YTICAROV AND HTWORG pili h.PRenaL and noraA llenroT noissucsiD repaP rebmuN7081 rebmevoN7991 DRAVRAH ETUTITSNI FO CIMONOCEHCRAESER DRAVRAHYTISREVINU ,EGDIRBMACSTTESUHCASSAM Voracity and Growth Philip R. Lane and Aaron Tornell* We analyze an economy that lacks a strong legal-political institutional infrastructure and is populated by multiple powe*rful groups. Powerful groups dynamically interact via a ,$scal process that effectively allows open access to the aggregate capital stock. In equzl~&ium, this leads to slow economic growth and a “‘voracity e,@ect,” by which a shock, such as a terms o,f trade wind- ,fall, perversely generates a more than proportionate increase in fiscal redistribution and reduces growth. We also show that a dilution in the concentration o,f power leads to ,faster growth and a less procyclical response to shocks. (JEL F43, 010, 023, 040) Two common characteristics of developing countries that have grown slowly in the last several decades are the absence of strong legal and political institutions and the presence of multiple powerful groups in society. In this paper, we analyze a dynamic model of the economic growth process that contains these features. We employ the model to ask three questions. First, why does the combination of a weak institutional structure and fractionalization inside the governing elite generate slow growth? Second, what is the relationship between the concentration of power (the number of powerful groups) and growth? Third, why do such countries not only grow slowly but also frequently respond in a perverse fashion to favorable shocks, by increasing more than proportionally fiscal redistribution and investing in inefficient capital projects? The importance of weak institutions and fractionalization in explaining poor growth per- formance has been highlighted in the empirical literature. Furthermore, country studies have recently emphasized these features in explaining procyclical fiscal policies and a decline in the quality of investment in response to favourable shocks. However, a theoretical analysis that explicitly jointly links these perverse responses and the prevalence of low growth to the primitive characteristics of an economy has been lacking.’ This provides the motivation for our paper. We focus on the fiscal process as an important arena in which powerful groups interact in a society with a weak legal-political infrastructure and emphasize discretionary fiscal redistribu- tion as a key mechanism by which such groups appropriate national resources for themselves. Examples of powerful groups are provincial governments that extract transfers from the center, strong unions and industrial conglomerates that seek protection, and patronage networks that obtain kickbacks from public works.’ We consider a two-sector economy. The formal sector employs the efficient production tech- nology but is subject to taxation; the shadow sector enjoys a less productive technology but is non-taxable. For instance, the shadow sector may represent the domestic informal sector, sectors sheltered from international competition or, if capital is mobile, secret overseas bank accounts nI that are out of the reach of the domestic fiscal authorities. each case, it should be clear that, the raw rate of return in the inefficient sector is lower than in the formal one, especially in the case of LDCs. We model the extent of discretionary fiscal redistribution as a variable that is endogenously determined by three primitives: the existence of powerful groups, raw rat,es of return and in- stitutional barriers that limit the power of groups to extract transfers from the rest of society. Countries where the extent of fiscal redistribution is limited are those in which powerful groups do not exist, groups coordinate and act as one agent, or where there are well developed legal and other institutional structures that make it impossible for powerful groups to extract transfers ar- bitrarily from the rest of society. These limits do not exist in countries with powerful groups and a lack of institutional barriers. Contrasting the behavior of the economy across these different regimes constitutes a major part of the paper. The idea is that when groups have the power to extract fiscal transfers, capital stocks in the formal sector of the economy are not truly private. Since transfers must be financed by some form of taxation, higher transfers to one group result in higher taxes for the entire formal sector of the economy. In order to protect their profits from arbitrary taxation, agents transfer part of their resources to where they are free from this taxation. Agents can do this by investing in the shadow sector which is out reach of fiscal authorities. Typically these investments yield a lower raw rate of return. We model the interaction of the powerful groups as an infinite horizon dynamic game. In this game each group hns open access, via the fiscal process, to the capital stocks that other groups have in the formal sector. In contrast, capital in the shadow sector is truly private. The solution concept we use is Markov perfect equilibrium, which restrict strategies to be functions of payoff relevant state variables only. History dependent strategies, such as trigger strategies are not permitted. Our first point is that if there do not exist institutional barriers to discretionary redistribution, the existence of powerful groups reduces the growth rate relative to an economy in which society is composed of a single group or where groups can coordinate. This is because the existence of non-cooperative powerful groups generates a redistributive struggle, and as a result a greater share of resources ends up in non-taxable inefficient activities. Second, we show that if there exist multiple powerful groups, a reduction in power concentra- tion (an increase in the number of groups) leads to better economic performance. This result is reminiscent of the result that in a market in which firms play Cournot, the outcome approaches the competitive one as the number of firms increases. Our third point is that if there do not exist institutional barriers to discretionary redistribu- tion, an increase in the raw rate of return in the formal sector reduces growth.3 The intuition is as follows. An increase in the raw rate of return in the formal sector unleashes two conflicting effects: a direct effect that increases the profitability of investment in the formal sector, and a voracity effect that leads each group to attempt to grab a greater share of national wealth by demanding more transfers. This is reflected in a higher tax rate in the formal sector, which 2 induces a reallocation of capital to the informal sector, where it is safe from taxation. This shift reduces the growth rate in the economy, counteracting the direct positive effect of an increase in the raw rate of return. We prove that redistribution increases more than proportionally to the windfall, and that the tlirect effect of the windfall is dominated by the voracity effect, so that the growth rate declines as the raw rate of return increases. Note that, in the absence of powerful groups, no such endogenous increase in discretionary transfers and taxation occurs. In this case, the rate of return improvement stimulates investment and hence growth. Finally, we want to note that our approach is very different from the “Dutch disease” analysis of country adjustment to terms of trade windfalls (see J. Peter Neary and Sweder Van Wijnber- gen, 1986). According to that literature, a positive terms of trade windfall leads to a contraction of the non-resource tradables sector, either due to the crowding out effect of an expansion in the natural resource sector or a positive wealth effect that raises demand for nontradables. In con- trast, according to our model, the sector experiencing the positive price shock actually shrinks and there is no positive wealth effect: the decline in growth arises from the endogenous increase in distortionary redistributive activity. Section 1 provides an overview of the model. Section 2 contains the model. In section 3, we analyze the relationship between power concentration and growth, the voracity effect and welfare. Section 4 discusses some related empirical evidence in the context of our model. Finally, in section 5, we present our conclusions. I. Overview of the Model In this section we make an intuitive presentation of the model. We consider an economy pop ulated by infinitely lived groups and formed by two sectors: a high-return formal sector and a less efficient shadow economy. The shadow economy can be identified with a foreign tax haven or with the domestic informal sector. Taxes can only be levied in the formal sector. If powerful groups exist, each group is able to extract fiscal transfers. The government in turn must finance such transfers by levying taxes on the formal sector. This interaction is repeated for an infinite horizon. We want to stress that this dynamic game is a minimal model to address the issues discussed in the introduction. First, we need a dynamic setup because otherwise groups would just try to appropriate as much as they can and there would be no role for productivity shocks to affect the intensity of rent-seeking activity. Second, we need more than one group to analyze redistribution. It is straightforward to see that if there is only one powerful group, all powerful spuorg can coordinate, or there are institution that prevent discretionary fiscal redistribution, then all capital will be allocated in the efficient formal sector and first-best outcome achieved. If the above is not the case, there are two types of Markov perfect equilibria in the dynamic 3 game among the powerful groups: extreme and interior. In an extreme equilibrium, groups transfer as much capital as possible from the formal to the shadow sector. This is a socially inefficient outcome because the return in the shadow sector is lower. In contrast, along the interior equilibrium, groups limit their demand for transfers. To illustrate the mechanism that underlies the anomalous response of growth to windfalls observed in the data (the voracity effect) we will use equations (1) and (2) below. Suppose that the rates of return in the formal and shadow sectors are o and /3 respectively (o > p), that there are only two groups and that each group extracts ziK(t) out of the aggregate formal capital stock. It follows that the aggregate capital in the formal sector evolves according to i-(t) = [ct - x1 - x2]K(t) Consider the case in which, along the interior equilibrium, capital is transferred from the formal to the shadow sector (the conditions under which this occurs are derived in Section 3). In this case each group demands transfers up to the point where the other group is indifferent between investing in the two sectors. That is, each group sets Z: so that (2) cr - ;x = p, i = 1,2 To illustrate the voracity effect consider an increase in the rate of return in the formal sector equal to ACY. In the interior equilibrium each group increases the transfer it demands up to the point where the net rate of return available to the other group is equal to /3. That is, Axi = Acu (by equation (2)). Since both groups behave the same way, the increased redistribution induced by the increase in the raw rate of return is 2Acr. This is the voracity effect. Since the voracity effect dominates the direct effect of the windfall, the rate of accumulation in the formal sector falls. From (1) we can see that A$ = Aa: - 2Aa < 0. The counterpart of higher voracity is a shift of capital to the inefficient shadow economy. The argument we have made is loose. First, we did not prove that agents will choose linear transfer policies as assumed in equation (1). Also, we should note that equation (2) is valid only when capital flows from the formal to the informal sector. To determine when this is the case we need to solve consumption-savings problems of the n groups and fully characterize the interior equilibrium. We do this in section 3. In that section we embed the argument just made in a two- sector growth model, and let the number of groups be arbitrary. We compute the consumption policies and the accumulation paths for both types of capital. We prove the following results. First, if there initially exists multiple powerful groups, a reduction in power concentration (an increase in n) reduces discretionary redistribution and raises the average rate of return in the On economy. As the number of powerful groups increases, there are two conflicting effects. the one hand, there are more groups with the ability to extract subsidies. On the other hand, each 4 group knows that it must ask for a smaller subsidy if the formal sector is to offer a satisfactory after-tax rate of return to the ot,her groups. In equilibrium, the second effect dominates.4 This result is analogous to the result in industrial organization that the Cournot oquilibrium converges to the perfect competition outcome as the number of sellers increases. Second, when there are groups with the power to extract subsidies and there do not exist institutional barriers to discretionary redistribution, a positive shock to the terms of trade induces an increase in redistribution and a reduction in the growth rate. Moreover, we show that this anomalous response is also operational in the presence of anticipated future shocks. II. The Model We consider a two-sector growth model. There is an efficient formal sector, and an inefficient shadow sector. Resources in the formal sector are susceptible to taxation whereas, although productivity is lower, resources in the shadow economy are free from taxation. This is a caricature of what goes on in many economies. In the real world, both sectors may be subject to some form of taxation but the formal sector is subject to higher rates and is less able to evade taxation. An important difference between our model and conventional growth models is that in our model the economy is populated by groups that have power to extract subsidies from the gov- ernment rather than by atomistic agents that behave competitively. This captures the fact that fiscal policies in many courttries are determined by powerful interest groups. Since we want to analyze the effects of shocks that change the productivity of the formal sector relative to informal one, we shall consider two goods: an exportable and an importable. The exportable is produced in the efficient formal sector and the importable in the inefficient shadow economy. The importable will be the numeraire. In this section we will solve the model for a given price of the exportable. In section 4 we will consider anticipated future shocks. The objective function of each group is the present value of utility derived from consumption of the importable good t Within each instant the timing is as follows. Each group i enters period t with a stock of capital .)t(ih in the formal sector Ici (t) and a stock of capital in the shadow economy The formal sector capital stock ki (t) is used to produce the exportable good with a constant returns technology )t(ih that is sold at a price p in terms of the importable good. The shadow economy capital stock is used to produce the importable good in the shadow economy, again using a linear technology. Next, group i requests a fiscal transfer ri(t). Lastly, puorg i pays a tax Ti (t) from its income in the efficient sector, and consumes c;(t). It follows that the accumulation equations for efficient- 5 taxable capital and for inefficient-nontnxable capital are given by ii (t) = pk,(t) -T,(t), k(t) = f%(t) +7-i(t) -&(t) Next, we describe the fiscal-redistribution process that links the ‘ri’s and t,he T%‘s. We assume that the fiscal authority has no objective of its own. It acts solely as the agent of powerful groups. Using the terminology in Per Krusell, Vincente Quadrini, and .Jose-Victor Rios-Rull (1994), the “fiscal constitution” of this economy is the following: Fiscal Constitution Transfers and taxes must satisfy the following conditions: l The fiscal budget has to be balanced during each instant: C,“=, r3 (t) = C,“=, Tj(t) l Only proportional taxes on income in the formal sector can be levied, and the tax rate must be equal across groups. l The fiscal transfer that a group can obtain is bounded by (5) ,ri(t) L: zC,“=, kj(t), yf=j<r<cc The last restriction precludes each group from appropriating the aggregate capital stock at once. The lower bound on z is equal to the appropriation rate in the interior equilibrium (see (17)). The fiscal constitution implies that the tax rate r(t) must be ad,justed continuously to ensure a balanced budget r(t) = c;=, r,(t) T,(t) = T(t)qki(t), , K z C,“=, k, )t(Kpa ’ This tax rule implies that if group i increases its subsidy by an amount Ari, its tax burden increases by only Ar,~. In effect, the subsidy is financed largely by other groups in the economy. In this way, each group’s ability to extract subsidies grants it “open access” to the other groups’ capital stocks in the formal sector. This implies that the capital held in the formal sector is not truly private. Only capital in the informal sector is truly private in that it enjoys “closed access.” Using the terminology of the introduction, we may say that there is on possibility of discretionary fiscal redistribution if there is no open access to the capital stocks in the formal sector. In contrast, if there is open access, fiscal redistribution nac occur. To finalize the description of the economy, we list the initial conditions and the restrictions we impose. First, initial conditions are r,i(O) = 0 and ki(O) = kio > 0 for all i. Second, we restrict all capital stocks to be non-negative (7) k,(t) 2 0, 0,(t) 2 0, i = 1, . . ..n. t 2 0 Third, the rate of return in the inefficient sector is lower thau in the efficient sector 6 We would like to point out that (3)-(8) .is in fact a minimal model to analyze the problem at hand. First, multiple agents that interact strategically are necessary to analyze redistribution. Second, a dynamic setup is needed in order to analyze investment and saving decisions. Third, we assume that agents can costlessly move resources between sectors. Including appropriation or adjustment costs would add nothing to the insights provided by the model. A. The First Best In this case the setup we described reverts to the standard one-sector representative agent growth model. The solution to this case will be a useful benchmark. The first best allocation obtains in the following cases: (i) if powerful groups can coordinate and act cooperatively; (ii) if there is just one group; or (iii) if there are several groups, but institutional barriers do not permit them to extract any fiscal transfers. The first two cases cover what Mancur Olson (1982) labels encompassing groups. The allocation in case (i) is the solution to the problem in which a central planner maximizes (3) subject to the accumulat,ion equations in (4), the fiscal constitution (5)-(C) and non-negativity constraint (7). Since fiscal transfers do not generate any externality, net transfers to each group should be zero. Moreover, since the rate of return in the formal sector is higher than in the shadow economy, the central planner would allocate all resources in the formal sector. In terms of our setup this entails setting the consumption of each group )t(ic equal to the transfer t-i(t) it receives, ri(t) and making equal to the tax paid by each group Ti (t). This implies that accumulation equations (4) and (11) nac I)e rewritten as 6-J) &(t) = typIci - cz(t), I,(t) = 0 It follows that the optimization problem solved by the central planner is the standard consumption- savings Ramsey problem (see Robert J. Barro and Xavier Sala-i-Martin, 1995). The solution is )t(bfr = cfb(t) = [ap(l -(T) + Gn\k,(t) = z(q)k,(t) (10) k/“(t) = ki(.+‘=[a”-6]~t-s], Q”(t) = 0 where the superscript f0 stands for first best. nI this case the transversality condition is satisfied if and only if z((yp) > 0. The consumption of each group is proportional to its nwo capital, and in the case of logarithmic utility consumption is equal to the familiar Gki(t). In the second case, in which there is only one group it is straightforward to see that the optimal allocation is given by (10) replacing ki(t) by aggregate capital. Lastly, in case (iii) when groups cannot extract transfers, the individual capital that each group owns in the formal sector is truly private. Thus, we may replace the fiscal constitution by the condition ri(t) = Ti(t). Therefore, in this case we may reinterpret Ti as the amount that group i. takes out from the 7 formal sector to either consume or invest in the informal sector. Of course, as in the central planner’s case since ,0 < ap each group will set )t(ih = 0 and )t(ic = r,(t). Hence, group i’s accumulation equations will be given by (9) and the equilibrium allocation will be given by (10). B. Solution Concept In the case where powerful groups do not coordinate we will use Markov perfect equilibrium (MPE) as the solution concept. This is the natural extension of the first best solution, which is standard in representative-agent macroeconomic models. In these models control variables such as consumption and investment are typically functions of the state variables: the capital stock, wealth, etcetera. Similarly, in a MPE strategies are just functions of payoff-relevant state variables, not of history (see Eric M,askin and Jean Tirole, 1994). This restriction captures the nI notion that bygones are bygones. particular, MPE rules out history-dependent strategies, such as trigger strategies. We consider that MPE is a more appropriate concept than trigger strategies to study the problem at hand for the following reason. Countries ‘with procyclical government spending or voracity effects are not countries with well established institutional arrangements, such as the congressional committee system in the US, that allow powerful agents to coordinate on specific agreements and to design the threats that support these agreements. Another reason why we consider MPE more appropriate is that it reduces considerably the multiplicity of equibria in dynamic games. As is well known, if the discouut rate is sufficiently low, trigger strategies can support virtually any outcome as an equilibriunl. In this model the payoff-relevant variables for group i are the aggregate capital stock in the formal sector )t(K and group i’s closed-access capital stock b,. To see why ,)t(K and not ki(t), is payoff-relevant for pUOTg i note that although the efficient capital stocks of the other n - 1 groups are nominally private, group i has open access to them via the fiscal process. Since the transfer appropriated by group i, ri, is financed by taxing income in the formal sector of all groups (not only that of i), it follows that by demanding ri group i appropriates ri(l - s) from the formal capital stocks of the other groups. This is because the fiscal constitution implies that the tax paid by i only h<as to finance a proportion s of the transfer it receives. Hence, the consumption possibilities of group i (and its payoff) depend on )t(K and not on ki(t). To obtain the accumulation equation for aggregate capital we substitute (6) in (4) (11) k(t) = paK(t) - c;=“=, T3 (t) To see why any b,(t) j # i is not payoff-relevant for group i note that since none of the capi- tal stocks groups hold in the shadow economy are subject to taxation, they are truly private. Therefore group i does not have open access to them. ,z})t(i)?( In the game we are considering a strategy of group i is a pair of sequences { = {Ci(t),ri(t)}E,. )t(i4 A strategy is Markov if is only a function of the payoff-relevant state 8 variables K(t) and hi(t). An n-tuple of Markov strategies ({(ii;(t)}ES,..., (~$:(t)}&) forms a EPM if it is a subgame-perfect equilibrium for every realization of the state (K(t) , bi (t), . ,6, (t)). That is, if )21( J(+r(t),@T,(t)) 2 J(h(t),@P*_i(t)), f o r a l l i,t where * denotes equilibrium value, @Li = (4;) . . . ,4:-i, a:+,, , . ..+z). and J(., .) is the value taken by payoff function (3). In order to be able to use optimal control methods we will allow groups to choose transfer policies from the class of differentiable functions of the payoff-relevant state variables. That is (13) c(t) = c(K(t), h(t)), G(t) = ci(lc(t),bi(t)) We will derive the equilibria of this game in three steps. First, we let each group choose its consumption and transfer policies taking as given the strategies of the other n - 1 groups. Second, we find a set of R transfer policies that are best responses to each other. Last, using the equilibrium transfer policies we derive the equilibrium paths of the capital stocks in the formal and informal sectors and the consumption policies. During each instant s, group isolves the following problem: Problem (P,(s)). Clioosf? a consumption policy { )t(iC &} and a transfer policy { ‘ri (t) } y=, in order to maximize payoff function (3) subject to accumulation equations (??) and (ll), restrictions (5) and (7), and the transfer policies of the other groups (13). The present value Hamiltonian associated with i’s problem is (14) Hi = ViU(ci) + A; poIi - 1’; - C 1‘3 (Ii, 6j) + ci [PO, + ri - ci] + [i [21i - ,ri] + /lib; i J#Z 1 n-t where U(q) = eci 0 The second and third terms correspond to accumulation equations (4) and (11). The fourth and fifth terms to restriction (5), and the last term corresponds to the second constraint in (7). WC have disregarded the first constraint in (7). It turns that it is not binding in equilibrium. Notice that in deriving the first order conditions for group i, Ti and ci are treated as control variables, while the other R - 1 transfers r;(K, bj) are treated as functions of the state. In fact, these functions are the equilibrium policies derived from analogous control problems. To find an MPE, it is necessary to find n transfer policies r;(K. bj), j = 1, . . . . R that simultaneously solve n, Hamiltonian problems like (14). There are two types of Markov perfect equilibria in the game we are considering: interior and extreme. In an interior equilibrium, the transfers demanded by all groups are within the bounds defined in (5) at all times. This is not true in an extreme equilibrium. C. Extreme Equilibrium :K(t) In the extreme equilibrium all groups set their appropriations equal to the upper bound TK(t), defined in (5). Itis straightforward to check that if n - 1 groups set ri(t) = the best 9