INTEREST RATES, RISK, AND IMPERFECT MARKETS: PUZZLES AND POLICIES

Transcription

1 INTEREST RATES, RISK, AND IMPERFECT MARKETS: PUZZLES AND POLICIES JOSEPH E. STIGLITZ The Wold Bank 1 Taditional theoy emphasizes the key ole that monetay policy can play though the manipulation of inteest ates. But thee ae seveal puzzles that cannot be econciled with standad models. These include: the appaent constancy in inteest ates ove extended peiods, and changes at othe times which appea unelated to changes in technology and demogaphy; the cyclical patten of movements in eal inteest ates; the impact of nominal not eal inteest-ate changes on eal vaiables; and the cyclical patten of movements in inteest-ate speads. This pape eaches beyond the standad competitive equilibium, pefect infomation, model of cedit makets towads impefect infomation models, paticulaly those that focus on the deteminants of bank behaviou. Of the standad models, the money demand model is most deficient in undestanding these puzzles. The loanable funds theoy and a genealized vesion of eal poductivity theoy can be econciled with impefect infomation, and makets and the consequent cedit and equity ationing egimes help to explain the puzzles. Specifically, banks may be insensitive to changes in monetay stance owing to isk avesion. Thee ae stong policy implications; it is agued, fo instance, that in East Asia aising inteest ates exacebated economic decline and, athe than contibuting to exchange-ate stability, may have induced capital flight as default isk inceased, loweing isk-adjusted expected etuns. I. INTRODUCTION Moden economic theoy has stessed the key ole that eal inteest ates play in economic behaviou: eal inteest ates affect investment, which, in tun, affects the aggegate level of economic activity. Monetay policy is given a cental ole in contolling the level of economic activity though its ole in contolling inteest ates. Monetay authoities ae hypothesized to change nominal inteest ates in esponse to a change in expectations concening inflation so 1 This pape extends Stiglitz (1995). The autho is Chief Economist and Senio Vice Pesident at the Wold Bank, on leave fom Stanfod Univesity. He is indebted to Maya Tudo and Nadia Roumani fo helpful assistance. The views expessed ae solely those of the autho, and not necessaily those of any institution with which he is o has been affiliated OXFORD UNIVERSITY PRESS AND THE OXFORD REVIEW OF ECONOMIC POLICY LIMITED 59

2 that the eal inteest ate adjusts in the desied way. 2 In spite of the cental ole the inteest ate plays, thee ae seveal long-standing puzzles. Fist, how can we econcile obseved movements in eal inteest ates with shifts in the demand and supply cuve fo funds? Why is thee a seeming constancy ove long peiods of time while, in some peiods, thee ae also damatic changes in eal inteest ates seemingly unelated to, say, lage changes in technology, undelying pefeences o demogaphy? How can we explain the cyclical movement of eal inteest ates? Second, why is it that empiical studies often seem to show that nominal inteest ates, not just (o in some cases, even) eal inteest ates matte? Thid, how do we explain the elationship among etuns, e.g. the equity pemium o cyclical movements in inteest ate speads? This pape agues that to find the answe to these puzzles, we must move away fom the standad competitive equilibium and pefect infomation model of cedit makets towads impefect infomation (and in some cases, impefect competition) models, paticulaly those that cente aound banks and the deteminants of thei behaviou. Moeove, I ague that this pespective has stong policy implications; that using the wong model not only may poduce the wong esults, but esults that ae countepoductive. The disastous couse of policies pusued in East Asia in esponse to the 1997 cisis illustates such danges. Standad economic theoy has employed thee distinct models of inteest-ate detemination: the money demand model, the loanable funds model, and the neoclassical eal poductivity theoy. Afte quickly disposing of the money demand model, I show how the impefect infomation/impefect makets appoach can be used to econcile the loanable funds and (a genealized vesion) of the eal poductivity theoy. II. THE COMPETITIVE EQUILIBRIUM MODEL AND THE EMPIRICAL PUZZLES The deficiencies in the money demand model, which has been the cente of attention at least since Keynes (see Keynes, 1936), ae inceasingly being ecognized. In that model, the demand fo money depends on the level of (money) national income and the inteest ate, which epesents the oppotunity cost of holding money. But today, most money is inteest beaing (see, fo example, Stiglitz, 1988b). Indeed, on lage accounts, the inteest ate is little diffeent fom that on Teasuy Bills (T-bills). (The spead is detemined basically by technology, by the elatively small tansactions costs, and has little to do with cyclical concens.) Hence, the oppotunity cost of holding an inteest-beaing money account (such as a cash management account) is negligible. To be sue, thee is an oppotunity cost fo holding cuency, but few have advocated a cuency theoy of inteest-ate detemination, fo obvious easons. Second, money is not needed fo most tansactions; cedit (including cedit cads) can be used and inceasingly is being used instead. Thid, most tansactions ae in fact not income-geneating, but ae exchanges of assets. Wee thee a stable elationship between these asset exchanges and income-geneating tansactions, it would make little diffeence; but thee is no eason to believe that the elationship between the two is stable, especially ove the business cycle. 3 In the 1930s, thee was a stong competing theoy, the loanable funds theoy, advocated, fo example, by Robetson (1936). In that model, the inteest ate is detemined as the intesection of a downwadsloping demand fo funds and an upwad-sloping supply cuve of funds (Figue 1). In accodance with standad economic pecepts, both depend on eal inteest ates: thee is no money illusion. As the economy moves into a ecession, the demand cuve (deived fom the demand fo investment goods) 2 Mattes ae consideably moe complicated than this sentence might suggest. What pesumably mattes fo investment ae longtem inteest ates (and the pice of equity); the actions of monetay authoities seem most diectly to elate to shot-tem inteest ates. Long-tem inteest ates ae elated to expectations of shot-tem inteest ates ove an extended peiod of time, and it is not obvious how actions taken by the monetay authoity today would have maked effects on these long-tem expectations, and theefoe on the long-tem inteest ate. This suggests, as we ague, that monetay policy may execise its effects not so much though intetempoal pice effects as though cash flow and cedit availability effects. 3 The money demand model detemines the nominal inteest ate; a sepaate equation (e.g. a vesion of the Phillips cuve) is equied to detemine the ate of inflation, and the two togethe detemine the eal inteest ate. 60

3 J. E. Stiglitz Figue 1 Loanable Funds Theoy of Inteest Rate Detemination The inteest ate is detemined at the intesection of the demand and supply cuves fo loanable funds (Panel A). The demand cuve is lagely detemined by the etun to investment. Panel B shows how the cuves move as the economy goes into a ecession. The demand cuve shifts down (and to the left), and the supply also shifts to the left. With a close to hoizontal demand cuve, the fome effect dominates, and inteest ates fall, though even with the fall of inteest ates, the equilibium level of investment falls. A Panel A L Panel B B S 1 S 0 D 0 D 1 L 1 L 0 L shifts makedly to the left, while the supply of funds (fom savings) also shifts to the left; but the latte shifts in popotion to the decline in GNP, while the fome shifts moe (investment is typically the most volatile component of aggegate demand). Hence, eal inteest ates should fall, moving in a countecyclical manne. Monetay policy loosening monetay policy in a ecession in ode to stimulate the economy einfoces these movements, leading to a futhe decline in eal inteest ates. The thid theoy focuses on the eal side of the economy. The eal inteest ate (ignoing isk fo the moment) should be equal to the maginal poductivity of capital. The maginal poductivity of capital can be ead off the poduction function. In the shot un, the capital stock is fixed. As employment (hous woked) inceases, the maginal poductivity of capital inceases. Hence, the eal inteest ate should incease in a boom, and should decease in a ecession: eal inteest ates should again move in a makedly countecyclical way (Figue 2). The magnitude of the effect can be seen by using a Cobb Douglas poduction function. 4 Witing Q = A K α L (1 α) 4 The discussion of the following paagaphs is taken diectly fom Stiglitz (1995). 61

4 Figue 2 Real Maginal Poductivity Theoy The eal inteest ate should be equal to the maginal etun to capital. In the shot un, with the stock of capital fixed, as employment inceases, the maginal poductivity should incease. Hence, the eal inteest ate should decease as the economy goes into a ecession and employment deceases. MPK = N whee Q = output, K = capital, and L = labou, we obtain, unde the hypothesis that the eal inteest ate,, is equal to the maginal poduct of capital =AαK (α 1) L (1 α) so that, assuming that the fom of the poduction function and the paamete α emain unchanged, dln = dln A + (1 α) (dln L dln K). Let P denote aveage labou poductivity, Q/L; then dln P = dln A + α(dln K dln L). Substituting, we obtain dln = dln P + (dln L dln K). As the economy comes out of a ecession, by Okun s law, output inceases 3 pe cent fo each 1 pe cent incease in labou. Conside a ecession with a 5 pe cent tough-to-peak change in employment. Fo ou shot-peiod analysis, changes in capital can be ignoed. Then, a 5 pe cent incease in employment should incease the eal etun to capital (the eal inteest ate) by 15 pe cent; by contast, as the economy emeged fom the Geat Depession and fom the ecession of the ealy 1980s, eal inteest ates actually fell. These esults would be stengthened if we had assumed an elasticity of substitution of less than unity (as most of the micoeconomic evidence suggests). The lowe elasticity of substitution means that the wage ental atio should fall moe damatically as the labou capital atio ises. We now need to econcile these two theoies with each othe, and with empiical data. If we could ignoe isk, econciling the two theoies would be simple enough. In the shot un, we can take the capital stock and employment as fixed, and additions to capital as being elatively small compaed to the stock. Investos will want to invest so long as the inteest ate chaged is less than the etun to capital. Hence, the demand fo loanable funds is close to a hoizontal cuve. (Thee is also some demand fo consumption puposes, which is pesumably downwad sloping; but the consumption demand is small compaed to the investment demand.) Hence, to a fist appoximation, egadless of the supply cuve of funds, the equilibium inteest ate is (appoximately) equal to the maginal poductivity of capital. As the economy goes into a ecession, then (continuing to ignoe isk), the maginal poductivity of capital falls, and the demand cuve fo loanable funds shifts down. The eal inteest ate falls. If eal capital becomes iskie elative to T-bills (a plausible assumption, to which I shall etun late), then the demand cuve fo 62

5 J. E. Stiglitz Figue 3 Movements in the Real Inteest Rates Both long-tem movements in the eal inteest ate and movements ove the business cycle ae had to econcile with simple vesions of eithe the loanable funds theoy o the eal maginal poductivity theoy. United States United Kingdom India loanable funds shifts down even moe, einfocing the countecyclical effects. In my ealie pape (Stiglitz, 1995), I pointed out the difficulties of econciling the esults with the eal poductivity theoy, and, since the two theoies seem to give paallel esults, by implication with the loanable funds theoy. Figue 3 shows the movement of eal inteest ates in the United States and seveal othe counties. None of them exhibits the stong cyclical patten pedicted by the theoy. The eal poductivity theoy also pedicts smooth changes in the eal inteest ate as the poductivity and the labou capital atio change. It would seem a fluke that, fo instance, fo nealy 20 yeas, between 1952 and 1969, the two just offset each othe, so that eal Fance Koea inteest ates fluctuated between 1 and 3 pe cent; then suddenly the maginal poductivity of capital plunged, even becoming negative. Just as suddenly howeve, it soaed to high levels not seen in a long time. Othe counties exhibit similaly anomalous movements. These numbes, howeve, ae eal inteest ates on T-bills. The maginal poductivity theoy focuses on the etun to capital. Thee ae impotant diffeences in isk between the two assets. In eal tems, neithe asset is cetain. While the isk of default on a T-bill is negligible, the ate of inflation is uncetain, so that thee is uncetainty about the eal etun. By contast, with physical capital thee is both poductivity isk (what will be the etun to capital) and isk 63

6 associated with the elative pice of capital goods and consumption goods. 5 Still, it seems implausible that changes in peceptions of the diffeential isk between the two assets, and in isk avesion, can econcile the theoies with the appaent movements in eal inteest ates. Indeed, as I have noted, the etun to equity (which should be moe diectly elated to the eal maginal poductivity of capital) and its elationship to the etun to debt, epesents a puzzle itself. Some have obseved that it is had to econcile the highe aveage etun (the equity pemium) with the highe vaiance of equities and with easonable paametes fo isk avesion, but the puzzle is deepe than that. Fo longe holde peiods, the etun to equities stochastically dominates that on debt, so that all individuals with long enough holding time hoizons should hold only equities but they do not (see MaCudy and Shoven, 1992a,b,c). 6 Befoe leaving the discussion of these puzzles, let me say a wod about the spead puzzle. Standad monetay theoy focuses on the T-bill inteest ate; but economic activity depends on the inteest ate that boowes (investos) must pay. The pesumption of the standad theoy has been that the two closely tack each othe, adjusting fo isk. The spead between the two, howeve, often moves in stange ways. Fo instance, in the 1991 ecession, the T-bill ate came down slowly, to between 3 and 4 pe cent while cedit cad ates emained at pe cent. It seems implausible that the isk adjustment just happened to change by the amount that the inteest ates fell. Movements in othe inteest ates (e.g. pime ates) exhibit similaly anomalous behaviou (though admittedly not as stiking). III. THE IMPERFECT INFORMATION/ IMPERFECT CAPITAL MARKET MODELS The impefect infomation/impefect capital maket models povide at least a famewok fo econciling the obseved pattens of inteest-ate movements with economic theoy. Thee ae thee undelying ideas: (a) Because of the pobability of bankuptcy (default), expected etuns on a loan ae lowe than pomised etuns; the pobability of bankuptcy may depend on the inteest ate chaged, so that beyond a point, inceases in the inteest ate chaged actually lead to lowe expected etuns (see Figue 4). As a esult, thee may be cedit ationing (Stiglitz and Weiss, 1981; Jaffee and Stiglitz, 1990). (b) Equity makets ae impefect (again because of impefections of infomation), so that fims cannot fully divest themselves of the isks they face; they must boow. 7 Thee is a pobability that they may not be able to meet thei debt obligations that they will go bankupt. Because the costs of bankuptcy ae high, fims act in a isk-avese manne. 8 How isk avesely they behave depends on thei net woth. When thei net woth deceases, thei willingness (and ability) to invest, to hie labou, to hold inventoies, even to poduce, deceases (see 5 In addition, individuals do not buy a divesified potfolio of K, of the capital stock, but athe equities, which ae claims on fims, which have cetain outstanding debt obligations. If the Modigliani Mille theoem wee coect, then, of couse, the capital stuctue of the fim the natue and amount of these claims would make no diffeence. But the conditions unde which the Modigliani Mille theoem is coect ae highly esticted and ae not satisfied in pactice (see Modigliani and Mille, 1958; Stiglitz, 1974); these limitations play a lage ole in the discussion that follows. 6 This is, of couse, not the only puzzle elating to the behaviou of stock-maket pices. A seies of aticles in the Jounal of Economic Pespectives in the late 1980s and ealy 1990s documents a vaiety of capital maket anomalies, and shows how the vaious explanations put fowad do not seem to go fa in explaining them. Othe anomalies elate to the fact that standad theoy suggests that the only pat of isk that should matte in valuing a fim and that manages should pay attention to is covaiance with the maket; but this does not seem to be tue (Stiglitz, 1989). Moeove, thee is a vaiety of aspects of fim and maket behaviou that seem inconsistent with fims maximizing stock-maket value; the anomalous behaviou is especially appaent in how fims espond to taxes (see Stiglitz, 1973, 1981, 1982a, 1983b). 7 Thee is a vaiety of models explaining the impefections of equity makets, based on models of advese selection, moal hazad, costly and impefect contact enfocement, and state-veifiability. See Myes and Majluf (1994), Benanke and Getle (1989), Geenwald et al. (1984), Stiglitz (1982a,b, 1988a,b, 1992). Fo empiical evidence in suppot of some of these hypotheses, see Asquith and Mullins (1986a,b). 8 Unless thei debt is so high that they gamble on esuection (Kane, 1989). Agency poblems may also give ise to isk-avese behaviou on the pat of fims; see Geenwald and Stiglitz (1990b). 64

7 J. E. Stiglitz Figue 4 Dependence of Expected Retun on Inteest Rate Chaged Because, as the inteest ate chaged inceases, the pobability of default inceases (eithe because of advese selection o moal-hazad effects), the expected etun inceases less than in popotion to the incease in the inteest ate chaged, and above a cetain point, may actually decease. Expected etun to the bank (%) Geenwald and Stiglitz, 1987,1993, 1995). (Fims may be both equity and cedit constained see Hellman and Stiglitz (2000) in which case thei investment is limited to thei own cash.) 9 Inceases in the inteest ate epesent a tansfe of wealth fom debtos to ceditos, theeby eoding fim net woth. (The impefections in equity makets imply that the fim cannot easily undo the damage by etuning to the maket to aise new equity.) 10 These effects ae paticulaly significant when the incease in inteest ates is unanticipated; fo then the fim is likely to have a highe level of indebtedness than it would have had had it known what inteest ates wee going to be. And given the highe level of indebtedness, the cost of the incease in inteest ates is geate. 11 (c) Banks can be viewed as fims that ae engaged in the lending/cetification business. Like * Quoted inteest ate (%) othe fims, they ae equity constained, and act in a isk-avese manne as a esult (see Geenwald and Stiglitz, 1990a, 1991a). Changes in thei net woth (e.g. as a esult of defaults) and/ o thei isk peceptions alte thei ability and willingness to lend. Inceases in inteest ates, while they may incease the etuns banks eceive, also incease its cost of capital. The dominant effect of lage, unanticipated inceases in inteest ates is thus induced bankuptcies and an incease in non-pefoming loans. The fist obsevation means that, in the ceditationing egime, inteest ates chaged ae detemined by factos othe than demand and supply fo funds (o the maginal poductivity of capital). Cedit ationing occus when the noncedit-ationed inteest ate is above the ate at which the expected etun is maximized. What mattes fo the latte is not the mean maginal 9 A lage numbe of studies have ecently veified the impotance of these cash-flow and balance-sheet effects; see, fo example, Deveeux and Schiantaelli (1990), Fazzai et al. (1988), Galeotti et al. (1990), Hoshi et al. (1988, 1991), Hubbad and Kashyap (1989), Gilchist (1989), and Calomiis et al. (1986, 1987). 10 Maye (1990) documents the elatively small ole that new equity issues play in financing investment. 11 The model can help explain why nominal, athe than just eal, inteest ates matte, in seveal ways. Fist, while eal inteest ates matte ex ante, the supise the nominal inteest ate ex post mattes as well, since that has effects on the liquidity and balance sheet of the fim. Moeove, with high levels of inflation the fim is, in effect, epaying a lage faction of any loan evey yea; all medium-tem loans become, in effect, shot-tem; the lack of long-tem cedit in tun has advese effects on the fim. To be sue, lendes could design contacts, commitments to elend in eal tems; but such contacts do not seem common. 65

8 poductivity of the economy as a whole, but athe the financial position of each fim. As the economy goes into ecession, the maket-cleaing inteest ate may fall below the inteest ate at which expected etuns ae maximized, which implies that the inteest ate may fail to fall and could even ise as the economy goes into a ecession. While the inteest ate chaged is that which maximizes the expected etun to the lende, the inteest ate paid to depositos, with competitive banking, is equal to the maximized expected etun on the money deposited; this is likely to fall as the economy goes into a ecession. Thus, while this theoy can explain seemingly anomalous aspects of movements in lending ates of the business cycle, it does less well in explaining deposit ates. On the othe hand, secula changes in technology, in the costs of monitoing and the economy s infomation stuctue, may affect the elationship between the inteest ate chaged and expected etuns, 12 so that the theoy has no clea pedictions concening the elationships between lending ates, deposit ates, and mean maginal etuns on capital. 13 A slight vaiant of the model explains the possibility of igidity of lending ates (see Rodiguez et al., 1992). The mix of boowes facing any lende depends on the inteest ate chaged by othe lendes. Assume fo a moment that all lendes ae symmetic (i.e. face an identical distibution of boowes). The heavy cuve (labelled RR) in Figue 5, Panel A, shows the expected etun if they all chage the same inteest ate; the light cuve (labelled ) shows the expected etun to any single lende, if he o she vaies the inteest ate fom that chaged by all othes. 14 Because as any single fim aises its inteest ate, it loses some of its best isks, the cuve is flatte than the RR cuve. If the lendes coopeated, they would all choose the inteest ate * at which expected etuns wee maximized. But the Nash equilibium without coodination is at c, whee each fim maximizes its own etun, given the inteest ate chaged by all othes. Now we need to examine an equilibium in contacts, whee lendes decide both on an inteest ate chaged, and whethe to match deceases in inteest ates chaged by ivals (the downwad matching clause) o match inceases in inteest ates chaged by ivals (the upwad matching clause). Lendes want to discouage ivals fom undecutting them, and the downwad matching clause does this, by, in effect, ensuing that they get no advese selection gain fom doing so. On the othe hand, they have no incentive to discouage ivals fom aising inteest ates, since such inceases in inteest ates confe on them a positive extenality, i.e. a positive advese selection effect. The equilibium in contacts will thus entails all fims having a downwad matching clause, but not an upwad matching clause. We now conside what this implies fo the natue of a symmetic equilibium (which must lie along RR). Assume that thee wee an equilibium, E 1, at an inteest ate below c (Panel B in Figue 5). Then, as depicted, it would pay any lende to incease the inteest ate chaged even if ivals did not match. Hence, E 1 cannot be an equilibium. But now conside a possible equilibium at E 2, at an inteest ate between c and * (Panel C in Figue 5). The modified cuve now coincides with the RR cuve to the left of E 2, but with the old cuve to the ight. Thus, E 2 is an equilibium. In shot, any inteest ate between c and * is an equilibium. Thee is what is sometimes called a convention equilibium. Any inteest ate any convention (within a ange) can be sustained. 12 A key vaiable is the pobability of default, which in tun depends on the leveage as well as the size of shocks facing fims within the economy; but leveage is itself an endogenous vaiable. As the size of shocks deceases, fims will undetake geate leveage, so that the net change in the pobability of bankuptcy may change elatively little, suggesting that the changes in the discepancy between expected etuns and inteest ate chaged may be less than changes in the undelying stuctue of the economy. (These issues can be analysed using the kind of model pesented in Geenwald and Stiglitz (1993).) 13 One might ague that, ove time, the economy becomes infomationally moe efficient and bette at isk divesification, so that the gap between the maximized expected etun on cedit contacts and the mean maginal etun should decease; but while the economy might get bette in pocessing infomation and divesifying isks, the magnitude of the infomation and isk poblems can clealy change ove time, ovewhelming the impovements in the economy s capacities. Thus, it seems difficult to make any genealization about secula tends. 14 The analogy to Chambelain s model of monopolistic competition (Chambelain, 1933), with his DD demand cuve (whee all fims change the pice togethe) and his dd demand cuve, whee a single fim changes its pice, given the pice chaged by all othe fims, should be clea. 66

9 J. E. Stiglitz Figue 5 Inteest-ate Detemination with Many Lendes and Advese Selection When thee ae many lendes, the expected etuns to each lende depend on the inteest ate chaged by all lendes. If they coodinate inteest-ate changes, and they all face symmetic economic envionments, then the equilibium inteest ate chaged will be *, but if they do not coodinate, then the equilibium inteest ate chaged will be c. If lendes have downwad matching clauses, but not upwad matching clauses, then any inteest ate between * and c can be an equilibium. Panel A Expected etun R R c * Panel B Expected etun R R E 1 c * Panel C Expected etun R R c E 2 * Thee also can be conventions about changes in the inteest ate chaged. So fa, we have assumed that all lendes ae symmetically identical. Within the same famewok, we can think of diffeent lendes as facing slightly diffeent cicumstances (say costs of funds may be a andom vaiable; symmety ex ante may be maintained by assuming that they all ae dawn fom the same distibution). If thee is a shock which changes the etuns cuve facing a single lende, the othes may not want to espond; but if thee is a shock which changes the etuns cuve facing all the lendes in the same way, they may want to espond in a coodinated way. Changes in the discount ate chaged by the Cental Bank 67

10 may be an example of a shock which affects all lendes, in oughly simila ways, so that they may want a coodinated esponse. If they have, in effect, matching clauses which say that they will match any inceases in inteest ates induced by an incease in the discount ate (but if ivals fail to incease the inteest ate, they will too), then thee will be a coodinated esponse which eliminates the incentive fo any lende to ty to take advantage of the othes though an advese-selection effect. (Typically, thee ae multiple convention equilibia. If all ceditos believe that all othe ceditos ae going to leave inteest ates unchanged in esponse to a loweing of the T-bill ate, then that is an equilibium. If they believe that all othe ceditos ae going to lowe inteest ates in esponse to a loweing of the T-bill ate, then that too is an equilibium.) In a sense, the poblem with these models is that they ae too ich ; that is, they ague that factos othe than the aveage maginal poductivity of capital detemine the eal inteest ate, but they do not give vey clea pedictions. The equity-constained models, on the othe hand, can be used diectly to analyse shifts in the demand and supply fo loanable funds, and to yield somewhat cleae pedictions. Recall the ealie discussion in which the equilibium inteest ate was basically detemined by the demand cuve fo loanable funds, so that (ignoing isk), as the economy goes into a ecession, the eal inteest ate falls. Once the isk-avese behaviou of fims and banks is taken into account, the demand cuve fo funds is downwad sloping (the demand fo funds fo investment and woking capital depends on the inteest ate they have to pay); and as the economy goes into a ecession, the demand cuve shifts makedly to the left, as fims become moe isk avese. But at the same time, banks willingness to lend also deceases. The supply cuve also shifts to the left. It may shift to the left moe o less than the demand fo funds. Paticulaly in sevee ecessions, as bankuptcies incease and the net woth of banks is eoded, the leftwad shift of banks is even geate than that of fims, so that the equilibium inteest ate actually inceases as the economy goes into a ecession (Figue 6, whee epesents the constaints-equivalent inteest ate). If the incease in eal inteest ates is lage enough, the economy may switch into a cedit-ationing egime, as illustated in Figue 7. Futhe leftwad shifts in the loanable funds cuve thus have no futhe effects on the inteest ate chaged, but do have lage negative effects on investment and poduction, as cedit availability is cutailed. The model makes the impotant point that tightness of monetay policy, in this context, may not be well gauged by looking at the inteest ate chaged. IV. ECONOMIC POLICY The model povides a famewok fo thinking about monetay policy which is makedly diffeent fom the standad IS LM model o the neoclassical model, fo that matte. The cental bank can be thought of as opeating on the banking system to shift the supply cuve of funds to the ight, loweing inteest ates and inceasing investment, thus stimulating aggegate demand. Conside, fo instance, the impact of loweing the discount ate, which lowes the ate at which banks can boow funds fom the cental bank (o the effective isk they face in falling shot of eseve equiements). 15 Given the lowe cost of funds, banks should, at any maket inteest ate, be willing to lend moe. But banks may exhibit a low elasticity of esponse: a lage loweing of the discount ate may have little effect on the equilibium lending ate (Panel A, Figue 8). Even if the loanable funds supply cuve shifts moe makedly, if thee is cedit ationing, the inteest ate chaged may not change much (Panel B). 16 And even if the inteest ate chaged is loweed significantly, it may have little effect on the equilibium level of lending (Panel C). Thee may be seveal easons fo the lack of esponsiveness of lendes. Banks may be vey isk avese, so that they ae unwilling to extend thei isk exposue much, even if the cost of capital is lowe. O the etuns (including peceived isks) on altenative investments, such as long-tem govenment bonds, may be such that banks simply put any additional funds they have available into govenment bonds, athe than making moe funds available fo lending. Such was the case, fo instance, in 15 Simila esults hold in the case of othe actions by monetay authoities, e.g. changing the eseve equiement. 16 But even though inteest ates have not changed, if the supply cuve shifts, the amount of lending and accodingly, the level of economic activity may incease makedly. 68

11 J. E. Stiglitz Figue 6 Genealized Loanable Funds Theoy The equity constaints facing fims imply that thei demand fo funds (both fo woking capital and investment) may be shaply deceasing with the inteest ate. As the economy goes into a ecession, fims net woth and cash flows ae advesely affected, shifting the demand cuve fo funds to the left. But banks, the majo supplies of funds, also become moe eluctant to lend, so that the supply cuve also shifts to the left. The eal inteest chaged may actually incease. S 1 S 0 D 1 D 0 Figue 7 Genealized Loanable Funds Theoy With Cedit Rationing It is possible, howeve, that as the economy goes into a ecession and the supply of loanable funds cuve shifts makedly to the left, the inteest ate chaged may not ise commensuately. It pays lendes to limit the inteest ate chaged and to impose cedit ationing. D 0 D 1 S 1 L * S 0 0 L 1 L 0 L the United States as it went into ecession in Though the Fedeal Reseve intevened to lowe inteest ates, the supply of loans did not incease. Indeed, the spead between shot-tem inteest ates (which banks had to pay on deposits) and longtem inteest ates inceased. Given that long-tem govenment bonds wee teated as safe assets (i.e. with no ecognition that pat of the highe etun on long-tem bonds was a isk pemium associated with the isk of a fall in the pice of the bond), given that many banks wee woied about satisfying capital equiements, given that the expeiences of the 1980s (and the egulatos esponses to those expeiences) had made them moe isk avese, and 69

12 Figue 8 Effect of Loosening Monetay Policy Loose monetay policy induces banks to incease the supply of funds available at any inteest ate. But if banks ae extemely isk avese (say because they have had a lage advese shock to thei net woth as a esult of inceases in non-pefoming loans), the shift in the loanable funds supply cuve may be limited, so the change in the inteest ate may be limited. Panel B shows that if thee is cedit ationing, inteest ates chaged may not change, even if the shift in the loanable funds cuve is moe maked. Panel C shows that if the demand fo funds cuve is highly inelastic, then, even if the inteest ate changes, the amount of lending (and, coespondingly, the amount of investment) may incease little. Panel A S 0 S D Panel B S 0 S 1 L * D Panel C L 0 L 1 L S 0 S D L 0 L 1 L 70

13 J. E. Stiglitz given that banks wanted to make it look as if they wee moe pofitable, investing in long-tem govenment bonds was viewed as pefeable to lending (which given that the economy was enteing a ecession, looked paticulaly isky) o to investing in othe instuments. The theoy thus povides an explanation fo a liquidity tap, but one which is makedly diffeent fom that discussed by Keynes (which, fo good eason, has fallen out of favou). Monetay authoities have a had time diving the elevant inteest ate, that on loans, down. (The theoy of the equity constained fim, at the same time, can povide pat of the explanation fo the inelastic demand cuve fo investment, so that even wee the monetay authoities successful in diving down the inteest ate, investment might not be inceased by much.) The model has anothe impotant implication. Taditional monetay theoy was based on a stable elationship between the money supply, income, and inteest ates. While expectations about futue economic pospects might have maked effects on the IS cuve, the LM cuve was stable. If the economy was in a ecession, it was easy fo the monetay authoities to figue out by how much they should lowe inteest ates to estoe the economy to full employment, povided they knew the inteest elasticity of investment. They calculated by how much investment would have to incease to estoe the economy to full employment, and given the inteest elasticity, they could easily calculate the equied loweing of inteest ates. Fom hee, it was easy to detemine the equisite incease in the money supply. But as the expeience in 1991 showed, even maked loweing of inteest ates did not have the anticipated effect in estoing the economy to full employment. While thee is some agument about the eason fo this, it is clea that the amount of funds lent did not incease in the way pedicted. Clealy, the elationship between money and cedit had changed; and it is cedit, not money, which mattes fo economic activity. One possible explanation fo the changed elationship is that noted above: banks wee eluctant to lend; they ationed cedit, so that investment did not incease in the way that it would have, given the fall in inteest ates, had thee not been cedit ationing. 17 V. INTERNATIONAL DIMENSIONS The discussion so fa has focused on closed economies. Thee ae additional inteest-ate puzzles in open economies paticulaly associated with inteest-ate paity. The diffeence between the inteest ates in two diffeent counties should be equal to the (isk-adjusted) expected depeciation; but it is had to econcile the data with such an inteest abitage elationship (see, fo example, Fankel, 1992). To be sue, since what mattes is expectations of depeciation, and thee is a futhe adjustment fo isk, thee ae tautological ways in which one can ensue the elationship is satisfied (isk avesion o isk peceptions changed in the equied ways), but these ae unconvincing. I do not popose hee to esolve that dilemma. But thee is anothe policy dilemma that has played an impotant ole in ecent policy debates. Accoding to the standad theoy, if a county inceases the inteest ate it pays, given expectations about depeciation of the cuency, and given isk peceptions, then it becomes elatively moe attactive to hold funds in the county. Thee should be a flow of funds into the county. The flow of funds into the county should seve to stengthen the cuency, i.e. lead to an appeciation. Given expectations about the futue couse of the county s economy, thee is a new, highe equilibium exchange ate. Thus, aising inteest ates is the standad pesciption fo stabilizing a county s exchange ate. In the East Asia cisis that began in July 1997, this standad pesciption failed to aest the decline in exchange ates, much to the constenation of the 17 Of couse, anothe possible explanation put foth by those who do not believe that cedit ationing played an impotant ole is that the elasticity of investment was smalle than anticipated; and that the speads inceased so that the decease in lending ates was not as much as the decease in T-bill ates. The inceased speads, in this view, ae not due to the theoy of banking put fowad hee, but athe to the inceased iskiness of boowes. But, pesumably, the vaiable which is of concen to boowes is not the nominal inteest ate, but the isk-adjusted inteest ate, and the isk-adjusted inteest ate should have gone down as the T-bill ate went down, and thus investment should have gone up. (Of couse, if thee ae diffeences in opinion about the isk, so that lendes view thee being geate isk than boowes, then lendes view themselves as paying moe than boowes view themselves as eceiving. An incease in this wedge, in the discepancy in peceptions, will then lead to a decease in boowing (see Stiglitz, 1972). 71

14 doctos who had pescibed it. 18 To be sue, they wee eady with explanations: the patient had not followed all the doctos odes, and given that, it was not supising that the medicine did not wok. (Had they not made such stong pedictions about what was supposed to have happened, they could have made anothe agument: but fo the inteventions, the exchange ate would have fallen even moe.) All sides in the policy debate ageed that what was cental wee expectations. Those aguing fo a highe inteest ate agued that not only would thee be a positive diect effect fom the highe etun to keeping funds in the county, but also that the fact the county was following the doctos odes would estoe confidence, i.e. pesumably educe the peception of isk. These pedictions wee not, howeve, based on a coheent model of the stuctue of the economy, e.g. based on ational expectations, o on a signalling equilibium. 19 By contast, citics of the inteest-ate policy agued that its advocates had failed to note fist, that most of the debt was pivate-secto debt; second, that a citical vaiable detemining the iskadjusted expected etun on pivate-secto debt was the pobability of bankuptcy; thid, that concens about bankuptcy wee of fist ode it was these concens that had led banks to efuse to oll ove the loans at any inteest ate; 20 fouth, that the pobability of bankuptcy is endogenous; and fifth, that the policies being pusued wee likely to incease that pobability, possibly so much as to educe the iskadjusted expected etun. If that wee the case, then aising inteest ates would actually contibute to the weakening of the exchange ate, by encouaging capital to flow out of the county; and even if the pobability of bankuptcy did not incease so much as to lowe the isk-adjusted expected etun, the incease in the isk-adjusted expected etun would be much lowe than the incease in the inteest ate chaged itself, so that the impact on the exchange ate would be much less than anticipated. The advese effects of aising inteest ates in East Asia wee paticulaly significant, since fims wee vey highly leveaged with shot-tem debt. In those cicumstances, the huge inceases in inteest ates quickly eoded the net woth of fims, and pushed many fims into bankuptcies, with all the pedicted advese effects both on aggegate demand and on aggegate supply. As the numbe of nonpefoming loans inceased, the balance sheet of banks became even wose. That by itself would have shifted the supply cuve of funds to the left. But given the initial weaknesses of the financial institutions, and given the magnitude of the incease in inteest ates so the numbe of non-pefoming loans became enomous a numbe of banks became insolvent. In some counties, athe than being ecapitalized, many banks wee shut down. This contibuted to the futhe contaction of the availability of funds shifting the loanable funds cuve futhe to the left. The weakening of fims meant, at the same time, that the demand fo funds fo investment also deceased; but the need fo funds just fo woking capital inceased. As the monetay authoities then attempted to ease, the damage had aleady been done: banks and fims had both been weakened to the point whee (at least in some counties) even as govenment inteest ates came down, the level of lending inteest ates did not. But even had lending inteest ates come down, the level of investment (and theefoe the level of economic activity) that it would have geneated was limited. Inteestingly, when we looked back ove the histoical expeience asking whethe inceasing inteest ates had had the desied o pedicted effect in counties facing cises the answe seemed to be ambiguous. Sometimes it had, but just as fequently it had not (see Fuman and Stiglitz, 1998; Kaay, 1998). On-going eseach is attempting to ascetain the cicumstances unde which inceasing inteest ates had the desied o pedicted effects. 18 Thee wee othe elements of the pesciption that wee supposed to einfoce the incease in exchange ates lage adjustments in fiscal policy and vey lage bail-outs. 19 See Stiglitz (1999) fo a moe extensive discussion of these points. Thee, I point out that the standad signalling agument that high inteest ates show that the monetay authoities have esolved to addess the undelying poblems may be tuned on its head. In the case of the Latin Ameican cisis, the monetay authoities did need to tighten, in ode to addess the undelying macoeconomic poblems. In the case of East Asia, aguably thee wee no undelying macoeconomic poblems. By using policies that wee appopiate fo a macoeconomic poblem fo addessing poblems of a quite diffeent natue, the cental bank could have been intepeted as signalling its inability to addess the elevant situation; the signalling effect thus could have been advese. 20 This suggests the impotance of the cedit-ationing model. 72

15 J. E. Stiglitz Given this histoical expeience, it is pehaps not supising that investos, seeing an incease in inteest ates, do not automatically believe that exchange ates will incease; thee is little ational basis fo them to have confidence, at least in the stengthening of the exchange. On the othe hand, thee was a lage body of liteatue showing that inceases in inteest ates paticulaly of the magnitude in East Asia would have advese effects on aggegate demand. The implication was clea: if, as seemed to be the case, the counties of East Asia wee in ough macoeconomic balance immediately befoe the cisis, then the lage inceases in inteest ate would clealy suffice to push them into a majo ecession, even if nothing else had been going on. But in fact, the fall in the stock maket and the decease in the exchange ate had maked advese wealth effects. In Thailand, the busting of the eal estate bubble shut down investment in that secto. Thus, even without the incease in inteest ates, the economies wee headed fo a sevee ecession. To make mattes even wose, the counties wee uged to have a fiscal contaction. Remakably, the Fist Deputy Manage of the IMF, Stanley Fische, said that all he was asking of them was to have a balanced budget not a stuctual balanced budget, but an actual balanced budget (see Fische, 1997). Data on falling consumption (especially ca puchases) quickly cooboated the pedictions. But given that the economies wee thus seen to be falling into sevee ecessions o depessions, it is not supising that ational investos would see a lage incease in the pobability of default, and it thus became less attactive to put money into the county and moe attactive fo those in the county to take thei money out. Given the known association between ethnic and political instability (especially in counties with a histoy of those instabilities) and economic downtuns, it is also not supising that investo confidence would be eoded on that account. The eosion of confidence in the economy can be explained without any efeence to the patient not fulfilling all the doctos odes, but athe simply, on the basis that the pesciption the high inteest ates and the fiscal contaction wee exacebating the economic decline. The doctos had one moe excuse. They claimed that had they not inceased the inteest ates, the exchange ates would have declined even moe, and given the fact that the counties wee net debtos, the advese effects on the county s balance sheet would themselves have led to a decline; though they admit the advese effect of the highe inteest ate, the advese effect of the lage decline in exchange ates would have been even geate, o at least so they ague without any empiical evidence eithe about the imputed tade-offs between exchange ates and inteest ates o of the effect of changes in eithe on aggegate economic activity. The agument, howeve, misses the details of the situation confonting each county, ignoes the potential ole of bankuptcy laws, and, most impotantly, fails to take into account the bankuptcy effects upon which ou discussion has cented. The advese effects on fims balance sheets putting them into distess and at, o ove, the vege of bankuptcy meant, as noted, that thee may not have been any tade-off. Raising inteest ates exacebated the decline in the exchange ates. By ignoing the effect on the bankuptcy pobability, they in effect assumed that aising inteest ates would make it moe attactive fo investos to put, o keep, thei money in the county. Had bankuptcy laws been invoked, the magnitude of the edistibutions to foeignes would have been limited; but, quite explicitly, even though the debt is pivate secto, one of the main objectives of the inteventions (acting on behalf of, o seemingly in the inteests of, the foeign ceditos) was to pevent such non-epayments. And, at least in some of the cases, even if the failue to aise inteest ates had led to slightly geate devaluations (a hypothesis fo which, as I have said, thee is little evidence), and even if moe widespead use of bankuptcy laws had not been undetaken, the advese effects on economic activity would have been limited, in at least some of the counties. Take Thailand, fo instance, whee the fims with majo exposues wee the eal estate fims (and the financial institutions that had lent to them) and expotes. The fome wee aleady dead as a esult of the cash of the eal estate boom; hence, futhe devaluations would have had little incemental effect on the level of economic activity. And the latte wee, fo the most pat, essentially coveed: as expotes they moe than gained what they lost in the eal value of thei indebtedness in tems of local cuency To be sue, thee wee some fims that had gambled, and had excessive foeign-exchange exposue. But thee is a fundamental moal-hazad issue: should govenment policy be moe diected at bailing out o potecting fims that have engaged in excessive isk taking (these fims should have puchased cove fo thei exposue), o should it put at isk well-behaved fims, that had simply boowed amounts that wee, o would have been, fully pudential, had inteest ates not been inceased to exobitant levels? 73

16 VI. ALTERNATIVE MODELS Befoe concluding, I should say a wod about some altenative models. Tobin s genealized potfolio appoach (Tobin, 1958) has inteest ates (and the etuns on all othe assets, including equity) detemined by a genealized set of demand and supply cuves fo assets. The appoach focuses on the demand cuves fo diffeent assets, with the monetay authoities, though open maket opeations, essentially changing the elative supply of two of the assets: money and T-bills. Genealizations of Bao s theoem on Ricadian equivalence (see Stiglitz, 1981, 1988a) suggest that such a switch should have elatively little effect (especially in an ea when money beas essentially the same inteest ate as T- bills) though household potfolios. But banks the institutions though which lending occus ae affected, and it is upon this channel of monetay policy that we have focused. Similaly, while the intedependence of demand cuves in the potfolio appoach does imply that such changes will affect the pice of equities, we have noted that elatively little of the funds fo new investment is aised in the equity maket; and Tobin s q-model, focusing on the pice of equity as a deteminant of investment, has not faed well (see, fo example, Abel and Blanchad, 1986). Monetay policy can have lage impacts though the banking system on the supply of loanable funds, and hence on maket ates of inteest. Changes in monetay stances may affect the shot-un isks facing investos in vaious instuments, and in the shot un, these isk peceptions may play a lage ole in detemining elative pices and etuns. Still, the potfolio appoach, in conjunction with a banking model, may help povide some of the explanation of the movements in shot-tem inteest ates duing at least some peiods. Fo instance, one hypothesis concening the high eal inteest ates duing the 1980s in the United States was that the negative o low net woth savings and loans institutions gambled on esuection, offeing depositos high govenment guaanteed inteest ates to attact funds; these deposits, since they wee govenment guaanteed, competed with T-bills, and T-bill ates accodingly had to incease to compete. 22 VII. CONCLUDING REMARKS The inteest ate is a key vaiable in most macoeconomic models; yet its movements, both in the shot tem and the long, seem had to account fo fully, paticulaly in the simple models which have long been fashionable. We have agued that the insights povided by the potfolio balance theoy, combined with the pespectives povided by ecent advances in ou undestanding of impefections in capital makets, povide the famewok of a theoy which goes a long way towads poviding an explanation of at least seveal of the cental phenomena. Seveal concepts have played pivotal oles: impefections of infomation, and the isks of default (which ae endogenous) with thei consequences, including cedit and equity ationing; the esulting capitalmaket impefections leading in tun to iskavese behaviou of fims and banks; the fact that as a esult, fim and bank behaviou ae affected by balance-sheet and cash-flow vaiables; and the cental ole played by banks in cedit makets. All of these togethe povide a fa iche model of the capital maket, one that cannot be well summaized in a money demand equation. Moe impotantly, policy based on that simple model is likely to and has in the past gone seiously awy. 22 An explanation sometimes put fowad fo the negative etuns in the 1970s is that the unexpected incease in inflation meant that ealized eal etuns wee lowe than expected eal etuns. This explanation, howeve, is not entiely satisfactoy; while the oil pice incease was a shock, it would have seemed that inflation expectations should have adapted quickly enough so that thee should not have been negative eal etuns thoughout the est of the decade. Moeove, the inceased peception of inflationay isk should have made these instuments less attactive than, say, equities, so that the eal etun should have even inceased. 74