BEYOND THE VEIL OF IGNORANCE: THE INFLUENCE OF DIRECT DEMOCRACY ON THE SHADOW ECONOMY

Transcription

1 XXIV CONFERENZA ECONOMIA INFORMALE, EVASIONE FISCALE E CORRUZIONE Pavia, Aule Storiche dell Univerità, ettembre 2012 BEYOND THE VEIL OF IGNORANCE: THE INFLUENCE OF DIRECT DEMOCRACY ON THE SHADOW ECONOMY DÉSIRÉE TEOBALDELLI, FRIEDRICH SCHNEIDER Società italiana di economia pubblica - c/o Dipartimento di cienze politiche e ociali dell Univerità di Pavia

2 Beyond the veil of ignorance: the influence of direct democracy on the hadow economy by Déirée Teobaldelli and Friedrich Schneider Firt Draft: February 2012 Thi Draft: May 2012 Abtract In thi paper we analyze the influence of direct democratic intitution on the ize and development of the hadow economie. The framework developed predict a negative relationhip between the degree of direct democracy and the ize of the hadow economy. Countrie where direct democratic intitution upport democratic life are expected to be characterized by a lower informal ector. Our model ugget that thi effect i likely to be nonlinear and to interact with other feature of the political ytem, uch a the ize of the electoral ditrict. The empirical invetigation of a ample of 57 democracie confirm our theoretical finding. JEL-Code: 017, D78, H11, H26. Key-Word: Shadow economy, direct democratic intitution, ditrict magnitude, good governance. Déirée Teobaldelli, Department of Law, Univerity of Urbino, Via Matteotti 1, Urbino, Italy; Phone-No: Fax-No: ; deiree.teobaldelli@uniurb.it. (correponding author) Friedrich Schneider, Intitute of Economic, Johanne Kepler Univerity of Linz, Altenberger Straße 69, A-4040 Linz, Autria, Phone-No: +43 (0) ; Fax-No: -8209, friedrich.chneider@jku.at, 1

3 1. Introduction Thi article i a firt attempt to analyze the influence of direct democratic intitution on the ize and development of the hadow economy. The baic motivation of our tudy i that undertanding the mapping from intitutional arrangement into policy outcome i an eential precondition for evaluating both economic performance and the attitude of citizen toward the tate and the law. Over the pat decade the analyi of the hadow economy ha attracted increaing interet. Variou tudie conider intitutional characteritic a a key factor in the development of informal ector (e.g. Schneider and Ente, 2000; Friedman et al., 2000; Torgler and Schneider, 2007; Schneider, 2010; Teobaldelli, 2011). In thee tudie the author argue that the inefficient and dicretionary application of tax ytem and regulation by government might play a crucial role in the deciion of operating unofficially, even more important than the burden of taxe and regulation. In particular, corruption of bureaucracy and government official eem to be aociated with larger unofficial activitie, while a good rule of law, by ecuring property right and contract enforceability, increae the benefit of being formal (e.g. Johnon et al., 1997; Johnon et al., 1998a,b). Our analyi i complementary and new to thi field of reearch. In trying to ae which variable play a role in the tate-ociety interaction underlying informality, we pecially paid attention to governance and intitutional quality meaure, like direct democracy and accountability indexe, rather than to traditional variable, a thi approach ha been neglected in previou related tudie. In particular and for the firt time, the aim of thi paper i to analyze theoretically and empirically, on a cro-country bai, the effect of direct democratic intitution on the ize of the hadow economy. In our opinion, the development of the informal ector can alo be conidered a a conequence of the failure of public intitution to upport an efficient market economy, through appropriate public good proviion. Thi can occur when the government i either wateful or corrupt, with great dicretionary power over the allocation of reource. Citizen who feel overburdened by the tate, do not perceive their interet and preference properly repreented in political intitution and loe their trut in the authority. They will chooe the exit option and decide to work in the informal ector a a reaction to inefficient government (Schneider and Ente, 2002; Hirchman, 1970). Direct democratic intitution provide citizen with the voice option over government performance. They have the potential to contrain, both directly and indirectly, the ability of politician to extract rent from 2

4 public pending, and therefore repreent a form of non inulation of politician. Moreover, direct democracy may act a a valid correction mechanim for the low accountability of government. A government which elf-impoe check and allow citizen to be actively involved in the policymaking proce, indicate that it trut it contituency and take into account it preference. A a conequence, citizen who perceive government a benevolent and recognize their interet properly repreented, identify with the tate and are more willing to comply. They become more conciou of the opportunitie available to eventually complain. Direct democracy might then help altering the incentive of behaving illegally. Our theoretical analyi build on the theory of direct democratic intitution a a dicipline device. We lay out a model linking the impact of direct democracy on the hadow economy through the fical policy choice made by elected politician. Direct democratic intitution can contribute poitively to the efficiency of political deciion-making proce in two way. Firt, direct democracy may exert a direct effect on policy, a referendum and initiative can override the deciion of policymaker by removing their dicretion. Second, direct democratic intitution may alo work indirectly, a the imple threat of a ballot propoition may be ufficient to induce elected official to chooe policie more cloe to the preference of the median voter (Matuaka and McCarty, 2001). In our framework, both effect are at work and dicued. The economy conit of individual who can allocate their labor between two ector, the formal and the informal one. Production in the formal ector alo require ome productive public ervice and i perfectly obervable by the tax authoritie. Production in the informal ector relie only on labor and i completely unobervable by the authoritie, which implie that it cannot be taxed. The revenue of the public ector can be ued to provide both productive public ervice and rent to politician. The politician ha to decide the level of taxation, proviion of public ervice and rent under the rik that citizen will promote a referendum to reject the reformed fical policy. The model predict that higher level of direct democracy favor the implementation of fical policie cloer to the preference of citizen; thee policie are more efficient and able to reduce individual incentive to operate in the informal ector. We alo find that direct democracy i likely to exert nonlinear effect on the ize of the hadow economy, i.e., more direct democracy reduce the hadow economy at low or intermediate level, while the effect of uch increae i likely to be limited when direct democratic intitution are already quite good. Our theoretical analyi alo ugget that direct democratic intitution 3

5 are likely to interact with other feature of the political ytem, uch a the ize of the electoral ditrict. In particular, direct democracy i expected to have a higher impact on the reduction of the hadow economy when the electoral ytem i characterized by a larger ditrict magnitude. We then develop an empirical invetigation on a ample of 57 democracie in order to tet our theoretical finding. The analyi implemented demontrate that the effect of direct democratic intitution on the hadow economy i negative, nonlinear and quantitatively important; the reult are robut and alo depend on the interaction of direct democracy with the ize of the electoral ditrict confirming the theoretical prediction. The layout of our paper i a follow. Section 2 preent a hort literature review. Section 3 preent a model that provide explanation of the tranmiion channel trough which direct democracy influence the ize of informal ector. In Section 4 we preent the empirical evidence, firt the data and etimation approach i decribed, and concrete hypothee are formulated and finally the empirical reult are hown. After that in Section 5 a ummary i given and three connection are drawn. 2. A Short Literature Review The exiting literature on the economic effect of direct democratic intitution follow two main trand: a number of empirical tudie, mainly baed either on Switzerland or the United State, evaluated the impact of direct democracy on fical policy and government efficiency (Pommerehne, 1978; Pommerehne and Schneider, 1978; Frey, 1994; Matuaka, 1995, 2005; Feld and Kirchgäner, 2001a,b; Feld and Matuaka, 2003; Blomberg et al., 2004; Blume et al., 2009), yet other tudie have focued on the effect that direct voter participation in political deciion may have on citizen attitude toward intitution, in term of tax morale and civic trut in government (Pommerehne and Weck-Hannemann, 1996; Frey, 1997; Alm et al., 1999; Schneider and Ente, 2002; Feld and Tyran, 2002; Torgler, 2005; Torgler and Schaltegger, 2005). The common theme of thi literature i that democratic participation poibilitie by taxpayer lead public pending to be more efficient and in line with the preference of citizen. In particular, Pommerehne (1978) and Pommerehne and Schneider (1978) demontrate that in Swi citie where citizen participation in public deciion-making i only indirect, the government and public bureaucracy prove to be unreponive to voter preference over public expenditure, though thee are taken more and more into conideration a 4

6 election approach. Specifically Frey (1994) dicue the propertie of direct democratic intitution and argue that popular referendum are effective mechanim by which the voter can control the policy choice of tate government and expre collective preference. He highlight that in 39 percent of the referendum that took place in Switzerland during the period from 1848 until 1990, the majority will wa different from the etablihed will of the Parliament. Feld and Kirchgäner (2001a,b), taking into account Swi municipalitie, how that mandatory budget referendum are aociated with both lower total expenditure per capita and reduced per capita debt. Feld and Matuaka (2003) etimate regreion for Swi canton uing panel data from 1990 to 1998 and find evidence that mandatory referendum reduce government pending by 19 percent. With regard to the United State, Matuaka (1995; 2005) analyze the impact of initiative on fical policy and find that tate initiative improve reource allocation. In particular, he concentrate on the effect of initiative over a 30 year period and oberve that initiative are employed to contrain tax burden a well a to lower the overall tate and local government pending. Blomberg et al. (2004) obtain that over the year , among 48 US tate, initiative tate were between 20 to 30 percent more effective in providing public capital than non-initiative tate, reaching a better economic outcome in term of higher GDP growth. Blume et al. (2009) try to evaluate the economic effect of direct democracy on a cro-country bai. Conidering a ample of 88 countrie, they find that the preence of direct democracy intitution i correlated with lower government expenditure, epecially in countrie with weak democracie. Although extenive literature on different apect regarding the hadow economy exit, only very few tudie focu on the implication of direct democratic intitution for tax evaion. It i argued that higher participation right can raie direct political control and boot tax morale: the potential to participate in collective deciion-making help to improve the legitimacy of the political ytem before citizen and lower their inclination to cheat on taxe. In particular, Pommerehne and Weck-Hannemann (1996) find noncompliance to be negatively correlated with direct control of citizen/taxpayer over government budget for Swi canton. Frey (1997) argue that direct citizen participation, via popular referendum and initiative, can enhance civic virtue and increae the intrinic motivation to pay taxe. Schneider and Ente (2002) tate that proper element of direct democracy, together with fical federalim, might trengthen public trut in political intitution and foter tax morality. Torgler (2005) uing urvey data for Switzerland find that direct democratic right have a ignificant poitive effect on tax morale. 5

7 Our paper i an attempt to bring together thee two field of reearch and for the firt time introduce an innovation by providing a poible theoretical explanation of the tranmiion channel trough which direct democratic intitution influence the ize of informal ector, and by teting the theoretical implication highlighted on a cro-country bai. In the next ection we develop a model that predict that higher level of direct democracy favor the implementation of fical policie more efficient, cloer to the preference of citizen and able to reduce individual incentive to operate in the informal ector. 3. The Model We conider an economy of a continuum of individual of meaure 1. There i a unique final good that can be produced by two ector, the formal and the informal one. Each agent i i a conumer producer who upplie inelatically 1 unit of labor, which he can allocate between the formal ector ( one ( l i, ) o that (1) l l 1. i, f i, l i, f ) and the informal The production function in the formal ector i Cobb Dougla with contant return to cale in labor and in the quantity of per capita public ervice g, 1 and it i defined a (2) y 1 i, f li, f g, where 0 1. Production in the informal ector doe not require the input provided by the public ector, o the hadow economy production function i given by (3) y i ali,,, where 0 1 and a>0. Each agent chooe the optimal allocation of labor between the two ector and conume all income produced net of taxe. Income in the formal ector i perfectly obervable by the tax authoritie and can be taxed at a contant rate t [0,1 ]. In contrat, production in the informal ector i completely unobervable, which implie that it cannot be taxed by the public authoritie. Therefore, from (1), (2) and (3) follow that 1 It i clear from the production function that the productive input provided by the public ector i eential for production and that there i congetion a what matter for production i the per-capita level of public ervice that are aumed to be not excludable in the formal ector. Thi formulation i alo ued in Teobaldelli (2011). 6

8 the dipoable income of agent i i (4) y 1 i, d ( 1t)(1 li ) g ali, where l i denote the amount worked by agent i in the unofficial economy. The revenue of the public ector are equal to (5) 1 R ty i0 i, f 1 di t i0 1 1 ( 1 li ) g di t(1 l) g, where we have ued the fact that l i l for all i ince all agent are identical and face the ame fical policy. The revenue can be ued to provide productive public ervice and/or to provide rent to politician. Let G denote the total proviion of public ervice and [0,1 ] the fraction of revenue R pent for public ervice, 2 o that 1 denote the fraction of revenue ued for politician rent. Thu, the government budget contraint i G R. If we take into account the unitary ma of the population and the expreion for R in (5), the government budget contraint become 1 g t( 1l) g and it can be rewritten a 1/ 1/ (6) g t (1 l). We hall refer to t and a the fical policy variable to be choen by the politician, given that g i uniquely determined by the government budget contraint (6). The total utility of the politician i (7) u ( 1 ) R p B, R where the firt component i the current monetary rent, B>0 are the (exogenou) future benefit from being in office and p [0,1 ] i the probability of being reelected. We aume that thi probability may be affected R by the fact that a referendum ha taken place and by ome other intitutional characteritic. Hence, we aume that { p, p}, with (0,1 ), where p i the probability of reelection and 1 repreent the p R reduction in the probability of reelection of the politician when a referendum ha taken place (ee below for more detail on the characteritic of ). We conider a one period economy with an incumbent politician. Nature chooe randomly the tatu quo fical policy t, ) at the beginning of the period and the politician decide whether changing it or not; ( Politician rent are modeled here a a cah tranfer, but one could alo aume the exitence of a nonproductive public good that give utility to politician only. 7

9 we denote with t, ) the reformed policy. 3 After the policy ha been choen, citizen may mobilize and ( c c promote a referendum againt the reform. 4 If thi take place, then citizen vote and decide whether the reform can be implemented or not. If the reform i rejected, then the tatu quo policy i implemented. Finally, election take place on the reappointment of the politician. We aume that promoting a referendum i cotly for the citizen, and that thi cot i equal for all citizen and depend on the level of direct democracy, i.e. i ( ) for all i, where [, ] i an index of the level of direct democracy. We alo aume that ( ) 0 which capture the idea that at higher level of direct democracy correpond a lower cot of mobilization to organize a referendum. Individual are aumed to enjoy a poitive nonmonetary benefit i from promoting a referendum if the tatu quo policy provide him an income higher than the reformed policy, i.e. if y t, ) y ( t, ). Thi payoff differ d ( 0 0 d c c acro agent and i drawn from a continuou and differentiable cumulative ditribution function F ( ), with F ( ) f ( ). A each individual ha zero ma, he will not conider the monetary payoff in deciding whether to mobilize or not for the referendum, o that thi will be promoted only by the agent with poitive net payoff, i.e. with i 0. We aume that the probability that a referendum will take place (when thi i convenient for the citizen) i linearly increaing in the number of individual that mobilize and, therefore, it i equal to (8) P( ) 1 F( ( )), which i increaing in the level of direct democracy ince ( ) 0. We alo want to analyze how ome intitutional feature interact with direct democracy and affect the level of the hadow economy. In particular, we here conider the effect of ditrict magnitude whoe importance for political rent extraction ha been emphaized by variou author (ee Peron and Tabellini, 3 In a dynamic etting, the tatu quo policy could be interpreted a the policy implemented in the previou period. 4 Direct democracy identifie a variety of political procee that aign to ordinary citizen the right to directly decide on certain political iue through popular vote. All form of direct democracy deal with the deciion of citizen on ubtantive law lited on the ballot, called ballot meaure or propoition. Ballot meaure can concern either the propoal of a new law or the abolihing of an old law. There are difference on how propoition come to the ballot. We can ditinguih between initiative, that allow the citizen to propoe a new law pertaining to different level of legilation (contitutional veru ordinary legilation) and related to a variety of cope; and referendum, that i a vote on a law already approved by the legilature. Both thee form of direct democracy allow the citizen to control the agenda and typically require a predetermined number of ignature from eligible voter to qualify for the ballot (Matuaka, 2005). In order to implify the analyi, we only refer to referendum a a mean that enable citizen to directly intervene in the political proce. 8

10 2000, and Peron et al., 2003). A higher ize of the electoral ditrict may in fact act a a barrier to entry, reduce electoral competition and the poibility of the voter to punih the politician. 5 To capture thi fact, we aume that the reduction in the probability of reelection of the politician due to a referendum, decreaing in the ize m of the ditrict, i.e., (m), with d dm 0. 1, i We now move to characterize the equilibrium of the model and determine the effect of direct democracy on the equilibrium policy and the ize of the hadow economy. 3.2 The Equilibrium We firt determine the individual optimal allocation of labor among ector for any given policy et by the politician. The reult i contained in the following lemma. Lemma 1. If the government taxe the income produced in the formal ector at rate t (0,1 ) and ue the fraction (0,1 ) of total revenue for the proviion of public ervice, then the amount of labor employed in the informal ector by each individual i 1 a (9) l ( t, ). (1 ) / (1 ) / (1 ) t t Moreover, for any given policy ( t, ), we obtain that l t (1 ) (10) l, t t(1 t)(1 ) and l 1 (11) l, (1 ) 1 with l l( t, ) defined by (9). Thi mean that l / 0 and that l / t 0 for t 1, and that l 1 for t 0, t 1 or 0. Therefore, the fical policy maximizing the citizen dipoable income, and therefore ocial welfare, i t 1 and 1. 5 Peron and Tabellini (2000) alo explain the link between ditrict magnitude and the electoral formula with larger ditrict aociated to more proportional electoral ytem and maller ditrict to majoritarian ytem. Majoritarian election reduce the politician rent becaue voter in marginal ditrict are more mobile and electoral competition i tiffer, which implie that citizen can punih politician more everely for wateful pending. 9

11 Proof. Each agent chooe the labor hare between the two ector maximizing hi dipoable income (12) 1 max yi, d (1 t)(1 li ) g ali. l i From the firt order condition of problem (12) with repect to l i, uing the government budget contraint (6) and the fact that all individual are identical (i.e., l i l for all i), we obtain that the optimal amount of labor employed in the informal ector by each agent i the one reported in (9). The expreion (10) and (11) are obtained from the differentiation of (9) and rearranging term. Expreion (9) make it clear that the amount worked in the informal ector i monotonically decreaing in and, therefore, take the minimum value at 1. Other thing equal, the higher i the fraction of revenue ued for the proviion of productive public ervice, the higher i the marginal productivity in the formal ector (relative to the informal one), and the lower will be the labor upply in the hadow economy. 0 mean that public ervice are not provided and that there cannot be production in the formal ector, o that l 1, and the ame i true when t 0. All labor i employed in the hadow economy if t 1 becaue all production in the formal ector accrue to the government revenue. The relationhip between l and t, for a given, i intead nonmonotonic: l i decreaing in t when t 1, it i at it minimum at t 1, and then it become increaing in t for t 1. The intuition for thi reult ha to do with the fact that the allocation of labor depend on it relative net marginal productivity acro the two ector, and that an increae in taxation generate two oppoite effect affecting the net marginal productivity of labor in the formal ector (while leaving unaffected the marginal productivity in the hadow economy). On the one hand, higher taxation ha a direct effect on the reduction of the net marginal productivity of labor in the formal ector where taxe cannot be avoided. On the other hand, more taxation mean more proviion of productive public ervice and, therefore, a higher marginal productivity of labor in the formal ector. When taxation i relatively low ( t 1 ), the increae in the proviion of public ervice due to an increae of the tax rate generate an increae in the marginal productivity of labor in the formal ector that more than compenate the increae of taxation. Thi mean that the net marginal productivity of labor in the formal ector increae, and the upply of labor in the informal ector goe down. The oppoite happen for relatively high level of taxation (i.e., for t 1 ). Thee conideration alo ugget the exitence of a 10

12 monotonic relationhip between the net dipoable income and the level of hadow economy, which i tated in the following corollary. Corollary 1. There exit a monotonic negative relationhip between the maximized dipoable income of the agent ŷ d and the labor employed in the informal ector, i.e., dyˆ d dl 0. Proof. From the optimal level of labor employed in the informal ector (9) follow that ( t ) (1 ) / (1 t) ( a ) l (1 ). Subtituting thi expreion in (12), and taking into account the government budget contraint (6), we obtain that the maximized dipoable income of each individual can be rewritten a (1 ) yˆ d ( a ) l (1 l) al. Taking the derivative of thi expreion with repect to l lead to (1 ) dyˆ d dl al 1 (1/ ) (1 )(1 l) l 0. Thi reult i important becaue it etablihe that any policy increaing the maximized dipoable income of the individual involve an unambiguou negative effect on the labor employed in the hadow economy. The following reult will alo be ueful a it etablihe the link between the individual hare of labor employed in the informal ector and the aggregate ize of the hadow economy. Corollary 2. The ize of the hadow economy i monotonically poitively related to the amount of labor employed in the informal ector by the repreentative agent. Proof. Taking into account the unitary meaure of the population, the ize of the hadow economy i given by al yˆ d. Since ŷ d i decreaing in l (ee Corollary 1), it i traightforward that thi ratio will be increaing in l. We now determine the fical policy choen by a politician maximizing hi own utility without the contraint that a referendum can take place. Thi problem involve the maximization of the utility in (7), ubject to the government budget contraint (6) and the individual reaction function (9). Subtituting (6) into (7), the politician maximization problem can be rewritten a (1 ) / 1/ (13) max u (1 ) t (1 l) p B, t, R 11

13 with p R p. The olution to problem (13) i ummarized in the following lemma. Lemma 2. The uncontrained optimal fical policy for the politician i etting taxation and the fraction of revenue pent for the proviion of the public ervice at level defined by the following equation 1 l l (14) t, 1 l (1 )(1 l ) (15), 1 l l where l l( t, ) i given by (9). Moreover, t t 1 and 1. Thi implie that the amount of per capita public ervice provided i le than optimal, g g, and the amount of labor allocated by each agent to informal activitie i higher than the level choen under the efficient fical policy, Proof. See Appendix A. l l. Lemma 2 tate that the uncontrained optimal policy of the politician implie a taxation and a proviion of public ervice that are repectively higher and lower than the optimal one. Thi in turn tranlate into a higher ize of the hadow economy. Given thee preliminary reult, we can now move to characterize the equilibrium of the model by determining the optimal policy reform of the politician when he take into account the poibility that referendum can be held. In particular, the politician ha two poible trategie. One i to elect the policy that maximize hi payoff bearing the rik of a referendum and the implementation of the tatu quo policy. We call thi the nonprevention trategy. The other poibility for the politician i chooing a policy reform that give no incentive to the citizen to promote a referendum. We refer to thi a the prevention trategy. Before proceeding in the analyi we need to point out that it i never optimal for the citizen to promote a referendum when the reformed policy t, ) i uch that the maximized dipoable income of the ( c c individual under thi policy i higher than the one correpondent to the tatu quo policy, i.e., if (1 ) / (1 ) / (16) yˆ ( t, ) ( t ) (1 t)(1 l ) al ( t ) (1 t )(1 l ) al yˆ ( t, ), d c c c c c d c c 12

14 where l l t, ) and l l, ) a defined in (9). 6 If the tatu quo policy i uch that contraint (16) i 0 ( 0 0 c ( t c c atified alo under the uncontrained optimal policy ( t, ) defined in Lemma 2, then the politician alway implement uch policy and citizen will never find optimal to promote a referendum. A the level of direct democracy i irrelevant in thi cae, we will retrict our attention to the ituation where contraint (16) i not atified under the policy contained in Lemma 2. The following aumption tate thi point. Aumption 1. The tatu quo policy t, ) i uch that contraint (16) i alway violated when the reformed policy i ( t, ) ( t, ). 7 c c ( 0 0 Let u tart with the analyi of the prevention trategy. In thi cae, the politician chooe the policy that maximize hi utility ubject to the contraint that citizen will have no incentive to promote a referendum, i.e. that the policy choen atifie contraint (16). The following lemma characterize the optimal reformed policy when the politician follow thi trategy. Lemma 3. The optimal policy reform t, ) for the politician under prevention olve problem (13) ubject ( c c to (16). The policy i t, ) with t [1, t ] and [,1], where at leat one of the following two ( P P P P inequalitie hold: tp t, P. The fical policy ( P, P) t i uch that contraint (16) will alway hold with the equality ign, i.e., y ( t, ) yˆ ( t, ), and it i independent on the level of direct democracy. ˆd 0 0 d P P Proof. A contraint (16) i binding at ( t, ), the politician need to implement a policy that provide a higher utility to the citizen. t t i never optimal becaue the utilitie of the politician and individual are both decreaing in t, and t t 1 i alo not optimal becaue both utilitie are increaing in t; therefore, t P [1, t ] where u and y d are repectively increaing and decreaing in t. Similarly, [ P,1] come from the fact that u and y d are both increaing in for all. tp t and/or P follow from 6 Note that in (16), we have ubtituted the government budget contraint (6) and ued the fact that all individual are identical a in the proof of Lemma 1. 7 Thi aumption implifie the preentation but ha no effect on the generality of the reult. 13

15 contraint (16) being binding under the uncontrained policy t, ) (ee Aumption 1). The reform ( t, ) i uch that contraint (16) hold with the equality ign, i.e. y ( t, ) yˆ ( t, ), becaue ( P P ˆd 0 0 d P P otherwie the politician could alway increae hi utility by increaing t or reducing without violating the contraint. The lat part of the lemma i immediate ince doe not appear in any part of the problem. Lemma 3 tate that the policy reform when the politician want to avoid a referendum take intermediate value between the optimal uncontrained policy of the politician, ( t, ), and the ocial optimal policy ( t,1). It i immediate that the higher the utility of the agent from the tatu quo policy ( t 0, 0), and the cloer the reformed policy will be to the optimal one. The level of direct democracy doe not have any effect on the reformed policy ince in thi cae the policy choen ha to be uch that it i never optimal for the citizen to promote a referendum. The politician ha alo the option to chooe hi preferred fical policy and face the rik of a referendum (nonprevention trategy). In thi cae, the politician maximize the following expected utility (17) max Eu( t,, ) P( ) u( t0, 0, ) [1 P( )] u( t, ), t, ubject the government budget contraint (6) and the individual reaction function (9), and where P ( ) i defined by (8). The utility of the politician u () i till given by the expreion in (13), but note that u t 0, ) ha p R ( 0 p. In word, the expected utility of the politician i now affected by the poibility that a referendum will take place, which can happen with probability P ( ). In thi cae, the tatu quo policy i implemented and the probability of the politician of being reelected i reduced by a factor The following lemma characterize the optimal reform under the nonprevention trategy. 1. Lemma 4. The optimal policy reform for the politician under nonprevention, t, ), correpond to the 14 ( N N uncontrained optimal policy ( t, ) defined in Lemma 2. A y ( t, ) yˆ ( t, ), contraint (16) i ˆd 0 0 alway violated and a referendum take place with probability P ( ). The policy implemented i ( t, ) with probability 1 P( ), and the tatu quo policy i with the complementary probability. The maximized expected utility of the politician in (17) i decreaing in the level of direct democracy. d

16 Proof. The proof i traightforward from the firt order condition of (17) and Lemma 2. The lat part of the lemma, i.e. the fact that u( t,, ) ( P( ) )[ u( t, ) u( t,, )] 0, follow from P ( ) 0 and u t, ) u( t, ). ( 0 0 E 0 0 Under the trategy of nonprevention, the politician implement hi preferred policy and take the rik that a referendum i promoted and the tatu quo policy implemented. In thi cae, higher level of direct democracy improve the quality of the fical policy a the maximized dipoable income under the tatu quo policy i higher than the one under the reformed policy. The following propoition characterize the optimal behavior of the politician. Propoition 1. There exit a level of direct democracy,, uch that the politician chooe the trategy of nonprevention decribed in Lemma 4 for all, and the trategy of prevention reported in Lemma 3 for all. Proof. The expected utility under nonprevention i decreaing in, while the utility from prevention i independent on and it i alway trictly lower than the utility from the uncontrained policy ( t, ). Therefore, there will exit a threhold level of direct democracy [, ] uch that Eu( t N,, ) u( t, ) for all N P P, and vice vera. Moreover, if the lowet level of direct democracy implie a ufficiently low probability P ( ) that a referendum will be held (due to the high cot ( ) of promoting it), then Eu( t,, ) u( t, ) and. It cannot be excluded that the nonprevention N N P P trategy alway dominate the prevention one, i.e. Eu( t,, ) u( t, ) for all [, ]. Thi happen when P ( ) i ufficiently low and we denote. N N P P The intuition for thi reult i that when direct democracy i low, it i optimal for the politician to take the rik that a referendum i held a thi allow him to implement hi preferred policy relatively often. When intead direct democratic intitution are good, the probability that the policy reform i cancelled i relatively high, and it become optimal for the politician to implement policy that will not be changed by the citizen. 15

17 We can now etablih the effect of direct democracy on the level of hadow economy. Propoition 2. The preence of direct democracy ha a global negative effect on the ize of the hadow economy and thi effect i nonlinear. When the degree of direct democracy i relatively low ( ), higher level of direct democracy reduce the ize of the informal ector. An increae in direct democracy when thi i relatively high ( ) ha no effect on the hadow economy. Proof. From Corollarie 1 and 2 follow a negative relationhip between the maximized dipoable income of the repreentative agent and the ize of the hadow economy. Under prevention ( ), higher level of do not change the reformed policy and the maximized dipoable income of the agent; therefore, the ize of the hadow economy i independent on and i related to the level of income of the tatu quo policy, i.e. l yˆ d ( t 0, )), ince y ( t, ) yˆ ( t, ) (ee Lemma 3). Under nonprevention ( 0 ˆd 0 0 d P P ( ), the policy implemented i the tatu quo t, ) with probability P ( ) and the reformed policy ( t N ( 0 0, ) ( t, ) with the complementary probability. Therefore, the expected ize of the hadow economy N i l( t,, ) P( ) l( yˆ ( t, )) [1 P( )] l( yˆ ( t, )). From y ( t, ) yˆ ( t, ) and Corollary 1 E N N d 0 0 d ˆd 0 0 follow that l yˆ ( t, )) l( yˆ ( t, )). Since dp ( ) d 0, the expected ize of the hadow economy will ( d 0 0 d be decreaing in, i.e. del( tn, N, ) d 0. It i alo immediate that El( tn, N, ) l(( tp, P)) independently on the value of. d The reult in Propoition 2 i that an improvement of direct democracy when thi i at low or intermediate level reduce the ize of the hadow economy. When direct democratic intitution are well developed, the ize of the informal ector i lower, but further increae in the quality of thi intitution are likely to have no effect on the hadow economy. The following propoition tate how direct democracy interact with ditrict magnitude in affecting the equilibrium level of the hadow economy. Propoition 3. The politician i more likely to follow the nonprevention trategy when the electoral ytem i characterized by a larger ditrict magnitude, i.e. m 0. Thi in turn implie that the effect of direct 16

18 democracy on the ize of the informal ector i higher when ditrict magnitude i larger. Proof. The threhold (18) Eu( t,,, ) u( t, ) 0, N N i implicitly defined (when it i interior) by the following equation P P where the firt term in (18) i given by the expreion in (17) and the econd term by (13). Applying the implicit function theorem to (18), we obtain that m Eu( t Eu( t N N, N,, ),,, ) N, m where Eu (,,, ) 0 from Lemma 4. Hence, from t N N Eu( tn, N,, ) u( t P( ) m N, N,, ) P( ) pb 0 m m follow that m 0. The econd part of the lemma come from the fact that higher level of direct democracy have a greater impact under the nonprevention trategy, i.e. when (ee Propoition 2). The intuition for the reult in Propoition 3 i the following. When the electoral ytem i characterized by larger ditrict, political competition i lower and it i more difficult for voter to punih the politician for bad fical policie. Thi make the trategy of nonprevention more convenient and, under thee condition, direct democratic intitution play a bigger role in correcting bad policie. To ummarize, the framework developed here ha three main prediction that will be econometrically teted. Firt, higher degree of direct democracy reduce the ize of the hadow economy, ceteri paribu. Thi i becaue direct democracy improve the fical policy implemented by making it more efficient and cloer to the preference of citizen. Thi in turn increae the net marginal productivity of labor in the formal ector and reduce the incentive of individual to operate in the informal one. Second, the model predict the exitence of nonlinear effect of direct democracy on the ize of the hadow economy. In particular, an increae in the level of direct democracy i likely to reduce ubtantially the ize of the informal ector when direct democracy i at low or intermediate level, while the effect of uch increae i more limited when direct democratic intitution are already well developed. Third, direct democratic intitution are likely to interact with other feature of the political ytem; we find that direct democracy ha a tronger effect on the reduction of hadow economy when the ize of the electoral ditrict i bigger. 17

19 4. Empirical Evidence 4.1 Data Decription and Etimation Approach We develop a cro-ectional analyi on a ample of 85 countrie (from Peron and Tabellini, 2003) that can be conidered democracie for the period We conider only democratic countrie ince direct-democratic intitution are likely to work only in relatively table democracie. Referendum and popular initiative, even if oberved, are expected not to produce any effect in authoritarian regime a they are likely to be heavily manipulated by the government. In order to reduce the poibility of omitted variable, we run everal regreion controlling for a wide range of variable which, according to the exiting literature, might have an impact on the ize of the hadow economy and alo be correlated with the country democratic intitutional organization. We avoid decribing all variable in detail here; the definition and ource are reported in the Data Appendix. 8 Data concerning the informal ector are drawn from the dataet of Schneider (2005), which provide the ize of the hadow economy a a proportion of official GDP for 145 countrie over the period Our dependent variable i an average of the three available obervation for the period indicated; thi allow u to cover up to 73 of the countrie conidered in the original ample. The index of direct democracy (DDI) come from Fiorino and Ricciuti (2007). They derive it from three different ource: Kaufmann (2004) for 43 European countrie, Hwang (2005) for 33 Aian countrie, and Madroñal (2005) for 17 Latin American countrie. Hence, we are left with 57 countrie for which we have both the ize of the hadow economy and the index of direct democracy. The DDI lie within a range of value from 1 to 7, with 7 being aociated to the countrie rated a radical democrat, and 1 to the countrie with the lowet level of direct democracy. 9 A pointed out by Fiorino and Ricciuti, the main advantage of uing thi index i that it provide both a qualitative and a quantitative aement of direct democracy. The index, in fact, reflect at the ame time the quality of democracy and it performance, a it focue on the two mot important and widely ued procee, initiative and referendum, a well a on the quality of the 8 The variable that meaure governance and intitutional quality come from different dataet (Kaufmann et al., 2005; Treiman, 2008; Blume et al., 2009). 9 The only country ranked 7 i Switzerland, while the larget group of countrie (twenty-five) take the value 1. 18

20 procee themelve. 10 However, the DDI ha ome diadvantage that have to be dicued. Firt, it link together the legal poibility of having referendum and initiative and the actual choice of uing them. Second, it doe not allow to identify the iue tackled by referendum and initiative. Finally, the index i a ubjective meaure of direct democracy, o the criteria followed to contruct it remain omehow vague. In teting the hypothei that a higher degree of citizen direct political participation lower ceteri paribu, the ize of hadow economie, we ue a baeline pecification that include the following control variable. We take into account the age of democracy a well a the quality of democracy (proxied by the PolityIV index), a both the quality and the conolidation of democracy go hand in hand with the quality of government intitution and influence the incentive of citizen to operate in the informal ector. The baeline regreion alo contain a meaure of ditrict magnitude that, a explained in the previou ection, i likely to affect the government accountability to the contituent (ee alo Blume et al., 2009). For the ame reaon, we add a meaure of the country ize and a variable for the federal tructure. In our opinion, larger countrie may encounter greater difficultie in controlling economic activity and thi could lead to an increae in the ize of hadow economy; while federal countrie appear to better encounter the preference of citizen and ecure a higher quality of public pending, being correlated with a lower ize of hadow economy (Torgler et al., 2010; Teobaldelli, 2011). We alo ue an index of ethnic fractionalization becaue a large literature indicate that ethnic heterogeneity i a determinant of economic performance both in term of output and quality of intitution (La Porta et al., 1999; Aleina et al., 2003). We then include the log of GDP in 1960 to take into account the level of economic development. 11 We employ alo the central government expenditure, including ocial ecurity, a a percentage of GDP to proxy for the ize of government, and the conolidated central government expenditure on ocial ervice and welfare a percentage of GDP a a control for the compoition of government expenditure. We include in the control et the burden of regulation that capture the intenity of regulation in the economic ytem and reflect the ability of government to implement 10 In other word, it allow taking into account even the quality of procedure underlying the actual ue of popular initiative and referendum to propoe, approve, amend and delete law. In order to clarify thi point, Fiorino and Ricciuti mention the cae of Belaru a an example: depite 9 referendum had been held in thi country from 1995 to 2004, Belaru i characterized by the lowet poible core in the range, a referendum were propoed and ued by Preident Lukahenko in order to amplify it power at the expene of the legilature and a poitive outcome wa ecured thank to arret of political adverarie and preure on citizen. However Belaru i not included in our ample. 11 The reaon for uing the GDP of 1960 i to avoid poible endogeneity problem with repect to the dependent variable. 19

21 market-friendly policie promoting private ector development. We alo control for the percentage of the population profeing the Protetant religion, ince religiou belief might affect people attitude toward the economic ytem, private property and tax compliance in particular (La Porta et al., 1997; Lande, 1998). Moreover, we add to the baeline pecification a number of other control variable. We conider the characteritic of the political ytem, by including a variable for the electoral ytem (proportional or majoritarian electoral rule) and a variable for the form of government (preidential or parliamentary regime). We alo conider a meaure of the operational (de facto) independence of the Chief Executive. The ue of thee variable might capture the extent to which political leader are inulated from citizen and can exert their dicretionary power at expene of voter welfare (Peron and Tabellini, 2003). In addition, we control for a meaure of labor market regulation and the level of education a thee may affect the incentive of the individual to operate in the informal ector. The latter i alo important in promoting civic participation and cooperation with other, o facilitating the upport for more democratic regime (Lipet, 1959; Glaeer et al. 2006). We include the demographic characteritic of population, given repectively by the percentage of the population between age 15 and 64, and the percentage of the population aged 65 and older, that can influence the total amount and the compoition of public expenditure and have an impact on fical policie. The degree of openne to international trade i alo taken into account ince the literature on hadow economy ugget that globalization of market and increaing competitivene of third world economie, which exhibit lower production cot, can affect firm deciion to operate in the informal ector (Gerxhani, 2004). We control for legal, hitorical, and geographical characteritic by including variable for legal origin (common law, French civil law, German civil law, Scandinavian civil law, and ocialit law), for colonial hitory (Britih, Spanih-Portuguee, or other colonial origin), and for geographical localization (Africa, Eat Aia, Latin America, Central America, or the Caribbean), that may be correlated with the efficiency of the government, the quality of public good, and the ize of government and political freedom. We ue an index of religiou fractionalization and income inequality a a further control for the heterogeneity in the ociety. We add a control et for the religiou affiliation (the percentage of the population belonging to the Roman Catholic religion in 1980, and an index on Confucian, Buddhit or Zen 20

22 religiou tradition) a a proxy for the dimenion of culture given that many tudie have treed the role of religion belief in haping individual attitude like ethic, trut, tolerance, and compliance (La Porta et al., 1997; Lande, 1998). An index for the civil libertie and political right i conidered a thee feature can be related to the working of direct democratic intitution. Finally, we control for the efficiency and the quality of public intitution by including an index of protection of property right, an index of government effectivene, that take into account the perception of the quality of public ervice proviion, an index of government anti-diverion policie, an index for the rule of law, and an index of corruption of government official. 4.2 Empirical Reult and their Interpretation Table 4.1 provide ome decriptive tatitic for the main variable employed. The mean ize of the hadow economy in the ample i about 31 percent of GDP and the average value of the direct democracy index i Table 4.2 report the correlation among ome variable and reveal that all meaure of quality of democracy ued are negatively correlated with the hadow economy. In particular, we oberve a high negative correlation ( 0.57) between direct democratic intitution and the hadow economy, a predicted by the model. Moreover, there i a poitive correlation among all meaure of democracy employed; the index of direct democracy i correlated (0.36) with the age of democracy and i highly correlated (0.60) with the quality of democracy. Table 4.3 report the average value of the direct democracy index for each quartile of the ditribution of the ize of the hadow economy. The average direct democracy index i 4.92 in the firt quartile of the ditribution and decreae continuouly until 2.07 in the lat quartile. Our empirical trategy i baed on two alternative pecification. The firt one take the following form: SE DDI DM DDI DM ) ' Z i ( 1 i 2 i 3 i i 4 i, where SE tand for the ize of the hadow economy, DDI i the index of direct democracy, DM repreent the ditrict magnitude, the vector Z i compoed of variou control variable and i i an error term. The etimation alo include the interaction term between the direct democracy index and the ditrict magnitude, DDIi DM i. The marginal effect of direct democracy, DDI, on the ize of the hadow economy i therefore given by SE DM ) DDI. The conjecture motivating the incluion of thi variable come (

23 from the reult of our model that the impact of direct democratic intitution on the ize of the hadow economy i bigger in countrie characterized by a larger ditrict magnitude. 12 It i therefore intereting to undertand if thee two political intitution interact each other and whether they are complement or ubtitute. Table 4.4 report the etimate for thi pecification. Column (1) of Table 4.4 how that an increae of one point in the direct democracy index reduce, on average, the hare of the informal ector by 4.63 percentage point. The etimate of our baeline pecification where we control for a number of variable i reported in Column (2). We obtain an etimated coefficient of direct democracy equal to 4.82 and an etimated coefficient of the interaction term of 4.4; the joint ignificance i at 1% level. Thi implie that the marginal effect of direct democracy i 4.82 for the larget poible ditrict magnitude (DM=0, i.e., of a ingle national ditrict) and 0.42 when the ditrict magnitude i the lowet poible (DM=1). In other word, the effect of direct democracy on the hadow economy increae with ditrict magnitude a predicted by our model. A the average value of ditrict magnitude in the ample i 0.367, the effect of direct democracy evaluated at the mean i Thi mean, for example, that an increae in the direct democracy index by 4 point, correponding to a hift from the firt to the third quartile of the ditribution, reduce the hadow economy by 13 percentage point, i.e., more than forty percent of the average ize of the informal ector in the ample. A the average ize of the direct democracy index i 3.46, the marginal effect at the mean of ditrict magnitude i equal to 6.2, which confirm the theoretical prediction that a lower ditrict magnitude i aociated with a maller hadow economy. Next we modify the baeline pecification by adding one control variable (or one group of uch variable) at a time to generate ixteen further pecification. Column (3) and (4) of Table 4.4 report the etimate of pecification where proxie for the political ytem (electoral ytem and form of government) and the inulation of policy maker are added to the baeline pecification; thee etimate confirm the importance of the direct democracy in explaining the ize of the informal ector a the ize of the direct democracy coefficient and interaction term are baically unchanged with repect to the baeline pecification and their tatitically ignificance i alway at tandard level. Including meaure of labor regulation, education, demography, and openne to international trade (Column (5), (6), (7), and (8), repectively) it 12 A dicued in the Peron and Tabellini book (ee Chapter 8 for detail), ditrict magnitude i highly correlated with the electoral ytem (thi correlation i 0.84 in our ample); the electoral ditrict are large in proportional ytem and mall in majoritarian one. 22