National Tax Journal, Vol. 42, no. 2, (June, 1989), pp. 191-203 MATCHING GRANTS VERSUS BLOCK GRANTS WITH IMPERFECT INFORMATIONT NIRVIKAR SINGH* AND RAVI THOMAS** ABSTRACT A matching grant for an input in public good production is compared with a block grant, where nonobservability of public good outputs and some inputs prevents the use of the optimal grant system. The welfare comparison is shown to depend on the technology of production. The second-best grant is also compared with the optimal (full-information) grant. 1. Introduction HE issue of design of intergovernTmental grants in a federal system of government is a complex one, encompassing political, administrative and economic considerations (see, for example, Break (1980), Aronson and Hilley (1986) and Petersen et al. (1986)). In this paper, we examine a particular type of granta matching grant for an input in the production of a public good-and provide an explicit economic rationale based on imperfections in information.' These imperfections, as we explain below, rule out the use of the first-best grant scheme, which would be an output-matching grant. The grantor government's objective is to increase output of this good beyond the level that would be chosen by the recipient goverrunent acting totally on its own. This could be for reasons of super-local spillover of benefits, or for merit good reasons (Schultze (1974)). Subsidizing an input in the good's production provides a secondbest way of achieving this goal. An alternative possibility is a more general lumpsum grant, which would also increase production of the 'desirable' good, as well as outputs of other goods. We compare the input-matching grant to this alternative 'block' grant. Both are imperfect ways of achieving the grantor government's goal and, as we show, either may be better, de*University of California, Santa Cruz, CA 95064. **Temple University, Philadelphia, PA 19122. pending on preferences, technology and prices. Before we turn to the specifics of the model, we briefly put the issue in context. Recent debate, and grant-in-aid reform proposals have focused on the need to consolidate categorical grants, both formula and project, within broad functional areas (see Quigley and Rubinfeld (1986) for a recent summary). The ability of the federal government to measure the fraction of local output that spills over, and to implement a variable matching rate that a Pigovian subsidy requires, has been questioned. This inability, along with the variety of options that recipients have to convert categorical grants to fungible resources (McGuire (1979)) has led to the following conclusion. Given that most categorical grants restrict expenditure to narrow program areas, the shift towards block grants is seen to increase efficiency through the elimination of the dead weight loss resulting from selective inpUt2 subsidies, or attempts to convert contingent funds to an income supplement.3 The debate, as well as the analysis of the existing categorical grant system, for the most part, ignores an important fact about such grants. That is, very little attention seems to be have been paid to the fact that a large proportion of categorical Lrrants are capital grants. As such, these -grants can be viewed as selective subsidies on intermediate outputs, or for specific inputs as we do in this paper. Project grants, for example, are often inherently grants for an intermediate output, such as, sewage treatment plants and urban mass transportation projects. Of course some grants are more obviously for inputs, rather than for an intermediate output, for example, grants for school buildings and audio-visual equipment. Such capital grants have accounted for between 21 to 26 percent of U.S. grants-inaid since 1975 (see Table 1). While the policy winds have been shifting in favor 191 National Tax Journal, Vol. 42, no. 2, (June, 1989), pp. 191-203 192 NATIONAL TAX JOURNAL [Vol. XLH Table 1. Composition of Grant-in-Aid (billions of dollars) Outlays 1975 1980 1985 1987 Payments to Individuals Grants for Physical Capital Investment* Highways Mass Transit and Airports Community and Regional Development+ Sewage Treatment Facilities All Other Other 16.4 10.9 4.6 1.0 2.5 1.9 0.8 22.4 31.9 22.5 9.0 2.6 5.8 4.3 0.8 37.0 48.1 24.9 12.7 3.2 5.0 2.9 1.1 33.0 56.4 23.8 12.5 3.5 4.0 3.0 0.9 28.2 TOTAL 49.7 91.4 105.9 108.4 33.0% 21.8% 45.1% 34.9% 24.5% 40.5% 45.4% 23.5% 31.1% 52.0% 22.0% 26.0% Payments to Individuals Capital Grants Other @ourc*:SpecialAnalyses,Budgetof the UnitedStatesGovernment,FiscalYearl9T7, 1982.1989. Excludescapitalgrantsthat are includedas paymentsto individuals. Dueto the,,froblems involvedin classifyingexpenditure, all ofthe outlaysin this categoryhave beenclassifiedas capital investment, a inputgroupin whichthe recipientsspendmostof the money. of fewer restrictions and broader grants, the past and continued use of input grants, as indicated by the share of capital grants' in total grant-in-aid, deserves closer scrutiny. Previous discussions have focused on administrative and political advantages (see, for example, Tye (1973) and RaE;mussen (1976)). Here we provide a complementary economic analysis. This paper seeks to explain the importance given to this form of input subsidy. In doing so our analysis differs from the standard intergovernmental grant model. Treatments of intergovernmental grants and their effects in terms of standard price theory work in terms of the welfare function of the receiving government and the effects on consumption of the public and private goods, but do not model the objective fimction of the grants-giver or the production of public goods (Rasmussen (1976) is an exception for the latter). We take account of both these aspects. In doing so, we treat governments or communities as aggregate individuals, with well defined objective functions and side-step issues associated with aggregation, or how government level and individual choices may differ; this is, of course, standard in this literature. We may also place our analysis in the context of the 'principal-agent' literature, which distinguishes two types of asym- metric information. First, a higher level government (the 'principal') may not observe preferences, technology or other characteristics at the local ('agent's') level.5 Second, and what we focus on, the higher level government may not observe some local inputs and outputs, i.e., actions or results of actions;6 this is in fact the novel aspect of our approach .7 For example, education may not be perfectly measurable, though there are achievement tests which provide some information. Some inputs, such as 'general administration' and education's share of it, are very hard to monitor. On the other hand, some capital inputs, such as school buildings, and audio-visual equipment, are very easily monitored. Hence, the federal government, seeking to encourage the local production of 'education,' may not be able to do so directly and optiMally.8 Instead, it might have a choice between a general monetary grant (revenue sharing or a block grant-the classification depends on the degree of aggregation of the public goods) which provides an income effect stimulus, but also encourages production of less desirable public goods, and a subsidy of observable inputs, which directly encourages output of the desirable good but distorts input choices. In Section 2, we set out the model and assumptions. In Section 3, we compare the National Tax Journal, Vol. 42, no. 2, (June, 1989), pp. 191-203 No. 2] MATCHING GRANTS VERSUS alternate grant schemes by analytic means or via numerical simulations. Section 4 summarizes our results. 2. Model We assume that there are two local public goods and a single composite private good. The quantities of these goods are denoted by yi, y2 and x, respectively. The lower level, local government has a well defined objective fimetion, uL(yly2,x), by assumption. Furthermore, we assume that the higher level government also has a welfare function, UH(YlY2,X) over the locality's consumption. Hence the overall social welfare is assumed to be separable in the welfare of the locality in question and that of the rest of society. A further simplifying assumption is Al. UH(Yla2,X) = U"(PY1@Y2,X),P > Al incorporates the idea that the higher level goverrunent values consumption of the first public good more than does the local government. Implicitly, there is some positive externality, associated with this good, that the local government neglects. The relative valuation of the externality is captured by the magnitude of p. Assuming p is constant permits tractablility and does not qualitatively affect our results. Different valuations for the other goods could be similarly incorporated, but we wish to focus on a single systematic divergence of objectives. It should be noted that Al also implies ulH > ul, where the subscript denotes the derivative with respect to the first argument. Since we shall be interested in the input choices in the production of public goods, we make some assumptions about this technology of production. Spelcifically, A2. Public goods are produced under constant returns to scale (CRS). We shall assume that the inputs are capital and labor, with quantities k and 1, respectively for public good i. We assume that inputs can be purchased competitively, at prices r and w respectively, and there are no distortions in these input prices. By virtue of this and A2, the BLOCK GRANTS 193 cost function for public good i is yic,(r,w). Hence ci(r,w) is the marginal cost given the input prices. The private good, we assume, can be purchased at a given price q, and the locality has a total income of I. Hence we assume that the locality's income is independent of its consumption decisions. We do not need to model the mechanism whereby funds are raised by taxation for public good expenditure. By A2 (CRS), ci is the price of one unit of the public good i, and the locality's decision problem is Max UL(YlY2,X) Y*la2,- s.t. cl(r,w)yl + C2(r,W)Y2 + qx With our formulation, we have reduced the locality's problem to a consumer's utility maximization problem, and therefore the result of solving (1) is an indirect utility function U"(Cl,C2,q,I), with all the standard properties. We next consider the higher level government's viewpoint of these local decisions, and its ability to observe and control the outcome. What follows is therefore the heart of our treatment of intergovernmental grants. First, in program (1), if we define yt yllp, then (1) becomes max yi-,x s.t. U(PYI*,Y2,X) cl(r,w)py* + C2(r,W)Y2 + qx But the objective function in (1') is U'I(YtY2,X), by Al, and the higher level government's welfare is thus V(PCI,C2,qJ). Hence the locality acts as if it were the higher level government faced with the higher price pcl for public good 1. As a result yt is lower than yl, the optimal quantity for the higher level government. The result is lower welfare, as viewed by the higher level government than if it could directly choose the levels of the public goods, since indirect utility decreases in price, i.e., V,(C,,C2@q,I)> U"(PCI,C2,q,I). National Tax Journal, Vol. 42, no. 2, (June, 1989), pp. 191-203 194 NATIONAL TAX JOURNAL What can the higher level government do to increase social welfare? We shall assume that it cannot directly choose the levels of the local public good. This is institutionally realistic in the case of the federal government because the U.S. constitution restricts the federal government from intervening in areas considered strictly in the state or local government's domain.' The state government is not constrained in the same manner, but in our model the inability of the higher level government to observe public good outputs and certain inputs effectively prevents it from choosing their levels. The feasible alternative is to affect the incentives of the local government so as to increase consumption of public good 1, by means of a grant. If the output of good 1 were observable, this would be straightforward: it could use a matching grant for good 1 up to the point where the higher level government's marginal benefit was equal to the marginal opportunity cost of the grant funds. However, it is in the nature of public goods, such as education, that their output may not be easily observable or measurable." In that case the above matching grant would not be feasible. Next, we explore the implications of nonobservability of y, for grant design. For simplicity we assume the extreme of nonobservability, that the local government is able to divert matching funds for one public good to another, without beinf detected, i.e., the grant is fungible.' Specifically, this is possible due to the nonobservability of a particular input expenditure, combined with the nonobservability Of OUtpUtS.12,13 With this asswuption, a matching grant for a single public good is equivalent in effect to a block grant. This is therefore one avenue open to the higher level government: it can only encourage production of good 1 by subsidizing both public goods. Hence the first-best will not be attainable in this case In contrast to the output, say, of education, inputs in its production are more likely to be measurable and observable. With a known technology, if all inputs are observable, output can be deduced from the inputs. However, it is realistic that a [Vol. XLII higher level government might not be completely informed about the technology; for example, it may not know the elasticity of substitution between the inputs. In this paper we assume the technology is fully known. Instead, we assume that only one input, capital, is observble. This gives the higher level government a way of increasing consumption of public good 1, by a matching grant for capital used in its production. This grant is also not fully optimal, since it distorts input choices. Hence we have two different grants, a block grant and an inputmatching grant, with different defects. We next compare the welfare effects of these grants. Our analysis proceeds by comparing the two types of grants under the assumption that the amounts transferred are equal for the two grants, without computing optimal levels. This avoids having to make any specific assumptions about the opportunity cost of grant funds. In order to proceed we need further notation for the two types of grants. Let G be the amount of a block grant, and let s be the matching rate for capital used in the production of good 1, i.e., 0 :s s < 1. With the matching grant, the marginal cost of good 1 is cl((l - s)r,w). We will use cl(s) as an abbreviation. Clearly cl is decreasing in s. Similarly, let kl(s) be the demand for capital in good l's production, given s, i.e., L kl(s) = kl((l - s)r, W yl(cl(s),c2,q,I)). Hence the condition for the two types of grants to involve the same dollar amount is A3. G = srkl(s). We now turn to the analysis and results. 3. Analysis and Results We proceed to compare welfare as perceived by the higher level government for a block grant and an input matching grant, under the assumption that the two types of grants involve the same dollar expen- National Tax Journal, Vol. 42, no. 2, (June, 1989), pp. 191-203 No. 21 MATCHING GRANTS VERSUS BLOCK GRANTS diture by the higher level government. That is, in terms of the notation introduced previously we compare V'(PCI(O),C2,q,I V'(PC1(S),C2,q,I) + G) and given A3. This is not possible for general welfare functions because without specific assumptions it is impossible to compare the two indirect utility functions. However, we are able to provide a general result for small grants. We are also able to compare marginal benefits with a mild restriction on the welfare functions. This is done in Section 3.1. In Section 3.2 we examine the case of Cobb-Douglas welfare fimetions and CES technology, and make comparisons via numerical simulations. In Section 3.3 we are able to get some analytical results assuming Cobb-Douglas technology. 3.1 General Comparisons The first result is for small grants. PROPOSITION 1. Starting from the situation of no grant, the marginal benefit of an input matching grant of the form described above is greater than the marginal benefit of a block grant that involves the same transfer of funds. PROOF: See Appendix. The intuition for the result is as follows. Starting at a point of no grant, the marginal distortionary effect on input choice of the matching grant is zero. The input matching grant is better than the block grant because it directly corrects the initial output distortion as perceived by the higher level government. Note that while in the proof we have assumed p constant for simplicity, the result is actually more general. We only need p to be a differentiable fimetion of the outputs and have a value greater than one in the nogrant situation. It may be argued that proposition 1 has restricted application because in practice matching rates are typically not small, ranging from 50 to 94 percent (Aronson and Hilley (1986)). Hence, our next result is not restricted to small grants. To do this we need to assume that the government 195 welfare function is quasi-linear, i.e., it has the form W(YIY2) + x." The next proposition also compares marginal benefits of the two types of grants. Hence, it does not provide complete information about the comparison of total benefits. We discuss this further after stating the proposition. We also need some additional notation. Let E be the price elasticity of demand for capital used in production of good 1. PROPOSITION 2. If the welfare function has the form 0(yly2) + x then the difference in marginal benefits for an input matching grant and a block grant is positive if s p > 1 - --E. 1 - s PROOF: See Appendix The comparison between marginal and total benefits is best understood through illustration. Figure 1 plots the difference in total benefits, um - uB, against the matching rate s for particular CobbDouglas utility functions and CES production functions." In figure 1-A there is only one turning point and proposition 2 identifies a subset of the region over which the input matching grant is better. In figure 1B, for a different elasticity of substitution between inputs, proposition 2 is less informative. Our numerical simulations with these fiinctional forms suggest that figures 1A and 1B encompass the only two possibilities. Note that proposition 1 H tells us that vm - VHB is always po8 itive near s = 0. With its limitations in mind, the condition in proposition 2 allows some conclusions about the relative desirability of an input matching and a block grant. For a given p and s, the inequality is more stringent the greater the magnitude of E, i.e., the more responsive is the demand for capital to its price. Also, the condition will not hold, for a given p and f, if s is sufficiently close to 1-unless of course E = 0, in which case it always holds since p is greater than 1 by assumption. Hence, if the subsidized input is very price sensitive but a high matching rate or a large National Tax Journal, Vol. 42, no. 2, (June, 1989), pp. 191-203 196 NATIONAL FigurelA: [Vol. XLII Difference in total federal benefits as a function of the subsidy rate ( (7=0.5, VI TAX JOURNAL r/w=1.1,p=1.5) - VB 0 Input-Matching Grants Better :> 0.6 Figure 1B: Difference in total federal benefits as a function of the subsidy rate ( a = 40, r/w = 1.1, p = 1.5) VM- VB 0 Block Grants Better 0.1\\\\/ 0.3 block grant is still optimal, depending on which type of grant is chosen, it is more likely that a block grant will be preferred by the higher level government. Finally, it is instructive to decompose the input price elasticity F. It is easy to show that E @ Ek, + Eky * Tly@ * Ok, (2) where the right hand side terms are, respectively, the conditional input price elasticity, the output elasticity of the input, the price elasticity of demand for the aided good, and the share of the input in total cost. Hence, for example, if the subsidized input is relatively more important, i.e., 0,, is higher, then, other things equal, the condition in proposition 2 is less likely to hold. 3.2 CES Technology In this section we work with a CES production function. We are particularly in- s terested in how the elasticity of substitution between inputs affects the comparison between an input matching grant and a block grant. Note that the limiting case of a zero elasticity of substitution is straightforward. In this case the input matching grant causes no distortion in input use, and so the input matching grant better achieves the higher level goverwment's goals. It might seem that as the elasticity of substitution increases, at some point the block grant would become preferred and stay so for further increases, but this turns out not to be the case. In this and subsequent sections we assume that the utilitk fimction has the Cobb-Douglas form, u = a, In y, + OL2 In Y2 + a3 In x, with (xl + (X2 + OL3 = 1. The corresponding indirect utility is vH = In I - CL1In pci - OL2In C2 - ot3In q. For ease of exposition and computation we assume the simplest form of CES pro- National Tax Journal, Vol. 42, no. 2, (June, 1989), pp. 191-203 MATCHING No. 2] GRANTS VERSUS BLOCK GRANTS duction function, with capital and labor entering symmetrically. Hence, y, = (kPi+ ll)'IP, with -- < p < 1. Therefore u = 1/(l - p) is the elasticity of substitution. For this production function, Cl = (rl-a + w 1-a)lll-a , and ki = (cllr)'yl. With some further substitution, simplification, we have UHM _ VB 'q 0" - or (1 In 1 + I + (wlr(l ( 1n 1 1 + (wlr(l and S 1 - s))' Otl s p (3) We would like to examine the behavior of the last expression as a function of the matching rate, input prices, and utility and production function parameters. This is not feasible by analytical techniques and we proceed by using numerical simulations. Figure 2 shows the combinations of the matching rate and the elasticity of substitution for which one or the other grant scheme is preferred by the higher level government. The graph is divided into two parts. In figure 2A, u exceeds 1, while in figure 2B it ranges from 0 to 1. This division allows us to clearly display the different behavior of em - uB' for low and high elasticities of substitution. The dividing line is, of course, the Cobb-Douglas case. For low elasticities of substitution the results agree with an obvious intuition: up to some level of the matching rate, the input distortion, which is negligible for small s, continues to be outweighed by the fact that the input matching grant directly affects the desired output. Hence, for these values of s, the input matching grant is preferred. For higher s, the input distortion is too great and the 197 block grant is preferred. In figure 2B the region to the left of each line represents values of u and s for which the input matching grant is preferable, for a particular factor price ratio. Each line is labeled with the associated factor price ratio. We may note two other features of figure 2B. First, as the elasticity of substitution approaches zero the critical value of s approaches 1. As noted earlier this is what we would expect since in this case the input distortion associated with the input matching grant becomes zero. More generally, the dividing value of s increases as the elasticity of substitution decreases. This also means that for a given matching rate there is a value of the elasticity of substitution such that the input matching grant is preferable only below that value. For high levels of the elasticity of substitution the situation is considerably more complex, as may be seen from figure 2A. There, we again display the regions over which one or the other grant scheme is better for the same three values of r/w as in figure 2B. For r/w = 1.1 the shaded side indicates the direction in which the input matching grant is preferable. For other values of r/w, the area to the left of the line represents values for which the input matching grant is better. When r/ w = 0.5, then for each value of the elasticity of substitution there is a critical value of s below which the input matching grant is preferable and above which the block grant is better. This is not true for r/w = 1.1. For high enough (Tthe input matching grant if; preferred for values of s very close to zero, and then again for intermediate values of s. The intuition for this phenomenon is as follows. Increases in s beyond zero very quickly lead to a substantial input distortion, which in turn makes the block grant preferable. Further increases in s, however, stimulate output of the desired good and the input matching grant becomes preferred again. For still higher s the increased output is valued less and the input distortion matters more. For r/w = 2 we again have a single dividing line between the two grant schemes. Now the initial factor price ratio is so unequal that the National Tax Journal, Vol. 42, no. 2, (June, 1989), pp. 191-203 NATIONAL 198 TAX JOURNAL Figure 2ADividing lines between grant types 80 rlw r/w 1.1 2.0 [Vol. XLII p = 1.5) r/w r/w 0.5 70 60 50 40 30 20 10 subsidy rate 0 0.4 0.2 Dividing 0.6 1.0 0.8 Figure 2B: line between grant types (p 1.5) Block Grants Better 0.6 0.5 Input -Matching Better o.4 0.3 0.2 rlw = 0.5 0.1 0 r/w = 1.1 r/w = 2.0 02 0.4 phenomenon described above disappears. If we see how the dividing line changes as the elasticity of substitution changes for a given matching rate, there are similarities for all three values of the factor price ratio. In each case, for s high enough, there is a value of cr below which the input matching grant is preferred, and above which the block grant is preferred. For 0.6 0.8 1.0 subsidy rate lower S, however, as cr increases further, the input matching grant may again become preferred. The intuition is again in terms of the relative importance of input distortions and effects on the desired output. From figure 2A, the effect of the factor price ratio on the relative desirability of the two grant schemes is also seen to be National Tax Journal, Vol. 42, no. 2, (June, 1989), pp. 191-203 No. 21 MATCHING GRANTS VERSUS BLOCK GRANTS complex. As r/w increases from 0.5 to 1.1 the region over which the input matching grant is better shrinks. However, this does not happen as r/w increases from 1.1 to 2.0 if (r is above a certain level. Finally, the effect of changes in p is more straightforward. In fact, by differentiating the expression in equation (3), it can be shown that as p increases, the region over which the input matching grant is better increases. 199 From proposition 1 we know this is positive for s close to zero. For this special case we can deduce more by differentiating the last expression. It is easy to show that a(vhmas vBH) > 0 if andonlyif (p - 1) s<(p - aotl) (5) Statement(5)allows us to assert that vB' > 0 for an interval of s of the form In this section we employ the Cobb- (0,9). 9 is not analytically obtainable, but Douglas technology to examine several it is implicitly defined by setting the difadditional issues. In addition to a further ference between the matching and block analysis of how the utility function and grants equal to zero. Hence we may exproduction function parameters affect the plore how it varies with the parameters relative desirability of equally costly in- p, a and a,. put-matching and block grants, we look The following result is stated without at the welfare loss due to the use of an proof input-matching grant rather than the fullPROPOSITION 3. For a Cobb-Douginformation optimal output-matching las welfare function and Cobb-Douglas grant. Also, we explore the issue of the technology in producing public good 1, the optimal grant for a particular specifica- range over which an input matching grant tion of the cost of grant fimds for the is better than a block grant increases with higher level government. the magnitude of the distortion, p. The efThe Cobb-Douglas technology, with fects on 9 of changes in the elasticity of constant returns to scale, is y, = k'111-4. output of good 1 with respect to the subThe technology of the other good need not sidized input, a, and of changes in the be specified, and can be very general. The elasticity of welfare with respect to conmarginal cost, conditional demand for sumption of good 1, ocl, are ambiguous. capital and the output functions are reWhile the direction of variation of 9 with spectively; respect to 'a' is ambiguous, numerical simulations suggest that in practice 'a@has (1-a) w little effect on 9. The conclusion for oL,is 1-a cl = Aew , ki = aA Yi the same since it enters (4) in the same manner as 'a'. Next, we examine the welfare loss due and y, = (xi to the use of the input matching grant pcl rather than the output matching grant, which would be optimal if the higher level where A = a-'(1 - a)'-'. Hence, the equi- government had full information. librium use of capital with no grant is k, Let t be the matching rate for the production of public good 1, if its output is a(ot'I observable.Then an input matching grant pr). involver, the same transfer of funds With some further derivation we have which satisfies V'M' - VB" = -cL,a ln(l - s) srkl(s) = t(rki(t) + wli(t)). s aoti). -In 1+-.(4) (1 - S) p Substituting for the input demands and 3.3 Cobb-Douglas Technology t#m - National Tax Journal, Vol. 42, no. 2, (June, 1989), pp. 191-203 200 NATIONAL TAX JOURNAL simplifying, we obtain s = t/(a + (1 - a)t). From this we can obtain the welfare loss from the lack of observability of outputs, and show that this is increasing in t. 16 Hence, as might be expected, the welfare loss due to the lack of information is greater for a larger grant. We can also examine the effect of an increase in 'a' on the welfare loss. It can be shown that an increase in the elasticity of output with respect to capital reduces the welfare loss from using the input matching grant. This is what we would expect. Finally, we explore the issue of the optimal input-matching grant. To do this, we must explicitly introduce the opportunity cost of grant funds. If the cost is linear, the higher level government will wish to increase the grant without limit given the Cobb-Douglas welfare and production functions. Therefore we let the cost be eG - 1, where G is the amount of grant funds. Even this simple case is not analytically tractable, but we may extract some further information from numerical simulations. Figure 3 plots the optimal input matching grant, characterized by s*, for various values of p and I (a and ot, constant). We observe that for sufficiently high p and I the optimal input matching [Vol. XLII grant lies in the region where the input matching grant is better than a block grant, the region to the left of 9 in figure 3. 4. Conclusion In this paper, we have provided an economic rationale for input matching grants as a second-best alternative in the face of certain kinds of imperfections in information. We have examined how the relative desirability of an input matching grant over a block grant depends on various factors. For example, an inputmatching grant is always preferable for small grants. Furthermore it is preferable over a wider range of technologies and subsidy rates if the higher level government has a strong relative preference for the good it wishes to aid. For low elasticities of substitution in production, this is also true as the elasticity of substitution decreases, or as the relative price of the matched input decreases. However, for high elasticities of substitution, such comparative statics are ambiguous. This is because of the differing importance of substitution and output/utility effects. Further comparisons are made in terms of various input and output elasticities. Figure 3: (a = 0.5, (XI= 0.5) p 1.9 s 1.8 (I= 0.9) s*(l 0.7) . s (I = 0.5) 1.7 I-s 1.6 1.5 1.4 400 1.2 ' o 0 0.2 0.4 0.6 0.8 1.0 subsidy rate National Tax Journal, Vol. 42, no. 2, (June, 1989), pp. 191-203 No. 21 MATCHING GRANTS VERSUS Hence empirical work which can estimate such quantities, combined with our analysis, may have implications for designing grants according to economic criteria. This remains a subject for further research. ENDNOTES tAn earlier version of this paper was presented at the Econometric Society Winter Meetings in New Orleans, December, 1986. We are grateful to Steven Craig for his extremely helpful comments as a discussant on that occasion. We also thank Robert Adams, George Break, Dan Friedman, Ron Grieson, Peggy Musgrave, Richard Musgrave, John Quigley and Don Wittman for useful comments and discussions on this and related work. Finally, the comments of two anonymous referees were very valuable. Remaining errors are ours. The first author acknowledges with thanks, financial support from NSF Grant SES 860643a ana various UCSC Faculty Research grants. 'Although the appropriateness of using inputmatching grants to correct for spillover benefits has been questioned in the literature (see Schultze (1974), McGuire (1971)), a satisfactory formal economic rationale has not been developed for its use. In fact, the inefficiency generated by distorting input prices argues against its use. 2We may think of grants for inputs as being an extreme case of narrowness. If one were to di8tin gtiish between final and intermediate outputs in the provision of a public good, then most categorical gra nts are associated with inputs (Bradford, Malt and Oates (1969)). Of course our analysis does not rely on this view of public goo&. sthrough the resale of aided goods at a disc ount, or entitlements at a discount. For example, smal ler communities in California sell Federal-Aid Urban System fxmds to large cities and counties at a discount. The smaller jurisdictions accept the discount because they don't have to match funds and save on processing ooets. In many instances the State has acted as an agent and had standardized the discount rate at 65 percent (OHP (1985)). Other examples are in McGuire (1978), (1979). 'Although only grants for physical capital inv estment are listed in this table we do not intend to imply that they are the only input grants possible. Neither do we mean to imply all capital grants are matching, open-ended, categorical grants. 'Such lack of information and its implications for decentralization in general are discussed in Holmstrom (1984) and Guesnerie and Laffont (1984). Specffic applications to the design of grant schemes are in Bohn (1984) and Singh and Thomas (1984a). An application to user charge requirements for grants is in Singh and Thomas (1986). 'This leads to moral-hazard-type problems, analyzed abstractly in Holmstrom (1979), Shavell (1979) and numerous other papers (references in Holmstrom). In the context of grants this is the fungibility problem (see McGuire (1978, 1979) and Singh and Thomas (1984b)). A discussion of the "agency approach" and its empirical application to intergovernmental grants is in Craig and Kohlase (1985). BLOCK GRANTS 201 7Rasmussen mentions some reasons for funding re'trict"' which he traces to administrative and political considerations. Some of these "noneconomic reasons" are control of grant spillovers and avoidance of indefinite commitment of funds. We suggest that these noneconomic reasons are really problems of imperfection of information, and in some cases can he e licitly modeled. XP,Here,and throughout the paper 'optimal' means according to the objective function of the federal government. 9However, no Court decisions have disputed the right of the federal government to involve itself in grantin-aid activity (see ACIR, A-62 (1978), for a more detailed discussion of the constitutional issues). "Dffficulties in isolating measures of output, and the resulting problems it posesfor deternuning unit costs are discussed in Brafford, Malt, and Oates (1969). "Grant moneys are considered to be 'Tungibl&' when the recipient is able to use the fimds for purposes other than those specified in the grant authorization (see ACIR, Summary and Concluding Observations, Report A-62, June 1978). We note that fxmgibility may anse even though the donor may observe the recipients' actions, i.e., through the inability to control. However, our concern is the conversion of categorical funds which arises through trading or reclassification. IzThe process by which a grant is made fungible is illustrated here (the example is from Friedman (1984)). A local government requiring medical equipment for its hospital could transfer hospital security guards to the police payroll. Then, it could apply for criminal justice funds. The federal funds plus local resources, would then be used to maintain police services plus the guards. The hospital would rind its costs decreased freeing the required funds for the desired ent. e'@@-Wp ' emneed nonobservability of output as well as certain inputs; otherwise, the federal government with knowledge of the technology may be able to infer the input use through the observation of output. "'This is a common simplifying assumption in theoretical treatments. Why it is necessary here will be clear from the proof of proposition 2 in the appendix. We should note this form implies zero income elasticity for the public good. "These are used in subsequent analysis and nunierical simulations. 'cdetails of this and other derivations are available from the authors. REFERENCES Advisory Commission on Intergovernmental Relations, I;ilmmary and Concluding Observations, A62 (U.S. Government Printing Office, 1978). Aronson, J. R. and J. L. Hilley, Financing State and Local Governments, The Brookings Institution, Washington, D.C., 1986. Bohn, H., "Taxation and Intergovernmental Grants in a Federal Governmental System," Graduate School of Business, Stanford University, April, 1984. Bradford, D. F., R. A. Malt, and W. E. Oates, "The Rising Cost of Local Public Services: Some Evi- National Tax Journal, Vol. 42, no. 2, (June, 1989), pp. 191-203 202 NATIONAL TAX JOURNAL denceand Reflections," National Tax Journal, 22:185-202,1969. Break, G. F., Financing Government in a Federal System, Washington D.C.: The Brookings Institution, 1980. Craig, S. G. and J. Kohlase, "Why There is not an Unified Welfare System: Fiscal Federalism from an Agency Approach," in Perspectives on Local Public Finance and Public Policy 11,J. Quigley ed., Grenwich, JAI Press, 1985. Friedman, L. S., Microeconomic Policy Analysis, McGraw Hill, N.Y., 1984. Guesnerie, R. and J.-J. Laffont, "A Complete Solution to a Class of Principal-Agent Problems with an Applicatiton to the Control of a Self-Managed Firm," Journal of Public Economics, 25:329-369, 1984. Hohnstrom, B., 'Moral Hazard and Observability," Bell Journal of Economics, 10:75-91, 1979. Holmstrom, B., "The Theory of Delegation," in M. Boyer and R. Kihlstrom, Bayesian Models of Decision-Making, North-Holland, Amsterdam, 1984. McGuire, M. C., "Cost Versus Performance Subsidies as Tools of Intergovenunental Finance," National T= Journal, 24:13-18, 1971. McGuire, M. C., "A Method for Estimating the Effect of a Subsidy on the Receiver's Resource Constraint," Journal of PublicEconomics,16:25-44, 1978. McGuire, M. C., "The Analysis of Federal Grants into Price and IncomeComponenta,"in Fiscal Federalism and Grants-In-Aid, P. Mieszkowski et al. Washington, D.C.: The Urban Institute, 1979. Oates, W., Fiscal Federalism, New York: Harcourt Brace Jovanovich, Inc., 1972. Office of Highway Planning, Review of Federal-Aid Urban System Program, U.S. Department of Transportation, Washington, D.C., January 1985. Petersen, P. E., B. G. Rabe and K. K. Wong, When Federalism Works, Washington, D.C.: The Brookinp Institution, 1986. Quigley, J. M., and D. L. Rubinfeld, "Budget Reform and the Theory of Fiscal Federalism," American EconomicReview,76:132-136, 1986. Rasmussen,J., "The Allocative Effectsof Grants-InAid," National Tax Journal, 29:211-219, 1976. Schultze, C. L., "Sorting Out the Social Grant Programs: An Economist's Criteria," American EconomicReview,Vol. 64, No. 2, May 1974. Shavell, S., "Risk Sharing and Incentives in the Principal and Agent Relationship," Bell Journal of Economics, 10, 55-74, 1979. Silkman, R. and D. R. Young, Subsidizing Inefficiency:A Study of State Aid and Local Government Productivity, Praeger, New York, 1985. Singh, N. and R. Thomas, "Intergovernment Grants and Information Asymmetries," unpublished, U.C. Santa Cruz, 19&4a. Singh, N. and R. Thomas, "Fungibility of Categorical Grants: Some Implications for Grant Design," unpublished, U.C. Santa Cruz, 1984b. Singh, N. and R. Thomas, "User Charges as a Delegation Mechanism," National TaxJourn4al, 39:log113,1986. Tye, W. B., "The Capital Grant as a Subsidy Device: The Case Study of Urban Mass Transportation," in Joint Economic Committee, 93rd Congress, Ist Ses- [Vol. XLII sion,TheEconomicsof FederalSubsidyPrograms, 1973. Wilde, J. A., "Grants-In-Aid: The Analytics of Design and Response," National Tax Journal, 24:143-155, 1971. 5. Appendix 5.1 Proof of Proposition I The marginal benefit of a block grant is aV(PCI(O),C2,q,I)IaI. From A3, an equivalent input matching grant satisfies rki(s) + sr ds - = 1. dG as 1 ds 0, this becomes- = -. dG rk,(O) The marginal benefit from the input matching grant is As s - auff ac, cls *p -- -(evaluated as dG a(pc,) at s = 0). Substituting in for ds/dG and using ac,las auff ac, p -rac,lar, this becomes Howa(pc,) ar k, * ever, from Shephard's Lemma, y,ac,lar k,. aul . p Substituting again, we get - Fia(pcl) Y, au"la(pc@) nally, from Roy's Identit,, , au-1ai This last substitution gives the following result for the marginal benefit of the input matching grant, p -aulai, which is greater than aumlai, since p > 1. Q.E.D. 5.2 Proof of Propostion The general ds dG expression rk, + sr(ak,las) ak, ak, - = -(-r). a(l - s)r as Hence 2 for National Tax Journal, Vol. 42, no. 2, (June, 1989), pp. 191-203 MATCHING No. 21 ak, (1 - s)r E = - k, GRANTS VERSUS BLOCK GRANTS (1 - s) ak, a(l- s)r k, as Hence 1, is therefore, auff pl-aI 1 se/(l s)' whilethe marginal ds dG Proposition - SE/(l benefit of theequally costly allH 1 rkl(l 203 blockgrant - 8))' The marginal benefit from the input matching grant, following the steps in the proof of ear,- auH a(I + G) is -. Since a(I + G) aum = -. Hence aI UH is quasi-lin- the result follows. Q.E.D.