_______________________ i' I '' '**I 'L / Policy, Research, and External Affairs WORKING PAPEJRS L Country Operations Country Department IV Asia Regional Office The World Bank July 1991 WPS 727 The Effects of Debt Subsidies on Corporate Investment Behavior Mansoor Dailami and E. Han Kim Credit subsidies are ineffective in stimulating business invest- ment in productive assets. Instead, they lead to an increase in corporate holdings of financial assets and real estate. ThePolicy.Resarch. and ExternalAffairs Cornplex distributesPRE WorkingPapers todissemrinatethofindings of work in progrss and to enoouragc the exchange of ideas among Bank staff and all others interested in devcloprnent issucs. These papers carry the names of the authors. Qflect oly their views, and should be used and cited accordingly. The findings, intpetations, and conclusions are the authors' own. They should not be auributed to the World Bank, its Board of Diretots, its management, or any of its member countries. Policy, Hesearch, and External Affaire Country Operations WPS 727 This paper- a product of the Country Operations Division, Country Department IV (India), Asia Regional Office - is the second in a planned series of research on the perormnance of capital markets and their role in providing risk capital to the corporate sector in India and the Republic of Korea. The research is funded by the Bank's Research Committee (RPO 675-84). Copies are available free from the World Bank, 1818 H Street NW, Washington DC 20433. Please contact Adala Bruce-Konuah, room DIO-079, extension 80356 (22 pages, with figures and tables). Dailami and Kim argue that credit subsidies are Their estimates indicate that -without interest ineffective in stimulating business investment in rate controls and o!- forms of subsidy, corpo- productive assets. Instead, they Iead to an rate holdings of speculative assets would have increase in corporate holdings of finanicial assets been one-sevenlth of observed levels. Moreover, and real estate. most corporate real estate holdings appear to be unrelated to production activities. For empirical verification, Dailami and Kim examined investment patterns in a sample of 241 They find little evidence that the Korean Korean corporations listed on the Korea Stock government's interest rate controls and credit Exchange between 1984 and 1988. They found allocation policy have accelerated expansion of a significant positive relation between corporate corporate investment. 11 anything, they are speculative asset holdings and access to subsi- partly to blame for the overheated Korean stock dized loans. market during 1986-88. The PRE Working Ilaper Series disseminates the findings of work under way in the Bank's Policy. Research, and External AffairsComplex. Anobjective ofthescries is to getthesefindingsoutquickly, even ifpresentations areless than fully polished. TThe findings, interpretations, and conclusions in these papers do not necessarily represent official Bank policy. i Produced by the PRE Dissemination Center - -l-| - i~~~~~~p- -- ! S-S* -- *- -- - Page No. TABLE OF CONTENTS I. Introduction . . . . . . . . . . . . . . . . . . . . . . . 1 II. Corporate Debt Subsidies and Investment Behavior . . . . . 3 III. The Hypothesis . . . . . . . . . . . . . . . . . . . . . . 3 IV. Data and Measurements . . . . . . . . . . . . . . . . . . . 12 V. Empirical Results . . . . . . . . . . . . . . . . . . . . . 18 VI. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . 27 References. . .. X so List of Figures and Tables Figure 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Figure 2 .6 Table 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Figure 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 Table 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Table 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Table 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Table 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 Table 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Table 7 .26 I. INTRODUCTION Govenmnent intervention in the pricing and allocation of credit remains an enduring feature of both developed and developing countries. Although considerable world-wide progress has been achieved in recent years toward financial liberalization and open capital markets, goverunents continue to deploy credit instruments to address a variety of social, political, and economic problems. In industrialized countries govenmments frequently intervene on a broad scale in efforts to increase the availability of loans to students, fanners, and home owners.I They also extend loan guarantees to exporters and tv large enterprises in financial distress.2 In developing countries government intervention figures more prominently. Both loans at subsidized interest rates and government loan guarantees are frequently used to encourage investrnent and foster industrialization. Goverunents also impose ceilings on interest rates and loan guarantee fees, provide cheap dirct credit to targeted industries, and bail out fims in financial distress. Two basic assumptions underlie developing countries' reliance on these strategies. First, extemalities in fmancial markets are presumed to follow from either market failure and structural weaknesses--e.g., the absence of weU organized equity and bond markets-and/or from severity of information asynunetries between lenders and borrowers.3 Second, it is widely perceived that the PFor instance, Bosworth, Canron, and Rhyne (1987), Gale (1990, 1991) and a report by the Congressional Budget Office (1981) describe practices in the United States. For other industrialized countries, see Teranishi (1990), Cox (1986), and the report of the Joint Economic Committee of the U.S. Congress (1981). 2The best known examples of federal loan guarantees in the United States are the Lockheed and Chrysler bailouts. See Moritz and Seaman (1981), Ho and Singer (1982), and Chaney and Thakor (1985). Govenmuent provisions of loan guarantees abound in other industrialized countries. See, for instance, Green (1985) for the case of France and Sakakibara and Feldman (1983) for Japan. 31t is often argued that financial internediaries, left alone, behave overly conservatively and deny crdit to sone creditworthy finns with positive net present value projects. The foregoing of such projects entails social welfare costs. This underinvestment problem can be alleviated if government iitervention fills the infonnation gap and facilitats the provision of long-term loans to the appropriate users. See Stiglitz (1991) for an in-depth discussion on the market failure due to informational asymmetry and moral hazzard problems and Berkovitch and Kim (1990) on the hiteraction between debt contracts and the under- and over-investnent incentves. -2- various debt subsidies will, by lowering the cost or increasing the supply of funds, induce fimns to expand capital holdings in productive assets such as plants and equipment. These investrnents are in tum anticipated to generate a higher rate of employmnent and economic growth. The objective of this paper is to question the validity of this second assumption. We suggest that access to low cost borrowing may not necessarily lead to higher investment in productive assets, but rather to investment in finanzial and speculative assets. We elaborate on the conditions under which such an adverse result may occur, and show that these conditions are of a sufficiently general natu:re to warrant serious attention by policymakers. For empirical verification we focus on investment pattems in Korea, which provides an interesting case study for several reasons. It is well known that Korea has relied on financial market intervention as an impoitant policy instrument for channeling resources to priority sectors and firms. This strategy, which involves directed lending through the Bank of Korea, subsidization of debt via interest rate controls, and provision of loan guarantees at below market rates, has been a hallmark of Korean industrialization policy since the 1960s. By exercising control over corporate funding, the Govenunent has played an active role in determining allocation of scarce capital. One aspect of this intervention is the provision of funds to priority sectors at preferential rates. A second aspect involves risk sharing in long-tem investnents. By invesdng in a project which had the govenmient's blessing, a firm can benefit from the guarantee of a stable and subsidized flow of credit, often irrspective of its economic and financial perfornance. The result of this policy is a significant reduction in the risk of bankruptcy, which in tum reduces the cost of capital for eligible firms. How effective has this strategy been? To address this question we proceed in the next section with a discussion of debt subsidies. We then provide a simple theoretical analysis of the effect of debt subsidies on corporate investment beha-ior. The analysis shows debt subsidies to be ineffective in increasing the stock of productive fixed assets. Instead, subsidies provide an incentive for fims to increase their holdings of speculative assets. Based on these theoretical results we develop a testable hypothesis in Section m, which is followed by empirical tests in -3- Sections IV and V. The results indicate a significant positive relation between the availability of subsidized loans and corporate speculative investment. Section VI contains concluding remaiks. H. CORPORATE DEBT SUBSIDIES AND INVESTMENT BEHAVIOR 1. Debt Subsidy The most obvious forn of debt subsidy is the provision of funds at below market interest rates. More subtle implicit subsidies arise when there are: (i) official ceilings on interest rates and loan guarantee fees and (ii) bail-outs of companies in fnancial distress. Effective ceilings on interest rates and loan guarantee fees create excess demand for credit and lead to rationing. If for some reason (e.g., persuasion by governmental agencies) banks are required to extend or guarantee loans to high risk finns within the ceilings, the firms that are fortunate enough to obtain such loans or guarantees receive an implicit debt subsidy. To illustrate, consider a bank that must eam a ten percent return on its loans to break even. The bank is contemplating a loan request from a firn that has a five percent probability of default, with twenty percent of the loan recoverable in the event of default. To break even the bank must charge at least [(1 + .1) - (.2)(.05)]A/1 - .05) - 1 = 14.7%.4 If this rate exceeds the interest rate ceiling, the optimal response for the bank is to deny the loan request. Suppose, however, tuit the bank is required to extend the loan and that the maximum rate it can charge is only 11.5 percent.5 At this pmise rate, the rate the bank eRt to receive is (1 + .115)(.95) + (.2)(.0S) - 1 = 4Let A R = the promised interest rate that fully reflects the default risk of the borrower; p = the probability of default; d = the percentage of the loan recoverable in the event of defimlt; R = the break-even return on the bank's loan portfolio. Then the promised rate R that the bank must charge to break even is: R = (1I + R) - p * d]/(l - p) - 1. 5The 11.5% used in this examplc was the actual interest rate ceiling on Korean bank loans during the mid 1980s, which is the sample period used for our empirical tests. -4- 6.925%. Thus, for every dollar loaned, the bank expects to lose 3.075%, and the borrower receives an equivalent implicit subsidy. A ceiling on loan guarantee fees has an identical effect. Suppose the borrower has instead requested a loan guarantee. Ignoring the costs involved in adninistering loan guarantees, the minimum guamantee fee that the bank must charge in order to break even is (1 - .2)(.05)/(l - .05) = 4.2% of the amount loaned.6 Suppose, however, that the ceiling on the guarantee fee is only 1.5% and the bank is required to guarantee the loan.7 Then for every dollar guaranteed, the bank expects to lose .015(1 - .05) - (1 - .2)(0.5) = -2.575%. In sum, ceilings on interest rates and loan guarantee fees in combination with the nonprice allocation of credits, provide implicit interest subsidies to high risk firms. A final category of debt subsidy arises from govenunent bailouts of fmancially troubled firns. These bailouts typically involve a restructuring of the fin's debt in which the govenmment provides new capital at a substantialy below-mrket interest rate. The new capital often takes the formn of mandated bank loans.8 2. Effects of Debt Subsidy on Corporate Investments To analyze the impact of these explicit and implicit debt subsidies on corporate investrnent behavior, we first consider the traditional approach embodied in both the Keynesian and the neoclassical models of investment. These models reduce the multitude of asset categories on a 6Let g be the loan guarantee fee per dollar of borrowing. Then for each dollar guaranteed the bank will eamn g if the firm does not default, and will lose (I - d) if the fim defaults. (See the preceding footnote for notational definitions.) Thus to break even, g must satisfy the following equation: g=(l -d)p/(l -p) 7The 1.5% in this example was the official ceiling for loan guarantee fees in Korea for several years during the 1980s. 8See Kim (1990) for an analysis of the effects of debt subsidies on the fmancing behavior of Korean corporations and Ter-nnishi (1990) regarding the nature of government bailouts during the industrialization of Japan. -5- companies' balance-sheet to a single item.9 By concentrating on one asset, which is conventionally taken to be "productive fixed capital", these models can describe the set of investment oppommities available to the firm by means of a single downward sloping marginal efficiency of capital schedule. Given such a schedule, a lower (marginal) cost of capital brought about, for instance, through interest rate subsidies, can readily be shown to induce a higher level of investment in proxuctive fixed capital. This argument is illustrated in Figure 1 which describes the opportunity set of investments facing a representative fimn Figure I depicts the marginal cost of capital (MCC) line and the marginal rate of return (MRR) curve. The marginal cost of capital should be constant in a competitive capital market. As is typically assuned, the fium is confronted with a decreasing marginai rate of return from incremental fixed investments. Without interest subsidies, the profit- maximizing finn will invest up to j* where the marginal cost of capital is equal to the marginal rate of retum. Suppose, however, that the fwm is given an oppommity to obtain a subsidized loan in the amount of X at the rate of MCC minus k. The firm's cost of capital will be reduced by k up to Xi. Iif X. is less than the profit maximizing level of investment, I*, the subsidy does not affect the marginal cost of capital at J* and hence wiU not increase the investment level. The subsidized loan will only enrich the owners of the firm by an amount kXl without achieving the goal of increasing the firm's investment in fixed assets. Figure 2 depicts the case in which the size of the subsidized loan (X2) is greater than I.* Even in this case, it is unlikely that the fmn will increase its investment in fixed assets. Note that the cost of capital is the opportunity cost that the owners of the firm forego by not investing elsewhere. In other words, the marginal cost of capital line not only represents the cost of obtaining funds, but also represents the investment opportunity set available to the owners of the 9Theoretically, such an aggregation is viable only if all assets on the company balance sheet are perfect substitutes. -6- Ralt MRR MCC ___\\ MCC MCC - k I I MRR I I I I I~ILVOSO InveMtents XI Figure 1: Th Impact of ubeldLzd Loan X St MMCC - k on Corpcorut Investmnts In Productve Aset: X I* MRR Rato MCC I - MCC MCCGk k- I - I'JIMRR I ,I Wo I ~~~~~~~Lewisof Investments I X2 Figure 2: The Impac of Subsldlud Loan X2 at MCC - k on Corporate Invesments In Produtve Asset: X2 > 1r -7- finn via "speculative" assets such as financial assets and real estate. Consequently, the optimal investment decision requires investment in fixed assets only up to the original I with the remaining amount of X2 diverted to speculative assets. This investment path is traced by the bold line in Figure 2. As in the previous case, the subsidized loan will only enrich the owners of the firm without increasing the firm's investment in productive assets. The objective of increasing the level of corporate investment beyond I*, say to X2 or I' in Figure 2, can only be achieved if there is an effective monitoring mechanism that prohibits finns from investing the subsidized loans in anything other than the fixed productive assets which yield rates of return below the firm's opportunity cost of capital. This would require that (1) the process of investment is verifiable at every stage throulgh its completion and (2) there is no collusion between the monitoring agent and the fimn. In practice governments monitor investments even if the process of investment is not veifiable at every stage, and penalze firms for diverting funds to other uses. While collusion is possible, monitoring agents will require adequate compensation for the risk of detection. Thus a profit maximizing firm will weigh the expected penalties and the cost of bribery against the difference in yields on productive and speculative assets. Consequently, the greater are the expected penalties and the cost of bribery, the greater will be the proportion of subsidized loans used to finance productive assets. In sum, the impact of debt subsidies on the investment behavior of the recipient fm is dependent on the monitoring effectiveness of the govenmment agency which is providing or mandating the subsidies. M. THE HYPOTHESIS 1. The Model The theoretical predictions in the preccding section can be fornalized by means of a switching regression model with a stochastic sample separation point. Let us define Yi and Xi respectively as the amount of investment in speculative assets by fim i and the net flow of subsidized loans received by firm i in a given year. Then Figures 1 and 2 imply that, absent any other sources and uses of funds, Yi will be equal to Xi - 1*i if Xi > .*i and zr-o otherwise. More generally, the relation between Yi and Xi can be stated as follows: Yi 8Po + 1xi + Uii, if > *i (1) Yi a + u2i, if XiSI*i where 3o. P1i and a are the estimation parameters. We assune that the error tenns u 2 2 2 and u2i satisfy the usual conditions of E(uli) = E(u2i) = 0, and E(uli) = E(u2i) = r. Equation (1) implies that the relation between a firm's investment in speculative assets and its access to subsidized loans depends on whether or not the fim's supply of subsidized loans exceeds its desired level of investment in productive assets. Thus, for furms with X, > 1*i, a positive fraction, ,I3, of the subsidized loan is used to finance speculative investment. In the extreme case in which govenment monitoring is either nonexistent or totally ineffective, profit maxumzng finns will divert all excess financing into speculative assets. For the group of firms for which Xi < 1*i, we postulate that P1 = 6. The sample separation between the two groups of fims occurs at the point where Xi = I*i, i.e., the nev supply of subsidized loans is equal to the desired level of investment in productive assets. Note that the location of this sample separation point is not readily observable as it depends on the determinants of optimal investment in productive assets. The procedure for estimating the switching regression model (1) is well known (See Kiefer (1980), Maddala (1983)]. Let the probability that firm i belongs to the group of firms with net flows of subsidized loan in excess of I be: Pit = Pr(i*< Xi] = F(iZt2t), (2) where Zit is a matrix containing observable determinants of each firm's optimal investment in productive assets and supply of subsidized loans, Ot is a corresponding vector of parameters, and F(-) is the standard nonnal distribution function. Potential candidates for inclusion in the matrix Z would include the firm's level of output and profits which may be related to the optimal level of investment in productive assets and the finn's access to subsidized loans. -9 . Assume that a proportion k of observations are generated by regime I and (1 - X) by regime II, where regime I represents the group of firms for which Xi > P*i, and regime H represents the remaining fiins. Then the likelihood function for an observation Yi can be written as: L(X,0,l01,A,ax2) = XLI(030 I 9 2) + (I-X)L2(x,a2), (3) where L and L2 are respectively given by I L1 =(2s) a Iexp{- I(Yi- 2Ao 21Xi) . (4) L1 (2n) 2 1 exp{- 2 (Yi - a)2/a2}. (5) Assuming that u i and u2i are normally and independendy distributed, the likelihood function for observations (Y1 ..YN) is given by 2 N 2 2 L(X,iP a4Pa,ci )f Li(%i,o ) +(l-X)L2i(a, ) (6) Maxinizing the log of likelihood function (6) with respect to its four relevant arguments, we obtain: A WiXiYj - (WX)(WY) WXi- ((7) Oo (Y) - p1WY (8) A X(l-W )Yi E(l - Wi) EW. whereW =W -p - is the condid probability of regime I given Yi; (WX) and AL1- + (1- X)L i ( ) ax 3respevely the we avemp of xi and Yi. -10- 2. Specification The estimation of the switching regression model described above involves specifying, first, the optimal level of productive investment (1*) and, second, the supply of subsidized loans (X). To estimate the desired level of investment in productive assets, we rely on the following standard model of corporate invesmennt behavior. It = )Yo + Yy12Qt - Kt-1) + 73Ft + vt , (11) where yo, y1, y2, y3 are paramneters to be estimated, Qt is a finn's level of output as measured by sales plus the change ii inventories of final goods, Kt_I is its capital stock of productive assets lagged one year, Ft is a financial variable altematively measured by either the finn's previous year profits or by the first difference in the firm's value as measured by the market capitalization of its equity. Finally, vt is a disturbance term. All variables are scaled by the finr's begining of year book value of total assets. Equation (11) combines the conventional accelerator model with the usual interemporal adjustment specification. It also contains a measure of profitability and stock maket performance designed to capture the firn's present and future investment oppo10nties. If the supply of subsidized loans to each firm is observable, it is possible to detrmine the probability, Pit, that firm i at time t belongs to regime 1. Using equation (11) we obtain: Pit =Prllit* < X,i1 =Prvt