries produce so much more output per workers than others?, Quarterly Journal of Economics, 114(1), 83-116. Hellman, J. S., Jones, G., Kaufmann, D. (2000) Seize the State, Seize the Day: State Capture, Corruption, and Influence in Transition, World Bank Working Paper 2444, World Bank. Huntington, Samuel P. (1968) Political Order in Changing Societies, New Haven, Yale University Press. Johnson, S. and Kaufmann, D. (2001) Institutions and the Underground Economy, pp.212228, in: O. Havrylyshyn and S. Nsouli (eds.): A Decade of Transition: Achievements and Challenges, Washington D.C: IMF. Johnson, S., Kaufmann, D. and Shleifer, A. (1997) The Unofficial Economy in Transition, Brookings Papers on Economic Activity, 2, pp.159-239, Brookings Institution Johnson, S., Kaufmann, D. and Zoido-Lobaton, P. (1998) Regulatory Discretion and the Unofficial Economy, American Economic Review, Vol.88, N.2, pp.387-92 Johnson, S., Kaufmann, D. and Zoido-Lobaton, P. (1999) Corruption, Public Finances and the Unofficial Economy, Policy Research Working Paper 2169, World Bank Johnson, S., Kaufmann, D., McMillanc, J., Woodruff, C. (2000) Why do firms hide? Bribes and unofficial activity after communism, Journal of Public Economics, vol. 76, 495–520.

24

Kaufmann, D., Kraay, A. and Mastruzzi, M. (2005) Governance Matters IV: Governance Indicators for 1996-2004, World Bank. Leff, N. H. (1964) Economic development through bureaucratic corruption, American Behavioural Scientist 8: 8-14. Loayza, N. V. (1996) The economics of the informal sector: a simple model and some empirical evidence from Latin America, Carnegie-Rochester Conference Series on Public Policy 45: 129-162. Loayza, N., (1994) Labour Regulations and the Informal Sector, Policy Research Working Paper 1335, World Bank. Mauro, P. (1995) Corruption and growth, Quarterly Journal of Economics 110: 681-712. Public Economics 76: 399-457. Rauch, J. (1991) Modelling the Informal Sector Formally, Journal of Development Economics, 35: pp. 33-47. Rose-Ackerman, S. (1999) Corruption and Government: Causes, Consequences and Reform, Cambridge University Press, Cambridge, UK. Schneider, F. and D. H. Enste (2000) Shadow economies: size, causes and consequences, Journal of Economic Literature, XXXVIII, 77-114. Schneider, F. (2005) Shadow Economies of 145 countries all over the world: Estimation results of the period 1999-2003, University of Linz: Department of Economics, Discussion paper Linz, Austria. Shapiro, C. and Stiglitz, J. (1984) Equilibrium unemployment as worker discipline device, American Economic Review, vol. 74, pp. 433-44 Shleifer, A. and Vishny, R.W. (1993) Corruption, Quarterly Journal of Economics, CVIII: 599-618. Tanzi, V. (1998) Corruption around the world-causes, consequences, scope and cures, International Monetary Fund Staff Papers 45: 559-594. Tanzi, V. and Davoodi, H. (1997) Corruption, Public Investment and Growth, IMF Working Paper 97/139 (Washington: International Monetary Fund) Treismann, D. (2000) The causes of corruption: a cross-national study, Journal of Public Economics 76, 399–457. World Bank (1998) Assessing Aid: What Works, What Doesn't and Why?, Oxford University, Press for the World Bank (SLC). World Bank (2005) World Development Indicators, World Bank.

25