\"PlT-h" m. w H -- ”1&5“. anus-4H :«4. “- $774.. w- r.- gm" «4 HM «m a . . 4 ‘~ -"' ‘ T‘n: .. ~ 4 , 4 4 v . {KN r :v ‘1’ 4 v of . . 74-5 "‘5'“? ~.§.{.§§ 4; fix. wr 444491" . ~*\ 45.44" v“. .7 ' '. :44,” ‘ Lfk‘fo ! w . n -. I n. u de‘n u .0411? 4 4p umwx v a ”4‘ u- $7143“, " ‘ . .:~‘" w. 4 - I‘ v- 9:- c. . fidéupm wt u ”'3"? - "#1?!”th U m. - "1:4“: .9 M2 . ”i .‘i $31.55;. '-(|~ '41: ' 4344 " 4‘: hfégié‘? 4‘s. “Raga z“: “""‘ militia 'fig;c.g U3 . .57 wag? 5 l4 433:3; 7515.534" Xx; Am 4 I I n '1 l. . ' ”:11"? “'f" ‘ 5.51.4.sz {‘3th: .v .c. (u ‘ 5:555" .33 {4 3;" 75 “353...:ng 7' 5. - ‘92“4 $575552; Maya .u‘ . ‘1 "h“..‘xfh: ‘ “71' 4 :‘th‘f‘i, " '3‘“ 4 Ii. 4 '7‘“ . '1. 4:25: ~ 4.” “ rifliwa- .I‘.’ ‘7”; '1' 1 u o . "7‘ «a». "3:1: .‘M’ux ; ... on ' \I ‘ 5‘ J“ "5.» .- n 4 '33va ~71 {‘5‘ '5 4 3" 0:.“ :‘w a!“ .4." .M :15 M‘ ”I (l hulk .5615 (Ougs‘m?lfi‘fl :1". 5.332;; “3’5 " 5*; 5‘ “GAR“; 5’“ . gfgkvm“ div “'1‘”: -I‘v~fl!'l" ~rr’41'uzrpec. If” ,t 0:? .4’.‘«;S_<‘4;" i}, 43m... w 4 p. 4 my.“ a: '4" .LV'! llIHHNUNIlllHlUIHIIIHHIHUWIIIHIIHHHIUM , ”Ema? i 00762 2115 Michigan State University This is to certify that the dissertation entitled DETERMINANTS OF STATE AND LOCAL GOVERNMENT BORROWING presented by JUDY ANN TEMPLE has been accepted towards fulfillment of the requirements for Ph-D- degree in W— W mic profess 02®§7 78/7 'T r e.— I PLACE IN RETURN BOX to remove We checkout from your record. TO AVOID FINES return on or before due due. DATE DUE DATE DUE DATE DUE usu I. An Affirmative Action/Equal Opponunny summon WM‘ DETERMINANTS OF STATE AND LOCAL GOVERNMENT BORROWING BY Judy Ann Temple A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Economics 1990 ABSTRACT DETERMINANTS or STATE AND LOCAL GOVERNMENT BORROWING By JUDY ANN TEMPLE The objective of this research is to explain the variation across states in the issuance of long-term bonds by state and local governments. These governments issue two types of bonds: general obligation bonds that typically are issued for traditional government or public purposes, and revenue bonds that are more likely to be issued in support of non- traditional government activity. I develop a model of state and local government borrowing in which three important decisions are made. First, the government official determines the optimal share of debt finance associated with any desired level of state and local capital expenditures. At the same time, the median voter selects the level of state and local capital spending. Finally, the government selects the optimal share of debt finance associated with the desired level of economic development activity. The quantity of general obligation bonds and revenue bonds issued depends on both the level of the activity being financed and the method of financing that level of activity. The purpose of this research is to identify and then estimate the significance of the proposed determinants of each of these decisions. The hypotheses generated by the model are tested using state and local government borrowing levels by state as units of observation. The research uses general obligation bond data from the Securities Data Company and newly-available data on private-activity bond issues from the U.S. Treasury for the sample years of 1983 and 1984. Income and past and future population growth are shown to be important determinants of the variation in general obligation bond issues. The positive relationship between income and the general obligation debt share contradicts the predictions of recent arbitrage models of government borrowing. .Although the variation in some types of revenue bonds cannot be explained by economic factors, a significant portion of the across-state variation in industrial development bonds can be explained by variables in the model. Copyright by JUDY ANN TEMPLE 1990 ACKNOWLEDGMENTS I wish to express my gratitude to my dissertation advisor, Professor Ronald Fisher, for his endless patience and enthusiasm. I also greatly appreciate the efforts of Professor John Goddeeris. Professors Larry Martin and Paul Menchik generously read a draft of my dissertation for the oral defense. Fellow students Tom Schuster and Pat Redmon provided valuable assistance at earlier stages of this research, while colleague Bill Sj ostrom provided assistance at a latter stage. For encouragement, advice and friendship, I want to thank Arthur Reynolds. Last but not least, I thank the Temple family for their support. TABLE OF CONTENTS LIST OF TABLES ................................................... viii LIST OF FIGURES .................................................. ix CHAPTER I. INTRODUCTION ............................................... l A. General Overview ........................................ 1 B. Related Studies ......................................... 5 C. Conclusion .............................................. 10 Footnotes ............................................... 12 II. STATE AND LOCAL BORROWING COSTS ............................ 13 A. Introduction ............................................ 13 B. General Obligation and Revenue Bond Interest Costs ...... 14 C. Interest Costs and the Level of Borrowing ............... 15 D. Other Factors Affecting Interest Costs .................. 20 E. Conclusion .............................................. 2O Footnotes ............................................... 22 III. THE OPTIMAL DEBT SHARE IN THE FINANCING OF STATE AND LOCAL CAPITAL EXPENDITURE .............................. 23 A. Introduction ............................................ 23 B. The Optimal Debt Share .................................. 24 C. The Importance of General Obligation Debt Limits ........ 29 D. Conclusion .............................................. 29 Footnotes ............................................... 31 IV. THE USE OF STATE AND LOCAL BONDS IN THE FINANCING OF NON-TRADITIONAL.GOVERNMENTAL ACTIVITY ................... 32 A. Introduction ............................................ 32 B. The Choice Between Revenue Bonds and Other Fiscal Incentives in the Support of Non-traditional Governmental Activity ................................... 33 C. Graphical Illustration of the Optimal Debt Share ........ 35 D. Algebraic Formulation of the Debt/Tax Choice ............ 41 E. The Relationship Between the Two Debt/Tax Choices ....... 43 F. Why Do State and Local Governments subsidize Private Activity? ....................................... 45 G. The Importance of General Obligation Debt Limits ........ 46 H. Conclusion .............................................. 47 Footnotes ............................................... 49 vi vii A MODEL OF STATE AND LOCAL BORROWING ....................... A. Introduction ............................................ B. The Individual's Utility Function ....................... C. The Individual's Budget Constraint ...................... D. The Jurisdiction's Budget Constraint .................... E. Derivation of the Median Voter's Demand Functions ....... F. The Level of General Obligation Bond Issues ............. G. The Level of Revenue Bond Issues ........................ H. Conclusion .............................................. Footnotes ............................................... VI. ESTIMATION ................................................. . Introduction ............................................ . General Procedure ....................................... . Data Description ........................................ . Estimation of the Demand for Capital Expenditures ....... Estimation of the Debt Share in the Financing of State and Local Government Capital Expenditures ......... Simultaneous Estimation of the Investment Demand and Debt Share Equations ................................ G. Estimation of the Determinants of Private-Activity Bond Issues ............................................. MUOU> u: VII. CONCLUSION ................................................. APPENDIX.A ....................................................... APPENDIX B ....................................................... BIBLIOGRAPHY ..................................................... 50 50 51 53 54 55 56 58 59 61 62 62 62 63 67 75 79 83 88 100 102 106 111 114 E E O‘ O‘C‘O‘O‘O‘O‘HH O‘UIPWNHNH O‘O‘O‘O‘ .10 .11 LIST OF TABLES State and Local Government Borrowing in 1983 . State and Local Government Borrowing in 1984 . Definitions of Variables ..................... State and Local Capital Expenditure Estimates State and Local Capital Expenditure Estimates State and Local Capital Expenditure Estimates Debt Share Estimates ......................... Volume of New Issue Private Activity Bonds, 1983 and 1984 ................................ Average Private Activity Bond Issues by State, 1983 and 1984 ................................ Private-Activity Bond Issue Estimates ........ Student Loan and Exempt-Entity Bond Estimates Industrial Development Bond Estimates ........ Small Issue and Industrial Park IDB Estimates viii 00000000000000 OOOOOOOOOOOOOO page 65 71 72 74 76 84 84 89 93 95 97 LIST OF FIGURES FIGURE Page 4.1 The iso-subsidy curve is convex ........................... 37 4.2 The effect of a change in h and/or I on the cost-minimizing combination of Z and RB ................... 40 ix CHAPTER I INTRODUCTION 1 , A, General Overview. Long-term bonds issued by state and local governments are of two types: general obligation bonds and revenue bonds. The former are usually issued by the state and local governments themselves and are backed by the "full faith, credit, and taxing power" of the issuing jurisdiction. General obligation bonds tend to be issued for what are typically referred to as "public purposes,” such as the construction of roads, bridges, correctional facilities, and elementary and secondary schools. Revenue bonds are issued by state and local governments and their special districts and statutory authorities to finance public utilities and a wide variety of nontraditional government activities such as aid for industrial development, hospitals and colleges, student loans, and mortgage subsidies. Unlike general obligation bonds, revenue bonds are not backed by the taxing power of the jurisdiction. Instead, the payment of interest and principal comes exclusively from the earnings of the particular investment project. In many cases, revenue bonds can be thought of as corporate bonds that are issued as municipal revenue bonds so as to qualify for the federal income tax exemption of municipal bond interest. The majority of research on state and local borrowing has concentrated on the determination of the yield differential between taxable and tax-exempt bonds. In many studies, the supply of state and local bonds is ignored because it is believed that only changes in the 1 2 demand for tax-exempthonds and/or the supply of taxable bonds will affect the yield ratio. Studies that have analyzed general obligation bond supply traditionally assume that long-term debt is issued to finance capital expenditures. Evidence reported in Tables 1.1 and 1.2 suggests, however, that actual borrowing levels are always well below 100% of capital expenditures and that this debt share varies dramatically by state. The data in the tables also show substantial interstate variation in per-capita borrowing levels for both general obligation and revenue bond issues, and per-capita state and local capital expenditures.1 Per capita long-term general obligation bond issues averaged $56 over the two-year period. State and local governments spent an average of $316 per capita on capital expenditures, and the share of debt in the financing these capital expenditures averaged .18. Interestingly, the variation in the level of general obligation bond issues is driven primarily by the 'variation in the debt share rather than the variation.in capital spending. The coefficient of variation (100 times the standard deviation divided by the mean) is 76 for per capita general obligation bond issues, 72 for the debt share, and only 47 for per capita capital spending. Per capita revenue bond issues averaged $204 with a coefficient of variation of 49.2 The purpose of this research is to examine possible explanations for the existence of differences in general obligation and revenue bond amounts by state with the intent that this framework can be used to explore these differences empiricallyu While the traditional view maintains that the demand for state and local capital investment will influence the level of borrowing, there has been no explanation of variation in.the debt share and hence the volume of new general obligation issues. The volume of general obligation bonds will be equal to the 3 Table 1.1 State and Local Government Borrowing in 1983 STATES GENERAL STATE AND SHARE 0F G0 REVENUE OBLIGATION LOCAL CAPITAL BOND ISSUES BOND ISSUES EXPENDITURES IN FINANCING ISSUES PER CAPITA PER CAPITA GOVT CAPITAL PER CAPITA Alabama $ 6.96 $ 262.34 .03 $ 168.26 Alaska 1827.00 2786.13 .66 494.82 Arizona 55.75 517.64 .11 492.78 Arkansas 0 150.43 .00 98.88 California 35.29 243.11 .15 146.91 Colorado 11.92 386.82 .03 217.92 Connecticut 104.35 189.14 .55 99.68 Delaware 76.07 332.67 .23 191.42 Florida 53.79 340.10 .16 222.08 Georgia 26.40 361.26 .07 187.24 Hawaii 220.80 402.55 .55 75.56 Idaho 0 285.83 .00 74.90 Illinois 79.07 239.87 .33 148.45 Indiana 12.11 187.58 .06 192.48 Iowa 15.54 282.55 .05 109.12 Kansas 38.32 313.51 .12 200.16 Kentucky 0 229.74 .00 160.43 Louisiana 128.02 388.14 .33 233.00 Maine 52.42 193.03 .27 42.72 Maryland 95.83 341.34 .28 223.31 Massachusetts 65.74 269.68 .24 270.05 Michigan 36.07 202.52 .18 82.82 Minnesota 55.53 331.19 .17 302.22 Mississippi 10.98 187.96 .06 90.59 Missouri 26.26 191.34 .14 212.69 Montana 49.79 332.23 .15 261.03 Nebraska 13.81 396.99 .03 78.90 Nevada 56.05 509.70 .11 208.47 New Hampshire 100.18 202.29 .50 256 52 New Jersey 79.56 224.14 .35 191.05 New Mexico 49.62 455.78 .11 175.46 New York 79.32 327.84 .24 97.51 North Carolina 34.13 184.25 .19 52.29 North Dakota 8.19 341.85 .02 180.62 Ohio 20.13 223.82 .09 126 79 Oklahoma 20.72 309.78 .07 118.96 Oregon 10.90 281.95 .04 45.49 Pennsylvania 40.22 216.55 .19 195.06 Rhode Island 6.18 160.00 .04 109.95 South Carolina 37.39 206.18 .18 148.48 South Dakota 0 324.75 .00 233.19 Tennessee 58.88 235.02 .25 188.02 Texas 76.16 355.00 .21 212.86 Utah 122.91 615.35 .20 267 54 Vermont 101.68 220.91 .46 201 52 Virginia 38.99 221.53 .18 259.40 Washington 96.70 650.70 .15 55.99 West Virginia 4.25 257.08 .02 107.43 Wisconsin 81.67 248.92 .33 62 75 Wyoming 67.49 917.64 .07 459 30 4 Table 1.2 State and Local Government Borrowing in 1984 STATES GENERAL STATE AND SHARE OF GO REVENUE OBLIGATION LOCAL CAPITAL BOND ISSUES BOND ISSUES EXPENDITURES IN FINANCING ISSUES PER CAPITA PER CAPITA GOVT CAPITAL PER CAPITA Alabama $ 2.84 $ 268.74 .01 $ 261.52 Alaska 1341.48 2577.03 .52 273.27 Arizona 106.64 491.51 .22 340.71 Arkansas 0 153.22 .00 93.74 California 58.04 252.28 .23 145.00 Colorado 99.27 435.33 .23 228.53 Connecticut 127.49 198.86 .64 152.24 Delaware 183.22 337.13 .54 395.77 Florida 29.99 373.82 .08 346.82 Georgia 48.71 295.47 .16 488.19 Hawaii 184.36 408.30 .45 153.47 Idaho 5.50 270.80 .02 72.00 Illinois 76.32 260.56 .29 201.32 Indiana 17.17 222.68 .08 171.13 Iowa 27.83 319.90 .09 105.72 Kansas 23.85 295.66 .08 195.33 Kentucky 0 263.26 .00 189.58 Louisiana 111.12 397.98 .28 332.36 Maine 27.55 196.63 .14 66.49 Maryland 118.93 362.88 .33 291.17 Massachusetts 109.49 266.63 .41 237.40 Michigan 33.13 220.28 .15 176.67 Minnesota 158.93 378.89 .42 291.31 Mississippi 4.68 188.23 .02 156.98 Missouri 18.07 214.59 .08 233.77 Montana 0 420.87 .00 389.56 Nebraska 17.20 390.16 .04 215.58 Nevada 63.69 470.34 .14 179.93 New Hampshire 35.79 192.43 .19 302.66 New Jersey 69.50 229.11 .30 254.49 New Mexico 55.49 498.95 .11 121.93 New York 89.16 353.44 .25 193.98 North Carolina 16.35 193.34 .08 129.98 North Dakota 9.44 392.29 .02 337.70 Ohio 25.85 225.04 .11 111.32 Oklahoma 121.24 335.60 .36 123.19 Oregon 167.07 302.24 .55 113.60 Pennsylvania 43.15 192.27 .22 319.76 Rhode Island 52.74 183.75 .29 447.92 South Carolina 44.74 192.69 .23 244.84 South Dakota 11.44 338.01 .03 164.54 Tennessee 33.89 249.10 .14 280.82 Texas 113.84 368.21 .31 264.57 Utah 106.09 836.08 .13 395.32 Vermont 64.22 251.51 .26 222.64 Virginia 31.20 205.41 .15 353.97 Washington 117.56 496.71 .24 108.97 West Virginia 7.14 214.55 .03 103.48 Wisconsin 95.99 273.17 .35 113.14 Wyoming 63.73 1013.25 .06 711.50 5 chosen debt share in the financing of capital investment times the quantity of capital investment. Similarly, the volume of revenue bonds will be a portion of the total amount of state and local aid in support of nontraditional government activity. In order to explain the determinants of both state and local general obligation and revenue bond supply, a newer view of the determinants omeunicipal financial policy is emphasized where the focus is on the relative attractiveness of tax and bond finance and the possibilities for substitution between them. 1,§, Related Studies. The first empirical analyses of the state and local bond market were done in the early 1970's by Galper and Petersen (1971) and Fortune (1973). Their objective was to examine Congressional proposals to subsidize the costs of state and local borrowing. Both studies incorporate aggregate measures of state and local general obligation bond issues into a demand and supply model in order to explain the behavior of tax-exempt yields over time. Galper and Petersen. assume that state and local borrowing is undertaken to finance public construction expenditures. They find that the amount of construction (and hence borrowing) depends negatively on the municipal bond rate (r5). Fortune also assumes that state and local bonds are issued to finance capital expenditures. He expresses municipal bond supply as a function of lagged interest rates and disposable income, where disposable income serves as a proxy for the demand for state and local capital goods. Fortune finds a positive relationship between municipal bond issues and disposable income over time. While both of these general obligation bond supply models assume that 6 the level of borrowing is related to the demand for state and local capital, neither of them attempt to estimate the determinants of the jurisdiction's demand for capital. Allman (1982) appears to be the only one who has incorporated a demand equation for state and local capital into a model of municipal bond supply. In his model, the median voter's demand for municipal capital investment is derived from the voter's demand for municipal goods and services. In his empirical work, however, Allman uses a measure of total national income (rather than median) as a determinant of capital demand. Allman also suggests that there is an optimal share of debt in the financing of capital investment that depends solely on voter ”tastes.” Unfortunately, his use of the ratio of total state and local government spending to voter incomes as a proxy for the preference for debt finance is done without explanation. In a somewhat different empirical analysis of the demand and supply of tax-exempt 'bonds, Hendershott and. Koch (1977) use a government accounting framework to explain tax-exempt bond issues. They utilize a sources versus uses statement where the sources of funds (bond issues, federal grants, tax revenues and the municipal surplus) must equal the uses of funds (capital purchases, other outlays, and financial asset purchases.) They express bond supply as a function of the exogenous sources and uses (capital outlays, other outlays, and grants.) Their time-series results for the sample period 1963-1974 are generally consistent with the standard view that state and local governments typically finance half of their capital outlays with debt. Research in this dissertation incorporates this sources versus uses framework as a description of the government's budget constraint. It is important to note that the objective of all four of the studies 7 described above is to explain the aggregate amount of tax-exempt bonds issued in the U.S. The variation in the volume of bond issues across jurisdictions is not addressed. Several recent studies have attempted to analyze the determinants of state and local borrowing behavior using a cmoss-sectional framework. Adams (1977), Asefa, et. a1. (1981), Gordon and Slemrod (1986) and Metcalf (1989) argue that municipal governments issue bonds in order to engage in arbitrage. One type of arbitrage is direct: jurisdictions issue tax-exempt bonds and invest the proceeds in higher-yield taxable bonds. This behavior, however, is limited by law. Indirectly, jurisdictions can engage in two other types of arbitrage. In the first, residents are expected to let their jurisdictions save for them because the jurisdiction can«earn.the pre-tax rate of return on taxable investments. Higher-income residents are expected to prefer that their governments collect greater tax revenues from them and use the proceeds to purchase taxable securities because these residents would earn a relatively low after-tax rate of return on their own investments. (Presumably, taxes will be less in the future.) The second type of arbitrage is more closely related to government borrowing behavior. Jurisdictions can take advantage of the differing pre-tax.yields on tax-exempt and taxable bonds by issuing tax-exempt bonds and using the proceeds to lower the current tax rate. Residents are then free to invest their higher after-tax income in taxable securities. This substitution of debt for tax finance is a form of arbitrage that is expected to be preferred by lower-income residents because they earn the greatest after-tax yield on taxable securities. While the traditional view of state and local borrowing assumes that 8 long-term debt is issued to finance capital expenditures, this type of bond supply model does not include capital spending as a determinant of bond issues. Instead, the emphasis is on the relative attractiveness of tax and bond finance and the possibilities for substitution between them. An important determinant of bond supply is the relationship between the federal marginal tax rate of the residents in the jurisdiction and the marginal tax rate implied by the tax-exempt/taxable bond yield ratio (t - l - rh/rt.) While the entire area of state and local borrowing has been relatively under-worked, especially little has been done to explain the decisions of state and local governments to issue revenue bonds. Descriptive analyses of revenue bond issues (for example, see Clark and Neubig (1984), Clark (1985) focus on the perceived costlessness to the jurisdiction of this type of borrowing. State and local governments are not responsible for the repayment of principal and interest. Rather, the income from the particular investment project is used to service the debt. The cost of these revenue bond issues is borne primarily by the U.S. taxpayer because revenue bond interest is not subject to federal taxation. Allman (1982) provides an interesting model of revenue bond supply in which the decision maker is an elected official who attempts to satisfy his constituency by providing as many services as possible. This official decides to issue revenue bonds only when he expects the revenues from future user fees to fully cover costs. In his empirical work, Allman assumes that the official examines recent trends in the amount of user fees collected. Because increasing user fees are viewed as a sign of increasing demand for (and hence profitability of) the revenue bond-financed projects, revenue bond supply is assumed to depend 9 positively on the recent trends in user fee collections. This assumption is supported by his empirical results. Allman makes a valuable contribution in this first attempt to model the aggregate level of revenue bond issues, but a problem with his use of user fees as a determinant of bond issues is that only a small fraction of revenue bond-financed projects use this type of user fees to service the debt. Finally, research in this dissertation draws on a variety of studies concerning the determinants of borrowing costs and the implications these costs have for the jurisdiction's debt/tax choice. Many researchers (Leonard (1983), Hendershott and Kidwell (1978), and others) have noted a positive relationship between bond issue volume and interest costs. Hendershott and Kidwell find that a change in the supply of tax-exempt bonds in a geographical region may affect the interest costs in that market relative to those nationwide. Kidwell, Koch and Stock (1984) find that jurisdictions in states that exempt from the state income tax the interest income from bonds issued within that state can issue bonds at a lower interest cost. This assumption that issuing governments are not price-takers in the market for loanable funds is an important component of the model developed in this dissertation. To an extent, the study of the determinants of the debt/tax decision in the financing of state and local capital is analogous to the debt/equity decision in the financing of the firm. While Miller (1977) argues that the debt-equity ratio is indeterminate for any particular firm, Auerbach (1979) and Feldstein, Green and Sheshinski (1979) claim that a unique optimal debt-equity ratio will exist if the cost of capital varies with the degree of debt finance, or "leverage." This is also the assumption made in more recent work such as Nadeau (1989). In this 10 dissertation, the fact that interest costs vary"with the level of borrowing may similarly serve to guarantee a unique optimal debt/tax choice for each jurisdiction. W. State and local government borrowing has been a relatively under-worked topic in public finance. Interestingly, while this dissertation is one of a relatively few studies that have analyzed the borrowing of state and local governments in a cross-sectional framework, this thesis topic has also been chosen by two other recent Ph.D students. Cunningham (1989) attempts to explain the determinants of state (not local) debt using primarily cross-sectional data from 1972. (He also derives some time-series estimates for 1940-1987 for a subsample of 21 states.) In his model, the equilibrium amount of debt is determined by equating the marginal welfare burdens of debt and taxes. He combines this argument with a political one in which high-income residents are assumed to prefer debt finance because these residents benefit most from the inclusion of tax-exempt bonds in their portfolios. He finds that his dependent variable (the ratio of outstanding state debt to personal income) is positively related to income and (less strongly) unemployment, but is not related to past capital spending and expected population growth. In a paper from his dissertation, Capeci (1990) examines the impact of local fiscal policy on the jurisdiction's borrowing costs. Using a sample of 243 bond issues made by New Jersey counties, towns and school districts in 1975-1977, Capeci finds that the amount borrowed per dollar of property value has a positive effect on the borrowing rate while the 11 level of discretionary revenues per dollar of property value has a negative effect. The relationship between his work and the research done in this dissertation will be discussed in the final chapter. In contrast to most of the other studies (including the work of Cunningham and of Capeci), the model used here incorporates directly the traditional assumption that state and local government general obligation bonds are issued to finance public capital expenditure. This study may be the first to use bond issue data separated by type (general obligation and revenue) in a cross-sectional analysis and may be the first to take advantage of a new revenue bond data set collected by the U.S. Treasury. CHAPTER I - FOOTNOTES 1. The data used in Tables 1.1 and 1.2 come from the following sources. The general obligation issues are from the files of the Securities Data Company. These issues are for "new money" so they do not include bonds issued for refunding. ‘Revenue bond issues are listed in "Private Activity Tax-Exempt Bonds" in.theIggag1§§19§_9j;lnggm§_§ullggig,'U.S. Department of the Treasury; Population figures are from the ati al bstract U.S. Capital expenditures are listed. in. nggrnm§n§__fiigaggg§, U.S. Department of Commerce, Census Bureau. 2. These summary statistics exclude Alaska due to its obvious outlier status. 12 CHAPTER II STATE AND LOCAL BORROWING COSTS II , A , Introduction . The effects of a jurisdiction's borrowing activity on its own costs of borrowing are described in this chapter. It is argued that because jurisdiction-specific factors may affect borrowing costs, the jurisdiction may not be a price taker in the market for municipal funds. It is also argued that the jurisdiction bases its borrowing decisions on a marginal cost of borrowing that may be greater than the observed or reported interest cost. In order to understand how a jurisdiction's financial policy can affect its costs of borrowing, it is important to distinguish between the general market effect of an increase in the quantity of state and local government bonds supplied and the additional effect of an increase in jurisdiction i's borrowing on its own cost of borrowing rm relative to the average tax-exempt rate rm. The general market effect occurs because an increase in the supply of municipal bonds increases borrowing costs for all municipal issuers. Evidence suggests that an increase in total state and local borrowing of $1 billion increases the tax-exempt rate by an amount ranging from a low of .37 basis points to a high of approximately 9.0 basis points. (Tuccillo and Weicher (1981), Kormendi and Nagle (1981), and Toder and Neubig (1985).) The magnitude of this market effect, however, suggests that a particular jurisdiction will not perceive its borrowing costs to be affected by its own borrowing in this manner 13 14 because the borrowing by any one jurisdiction will have such a small impact on the average rm. Recent evidence supports this assumption that the effect of a jurisdiction's own amount of borrowing on its own cost of borrowing dominates the general market effect. Capeci (1990) finds that a million dollar increase in the size of a particular bond issue is associated with an increase in borrowing costs of 3.5 basis points.1 (The previously mentioned estimates of .37 to 9.0 basis points were from a 2111193 dollar increase in total borrowing.) In effect, I claim that the ratio of jurisdiction i's interest cost rm to the average state and local government rate of rIII is a positive function of the level of borrowing undertaken by jurisdiction 1. In the next section, however, I make the simplifying assumption that the average rate rIII is unaffected by an increase in jurisdiction i's borrowing so that the ratio of borrowing costs (rm/rm) can be replaced by rm alone. I B Genera Obli ation a Reve ue n Interest Costs. To show clearly the manner in which the jurisdiction's borrowing costs are affected by its own bond-financing policies, I write the tax- exempt rate rm facing the ith jurisdiction as: Tm,Go - Tm,Go (G0,, R3,, w,) and rmiRB " mi.RB (601, R31: wi') where rum and rm” are jurisdiction i's interest costs on its general obligation issues and revenue bond issues, respectively. G01 is the level of general obligation bonds issued by jurisdiction 1 during a particular time period, RB, is the level of revenue bond issues, W1 is a vector of 15 variables representing the jurisdiction's credit worthiness and W1' consists of factors reflecting the likely profitability of the projects being financed.2 The effects of macroeconomic variables are less important in a cross-sectional study such as this one. Indeed, if the relevant interest cost variable were (rm/rm) as discussed above, then an increase in the general level of interest rates would leave the ratio unchanged. One important macro variable that will be included in W, and W1' is the state unemployment rate, because variations in the health of regional economies may have an important impact on run (as well as (rm/ta.) I assume that the costs of each type of borrowing (GO and RB) are positive functions of both types of borrowing. The next section is devoted to explaining these assumptions in detail. I nt e v w I claim that rum is a positive function of G01 and that rm“ is a positive function of RB,. It may also be true that rm°° depends positively on RB, and that run” may depend positively on G0,. I next explain each of these assertions in turn. Numerous studies (Hendershott and Kidwell (1978) , Leonard (1983) , and others) have found that interests costs are positively related to the size of the particular bond issue.3 This suggests that the jurisdiction is not able to issue any chosen level of bonds at a given market rate. One reason is the possible existence of regional segmentation in the market for state and local bonds. Hendershott and Kidwell find that an increase in bond issues from a particular state is associated with an increase in the interest costs on those bonds relative to bonds issued from other jurisdictions. Hendershott and Kidwell suggest that bonds that are 16 marketed regionally and bonds that are marketed.nationally are not perfect substitutes in the portfolios of investors. The variation across states in the tax treatment of state and local bond interest income may also serve to segment the market. Many states exempt from state taxation.the interest income that residents earn.on that state's bonds while at the same time taxing the interest income earned by the state's residents on.out-of-state (or "foreign") bonds. Kidwell, Koch and Stock (1984) find that this discriminatory taxation allows jurisdictions to issue bonds to state residents at a lower interest cost. Interest cost savings are more likely to be realized when the bond issue is relatively small due to the requirement that the marginal bondholder be a resident of the particular state. This is less likely to be true for large bond issues that are marketed nationally. If markets are segmented by state so that the pool of potential investors is small relative to that for bonds that are sold nationally, then an increase in the level of bonds issued may require that the jurisdiction increase the interest rate in order to induce more in-state buyers to hold the bonds. According to the above discussion of segmentation, however, the positive relationship between borrowing levels and interest costs that is caused by the state's tax policy is expected to exist only for regionally-marketed bond issues.‘ A second reason interest costs may depend positively on bond issue size has to do with the nature of the criteria used by the rating agencies. While the regional segmentation argument is relevant for regionally marketed issues that tend to be fairly small in size, this credit rating explanation applied to large, nationally marketed issues as well. Independent agencies such as Moody's Investors Service and Standard and Poor's Corporation provide ratings for virtually all municipal bond 17 issues. The ratings are intended to reflect probability of default and are based on the jurisdiction's willingness and capacity for timely repayment of principal and interest. The better the rating, the lower the jurisdiction's borrowing costs. While the rating process takes into account many diverse factors (some are included in the vector W, or Wi' to be discussed below), an important determinant of the credit rating will be the level of borrowing undertaken by the jurisdiction. Greater levels of new bond volume will be associated with lower credit ratings as the rating agencies express concern over the jurisdiction's capacity for servicing the debt. A third explanation for the positive relationship between borrowing costs and bond issues involves agency costs and/or reputation costs that may be associated with state and local debt financing. Gordon and Slemrod (1986) suggest that agency costs may serve to limit state and local borrowing. While they do not elaborate, it seems likely that they are referring to the fear bondholders may have that voters or bureaucrats may operate in a manner that will adversely affect the value of their claims.5 If agency costs are recognized by potential investors and these costs are a positive function of the level of borrowing, then jurisdictions will have to pay higher interest rates as the level of borrowing increases. A second way in which agency costs or reputation costs may affect borrowing decisions is more complicated. It is possible that the jurisdiction bases its borrowing decisions on a cost of borrowing that is comprised of the market interest rate plus a premium representing the marginal agency and/or reputation cost to the jurisdiction of additional debt finance. This premium is perceived only by the jurisdiction, and it causes the level of borrowing to be less than it would otherwise be at the 18 market rate of interest. This idea has been used in the analysis of corporate financial policy by Barnea, et a1. (1981). In contrast to the standard Miller (1977) analysis in which the firm is assumed to be able to supply any quantity of debt at a particular interest rate, Barnea, et a1. maintain that firms base their borrowing decisions on the sum of the actual rate of interest and a differential agency cost which is an increasing function of the volume bonds issued. In Barnea, et. al.'s work, the agency cost is not part of the actual interest to be paid. Instead, the firm is assumed to base its borrowing decision on the actual interest rate plus this added agency cost. Another factor that may play a part in increasing the jurisdiction's perceived cost of borrowing involves the importance of the jurisdiction's reputation. A bond default (such as the one by the Washington Public Power Supply System (WPPSS) in 1984) will certainly damage the jurisdiction's reputation and may increase current and future borrowing costs. Even though the jurisdiction may not be liable in case of a revenue bond default, a default (either on general obligation or revenue bonds) may be viewed by bondholders as revealing new information about the credit-worthiness of the jurisdiction. Epple and Spatt (1986) suggest that jurisdictions face rising reputation cost schedules as a function of the level of borrowing. Although the actual interest rate itself contains a default premium that investors require in order to hold the bonds, this reputation cost (like the agency cost in Barnea, et. a1.) is viewed by the jurisdiction as an additional cost of borrowing. Hence the jurisdiction.bases its borrowing decision on a cost of borrowing that includes both the actual rate of interest and a perceived reputation cost which is a positive function of 19 the level of borrowing. The discussion above suggest numerous explanations for the positive relationship between rm,"0 and G0,, and rm,RB and RB,. Several of these factors can also explain why rhf” may be a positive function of RB, and rmco a positive function of G0,. The segmentation arguments presented by Hendershott and Kidwell (1978) and Kidwell, et. a1. (1984) may imply that the interest costs associated with issuing general obligation bonds will increase with increase in the level of revenue bond issues and vice versa. If regional market segmentation exists, then the two types of'bonds (RB and.GO) issued by a jurisdiction may be viewed by the bondholder as closer substitutes for each other than are bonds issued.by other jurisdictions (particularly those in other states.) If so, then local holders of jurisdiction i's GO and/or RB bonds may have to be offered higher RB and/or GO yields in order to increase their holdings. Reputation effects also may be important. Epple and Spatt cite evidence that the WPPSS's default on $2.5 billion of revenue bonds has had an adverse effect on that state's GO borrowing costs. Because the jurisdiction. typically is not responsible for repayment of principal and interest in case of a revenue bond default, a default of this type should not have any impact on the jurisdiction's balance sheet and hence its credit worthiness. But because the revenue bond-financed project was approved by a jurisdiction official, investors may view the revenue bond default as evidence that the jurisdiction is being run by officials with poor financial judgement and/or as evidence of a downward turn in economic conditions.‘ The jurisdiction's concern for its good name in the market for general obligation bonds may serve to limit its revenue bond issues, and vice versa. 20 11,2, Other Fnctogg Affecting Integgst Qgsgg. The cost of borrowing also depends on jurisdiction-specific factors other than current borrowing levels. The financial health of the jurisdiction will also be important, because jurisdictions that are better credit risks can borrow at a lower interest rate. Potential bondholders require information on the overall ability of the jurisdiction to repay the interest and principal. Because it obviously would be too costly for each individual investor to collect the relevant information about all state and local debt issues, rating agencies help investors by providing information about the credit-worthiness of jurisdictions.7 The factors affecting borrowing costs are contained here in W, for jurisdiction i's general obligation issues and in W,’ for the jurisdiction's revenue bond issues. While many of these factors will be important for both types of borrowing, some of the factors may differ by borrowing type due to the nature of the borrowing contract. Because the jurisdiction itself is responsible for the repayment of the general obligation issues, the credit worthiness of the jurisdiction is an important determinant of the costs of borrowing. W, includes the factors in which rating agencies are particularly interested, such as the ratio of total general obligation debt to the taxable wealth in the jurisdiction, GO debt per capita, GO debt as a percentage of personal income. Because it is the proceeds of the revenue-bond financed project itself that are pledged to repay revenue bond debt, W,’ includes information about the expected profitability of the project. 11,5, Conclusion. Numerous explanations for the positive relationship between the jurisdiction's borrowing costs and its level of bond issues are presented 21 in this chapter. The borrowing cost facing the jurisdiction depends on its own level of borrowing in addition to the factors affecting the jurisdiction's credit worthiness. In corporate finance research, the introduction of endogenous borrowing costs serves to limit borrowing and results in a optimal debt/equity ratio for individual firms as well as for the corporate sector as a whole (Barnea, et. a1. (1981)). Similarly, the assumption that state and local borrowing costs are endogenous will help explain why the optimal share of debt in both the financing of state and local capital expenditures and private economic activity will tend to be less than 1001. The determination of the optimal debt share in the financing of capital expenditures will be discussed in the next chapter. CHAPTER II - FOOTNOTES 1. Actually, Capeci finds that an increase in the amount borrowed of $6.3 million is associated with an increase in borrowing costs of approximately 22 basis points. 2. The credit worthiness information in W is from the recent past. W represents the component of a jurisdiction's credit rating that is independent of current borrowing. It is W,’ rather than W, that affects nmfn because the potential bondholders are assumed to look through the issuing ,jurisdiction. to the credit-worthiness of the project being financed. The repayment of principal and the payment of interest on revenue bonds come from the proceeds of the project being financed. 3. Capeci also surveys studies that find an empirical relationship between bond issue size and interest costs. This has recently become a popular assumption as it is used by both Metcalf (1988) and Capeci (1990) . Metcalf takes the existence of this positive relationship as a given, while Capeci explicitly tests for it. 4. Cunningham (1989) repeatedly argues that the existence of this discriminatory tax policy implies that there are 50 separate markets in the U.S. for tax-exempt bonds. I claim that this is not necessarily the case, and in Chapter VI I will test the effect of this state tax policy on state and local government borrowing behavior. 5. For example, bondholders may fear that the jurisdiction may increase its reliance on debt financing in the future, thereby reducing the value of the current bondholders' claims. Jensen and Meckling (1976) explain that it is possible to write bond covenants to restrict future excessive borrowing (hence reducing the agency costs) but then the associated contract costs and subsequent monitoring costs may themselves be considered agency costs. 6. Actually, the State of Washington was later found to be liable for a small portion of the damages in the WPPSS case. Hence the increased general obligation borrowing costs that accompanied the WPPSS default may have been due to the ability of the investors to foresee the state government's financial loss. In general, however, the jurisdiction is not liable in the case of a revenue bond default. 7. R. Lamb and S. Rappaport (1987) provide a good discussion of the ratings process. 22 CHAPTER I I I THE SHARE 0F DEBT IN THE FINANCING OF STATE AND LOCAL CAPITAL EXPENDITURES I II , A , Introduction . In this section, I discuss the determinants of the portion of state and local capital expenditures financed through the issuance of general obligation bonds. Capital expenditures are financed through a combination of the proceeds from long-term bond issues, current taxation, and intergovernmental grants. I assume that a government official establishes the optimal debt share schedule as a function of current and expected future tax prices and the jurisdiction's cost of borrowing. The optimal debt share function will be announced to the residents of the jurisdiction who then use this information in making their spending decisions. The residents are assumed to act as though the specific debt share function established by the government official will remain in effect indefinitely. It is useful to think that this debt/tax choice is determined in a government agency where a government official selects the mix of financing methods that minimizes the cost of a dollar of per capita public expenditure to the residents of the jurisdiction. In this research I assume that the official operates as a dedicated civil servant who provides technical expertise in order to contribute to the efficient operation of the public sector. Because the residents are not homogenous with respect to their current and expected future tax prices, the official is assumed to establish the debt share schedule that minimizes the cost of 23 24 government spending to the median voter. The notion that residents will have preferences for either debt or tax finance assumes that capitalization is imperfect. This assumption is consistent with the summary of the evidence regarding the effect of local property taxes presented by Bloom, et a1. (1983). Many jurisdictions in the sample used in the current research rely on sources of revenues other than property taxes. Most importantly, it seems logical that parents living in.a multi-jurisdictional world are less likely to feel the need to make Barro-type bequests because their children. may live in other jurisdictions. I I e O t ma eb ar . I assume that the government official chooses the debt share that minimizes the median voter's price (at the margin) of the jurisdiction's capital expenditures. In a two-period model, the price P5 of a dollar of public capital expenditure I to a representative individual residing in jurisdiction 1 can be written: P1 - (l-h)tc + 11(tf)l>[1+I-‘...1](1H1).1 (1) where h - the bond-financed share of capital spending tc - the net cost (after deductibility) to the individual of a dollar of tax-financed capital expenditure tf - the net cost facing the individual in the future of a dollar of tax-financed capital expenditure p - the probability that the individual will be a resident of the jurisdiction in the future nu_- jurisdiction i's cost of borrowing 25 (l+d)’1- the discount factor used to calculate the present value of the individual's expected future tax liability due to debt finance As written, equation (1) represents the individual's average price of capital. .Assuming, however, that the jurisdiction issues its bonds all at one time rather than throughout the year, the average price will equal the marginal price. As discussed in the previous chapter, the interest rate nu_on jurisdiction i's general obligation bonds will be affected by the jurisdiction's borrowing activity so that the interest cost is a positive function of both hI and RB, where hI represents the level of general obligation bonds and RB represents the level of revenue bonds issued by the jurisdiction. The interest rate is also a function of W, a vector of attributes reflecting the jurisdiction's credit-worthiness. More generally, W can be thought of as the component of the jurisdiction's credit-rating that is unaffected by current borrowing. The expected future tax price is expressed as the product of p and t‘. It is assumed that p is exogenous - specifically, that residents do not.move in response to the debt/tax choices made in the various jurisdictions. Conversely, because the optimal debt share schedule is established once and then is assumed to remain in effect forever, a change in the level of migration into or out of the jurisdiction is not assumed to affect the debt/tax choice. I assume that the official chooses the value of h that minimizes the average cost of capital expenditures P, from equation (1) by equating the marginal costs of debt and tax finance: t‘p(l+d)'1[l + rm, + (6m,/6h)h] - cc . (2) 26 The marginal cost of debt finance (the left-hand side above) incorporates the fact that the interest rate increases as the reliance on debt increases. As previously discussed, the interest cost rm, is a function of hI, RB, and W. It will be useful to specify an explicit functional form for the jurisdiction's cost of borrowing. I make the assumption that rm, is a linear function of these three variables, so that the interest cost can be written rm, - ehi + fRB + gW.1 Because W consists of all determinants of the jurisdiction's cost of borrowing other than the current volume of bond issues, a jurisdiction that issues no bonds faces an interest rate of gW. Equation (2) can be rewritten as follows: t‘p(1+d)'1[1 + ehI + fRB + gW + ehI] - to (3) The optimal debt share is the debt share h that satisfies equation (3) . It can also be found by differentiating equation (1) with respect to h and setting 6P1/6h equal to zero. The optimal debt share can be written: cc - t‘p[1+fl?.B+gW](1+d)’1 h' - (4) 2t‘peI (1+d) '1 The official is assumed to communicate to the residents the optimal debt share for all possible levels of the tax prices and other parameters. The optimal debt share necessarily will be between zero and one, inclusive.z The formulation of the optimal debt share in equation (4) can be examined using comparative statics in order to generate testable hypotheses. The comparative static results are found in Appendix A. These results suggest 27 that differences in the debt share across jurisdictions will depend upon the following factors, where the signs in parentheses represent the effect of the variable on the optimal value of h: 1. the current tax price tc (+) 2. the future tax price tf (-) 3. the probability of the resident remaining in the jurisdiction p (-) 4. the amount of revenue bonds issued RB (-) 5. the level of state and local capital investment I (-) 6. the tax-exempt rate Em.(‘) 7. credit worthiness W (+) 8. the rate of discount d (+) Residents are expected to compare their current and expected future tax prices, where the term "tax price" refers to the price that an individual resident must pay for a dollar of per-capita state and local expenditure. Residents who must pay a relatively higher tax price t9 for a dollar of per-capita state and local capital expenditure in the current period are expected to prefer postponement of their tax liability until the future and consequently a higher debt share. On the other hand, residents who face a ‘higher future tax price t4 .and/or a greater probability of remaining in the jurisdiction in the future are expected to prefer a lower debt share. The probability p depends on the resident's future plans regarding geographic mobility and on the resident's view of his or her own life expectancy. The product of t‘ and p represents the resident's expected future tax price. Factors influencing the cost of borrowing are also expected to have 28 an effect on the debt/tax choice. The chosen debt share should be inversely related to the interest rate Em paid by jurisdiction 1 on its general obligation bond issues. As discussed in Chapter II, it is expected that the cost of issuing general obligation bonds Em will be a positive function of the quantity of both jurisdiction i's general obligation issues and its revenue bond issues. Because the level of general obligation bonds is equal to the product of the debt share and the level of the jurisdiction's capital investment, the cost of borrowing is assumed to be a positive function of the level of capital investment. Hence the debt share is expected to be inversely related to the level of public capital investment and to the level of revenue bonds issued by the jurisdiction. Other variables representing the credit-worthiness of the jurisdiction.are expected to have an impact on the chosen.debt share. The unemployment rate, for example, may be inversely related to the jurisdiction's ability to fulfill its debt-service requirements. Another example of'a credit-worthiness measure is the level of debt outstanding to the total wealth of the jurisdiction. This ratio can be expected to be negatively related. to jurisdiction's credit rating and. consequently positively related to am, Finally, the resident's rate of discount d is expected to have an effect on the chosen debt share. A higher discount rate implies that the future tax liability incurred because of debt financing of current expenditure will appear relatively ”smaller” to current residents, implying that a higher discount rate will be associated with a greater debt share . 29 I e I ta ce 0 Ob atio b ts. It is possible that the existence of binding statutory debt limits may affect the official's ability to select the optimal debt share according to equation (2). A common restriction is that debt is to be used only to finance capital expenditures. The evidence presented in Chapter I suggests that this type of constraint is not binding.3 Another common type of restriction is that debt should be limited to a certain percentage of assessed value in the jurisdiction. The assumption made in this research is that debt limits are not binding, an assumption that has also been made recently by Gordon and Slemrod (1986). The fact that borrowing levels doubled in 1985 before the new tax law took effect provides additional evidence that overall debt limits could not have been binding in the sample years of 1983-84. 111,2, anclngion. A common assumption in studies of state and local government borrowing is that jurisdictions finance capital expenditures through the issuance of general obligation bonds. Indeed, previous state and local public finance writers have noted a traditional "rule of thumb" that suggests that approximately 502 of state and local capital expenditures are to be financed through the issuance of long-term bonds (Peterson (1984) and Petersen (1981)) . Data presented in this dissertation suggests that the average debt share of state and local governments by state in 1983 and 1984 was less than 501 and that this debt share varied dramatically across states. The model of optimal debt financing developed in this chapter generates several testable hypotheses regarding the role of various 3O economic factors in the determination of the debt share. These hypotheses will be tested in Chapter VI. The level of general obligation bonds issued will be equal to the debt share times the level of capital to be financed. The determination of the level of capital investment will be discussed in Chapter V. It is important to note at this time that the chosen level of capital expenditure may depend on the debt share, just as the debt share depends on the level of capital expenditure. The next chapter focuses on the determination of the share of debt in the financing of state and local government support of non- traditional activity. In that chapter, the government official seeks to find the optimal mix of financing methods that will minimize the cost of providing a certain level of private-sector support. CHAPTER III - FOOTNOTES 1. The additive form of the interest rate equation is obviously a simplification. It will be suggested in this research that hI and RB are not independent of each other. An.increase in hI, for example, may lead to a decrease in.the chosen level of revenue bond issues. Because this effect is second-order in nature, the overall effect of an increase in hI on the interest cost Em. is still expected to be positive. This linear approximation would not be appropriate if the goal were to estimate the parameters e and f. That, however, is not the objective here. 2. An early explanation of the desirability of an interior solution for the optimal debt share problem is provided by Buchanan (1967). Buchanan argues that risk aversion on the part of the taxpayers will prevent the jurisdiction from selecting an all-debt financial policy. Buchanan's reference resident fears that other residents will fail to accumulate enough savings over time to pay off the jurisdiction's debt when it comes due. The awareness of this contingent liability leads the residents to prefer at least some tax finance. Buchanan also suggests that residents' uncertainty about their future income (and future tax bracket) will serve to limit borrowing. Buchanan also believes in the existence of ”asset illusion" under which residents systematically undervalue a long-lived capital investment. If asked to finance capital entirely in the current period, the residents would select an amount of capital investment that would be "too low.” Although Buchanan does not seem to believe so, it is possible that the bias caused by asset illusion could work in the opposite direction. If residents tend.to overvalue their future benefits from long-lived.capital, then 1001 tax finance might lead them to prefer too much. 3. Unfortunately, it is impossible to tell from the aggregate data if the constraint is binding in an individual jurisdiction. While Gordon and Slemrod (1986) have this same problem, they suggest that it is unlikely that any jurisdiction would be constrained by the debt limit. 31 CHAPTER IV THE USE OF STATE AND LOCAL REVENUE BONDS IN THE FINANCING OF NON-TRADITIONAL GOVERNMENTAL ACTIVITY IV ,A, Introduction. The use of revenue bonds in the financing of private sector development is discussed in this chapter. Because the jurisdiction can support private economic activity by offering a variety of fiscal incentives, the emphasis here is on the determination of the chosen composition of investment incentives. An extensive literature exists regarding the effects of state and local fiscal incentives on economic activity. For example, many researchers (including Bartik (1985) , Carlton (1979) and (1983), Papke (1987) and Sullivan and Newman (1988)) have examined the impact of the state tax climate on business location decisions. To date, the positive analysis of the jurisdiction's decision to provide investment incentives has not been emphasized. The analysis in this chapter attempts to provide an economic explanation of the chosen mix of investment incentives offered by state and local governments. While the chosen mix of financial incentives is one of several decisions that are made simultaneously, in this research the jurisdiction is assumed to make its tax abatements versus revenue bond issues choice as the third decision in a three-stage sequential choice process. The first choice is the optimal share of general obligation debt h in the financing of state and local capital expenditure. The second choice is the level of state and local capital expenditures 1. Finally, the jurisdiction is 32 33 assumed to take the information regarding the level of h and I into account while selecting the optimal combination of investment incentives to offer to potential investors. V e Choice tween Reven e Bo ds d Ot er Fiscal Incentives n e u art of n- ad t o Gove enta Activit Jurisdictions attempting to promote private sector development may offer a variety of incentives to potential investors. These development incentives include low-cost tax-exempt financing through the issuance of government revenue bonds, tax reductions such as property tax abatements and investment tax credits, direct grants and loans, and subsidies in the form of worker training programs and management assistance. The objective of jurisdictions offering development incentives such as revenue bonds and tax abatements is to induce firms, organizations or individuals to locate new facilities or expand old ones within the jurisdiction. The choice among the different incentive tools is essentially a debt/tax choice analogous (to an extent) to the one involved in the financing of state and local public capital investment. Tax abatements, direct grants and loans, and management and training programs each impose costs on taxpayers in the year in which they are granted. As a result, residents end up paying higher taxes in order to pay for these incentives. Because of the similar nature of the costs these incentives impose on taxpayers, in the discussion that follows I will combine all of these tools together and refer to them as ”tax abatements." The choice to be made by the jurisdiction is the share of revenue bonds and tax abatements in the provision of the subsidy. While residents pay directly for tax abatements in the form of higher 34 current taxes, revenue bond issues do not lead to a reduction in tax revenues collected nor do they have a direct impact on government expenditures.1 By issuing revenue bonds, the jurisdiction serves as a conduit for funds to the private sector. The jurisdiction is not responsible for the payment of the interest and principal on revenue bond debt. Instead, the profits or the proceeds of the project being funded are used to pay the debt service requirements. In this chapter, I assume that the desired level of investment Aincentives is exogenous. .As mentioned.previously, the emphasis here is on the determinants of the mix of incentives chosen by the jurisdiction. There are two main issues to address. The first is the nature of the production function for state and local government support of private economic activity. Jurisdictions tend to offer a package of various fiscal incentives to potential investors, and I suggest that there may be some degree of substitutability among incentive types. The second issue is the nature of the relative prices to the jurisdiction of the different types of incentives. I claim that all incentives are costly, and the chosen mix of incentives depends on their relative prices. The standard view that tax abatements are costly while revenue bond issues are costless would lead one to expect jurisdictions to place little or no emphasis on the use of incentives other revenue bond issues. This is not what is observed, and I suggest that the issuance of revenue bonds does impose a cost on the jurisdiction. The chosen mix of tax abatements and revenue bond issues is modeled as though the jurisdiction is a cost-minimizing firm choosing the optimal combination of inputs to produce a certain output level. The financial officer of the jurisdiction is assumed to operate in the interest of the 35 median voter. The chosen mix can be shown graphically as the point of tangency between an iso-subsidy curve and an isocost line. IVC G ca 11 s t n fte t e Share. The subsidy provided by the jurisdiction to the private sector is proportional to the value of the revenue loss estimates of the tax abatements Z and to (r,, - rm) RB. This latter expression represents the interest cost saving in jurisdiction 1 from borrowing an amount RB at the tax-exempt rate rather than at the higher taxable rate r,,. The level of the subsidy S provided by jurisdiction 1 is written: 31 ' S (Zi,(rti ' mi)RBi) (1) Because the tax-exempt rate r,,,, is an increasing function of RB,, an increase in revenue bond issues will have the effect of reducing the interest cost savings associated with revenue bond issues. Hence the iso- subsidy curve showing possible combinations of tax abatements and revenue bonds is convex to the origin. The convexity is due also to the varying marginal rate of substitution between the two inputs. At a point on the iso-subsidy curve where the jurisdiction grants a large amount of tax abatements (2) relative to its issuance of revenue bonds (RB), it is possible to trade a large amount of Z for an additional unit of RB holding constant the level of the subsidy. As the jurisdiction begins to rely more heavily on revenue bonds, the issuance of additional revenue bonds does not permit a very large reduction in the use of tax abatements. The diminishing rate of substitution of RB for 2 will arise if the different incentives are not equally valued by the recipients of the subsidy. For 36 example, only firms with taxable income will have any use for tax abatements. Similarly, revenue bonds will be of greater use to firms which finance their activities by borrowing. In Figure 4.1, the subsidy curve S shows the combinations of tax abatements and revenue bonds that could be offered by the jurisdiction to provide a private-sector subsidy of $100,000. Assuming an initial yield differential of 21, a jurisdiction interested in financing private development solely through the use of revenue bonds would issue $5 million worth of revenue bonds. If half of the subsidy ($50,000) were to come from the use of tax abatements, the jurisdiction would need to issue a quantity of revenue bonds less than $2.5 million to provide the additional subsidy equal to $50,000. This is becauseinn_is assumed to fall as revenue bond issues fall, so the yield differential will increase. If the yield differential increased from 21 to 2.5 1, then the quantity of revenue bonds required to provide a $50,000 subsidy would equal $2 million. As shown, in. Figure 4.1, the iso-subsidy curve 8 showing the possible combinations of tax abatements and tax-exempt revenue bonds is convex to the origin. The total cost to the jurisdiction of granting the fiscal incentives can be represented by an iso-cost line. The cost of tax abatements Z is merely the revenue loss estimates of these abatements. Residents incur this cost in the form of higher taxes. The cost of revenue bond issues is the increase in borrowing costs for the jurisdiction's general obligation bond issues. Although increased revenue bond issues reduce the interest cost saving (r3, -‘nu) to revenue bond financed projects, this effect of increased revenue bond issues on the subsidy is already incorporated into the iso-subsidy curve. 37 Tax Abatements 2 $100,000 50,000 $2 2.5 5 million Revenue Bonds RB Figure 4.1 The iso-subsidy curve is convex. 38 The total cost of the fiscal incentives granted by jurisdiction 1 is written: C1 - [(6rm/5RBi)G01] ”1 + Z, (2) where (6rm/6RB,)GOi is the cost of a dollar of jurisdiction i's revenue bond issues. As discussed in Chapter III, it is convenient to represent the jurisdiction's cost of borrowing by the expression rm,(hI,RB,W) - ehI+fRB+gW, so that (6rm/6RB) is equal to f. The slope of the iso-cost line is equal to -fGO,, or -fh,I,. As discussed in Chapter II, an increase in RB, will increase the cost of jurisdiction i's general obligation bond issues. While this specification does not include the administrative costs of the revenue bond issues or tax abatements, these costs could be incorporated easily. The cost-minimizing framework for analyzing the determinants of the optimal share of revenue bonds in the financing of private-sector development assistance is useful because some of the determining factors will vary by jurisdiction. The main factor that is expected to vary by jurisdiction is the level of general obligation issues h,I,. One prediction is that the share of (revenue bond) debt in the financing of the subsidy will be inversely related to the share of (general obligation) debt in the financing of the public capital expenditures and also to the level of public capital expenditures. This is due to the effect of the debt share h and the level of capital spending I on the jurisdiction's cost of borrowing. Recall from Chapter II that an increase in the quantity of general obligation bonds (hI) issued by the jurisdiction is expected to have a positive effect on the interest rate on 39 the jurisdiction's revenue bond issues. In order to illustrate the effect of an increase in the debt share h and/or the level of public capital I on the jurisdiction's chosen mix of investment incentives graphically, let C0 and So in Figure 4.2 be the original iso-cost and iso-subsidy curves. Point A shows the cost-minimizing mix of investment incentives to achieve the level of subsidy S°. An increase in h, (or 1,) will affect the slope of both curves. C1 and S1 are the new iso-cost and iso-subsidy curves. Point B shows the point of tangency between the new curves. The iso-cost line becomes steeper as h, (and/or 1,) increases because the revenue bond issue price depends positively on the level of general obligation bond issues. An increase in h, or I, also affects the slope of the iso-subsidy curve by making it flatter. Both of these effects work in the same direction so that an increase in h, and/or 1, will unambiguously lead to an increase in the share of tax abatements in the financing of the subsidy activity. In addition, it is also possible that the current tax price will also be positively related to the share of debt in the financing of the subsidy. Because the tax price is really the price to the residents of the jurisdiction of a dollar of per capita tax abatements, an increase in the tax price will make the isocost line flatter and will lead to an increase in the share of revenue bond issues in the financing of the subsidy. Finally, it should be noted that the derivative of the tax-exempt rate with respect to the level of revenue bond issues is expected to be inversely related to the share of debt in the financing of the subsidy. If the tax-exempt rate is very sensitive to the level of revenue bonds issued by the jurisdiction, then a jurisdiction hoping to keep borrowing 40 Tax Abatements Z Revenue Bonds RB Figure 4.2 The effect of a change in h and/or I on the cost-minimizing combination of Z and RB 41 costs low will want to rely more heavily on tax abatements in the financing of the subsidy. Lacking information to the contrary, however, this factor is not assumed to vary by jurisdiction. IV,D, Algebraic Eonnulagign of the QeQEZIax Choice. The economic factors affecting the jurisdiction's reliance on debt financing rather than tax financing of non-traditional government activity can 'be illustrated algebraically in the standard cost-minimization framework. Consider the problem where the level of the subsidy S is modeled using a Cobb-Douglas production function: S - Z°‘[(RB)(r,,-rm,)]1'°‘ (3) The graphical analysis in the previous section illustrates that the cost- minimizing combination of tax abatements Z and revenue bonds RB occurs at the tangency of the iso-subsidy curve and the iso-cost line. At that point, the marginal rate of technical substitution of RB for Z is equal to the ratio of the relative prices of RB and 2. Hence the cost-minimizing combination of Z and RB can be expressed by the following equation: (l-a)Z(r,, - ehI - gW - 2fRB) — m (4) aRB(r,, -ehI - gW - fRB) The left-hand side of the expression above represents the MRTS of RB for 2. It incorporates the assumption made in the previous chapter that the interest rate rm, is equal to ehI+fRB+gW. The right-hand side represents 42 the ratio of relative prices th (the price of one dollar of RB) and one (the price of a dollar of 2.) It is possible to derive factor demand functions for RB and 2. These demand.functions can be obtained by minimizing the total cost of providing a certain level of the subsidy. The problem is as follows: minimize C(RB,Z) - (RB)th + Z subject to: S - 2"‘[(R13)(rt1 - rm)]1-a After replacing r,,,, by ehI+fRB+gW and then substituting the term m for r,,-ehI-gW, the subsidy can be written as S - 2°‘[ (RB)m- (RB)2f]1’“. A consequence of this cost-minimizing behavior is that the first partial derivatives of the following Lagrangean function equal zero. L - (RB)th + z + u(S - Z“[(RB)m-(RB)2f]1-a (5) The following first-order conditions can be obtained by partially differentiating the Lagrangean function by RB, Z and the Lagrangean multiplier u. Lz - 1 - uaZ“’1[(RB)m- (RB)2f]1‘°‘ LR, - th - uZ“(l-a)[(RB)m-(RB)2f]'°‘[m-2(RB)f] L. - s - Z“[(RB)m - (mm.-. 43 Given that the sufficient second-order conditions for a minima are satisfied, the factor demands for revenue bonds RB and tax abatements 2 can be obtained from the first-order conditions.2 The factor demand for revenue bonds is : RB' - (2f)'1{m + 2fQ - [mz+(2fQ)2]'5} (6) where as previously stated m - r,,-ehI-gW. The expression Q is written as Q - S(l-a)“(tha)”“. The comparative static results are shown below for the following factors, where the signs in parentheses represent the effect of the variable on the jurisdiction's demand for revenue bonds: 1. the level of state and local public capital investment I (-) 2. the share of general obligation debt h in the financing of I (-) 3. the credit-worthiness of the jurisdiction W (+) Of course, these results are the same as those obtained in the graphical analysis. Holding the level of the subsidy S constant, the comparative static results indicate that an increase in I will lower the chosen share of RB debt in the financing of non-traditional government activity. Similarly, an increase in the general obligation debt share will lower the RB debt share. Finally, jurisdictions that are a better credit risk are expected to issue relatively more revenue bonds. IV . e elation h etwee he Two Debt ax C o ces. If the total amount of subsidy activity is assumed to be constant, then the share of debt in the financing of state and local capital 44 expenditure and the share of debt in the financing of state and local government assistance in support of private sector development will be inversely related. But to the extent that both debt/tax decisions represent preferences for paying now for government spending versus paying later, it is possible that the two debt shares may be positively related. In the case of the financing of public capital expenditure, residents compare the current tax liability' with their expected future tax liability. The future tax liability depends on the probability that the resident will be a resident of the jurisdiction in the future. Assuming imperfect capitalization, residents who do not have an operative bequest motive or do not have children who will be living in the jurisdiction in the future can escape payment of the future debt service by dying or moving to other jurisdictions. The nature of the expected future liability from debt finance is somewhat different in the case of the financing of state and local aid to the private sector because the repayment of the revenue bond debt service does not come from the general pool of tax revenues. Many revenue bond- financed projects use user fees such as medical service fees in the case of hospitals, tuition and board in the case of colleges, and tolls and entrance fees in the case of roads, bridges, stadia and parks to pay the debt service requirements. The resident's liability will depend on his usage of these bond-financed goods and services. For bonds issued to finance private business, the debt service payments may be passed on to consumers of the good in the form of higher prices, the workers in the form of lower wages, or the shareholders. In many cases the economic incidence will not be clear. Because the nature of the taxpayer's future liability for revenue 45 bond debt service is fundamentally different from the taxpayer's future liability for general obligation debt service, the analogy between the two debt share decisions is not a close one. To the extent, however, that taxpayer preferences for current financing rather than future financing of government activity may apply to both the financing of public capital and the financing of aid to economic development, then the preferences for debt versus tax finance may be positively rather than negatively related. IV F Do State and Local overnme ts ubsidize Private Activ t ? I assume that the jurisdiction's objective in pursuing economic development activities is to increase the income of the median voter of the jurisdiction. Presumably the individual's income will be affected positively by these development activities, and can be written: Y - f(S), where S - s[Z,(r,, - rm,)RB] (7) Z is the amount of tax abatements granted by the jurisdiction, and (rt, - rm,)RB is the subsidy provided to those entities that borrow an amount RB at the tax-exempt rate rm, rather than the higher taxable rate r,,. It is interesting to note that while increased revenue bond issues will increase the income of the median voter, an increasing quantity of revenue bonds supplied also will increase the jurisdiction's borrowing costs on both GO and RB issues. The jurisdiction is then faced with a tradeoff between its various goals: its desire to increase incomes, to keep rm, low in order to provide a sizeable subsidy (through the issuance of revenue bonds) to non-public economic activities, and to keep the costs of borrowing for traditional public purposes low as well. 46 An assumption made in this research is that the level of subsidy activity undertaken by each jurisdiction is exogenous. In equilibrium, the desired level of subsidy activity can be thought of as being beyond the control of the individual jurisdiction. For example, suppose that in equilibrium all jurisdictions experience the same level of economic growth from one year to the next. In order for this equilibrium to be attained, some jurisdictions are going to have to offer a larger amount of private- sector subsidies than are other jurisdictions. Jurisdictions that have greater amenities (such as a cheaper labor force or closer proximity to transportation centers) will be able to achieve this equilibrium level of economic growth by offering little or no subsidies. While different jurisdictions will engage in different levels of subsidy activity, the level of this activity chosen by the jurisdiction can be viewed as exogenous and will be denoted as 8*. As a consequence, the level of private incomes in the jurisdiction also can be viewed as exogenous. These assumptions will be used in the complete model of borrowing presented in Chapter V. IV,G, The Innogtange of Genernl Obligation Debt Linits. As discussed in.Chapter III, jurisdictions may issue revenue bonds in order to circumvent the general obligation debt limits if state and local constraints on general obligation bond issues are binding. While the importance of these borrowing limits is debatable,3 if constraints are binding there may be a positive relationship between the level of general obligation issues and. revenue ‘bond. issues, instead. of' the 'negative relationship suggested above in Section C. 47 IV,H, Conclusion. By issuing revenue bonds, the jurisdiction is often serving as a conduit for funds to the private sector. Revenue bonds, however, are just one of several fiscal incentives that can be offered by jurisdictions attempting to promote economic development. Although many researchers have investigated the effect of state and local fiscal incentives on private economic activity, the analysis presented in this chapter may be the first to provide an economic explanation of the jurisdiction's chosen mix of incentives. I suggest that the jurisdiction's choice can be modeled in a manner analogous to the firm's cost-minimizing combination of inputs. Assuming that there is some substitutability between the incentives, an important economic determinant of the jurisdiction's chosen mix of incentives will be the relative costs of each incentive type. It is commonly observed that jurisdictions offer a package of fiscal incentives to potential investors. This observation, however, seems inconsistent with the claim often made that revenue bonds can be offered at no expense to the jurisdiction itself. If tax abatements are costly but revenue bonds are virtually without cost, one would expect that jurisdictions would place little or not emphasis on the use of tax abatements. Moreover, the view that jurisdictions perceive revenue bonds as costless seems to imply that revenue bond issues should be at much greater levels than are observed. The logical question to ask is, "What is limiting the use of revenue bonds in equilibrium?” I argue that the issuance of revenue bonds does entail a cost to the jurisdiction. An increase in revenue bonds may increase the cost of issuing general obligation bonds and hence the cost of state and local government capital expenditures, and an increase in revenue bonds reduces 48 the effectiveness of the revenue bond as a subsidy because the yield differential (rt, - rm) will decrease. CHAPTER IV - FOOTNOTES 1. The issuance of revenue bonds is not done without cost. In addition to the administrative costs of the bond issue, it has been suggested in Chapter II that an increase in revenue bonds does increase the cost of issuing general obligation bonds and consequently the cost of public capital. I view this interest rate effect as an indirect cost. 2. The expression for the demand for RB actually consists of two equations because solving the first-order conditions for RB involves a quadratic equation. Only the demand equation presented in the text represents a solution to the cost-minimization.problem. The existence of two possible solutions is due to the specific functional forms chosen to represent S and r5,. The subsidy curve starts to slope up at high levels of revenue bond issues when (m-2RBf) switches sign from positive to negative. 3. Gordon and Slemrod (1986) suggest that it is unlikely that debt limits based on the level of capital expenditures are binding. While direct evidence is difficult to obtain, the fact that borrowing levels doubled.in 1985 before the tax law took effect suggests that overall debt limits were not binding in the sample years of 1983-84. 49 CHAPTER V A MODEL OF STATE AND LOCAL BORROWING V,n, Introducgion. Because borrowing is only one method of financing capital expenditures and economic development assistance, the level of borrowing undertaken by the jurisdiction is determined by the chosen mix of financing methods and the chosen level of the activity being financed. This chapter incorporates the discussions of the interest cost and.optimal debt shares from.the three previous chapters into a comprehensive model of state and local borrowing. In this model, three important decisions are made. First, the government (acting on behalf of the median voter) determines the optimal share of debt finance associated with any desired level of state and local capital expenditures. This process was discussed in Chapter III. At the same time, the median voter selects the level of state and local capital spending. In addition, the government also makes a choice between tax abatements and revenue bonds in the financing of economic development activity. The nature of this decision was discussed in Chapter IV. While the quantities of general obligation bonds and revenue bonds issued are determined simultaneously, I approximate this decision-making process with a sequential choice framework. The optimal debt share for capital expenditures is determined in advance for all possible levels of prices, income, capital spending, revenue bond issues, and other parameters. With this information available, the median voter then 50 51 expresses his or her desired level of capital spending as a function of the parameters of the model. Finally, the optimal mix of financial incentives in support of private activity is determined, In this chapter, the median voter expresses his or her preferences for state and local capital investment taking into account the optimal debt share information presented in Chapter 111. As the optimal debt share and the voter's demand for capital are both a function of the jurisdiction's revenue bond issues, the final stage of the model involves choosing the level of revenue bonds using the cost-minimizing framework outlined in the previous chapter. V,§, Inn Individual's Ugility Function. A common approach to modeling voters' demands for state and local expenditures is to assume that the preferences of a particular individual determine the level of spending in a given jurisdiction. Typically, the decisive voter is assumed to be the voter whose quantity demanded of publicly-provided goods and services is the median quantity demanded. The level of spending preferred by the median voter will (under certain restrictive assumptionsl) defeat any other level of spending in a majority-rule election. In this research, the individual's utility is represented by a utility function: U - U( K, EXP, X), (1) where the terms in the utility function are defined as follows: KL - the flow of services from the stock of tangible capital possessed.by states and localities 52 EXP - the level of state and local current expenditures X - a composite bundle of private and federal goods and services consumed by the individual. The specification of the individual's utility function used here is similar to the government preference function used by Gramlich and Galper (1973). An alternative specification of the utility function might include "state and local goods and services," which would then be produced with both state and local capital and.non-capital expenditures as inputs. The approach used here is more direct. I next describe the terms in the utility function in more detail. - Inn stock of tangible state and local capital K, The individual prefers a greater flow of services from the stock of state and local capital. If this flow of services is proportional to the actual stock of capital, the utility from this source is also a proportional function of the stock. The capital term in the utility function is proportional to: K - (l-d)K_,-+ I, (2) where I is the level of current capital expenditures, d is the rate of physical depreciation, and K_, is the capital stock that existed in the previous period. K represents the desired level of the capital stock in the current period. It will be assumed that jurisdictions can adjust the size of their capital stock:within.a year's time so that the desired level K can always be attained. 53 - Statg and local nunrent expenditnneg EXP. The individual prefers a greater level of non-capital state and local expenditures denoted as EXP. These expenditures include service on past debt as well as any type of expenditure other than current investment expenditures I. - Enngndiguges on otne; goods X. The individual prefers a greater level of consumption of private and federally-provided goods and services. V Th nd vidua '3 ud e onstra n . The individual chooses the amount of each good. to consume ‘by maximizing utility subject to the following budget constraint: Y -' P11 + PzEXP + P3X , (3) where Y - the individual's pre-tax income P,=- the price to the individual of a dollar per-capita addition to the stock of state and local capital P2 - the price to the individual of a dollar per-capita state and local current expenditures P3 the price to the individual of a dollar of private and federal goods and services (assumed to equal one.) As discussed in Chapter IV, income can be affected positively by the jurisdiction's economic development activity. In equilibrium, however, the level of subsidy activity undertaken by the jurisdiction is exogenous. (The jurisdiction does control the composition of the subsidy between revenue bonds and tax abatements.) As a consequence, the level of incomes in the jurisdiction is exogenous. 54 V u tio ' e o The government's budget constraint also can be expressed in a sources versus uses framework. The sources of funds include total tax revenue T, grants G, and the proceeds from general obligation borrowing hI. Government funds are used for current expenditures EXP, capital investment I, and the provision of tax abatements Z.2 The government's budget identity can be written as follows: T + hI + c - EXP + I + z (4) The level of tax revenues that can be spent on current expenditures EXP and capital investment I is actually T-Z. Hence an increase in 2 requires reduced expenditures on EXP and I. The individual and governmental budget constraint can be combined after recognizing that the individual's share of per capita tax revenues T is tCT, which is equal to the sum of P,I and PZEXP. The individual's budget constraint becomes: Y-X+t°T (5) Incorporating the budget constraint shown in equation (4) into equation (5), the combined budget constraint is written: Y - x + t°[EXP + I(1-h) - c + 2] (6) Income is spent on private goods X, current government expenditures EXP, the tax-financed portion of public capital investment I, and the provision of tax abatements Z. 55 V. er ti n the Medi Vot r' a tions. The median voter maximizes his or her utility subject to the two budget constraints (6) and (7). The utility function used here is a Cobb-Douglas utility function: U ( K, EXP, X ) - jan + mlnEXP + nlnX (7) where j, m and n are demand parameters (j+m+n~l.) The budget constraint does not include the price of purchasing the entire desired capital stock K -- rather, the amount of capital that must be purchased each period so that the actual capital stock equals the desired capital stock is I - K - (l-d)K.,, where K., is the capital stock that existed in the previous period and d is the constant rate of depreciation. The individual is assumed to choose his or her desired level of capital K, EXP, and.X, knowing the particular values of'h and.RB that will be associated with these demands. The demand equations for K, EXP, and X can be obtained from the first-order conditions given that the sufficient second—order conditions for a maxima are satisfied. The equation that is of most interest for determining general obligation borrowing levels is the demand for investment I, which is actually a demand derived from the individual's demand for capital K. I assume that the median voter maximizes utility shown in equation (7) subject to the following constraints. (V.El) Y - x + t¢[EXP + 1(1-h) - G + 2] (V.E2) h - h' (from Chapter III) (V.E3) s - s* (from Chapter IV) 56 (V.E4) MRTSRB’Z - Pius/P2 (from Chapter IV) (V.ES) K - I + (1-d)K-, This maximization problem can be simplified by substituting the K term in the utility function by constraint (V.E5). The details of the maximization are found in Appendix A where the first-order conditions are derived. It is theoretically possible to solve the system of equations for the demands for I, EXP, X, RB, and 2. As mentioned above, the level of general obligation bond issues is a function of the demand for investment I. The demand for I, however, is a cubic equation. Due to the complexity of the expression for I, the investment demand equation estimated in Chapter VI will not be the actual non-linear demand function derived from the maximization process. Instead, I let the investment demand portion of the model suggest which variables are the likely determinants of investment decisions and I test for their significance. VI vel e rs b at ssue . The level of general obligation bonds issued by the jurisdiction is equal to the product of the level of public capital investment I and the debt share h. The economic factors influencing the level of investment I are as follows: 1. the income of the median voter 2. the cost of borrowing rh, 3. the size of the capital stock in the previous period 4. the current tax price 5. the probability that the median voter will be a resident of the jurisdiction in the future 57 6. the future tax price 7. the quantity of tax abatements offered 8. the quantity of revenue bond issued 9. the optimal debt share As stated above, the complexity of the expression of the demand for new investment prohibits the use of comparative statics to derive hypotheses regarding the effect of a change in a particular factor on the quantity of investment demanded. It seems logical, however, to make the following conjectures. New investment is expected to be positively related to income, assuming capital is a normal good. Investment is also assumed to be negatively related to the cost of borrowing. Investment should also be negatively related to current and future tax prices and the optimal debt share. An increase in the debt share will increase borrowing costs and hence tend to reduce the quantity of I demanded. Finally, it is interesting to examine the likely effects of an change in the mix of revenue bonds RB and tax abatements 2 used to finance the equilibrium level of subsidy activity. An increase in the reliance on RB relative to 2 will tend to increase the cost of issuing general obligation bonds to finance public capital investment, thereby reducing the quantity of investment I demanded. However, this negative effect will be mitigate somewhat because the resulting reduction in Z (holding the level of subsidy activity constant) will tend to increase the amount of funds that the jurisdiction has available to spend on both I and EXP. As shown in Chapter I, the variation in general obligation bond issues across states in the sample period 1983-1984 is driven largely by 58 the variation in the debt share h. The factors influencing the debt share are derived in Chapter III from a cost-minimization process where the debt share that minimizes the price to the median voter of a dollar of capital expenditure is determined for each level of capital expenditure. Here, the hypothesized effect of a change in each factor on the optimal debt share in the result of applying comparative statics to the optimal debt share equation. Once again, the factors and their expected signs are: 1. the current tax price tc (+) 2. the future tax price tf (-) 3. the probability of the resident remaining in the jurisdiction p (-) 4. the amount of revenue bonds issues RB (-) 5. the level of state and local capital investment I (-) 6. the tax-exempt rate r5, (-) 7. credit worthiness W (+) VI G v o Reve u ue . Just as the volume of general obligation bonds is equal to the chosen debt share in the financing of capital investment times the quantity of capital investment, the volume of revenue bonds is equal to the chosen debt share in the financing of nontraditional government activity times the quantity of this activity; Although the analyses of the two types of ‘borrowing appear conceptually to be symmetrical, in.this research they are IKflL One assumption made is the level of nontraditional governmental activity S is exogenous while the level of public capital investment is exogenous. Data-availability problems also are to blame for the asymmetrical treatment of the two borrowing decisions. As will be 59 discussed.in Chapter VI, no data exist regarding the level of S undertaken by the jurisdiction. As a result, the dependent variable of interest in the empirical analysis will be the level of revenue bond issues rather than the components (the debt share and S) individually. The level of revenue bond issues will depend on factors affecting the revenue bond debt share and the level of S. In contrast to the study of general obligation issues, however, both the revenue bond debt share and the level of S are unobserved. Variables that are assumed to affect the level of revenue bond issues are those that influence the jurisdiction's reliance on revenue bonds versus tax abatements as well as factors that are correlated with the level of the activity being financed. The model suggests that two of the factors affecting revenue bond issues will be the general obligation debt share h and the level of capital investment I. V u o . This analysis in this chapter combines the various aspects of state and local borrowing decisions presented in Chapters II-IV. The decisions made by the jurisdiction to issue general obligation bonds and revenue bonds depend on both the level of the activity being financed and the method of financing that level of activity. The purpose of this research is to identify and then estimate the significance of the various determinants of each of these decisions. I suggest that these decisions are linked in such a way that the method of finance (i.e. the debt/tax choice) may depend on the level of the activity being financed and vice versa. I also suggest that the jurisdiction's decision to issue one type of bond is related to its decision to issue the other type. The main reason 60 these decisions are linked is due to the nature of the jurisdiction's borrowing costs. An increase in revenue bond issues, for example, may increase the interest cost associated with issuing general obligation bonds. Chapter' II provides several explanations for this positive relationship between the level of one type of bond issues and the cost of issuing the other type. Chapter'VI investigates the importance of the various determinants of state and local government borrowing using the framework presented in this and.previous chapters. Equations for the debt share h in.the financing of capital investment, the quantity of capital investment I, and.the level of revenue bonds will estimated in order to investigate empirically the significance of the hypothesized determinants. CHAPTER V - FOOTNOTES 1. The fact that three decisions are being made may suggest that the median voter model is not applicable due to its requirement that the choice being made is uni-dimensional (i.e. more or less of a particular good.) My assumption that the choice process is sequential rather than simultaneous will alleviate this problem. Further evidence of the appropriateness of applying the median voter model to state and local borrowing decisions is provided by DeBartolo and Fortune (1982). They find that the level of general obligation bonds issued in their sample of Ohio communities is consistent with the level preferred by the median voter. 2. The jurisdiction's use of funds also could be expanded to include additions both to the budget surplus SURP and to the stock of state and local financial capital FIN. For simplicity, I consider both variables to be part of current expenditures EXP. 61 CHAPTER VI ESTIMATION VI A Introduction. The model provides a useful framework for analyzing the determinants of the variation across jurisdictions in the levels of both general obligation and revenue bond issues. In this chapter, the procedure and the data used to estimate these determinants are described in detail. The economic factors affecting the capital investment decision are estimated first. In the following section, estimates of the economic determinants of the share of' debt in the financing 0 ate 2t£peI 6h - - t°(1+d) < 0 6tf 2 ( t‘)2pe1 106 107 6h - - t°(i+d) < 0 6p 2t‘p2eI db = tc > 0 6d 2t‘peI 6h - -f < O ORB 2eI __5.h_ = _;g._ > 0 6W 2eI 6h - - [tc - t‘p(l+fRB+gW) (I+d)'1] < 0 (SI 2t‘peIz(l+d)'1 The first five results are obvious, but the last two results warrant additional discussion. An increase in the credit—worthiness of the jurisdiction W is expected to lead to an increase in the chosen debt share. The comparative results show that the effect of a change in W on the chosen h depends on g. The term g represents the effect on the cost of borrowing of an increase in W. The effect on h of a change in W is positive because the term g is negative. The effect of an increase in the level of capital investment I will be inversely related to the chosen debt share if the term in brackets in the numerator in the comparative statics expression above is positive. This expression in brackets is actually the numerator in equation (A.2) above. The fact that h is required to be a non-negative number and is typically observed to be non-zero as well implies that term in brackets is positive. Hence, the effect on h of an increase in I is expected to be negative. Chapter V describes the median voter's selection of his or her utility-maximizing bundle consisting of three goods: the desired public capital stock K, the level of current (non-capital) expenditures EXP, and a composite good X. The problem is as follows: maximize U (K, EXP, X) - KJEXPmXu 108 subject to: (A.4) Y - x + c°[EXP + I(1-h) - G + 2] (A.S) h - h' (as shown in equation A.2) (A.6) s - S" (A.7) MRTSRBJ - PRB/Pz (A.8) K - I + (1-d)1(-1 The first constraint represents the combined budget constraint.of the individual and the jurisdiction. The individual spends his or her income Y on the goods EXP and I that are provided by the state or local jurisdiction and also on the composite good X. The jurisdiction's budget constraint is written T + hI + G - EXP + I + 2, which states that the jurisdiction's sources of funds T, hI, and G must equal its uses of funds EXP, I, and Z. The tax revenue T received by the jurisdiction is paid by the residents for their consumption of EXP and I. The individual's share of the required.taxes equals tqr. Hence the individual's budget constraint can be written as Y - X + tCT. Equation (A.4) combines the two constraints. The individual chooses the utility-maximizing combination of K, EXP and X knowing that any addition to the capital stock.will be financed.with the optimal mix of debt and tax finance as represented by h'. Hence the h term found in (A.4) will be the optimal hf as shown in (AA2). The cost of borrowing rm can be replaced by ehI+fRB+gW. Because the model suggests that borrowing in a particular period will depend on the level of new investment I rather than the capital stock.K, constraint (A.8) is used to replace the K term in the utility function with a function of investment I and the exogenous level of the past capital stock. Because the desired.demand.expressions for the choice variables I, X, RB and 2 will be functions of the exogenous variables in the model, it is necessary to incorporate the exogenous level of the subsidy activity 8 into the maximization problem. Constraints (A.6) and (A.7) require that the level of subsidy S offered by the jurisdiction is the equilibrium level 8* , and that the optimal mix of financing methods occurs when the marginal rate of technical substitution of revenue bonds for tax abatements equals the ratio of the prices of RB and Z. The technology used to produce the subsidy S is assumed to be Cobb-Douglas. Constraint (A.7) gives the following relationship between the inputs 2 and RB: AthRB (rt, - ehI - gW - fRB) Z - where A - a/(l-a) (A.9) ru ' ehI ' 3W ' ZfRB 109 The production function for 8 also can be used to provide additional information on the relationship between RB and Z. S - Z“[(RB)(ru - ehI - fRB - gW)]1'°‘. (A.10) An expression for RB in terms of'S' (called S for ease of notation) can be obtained by solving the two equations (A.9) and (A.10) for Z and setting them equal. mum - sl/“[RB(r,,-ehI-fRE-gw)]/<- rti ' ehI ‘ 8w ' ZfRB With difficulty, this equation could.be solved for RB. An expression for 2 as a function of S can be obtained in a similar manner, with the resulting expression being even more complicated. In order to solve the utility-maximization problem, the utility function and the budget constraint can be rewritten in order to take constraints (A.5) - (A.8) into account. The new problem is as follows: maximize U (I+(1-d)K_1, EXP, X) - [I + (1-d)1(_1]JEXP“'xu subject to: Y - x + tC{EXP + I - I{t° - t‘p(1+d)‘1 [1+fRB+gW]] - C + Z} 2t‘peI (1+d) ‘1 Z is the expression for tax abatements Z that can be obtained from equations (A.9) and.(A.10). For simplicity, I do not write the expression for Z in an explicit form. The expression for Z is terms of RB, S, and the other parameters in the model. A consequence of this utility-maximizing behavior is that the first partial derivatives of the Lagrangian function incorporating the utility function and the budget constraint equal zero. 'The first-order conditions are derived by taking the partial derivative of the Lagrangean function with respect to I, EXP, X, RB and the Lagrangian multiplier. (Z does not actually appear, as it has been replaced by a function of RB and 8.) Due to the great complexity of the problem, the first-order conditions will not be shown here. The five equations can be solved for the five unknowns to obtain demand functions for I, EXP, X and RB. Once RB is known, the demand for 2 can be obtained. The demand for investment I is of particular interest here. Unfortunately, the expressions for I, RB, and 110 the other variables are very complicated. As a result, the investment demand portion of the model that is estimated in Chapter VI is not the actual demand function derived from the maximization process. Due to the complexity of the problem, comparative static results will not be derived. Instead, I let the maximization problem suggest which variables are the likely determinants of investment decisions and I test for their significance. APPENDIX B This appendix contains further information about the data used in Chapter VI. The description of each data series is included along with its source. GO - State and local government issues of long-term new money (non - refunding) general obligation bonds per capita. As used here, G0 is the average of the 1983 and. 1984 per' capita general Obligation. issues determined by dividing the total borrowing levels in each state for both 1983 and 1984 by the total of the 1983 and 1984 state populations. These data are from the files of the Securities Data Company. Previous studies examining borrowing differences across jurisdictions (Metcalf (1989), Gordon and Slemrod (1986) and Asefa, et. al. (1981)) have used Census data on the change in debt outstanding or on long-term bond issues. Although there seems to be a general consensus among researchers that the Census bond data are unsatisfactory, no other data are publicly available. The Bond Buyer does not publish state level data, and the Public Securities Association does not distinguish between bonds issued to refund earlier obligations at a lower interest rate ("refundings") from "new-money” bonds issued to finance capital expenditures. RB - state and local government issues of long-tern! revenue bonds (specifically "private-activity bonds") per capita determined by dividing the 1983 and 1984 totals by the total population in both years. Although revenue bonds have surpassed general obligation bonds in new issue volume, it was only recently that comprehensive data have been collected. The IRS's SOI Bulletin presents detailed data on three major categories of revenue bonds: a broad category of industrial development bonds, student loan bonds, and exempt-entity bonds. Starting in 1985, mortgage subsidy bonds will also be included in the SOI reports. RB refers to the 1983-84 average per capita revenue bond amounts. More information about the 801 data can be found in ACIR (1990). I - state and local capital expenditure per capita. Capital spending can be found in the 1983 and 1984 Govetnmental Finances in a table entitled "State and Local Governmental Expenditure for Capital Outlay, by Function and States." Per capita figures were obtained by using 1983 and 1984 state population data from the 1988 Statistical Absttact of the U,S. h - the share of long-term general obligation in the financing of state and local capital expenditure. This is simply the ratio of GO issues to state and local capital expenditure. The debt share is the average of the 1983 and 1984 debt shares. TPl, TP2, TP11, and TP22 - tax prices representing the cost to the decisive voter of a $1 increase in per capita tax-financed state and local 111 112 expenditure. Using 1982 data on the proportion of returns filed by itemizers M (joint returns counted twice) and.the average federal marginal tax rate t faced by these itemizers, the first two tax prices are defined as: TPl - (l—M) + M(l-t), and TP2 - (l-t). The last two tax prices attempt to incorporate the effect of the reciprocal deductibility of state and local taxes from the federal income tax and vice versa. TPll is equal to (l-M) + M(l-t') where t' is a measure of the federal marginal tax rate for households earning $10,000 - $20,000 in 1982 calculated by Feenberg and Rosen (1986). The federal tax rate measure t' takes into account the reciprocal deductibility of federal, state and local taxes. Finally, TP22 is equal to (l-t'). FED GRANTS - federal grants to state and local governments per capita, computed as the average of the 1983 and 1984 levels. Per capita federal aid is listed in the 1986 and 1987 Ststistissl Abstract. MED INCOME - per capita median "effective buying income.” This is from the annual Survey of Buying Power in Sslss snsi Marketing Management Msgazins. Median income levels were converted into per capita amounts by dividing by the number of people per household, using information listed _ in the 1988 Statistical Abstract. MED INCOME is the 1983-84 average. MATCH - a proxy for the average matching rate to states receiving federal matching grants. This measure is equal to the ratio of federal aid to state and local governments in each state for highways (found in the 1982 Ssnsus sf Governments) divided by the total amount of federal aid. ENROLL - the percentage change in public elementary and secondary school enrollment in the years 1980-1985, as listed in the 1988 MM Abstrsst. MIGRATION - 1980-84 net total migration to a state as a percentage of the state's 1980 population. Net total migration includes net immigration from abroad and net interstate migration. These figures were obtained from the 1986 Statistical Absttast. FUTURE POP - projected percentage increase in state population to the year 2000. FUTURE POP is the average of the projected.population increase from 1983 to 2000 and the 1984-2000 projected increase. The percentage population increase was calculated from data on projected population levels in the 1988 Statisticsl Abstract, converted to percentage terms using 1983 and 1984 population figures. OLD - the percentage of the state population 65 years of age and older as of 1986, as reported in the 1988 Ststistisal Abstract. U RATE - the average of the 1983 and 1984 state unemployment rates, as a percentage of the civilian labor force. From the U.S. Department of Commerce Stats snd Msttsnolitan Ares Qsta S993, 1986. INCENTIVES - the number of state tax expenditures plus the number of ”special services" offered in support of industry as catalogued by the lndusttial Qevelonnent sng Site Selection fiandbook (1985). 113 GOVTS - the number of local (sub-state) governments in each state in 1982 per 1,000 inhabitants, from the 1982 Ssnsns of Sovetnnents. DISCRIM - a dummy variable set equal to one if states exempt interest on in-state bonds from state taxation but tax all out-state bond interest. The list of states which engage in discriminatory taxation comes from Kidwell (1984). I combine his Group 3 (states which exempt all in-state and tax all out-of state bonds) and Group 4 (states which exempt some in-state and tax all out-of-state bonds.) If DISCRIM - 1, then the jurisdictions in.that state may be able to issue bonds at a lower interest cost if the marginal bondholder is a resident of that state. TREASURER.- a dummy variable set equal to l for states that have appointed treasurers and equal to 0 for states with elected treasurers. From the Book of the States (Council of State Governments 1982-83.) DENSITY - Persons per square mile in 1980. From the 1988 Statistical Abstract 9i tne U,S. POPULATION - the average 1983 -l984 population in each state in thousands. From the 1988 Statisticsl Absttsst sf tns Q,S. CONTIG/STATE POP - contiguous state population divided by the state’s population. Determined using a Rand-McNally road atlas and 1983 population figures. MANUF CAP - per capita capital expenditures by manufacturing firms by state in 1983 from the 1986 Annual Sutgsy sf Mannfactutss, Bureau of the Census. CHANGE IN MANUF CAP - Per capita change in manufacturing capital expenditures from 1977 to 1983 using 1983 population from the 1986 Annnsl Suggey st Manufactntes, Bureau of the Census. DEBT/V - total state and local debt outstanding in 1982 divided by the total gross assessed value of property subject to local general property taxes in 1981. Both figures are from the 1982 gensus of Governments. BIBLIOGRAPHY BIBLIOGRAPHY Adams, Roy (1977) "Individual Preferences as Supply Determinants in the Municipal and Federal Bond Markets." Sublic Finance Qnarterly, vol. 5, pp. 175-202. Advisory Commission on Intergovernmental Relations, The Volume an to; -E em t Private v t d ' cal x erience 89, Washington, D.C. 1990. Allman, Phil E. (1982) ”A Neoclassical Model of the Tax-exempt Bond Market.” Unpublished Ph.D. dissertation, Michigan State University. Almanac of the 50 AStates. (1985) A. N. Garwood, ed. Information Publications. Burlington, VT. Asefa, Sally et.al. (1981) ”Municipal Borrowing: Some Empirical Results." Public Einance Qnarterly, vol. 9, pp. 271-80. Auerbach, Alan J. (1979) "Wealth Maximization and the Cost of Capital." uarter Journal 0 Economics, pp. 433-445. Barnea, Amir, Haugen, Robert, and lemma Senbet (1981) ”An Equilibrium Analysis of Debt Financing under Costly Tax Arbitrage and Agency Problems." ournal of Einance, vol. 36, pp. 569-581. Barro, Robert. (1974) ”Are Government Bonds Net Wealth?" Joutnal of Political Economy, pp. 1095-1117. Bartik, Timothy J. (1985) "Business Location Decisions in the United States: Estimates of the Effects of Unionization, Taxes, and Other Characteristics of States.” u 1 Si ss tat c , vol. 3, pp. 14-22. Benker, Karen M. (1986) "Tax Expenditure Reporting: Closing the Loophole in State Budget Oversight.” National Is; antnsl, vol. 39, pp. 403-417. Bergstrom, Theodore C. and Robert P. Goodman, (1973) "Private Demands for Public Goods." American Econonic Review, vol. 63, pp. 280-296. Bloom, H.S., H. Ladd, and J. Yinger, (1983) "Are Property Taxes Capitalized into House Values?" in Lscal £rovision.of Public Services; The Tiebout Model after Twenty-Fivs Years. G.R. Zodrow, ed. Academic Press, New York. Bond Buyer 1988 Yearbook. The Bond Buyer. New York. 114 115 Rook 9i the StstesI l982-l983, Council of State Governments. Borcherding, Thomas E. and Robert T. Deacon, (1972) "The Demand for the Services of Non-Federal Governments." Anetissn Econgnis Review, vol. 62, pp. 891-906. Buchanan,.James (1967) ub ance e cratic Process. University of North Carolina Press, Chapel Hill. Capeci, John. (1990) "Local Fiscal Policies, Default Risk and Municipal Borrowing Costs," Working Paper NO. 259, Brandeis University. Carlton, Dennis W. (1979) "Why New Firms Locate Where They Do: An Econometric Model," in lntertegignal Movement and Sconomis Gtowtn. W.C. Wheaton, ed. Urban Institute. Carlton, D.W. (1983) "The Location and.Employment Choices of New Firms: An Econometric Model with Discrete and Continuous Endogenous Variables," The Review of Economics and Statistics, vol. LXV, pp. 440-449. Clark, Phil and Tom Neubig (1984) "Private-Activity Tax-Exempt Bonds, 1983." 801 Sulletin, vol. 4, pp. 97-107. Clark, Phil (1985) ”Private-Activity Tax Exempt Bonds, 1984." _Ql Rulletin, vol. 5, pp. 55-64. Courant, Paul N. and Daniel L” Rubinfeld. (1987) "Tax.Reform: Implications for the State-local Public Sector." isn;nsl_sflfisnnnnis_2stsnsstixss.‘vol. 1, no. 1, pp. 87-100. Cooner, James J. (1987) Investing in.annisipsl Ssnds-Sslsncing RisRs snd Rewards. John Wiley and Sons, New York. Cunningham,.James P. (1989) "Determinants of U.S. State Debt.” Unpublished Ph.D. dissertation, University of Chicago. DeBartolo, Gil and Peter Fortune. (1982) "The Demand for Public Services: Inferences from Municipal Bond Referenda.” Rational Ts; Journal. vol. 35, pp. 55-69. Eisinger, Peter K. (1987) "The Rise of the Entrepreneurial State in Economic Development." Rroceedings ot the Seventy-ninth Annual Sonference o h t nal a Jou l-- t e a. pp. 34-44. Epple, Dennis and Chester Spatt. (1986) "State Restrictions on Local Debt.” Jonrnsl of Runlis Ssononiss, vol. 29. pp. 199-221. Feenberg, Daniel and Harvey Rosen. (1986) ”The Deductibility of State and Local Taxes: Impact Effects by State and Income Class," Srswtn sng Change, vol. 17, pp. 11-29. 116 Feldstein, Martin” (1985) ”Debt and. Taxes in. the Theory of Public Finance," Journal of Public Ssongnics, vol. 28, pp. 233-245. Feldstein, Martin, Green, Jerry, and Eytan Sheshinski. (1979) "Corporate Financial Policy and Taxation in.a Growing Economy,” Qnarterly Ssutnal oi Esononics, pp. 411-432. Feldstein, Martin and Gilbert Metcalf. (1986) ”The Effect of Federal Tax Deductibility on State and.Local Taxes and Spending," Working Paper 1791, NBER, Boston. Fisher, Ronald C. (1988) Stats and Lnsal Publis Rinance. Scott, Foresman and Company. Glenview, IL. Forbes, Ronald W. (1988) "The Tax-Exempt Securities Market," ms investment Banking Rangbook. J.P. Williamson, ed. Wiley and Sons, NY. Fortune, Peter (1973) "The Impact of Taxable Municipal Bonds: Policy Simulations with a Large Econometric Model.” Rational Tax Joutnal, vol. 26, pp. 29-42. Galper, Harvey and John Petersen. (1971) "An.Analysis of subsidy Plans to Support State and Local Borrowing." National st Joutnsl. vol. 24, pp. 205-34. Gordon, R.H and. Joel Slemrod. (1986) "An Empirical Examination of Municipal Financial Policy.” Studies in State and Lssal Rinanse. 1L Rosen, ed. NBER. Gramlich, Edward M. and Harvey Galper. (1973) "State and Local Fiscal Behavior and Federal Grant Policy." StssRings Raters 9n EcsnsnisiAstivity, pp. 15-65. HenderShott, Patric and David Kidwell. (1978) ”The Impact of Relative Security Supplies.” Jsutnsl sf Money, Stegit, snd SsnRing, vol. 10, pp. 337-347. Hendershott, Patric and Timothy Koch. (1977) Empirical Analysis of the MarRet for Iax-sxemnt Secntities. Monograph Series in Finance and Economics, New York University. Holtz-Eakin, Douglas. (1986) ”Unobserved Tastes and the Determination of Municipal Services.” t a T J r , vol. 39, pp. 527-532. Holtz-Eakin, Douglas and Harvey S. Rosen. (1988) "Tax Deductibility and Municipal Budget Structure,” Eiscsl Eedstalisn; Qnsntitativs Stugies. H. Rosen, ed. NBER. Hsieh, David.D. (1983) "A.Heteroscedasticity-Consistent Covariance Matrix Estimator for Times Series Regressions." Joutnal sf Economstrics, vol. 22, pp. 281-290. 117 du t a Develo me and S te election H bo . (Jan.- Feb. 1985) pp. 52-54. Inman, Robert P. (1978) "Testing Political Economy's 'As If' Proposition: is the Median Income Voter Really Decisive?" Rnblic Choise, vol. 33, pp. 45-65. Jensen, Michael C. and William H. Meckling, (1976) "Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure." Joutnsl pf W. vol. 3. PP- 305-360- Kaufman, George (1981) f cie 11 M nic 1.Bo Market' The U e of a - e na c n o " v t ” o e . JAI Press. Kidwell, David S. et. a1. (1984) "The Impact of State Income Taxes on Municipal Borrowing Costs.” Natipnal Is; qurnal, vol. 37, pp. 551-561. Kormendi, Roger and Thomas Nagle (1981) "The Interest Rate and Revenue Effects of Mortgage Revenue Bonds." in Kaufman (1981). Lamb, Robert and Stephen P. Rappaport. (1986) Municipsl Bonds, 2nd. edition, McGraw-Hill Book Co., New York. Lindsey, Lawrence B. (1988) "Federal Deductibility of State and Local Taxes: A Test of Public Choice by Representative Government,” Eiscsl de m° a t t v u s. H. Rosen, ed. NBER. Maddala, G.S. (1977) Enpnpnsttiss. McGraw-Hill Book Co. New York. Maxwell, James and J.R. Aronson (1977) Rinsncing Ststs ans Lssal vaetnnents. The Brookings Institution, Washington, D.C. Metcalf, Gilbert E. (1989) "Federal Tax Policy and Municipal Financial Behavior." P o eed o h i - st u 0 er ce 0 t e Nstionsl Ia; Joutnsl--IaR institute of Anerisa. pp. 109-114. ,(1988). ”Essays on.the Financial Structure of State and Local Governments.” Unpublished Ph.D. dissertation, Harvard University. Nadeau, Serge. (1988) "A.Mode1 to Measure the Effects of Taxes on the Real and Financial Decisions of the Firm,” Ratignal Tax Jonrnal. pp. 467-481. National Association of State Development Agencies (1986) Directoty of nce ti 0 Bus nes n a d ve o e t the ' S ate b State Guide, The Urban Institute Press, Washington, D.C. Netzer, Richard (1985) "The Effect of Tax Simplification on State and Local Governments,” cononis Sonseguenses oi Tax Simplification, Federal Reserve Bank of Boston. pp. 222-258. Papke, Leslie (1987) "The Location of New Manufacturing Plants and State Taxes: Evidence from Panel Data," c e ev - a gonferense of ths Rational Tax Joutnsl-st Institute of America. pp. 44-55. 118 Peterson, George. (1984) ”Financing the Nation’s Infrastructure Requirements." Petspectivss on Utpsn lnftasttustute. National Academy Press. Petersen, John E. (1981) ”The Municipal Bond Market: Recent Changes and Future Prospects," in Financi t Lo Governments in Walzer, N. and D.L. Chicoine, eds., Oelgeschlager, Gunn and Hain. Phares, Donald. (1980) Eng Rsys State and Lpsal Iaxes. Cambridge, MA. Oelgeschlager, Gunn and Hain. Public Securities Association (1983) Statistical Yearb00k of Municipsl Rinance. New York. Ott, David J. and Allan Meltzer. (1963) Eessral Tax Tteatnent of State and chal Securities. The Brookings Institution. Washington, D.C. Rubinfeld, Daniel. (1973) "Credit Ratings and the Market for General Obligation Municipal Bonds." Rstisnsl_IsR_antnsl, pp. 17-27. Sales and Marketing Management Magazine, e o Bu w . (1984) New York: Bill Publications. Sjoquist, David L. (1981) ”A Median Voter Analysis of Variations in the Use of Property Taxes Among Local Governments." Rnblic Snpice, vol. 36, pp. 273-285. Sullivan, Dennis, and Robert Newman. (1988) ”Econometric analysis of business tax impacts on industrial location: What do we know, and how do we know it?" Journal st Rtssn Econoniss, vol. 23, pp. 215-34. Toder, Eric and Thomas Neubig (1985) "Revenue Loss Estimates: The Case of Tax-Exempt Bonds.” Rstisnsl_IsR_;sntnsl. vol. 38, pp. 395-414. Tuccillo, John and John Weicher (1981) "The Impact of Mortgage Revenue Bonds on Securities, Markets, and Housing Policy Objectives.” in Kaufman (1981). U.S. Department of Commerce, Bureau of the Census (various years) Gsvstnnentsl Einances, Washington, D.C. U.S. Department of Commerce, Bureau of the Census (1987, 1988) U,S. Statistisal Abstrast, Washington, D.C. U.S. Department of Commerce, Bureau of the Census, (1986) Stste and Metropslitan Arss Rsts SppR, Washington, D.C. U.S. Department of the Treasury, Internal Revenue Service. (1984, 1985, 1986) "Private Activity Tax-Exempt Bonds,” Statistics pi lnspms Bnlletin. Washington, D.C. 119 U.S. Office of Management and Budget, e a Anal sis H ud o the U,S. Govetnnent, various years, Government Printing Office, Washington, D.C. Vitaliano, Donald F. (1988) "Evidence About Debt Equivalence From the Local Public Sector." Runlis Einsncsz Rinsnces Publigues. vol. 43, pp. 295-309.