《货币经济学》课程文献阅读：Credit Channels of Monetary Transmission（Bernanke and Gertler,1995）.pdf_大学文库

Journal of Economic Perspectives-Volume 9, Number fAll 1995-Pagres 27-48 Inside the black box: The Credit Channel of Monetary Policy Transmission Ben s. bernanke and mark gertler M ost economists would agree that, at least in the short run, monetary policy can significantly influence the course of the real economy. In deed, a spate of recent empirical research has confirmed the early find- ing of Friedman and Schwartz(1963)that monetary policy actions are followed by movements in real output that may last for two years or more(Romer and Romer, 1989 Bernanke and Blinder, 1992; Christiano, Eichenbaum and Evans, 1994a, b) There is far less agreement however about exactly how monetary policy exerts its influence: the same research that has established that changes in monetary policy are event tually followed by changes in output is largely silent about what happens in the interim. To a great extent, empirical analysis of the effects of monetary policy has treated the monetary transmission mechanism itself as a"black box Of course, conventional views of how monetary policy works are readily avail- able. According to many textbooks, monetary policymakers use their leverage over short-term interest rates to influence the cost of capital and, consequently, spending on durable goods, such as fixed investment, housing, inventories and consumer durables. In turn, changes in aggregate demand affect the level of production. But the textbook story is incomplete in several important ways One problem is that, in general, empirical studies of supposedly"interest sensitive"components of aggregate spending have in fact had great difficulty in identifying a quantitatively important effect of the neoclassical cost-of-capital vari- able.Indeed, the most common finding is that nonneoclassical factors--for examp pected price changes, is(r+ d)Pe, where ris the required real return to lenders(which could include sk), d is the depreciation rate and pa is the price a Ben S. Bernanke is Class of 1926 Professor of Economics and Public Affairs, Princeton University, Princeton, New Jersey (e-mail: bernanke@wws. princeton. edu). Mark gertler is Professor of Economics, New York University, New York, New York (gertler@ fasecon. econ. nyu. edu)

Journal of Economic Perspectives-Volume 9, Number 4-Fall 1995-Pages 27-48 Inside the Black Box: The Credit Channel of Monetary Policy Transmission Ben S. Bernanke and Mark Gertler M s ff ost economists would agree that, at least in the short run, monetary policy can significanty influence the course of the real economy. Indeed, a spate of recent empirical research has confirmed the early finding of Friedman and Schwartz (1963) that monetary policy actions are followed by movements in real output that may last for two years or more (Romer and Romer, 1989; Bernanke and Blinder, 1992; Christiano, Eichenbaum and Evans, 1994a,b). There is far less agreement, however, about exactly how monetary policy exerts its influence: the same research that has established that changes in monetary policy are eventually followed by changes in output is largely silent about what happens in the interim. To a great extent, empirical analysis of the effects of monetary policy has treated the monetary transmission mechanism itself as a "black box." Of course, conventional views of how monetary policy works are readily available. According to many textbooks, monetary policymakers use their leverage over short-term interest rates to influence the cost of capital and, consequently, spending on durable goods, such as fixed investment, housing, inventories and consumer durables. In turn, changes in aggregate demand affect the level of production. But the textbook story is incomplete in several important ways. One problem is that, in general, empirical studies of supposedly "interestsensitive" components of aggregate spending have in fact had great difficulty in identifying a quantitatively important effect of the neoclassical cost-of-capital variable.' Indeed, the most common finding is that nonneoclassical factors-for example, ' In a frictionless market, the usual Jorgensonian formula for the neoclassical cost of capital, ignoring expected price changes, is (r + d)pk, where ris the required real return to lenders (which could include a premium for systematic risk), d is the depreciation rate and pk is the price of a new capital good. * Ben S. Bernanke is Class of 1926 Professor of Economics and Public Affairs, Princeton University, Princeton, New Jersey (e-mail: bernanke@wws.princeton.edu). Mark Gertler is Professor of Economics, New York University, New York, New York (gertlerm@ fasecon. econ. nyu. edu)

28 Journal of Economic Perspectives accelerator"variables such as lagged output, sales or cash flow-have the greatest impact on spending. (For recent surveys on the determinants of spending components, see blinder and maccini. 1991. on inventories: Chirinko. 199%. on business fixed in- vestment; and Boldin, 1994, on housing. ) Further, in the relatively few cases in which evidence for a cost-of-capital effect is found, the nonneoclassical determinants of spend ing typically remain significant (for example, Boldin, 1994; Cummins, Hassett and Hubbard, 1994). Empirical studies that have tested the neoclassical model in its equi alent "Tobins q formulation have generally been no more successful Beyond the problem of weak cost-of-capital effects in estimated spending equa- tions, there is a presumption that monetary policy should have its strongest infl ence on short-term interest rates. For example, the federal funds rate, the most closely controlled interest rate, is an overnight rate. Conversely, monetary poli should have a relatively weaker impact on long-term rates, especially real long-term purchases of long-lived assets, such as housing or production equipment, which 9 rates. It is puzzling therefore, that monetary policy apparently has large effects to the extent that they are sensitive to interest rates at all-should be responsive primarily to real long-term rates These gaps in the conventional story have led a number of economists to ex- plore whether imperfect information and other"frictions"in credit markets might help explain the potency of monetary policy. Collectively, the mechanisms that have been described in this literature are known loosely as the credit channel of mon- etary transmission. In this article, we summarize our current view of the credit channel and its role inside the "black box"of monetary policy transmission We don' t think of the credit channel as a distinct, free-standing alternative to the traditional monetary transmission mechanism, but rather as a set of factors that amplify and propagate conventional interest rate effects. For this reason, the term credit channel" is something of a misnomer; the credit channel is an enhance- nent mechanism, not a truly independent or parallel channel. Moreover, this no- menclature has led some authors to focus-inappropriately, in our view-on the behavior of credit aggregates, a point that we discuss later. However, it is probably too late to change the terminology now According to the credit channel theory, the direct effects of monetary policy on interest rates are amplified by endogenous changes in the external finance premium, which is the difference in cost between funds raised externally(by issuing equity or debt) and funds generated internally(by retaining earnings). The size of the external finance premium reflects imperfections in the credit markets that drive a wedge be- tween the expected return received by lenders and the costs faced by potential bor owers.According to the credit view, " a change in monetary policy that raises or lowers open-market interest rates tends to change the external finance premium in the same direction. Because of this additional effect of policy on the extermal finance 2 Although we focus on the external finance premium in this article, it should be clear that credit market imperfections may also affect the safe real interest rate in general equilibrium(for example, Stein, 995b). In emphasizing the external finance premium, we implicitly assume that, over the time span under consideration, the Fed can set the short-term safe rate at the level that it desire

28 Journal of Economic Perspectives "accelerator" variables such as lagged output, sales or cash flow-have the greatest impact on spending. (For recent surveys on the determinants of spending components, see Blinder and Maccini, 1991, on inventories; Chirinko, 1993, on business fixed investment; and Boldin, 1994, on housing.) Further, in the relatively few cases in which evidence for a cost-of-capital effect is found, the nonneoclassical determinants of spending typically remain significant (for example, Boldin, 1994; Cummins, Hassett and Hubbard, 1994). Empirical studies that have tested the neoclassical model in its equivalent "Tobin's q" formulation have generally been no more successful. Beyond the problem of weak cost-of-capital effects in estimated spending equations, there is a presumption that monetary policy should have its strongest influence on short-term interest rates. For example, the federal funds rate, the most closely controlled interest rate, is an overnight rate. Conversely, monetary policy should have a relatively weaker impact on long-term rates, especially real long-term rates. It is puzzling, therefore, that monetary policy apparendly has large effects on purchases of long-lived assets, such as housing or production equipment, which to the extent that they are sensitive to interest rates at all-should be responsive primarily to real long-term rates. These gaps in the conventional story have led a number of economists to explore whether imperfect information and other "frictions" in credit markets might help explain the potency of monetary policy. Collectively, the mechanisms that have been described in this literature are known loosely as the credit channel of monetary transmission. In this article, we summarize our current view of the credit channel and its role inside the "black box" of monetary policy transmission. We don't think of the credit channel as a distinct, free-standing alternative to the traditional monetary transmission mechanism, but rather as a set of factors that amplify and propagate conventional interest rate effects. For this reason, the term "credit channel" is something of a misnomer; the credit channel is an enhancement mechanism, not a truly independent or parallel channel. Moreover, this nomenclature has led some authors to focus-inappropriately, in our view-on the behavior of credit aggregates, a point that we discuss later. However, it is probably too late to change the terminology now. According to the credit channel theory, the direct effects of monetary policy on interest rates are amplified by endogenous changes in the extemnal finance premium, which is the difference in cost between funds raised externally (by issuing equity or debt) and funds generated internally (by retaining earnings).2 The size of the external finance premium reflects imperfections in the credit markets that drive a wedge, between the expected return received by lenders and the costs faced by potential borrowers. According to the "credit view," a change in monetary policy that raises or lowers open-market interest rates tends to change the external finance premium in the same direction. Because of this additional effect of policy on the external finance 2 Although we focus on the external finance premium in this article, it should be clear that credit market imperfections may also affect the safe real interest rate in general equilibrium (for example, Stein, 1995b). In emphasizing the external finance premium, we implicitly assume that, over the time span under consideration, the Fed can set the short-term safe rate at the level that it desires

Ben s. bernanke and mark gertler 2 premium, the impact of monetary policy on the cost of borrowing broadly defined- and, consequently, on real spending and real activity-is magnified Why should actions taken by the central bank have any effect on the external finance premium in credit markets? In this article we describe two possible link ges. The first of these, the balance sheet channel, stresses the potential impact of changes in monetary policy on borrowers balance sheets and income statements, including variables such as borrowers' net worth, cash flow and liquid assets. The second linkage, the bank lending channel, focuses more narrowly on the possible effect of monetary policy actions on the supply of loans by depository institutions In our view, the existence of a balance sheet channel seems fairly well established e bank lending channel is more controversial, Institutional changes during the past 15 years or so have rendered the bank lending channel, at least as traditionally con- ceived, somewhat less plausible. On the other hand, certain other developments may increased the importance of bank lending in monetary transmission. In this paper we do not attempt to draw strong conclusions about the relative importance of the balance sheet and bank lending channels. Instead, we try to make the case for the broader view that allowing for a credit channel of some type is important for under- standing the response of the economy to changes in monetary policy How the Economy Responds to Monetary Policy Shocks: Facts and Puzzles To set the stage for our discussion of how monetary policy works, we first fill in some of the details about what happens in the economy after a change in mon etary policy (a tightening, say)occurs. We do so by extending recent empirical work on the effects of monetary policy to consider its impact on some of GDP. We emphasize four basic facts about the response of the economy to mon- etary policy shocks Fact I: Although an unanticipated tightening in monetary policy typically has only transitory effects on interest rates, a monetary tightening is followed by sustained declines in real GDP and the price level Fact 2: Final demand absorbs the initial impact of a monetary tightening, falling relatively quickly after a change in policy. Production follows final demand downward, but only with a lag, implying that inventory stocks rise in the short run Ultimately, however, inventories decline, and inventory disinvestment accounts for a large portion of the decline in GDP. Fact 3: The earliest and sharpest declines in final demand occur in residential investment, with spending on consumer goods(including both durables and non- durables)close behind nkages have been extensively discussed in the literature. Surveys of related material include others) Bernanke (1993a), Kashyap and Stein(1994), Hubbard(1994)and Bernanke, Gertler arist(forthcoming)

Ben S. Bernanke and Mark Gertler 29 premium, the impact of monetary policy on the cost of borrowing broadly defined and, consequently, on real spending and real activity-is magnified. Why should actions taken by the central bank have any effect on the external finance premium in credit markets? In this article we describe two possible linkages.3 The first of these, the balance sheet channel, stresses the potential impact of changes in monetary policy on borrowers' balance sheets and income statements, including variables such as borrowers' net worth, cash flow and liquid assets. The second linkage, the bank lending channel, focuses more narrowly on the possible effect of monetary policy actions on the supply of loans by depository institutions. In our view, the existence of a balance sheet channel seems fairly well established. The bank lending channel is more controversial. Institutional changes during the past 15 years or so have rendered the bank lending channel, at least as traditionally conceived, somewhat less plausible. On the other hand, certain other developments may have increased the importance of bank lending in monetary transmission. In this paper we do not attempt to draw strong conclusions about the relative importance of the balance sheet and bank lending channels. Instead, we try to make the case for the broader view that allowing for a credit channel of some type is important for understanding the response of the economy to changes in monetary policy. How the Economy Responds to Monetary Policy Shocks: Facts and Puzzles To set the stage for our discussion of how monetary policy works, we first fill in some of the details about what happens in the economy after a change in monetary policy (a tightening, say) occurs. We do so by extending recent empirical work on the effects of monetary policy to consider its impact on some key components of GDP. We emphasize four basic facts about the response of the economy to monetary policy shocks. Fact 1: Although an unanticipated tightening in monetary policy typically has only transitory effects on interest rates, a monetary tightening is followed by sustained declines in real GDP and the price level. Fact 2: Final demand absorbs the initial impact of a monetary tightening, falling relatively quickly after a change in policy. Production follows final demand downward, but only with a lag, implying that inventory stocks rise in the short run. Ultimately, however, inventories decline, and inventory disinvestment accounts for a large portion of the decline in GDP. Fact 3: The earliest and sharpest declines in final demand occur in residential investment, with spending on consumer goods (including both durables and nondurables) close behind. 3Both linkages have been extensively discussed in the literature. Surveys of related material include (among others) Bernanke (1993a), Kashyap and Stein (1994), Hubbard (1994) and Bernanke, Gertler and Gilchrist (forthcoming)

30 Jounal of Economic Perspectives fact 4: Fixed business investment eventually declines in response to a monetary tightening, but its fall lags behind those of housing and consumer durables and, ndeed, behind much of the decline in production and interest rates Responses to Policy Shocks: Evidence from Vector Autoregressions These four facts are illustrated by Figures 1-3, which show the dynamic re- sponses of various economic aggregates to an unanticipated tightening of monetar policy. Figures 1-3 are generated by the technique of"vector autoregression. A vector autoregression, or VAR, is a system of ordinary least-squares regressions, in which each of a set of variables is regressed on lagged values of both itself and th other variables in the set. VARs have proved to be a convenient method of sum- marizing the dynamic relationships among variables, since, once estimated, they can be used to simulate the response over time of any variable in the set to either an"own"disturbance(that is, a disturbance to the equation for which the variable is the dependent variable) or a disturbance to any other variable in the system The VARs that we employ here include various combinations of macroeconomic variables and, additionally, the federal funds interest rate. Following Bernanke and Blinder(1992), Christiano, Eichenbaum and Evans(1994a, b)and others, we em- ploy the federal funds rate as an indicator of the stance of monetary policy; this means that we identify the disturbances to the funds-rate equation in the vaR as shocks to monetary policy, and we interpret the responses of other variables in the system to a funds-rate shock as the structural responses of those variables to an unanticipated change in monetary policy. Because we are interested in observing ine timing of responses to monetary shocks, we use monthly data. The sample period on which Figures 1-3 are based is January 1965 through December 1993 (subsample results are similar) GDP"igure l is based on a VAR system that includes the log of real GDP, the log of the deflator, the log of an index of commodity prices and the federal funds rate(in ercentage points), in that order. Real GDP and the gDp deflator are included as broad sures of economic activity and prices, and the commodity price index is intended to control for oil price shocks and other supply-side factors influencing output and The use of VARs in macroeconomics was pioneered by Sims(1980); for a comprehensive recent di cussion, see Watson(1994) "Bernanke and Blinder(1992)argue that the Fed has often used the funds rate which is the interest te prevailing in the market for bank reserves, as its primary policy indicator(particularly before 1979) Bernanke and Mihov(1995)estimate a model of the Feds operating procedures and find that funds. te targeting describes Fed behavior particularly well prior to 1979 and from 1988 to the present. In the results discussed here are not dependent on using the funds rate as the monetary policy indicator; similar results are obtained when using reserves-based indicators(see, for example, Strongin 1992)or indicators developed through historical analysis(for example, Romer and Romer, 1989) t We constructed monthly data for real GDP and the GDP deflator by interpolation methods, using a f monthly series to provide the within-quarter information. Bernanke and Mihov( 1995)offer Results using noninterpolated monthly output and price data-for example, the industrial pr index and the CPI (excluding shelter)-yield very similar results. Twelve lags of each variable onstant term are included in each equation of the VAR

30 Journal of Economic Perspectives Fact 4: Fixed business investment eventually declines in response to a monetary tightening, but its fall lags behind those of housing and consumer durables and, indeed, behind much of the decline in production and interest rates. Responses to Policy Shocks: Evidence from Vector Autoregressions These four facts are illustrated by Figures 1-3, which show the dynamic responses of various economic aggregates to an unanticipated tightening of monetary policy. Figures 1-3 are generated by the technique of "vector autoregression." A vector autoregression, or VAR, is a system of ordinary least-squares regressions, in which each of a set of variables is regressed on lagged values of both itself and the other variables in the set. VARs have proved to be a convenient method of summarizing the dynamic relationships among variables, since, once estimated, they can be used to simulate the response over time of any variable in the set to either an "own" disturbance (that is, a disturbance to the equation for which the variable is the dependent variable) or a disturbance to any other variable in the system.4 The VARs that we employ here include various combinations of macroeconomic variables and, additionally, the federal funds interest rate. Following Bernanke and Blinder (1992), Christiano, Eichenbaum and Evans (1994a,b) and others, we employ the federal funds rate as an indicator of the stance of monetary policy; this means that we identify the disturbances to the funds-rate equation in the VAR as shocks to monetary policy, and we interpret the responses of other variables in the system to a funds-rate shock as the structural responses of those variables to an unanticipated change in monetary policy.5 Because we are interested in observing the fine timing of responses to monetary shocks, we use monthly data. The sample period on which Figures 1-3 are based is January 1965 through December 1993 (subsample results are similar). Figure 1 is based on a VAR system that includes the log of real GDP, the log of the GDP deflator, the log of an index of commodity prices and the federal funds rate (in percentage points), in that order. Real GDP and the GDP deflator are included as broad measures of economic activity and prices,6 and the commodity price index is intended to control for oil price shocks and other supply-side factors influencing output and 'The use of VARs in macroeconomics was pioneered by Sims (1980); for a comprehensive recent discussion, see Watson (1994). 5Bernanke and Blinder (1992) argue that the Fed has often used the funds rate, which is the interest rate prevailing in the market for bank reserves, as its primary policy indicator (particularly before 1979). Bernanke and Mihov (1995) estimate a model of the Fed's operating procedures and find that fundsrate targeting describes Fed behavior particularly well prior to 1979 and from 1988 to the present. In any case, the results discussed here are not dependent on using the funds rate as the monetary policy indicator; similar results are obtained when using reserves-based indicators (see, for example, Strongin, 1992) or indicators developed through historical analysis (for example, Romer and Romer, 1989). 6 We constructed monthly data for real GDP and the GDP deflator by interpolation methods, using a variety of monthly series to provide the within-quarter information. Bernanke and Mihov (1995) offer details. Results using noninterpolated monthly output and price data-for example, the industrial production index and the CPI (excluding shelter) -yield very similar results. Twelve lags of each variable and a constant term are included in each equation of the VAR

Ben S. Bernanke and Mark Gertler 31 Figure I Responses of Output, Prices and Federal Funds Rate to a Monetary Policy Shock 0.008 - Real GDP 0.006 - - - GDP Deflator /I\ - - Funds Rate 0.004 0.002 \ /\ _ 0.000 e -0.002 0 4 8 12 16 20 24 28 32 36 40 44 48 Months inflation.7 The figure shows the estimated dynamic responses of log real GDP, the log of the GDP deflator and the funds rate to a positive, one-standard-deviation shock to the funds rate (which we interpret as an unanticipated tightening of monetary policy). As output and prices are measured in logs, the responses can be interpreted as proportions (that is, .001 = 0.1 percent) of baseline levels. According to the estimated response patterns shown in Figure 1, GDP begins to decline about four months after a tightening of monetary policy, bottoming out about two years after the shock. The price level remains inert for about a year, then begins to decline, well after the drop in GDP begins. (The commodity price index, not shown, drops more quickly.) Finally, after rising sharply initially, the funds rate begins to fall precipitously after three to four months. Nine to 12 months after the policy innovation, the deviation of the funds rate from its baseline path is only a quarter or less of the initial shock; at two years out, the funds rate is essentially back to trend. These patterns are summarized above as Fact 1. These results in monthly data are quite consistent with the patterns observed by others in quarterly data and for various shorter sample periods. Although we do not show standard error bands in the figures, these responses and all that we report subsequently are statistically significant at conventional levels. 7 Sims (1992) and Christiano, Eichenbaum and Evans (1994b) discuss in detail the rationale for including the index of commodity prices; see also Sims and Zha (1993). Inclusion of this variable, along with measures of output and the general price level, has become conventional in the recent VAR-based literature on monetary policy

32 Journal of Economic Perspectives Figure 2 Responses of Final Demand and Inventories to a Monetary Policy Shock 0.0008 I t I I 0.0004 -0.0004 \ -0.0008 - -0.0012 0 4 8 12 16 20 24 28 32 36 40 44 48 Moniths To allow us to look at the economy's response to a monetary shock in closer detail, Figure 2 replaces log real GDP from the first VAR with two variables that sum to GDP, final demand and inventory investment. We measure both final demand and inventories relative to trend GDP, proxied by a six-year moving average of GDP; this normalization makes the magnitudes of the changes in the two variables comparable (both can be interpreted as fractions of trend GDP) and avoids taking the log of a series (inventory investment) that is sometimes negative. Note from Figure 2 that final demand drops quickly following an unanticipated tightening of monetary policy. In contrast, inventories build up for a period of several months before beginning to decrease, implying that the fall in final demand leads the decline in aggregate production (real GDP). The fall in inventories, when it occurs, appears to account for a substantial portion of the initial drop in output, which is consistent with Blinder and Maccini's (1991) evidence on the importance of inventory disinvestment in recessions. These results are summarized as Fact 2. Next, we explore what happens when we include various components of GDP in the VAR used in Figure 1. These series are added one at a time to the base VAR, although adding them in combinations gives very similar results. To make the magnitudes of changes comparable, the GDP components are also left in levels and normalized by trend GDP.8 Figure 3 shows the responses to a monetary contraction of 8We continue to order the funds rate last, which has the effect of assuming that the Fed uses contemporaneous economic information, but that innovations in monetary policy do not feed back to the rest of the economy until the next month (Bernanke and Blinder, 1992; Bernanke and Mihov, 1995)

Ben S. Bernanke and Mark Gertler 33 Figure 3 Responses of Spending Components to a Monetary Policy Shock 0.0002 0.0001 0.0000 l -\-/t -0.0002 \_ / -0.0005 \2 - Consumer Durables \. _/ -- - Nonidurable Conisumption -0.0006 - - Residential Investment Rusincss Fixcd I-ivestmcn-t -0.0007 0 4 8 12 16 20 24 28 32 36 40 44 48 Moniths some important components of private domestic spending. As noted in Fact 3 above, residential investment drops sharply following a monetary tightening and accounts for a large part of the initial decline in final demand. Next in importance are consumer durables and nondurables, which also contribute significantly to the fall in final demand. (Nondurables react by much less in percentage terms than durables do, but they make a similar total contribution to the downturn owing to their larger share in overall economic activity.) Figure 3 shows that business fixed investment also declines following a monetary tightening, but with a greater lag than other types of spending (Fact 4). An interesting result (not shown in the figure) is that equipment investment accounts for nearly all of the decline in fixed investment; structures investment by businesses appears to respond very litfle to a monetary policy shock. Does the Conventional Story Fit the Facts? In a number of ways, the behavior of the economy shown in Figures 1-3, and as summarized by Facts 1-4, is consistent with the conventional analysis of monetary policy transmission. According to the standard story, the Fed has leverage over the short-term real rate because prices are sticky. In turn, the change in real rates affects aggregate demand. The slow response of the GDP deflator in the wake of the monetary contraction (Figure 1) and the quick response of final demand (Figure 2) are consistent with this scenario. Apparently, so too is the fact that durables spendingtraditionally thought to be the most interest-sensitive part of aggregate demanddisplays large responses to monetary policy shocks. But there are some important puzzles. First among these is the magnitude of

34 Journal of Economic Perspectives the policy effect. We and many other researchers have found that the real economy is powerfully affected by monetary policy innovations that induce relatively small movements in open-market interest rates. But as we discussed earlier, empirical studies have not typically found commensurately strong cost-of-capital effects on the various components of private spending Second is the issue of timing. As shown by Figure 1, the interest-rate spike associated with an unanticipated monetary tightening is largely transitory; the fed eral funds rate is virtually back to trend at eight to nine months after the shock. Yet some important components of spending do not begin to react until after most of the interest-rate effect is past. For example, the bulk of the response of business fixed investment occurs during the period between six and 24 months after the shock(Figure 3). Inventories actually rise during the first three to four months after a monetary tightening; they begin to decline only in the period during which interest rates are falling sharply back to trend(Figure 2). The poor correspondence in timing between changes in interest rates and movements in some components of spending observed in Figures 1-3 no doubt helps to explain why robust effects of interest rates on spending have been hard to pin down empirically a third issue is the composition of the spending effects Because monetary policy as its most direct effects on short-term rates. it would seem that it should have its most significant impact on spending on assets with shorter lives-some types of nventories and consumer durables, for example. Yet the most rapid (and in per- centage terms, by far the strongest) effect of monetary policy is on residential in- vestment(Figure 3). This finding is puzzling because residential investments are typically very long-lived and thus(according to the conventional view) should be most sensitive to long-term real interest rates, not the short-term rates most directly nfluenced by the Fed. At the same time the other major type of long-lived invest ment, business structures investment, does not seem to be much affected by mon etary policy actions. It is not immediately obvious why residential and business struc- tures investment behavior should differ in this way The Credit Channel of Monetary Transmission We have argued that it is difficult to explain the magnitude, timing and com- position of the economys response to monetary policy shocks solely in terms of conventional interest-rate(neoclassical cost-of-capital)effects. The mechanisms col- lectively known as the credit channel help to fill in the gaps in the traditional story. Underlying our conception of the credit channel is the following basic premise whenever frictions-such as imperfect information or costly enforcement of Adjustment costs are an alternative potential explanation of the lagged response of investment terest rates.h most attempts to fit neoclassical investment equations with adjustment costs have estimated those costs to be implausibly large. If adjustment costs are important, there remains the que tion of why managers should respond at all to interest-rate fluctuations that they should expect to be

34 Journal of Economic Perspectives the policy effect. We and many other researchers have found that the real economy is powerfully affected by monetary policy innovations that induce relatively small movements in open-market interest rates. But as we discussed earlier, empirical studies have not typically found commensurately strong cost-of-capital effects on the various components of private spending. Second is the issue of timing. As shown by Figure 1, the interest-rate spike associated with an unanticipated monetary tightening is largely transitory; the federal funds rate is virtually back to trend at eight to nine months after the shock. Yet some important components of spending do not begin to react until after most of the interest-rate effect is past. For example, the bulk of the response of business fixed investment occurs during the period between six and 24 months after the shock (Figure 3).9 Inventories actually rise during the first three to four months after a monetary tightening; they begin to decline only in the period during which interest rates are falling sharply back to trend (Figure 2). The poor correspondence in timing between changes in interest rates and movements in some components of spending observed in Figures 1-3 no doubt helps to explain why robust effects of interest rates on spending have been hard to pin down empirically. A third issue is the composition of the spending effects. Because monetary policy has its most direct effects on short-term rates, it would seem that it should have its most significant impact on spending on assets with shorter lives-some types of inventories and consumer durables, for example. Yet the most rapid (and in percentage terms, by far the strongest) effect of monetary policy is on residential investment (Figure 3). This finding is puzzling because residential investments are typically very long-lived and thus (according to the conventional view) should be most sensitive to long-term real interest rates, not the short-term rates most directly influenced by the Fed. At the same time, the other major type of long-lived investment, business structures investment, does not seem to be much affected by monetary policy actions. It is not immediately obvious why residential and business structures investment behavior should differ in this way. The Credit Channel of Monetary Transmission We have argued that it is difficult to explain the magnitude, timing and composition of the economy's response to monetary policy shocks solely in terms of conventional interest-rate (neoclassical cost-of-capital) effects. The mechanisms collectively known as the credit channel help to fill in the gaps in the traditional story. Underlying our conception of the credit channel is the following basic premise: whenever frictions-such as imperfect information or costly enforcement of 9Adjustment costs are an alternative potential explanation of the lagged response of investment to interest rates. However, most attempts to fit neoclassical investment equations with adjustment costs have estimated those costs to be implausibly large. If adjustment costs are important, there remains the question of why managers should respond at all to interest-rate fluctuations that they should expect to be temporary

Inside the black Box: The Credit Channel of Monetary Policy Transmission 35 contracts-interfere with the smooth functioning of financial markets, we expect to observe a wedge between the cost of funds raised externally (for example, hrough the issuance of imperfectly collateralized debt) and the opportunity cost of internal funds. This wedge, which we call the external finance premium, reflects the deadweight costs associated with the principal-agent problem that typically ex- ists between lenders and borrowers. Among the factors reflected in the external finance premium are the lender's expected costs of evaluation, monitoring and collection; the"lemons"premium that results from the fact that the borrower inevitably has better information about its prospects than does the lender; and the costs of distortions in the borrower's behavior that stem from moral hazard or from restrictions in the contract intended to contain moral hazard (for example, restric- tive covenants or collateral requirements According to advocates of the credit channel, monetary policy affects not only the general level of interest rates, but also the size of the extermal finance premium. This complementary movement in the extemal finance premium may help explain the strength, timing and composition of monetary policy effects better than is possible by reference to interest rates alone. Two mechanisms have been suggested to explain the link between monetary policy actions and the extermal finance premium: the balance sheet channel(sometimes called the net worth channel) and the bank lending chan- nel. In describing how these channels work, in the rest of this section, we focus on the behavior of firms. SI be extended to consumer behavior, like spending on housing and consumer durable The balance sheet Channel The balance sheet channel is based on the theoretical prediction that the ex ternal finance premium facing a borrower should depend on borrower's financial position. In particular, the greater is the borrower's net worth-defined operation- finance premium should be. Intuitively, a stronger hina the lower the external ally as the sum of her liquid assets and marketable collateral-t ion(greater net worth)enables a borrower to reduce her potential conflict of interest with the lender, either by self-financing a greater share of her investment project or pur chase or by offering more collateral to guarantee the liabilities she does issue. This asic insight underlies many real-world financial arrangements, such as the require- ment that borrowers meet certain financial ratios, that they post collateral andyor that they make down payments Since borrowers' financial positions affect the extermal finance premium, and thus the overall terms of credit that they face, fluctuations in the quality of borrowe balance sheets similarly should affect their investment and spending decisions.An extensive theoretical literature has exploited this idea to argue that endogenous pro- cyclical movements in borrower balance sheets can amplify and propagate business cycles, a phenomenon that has been referred to as the"financial accelerator. This approach has been supported by a wide range of empirical work linking balance sheet and cash flow variables to firms' decisions concerning fixed investment, inventories and other factor demands, and to household purchases of durables and housing. Ber- nanke, Gertler and Gilchrist(forthcoming) offer a discussion and references

Inside the Black Box: The Credit Channel of Monetary Policy Transmission 35 contracts-interfere with the smooth functioning of financial markets, we expect to observe a wedge between the cost of funds raised externally (for example, through the issuance of imperfectly collateralized debt) and the opportunity cost of internal funds. This wedge, which we call the external finance premium, reflects the deadweight costs associated with the principal-agent problem that typically exists between lenders and borrowers. Among the factors reflected in the external finance premium are the lender's expected costs of evaluation, monitoring and collection; the "lemons" premium that results from the fact that the borrower inevitably has better information about its prospects than does the lender; and the costs of distortions in the borrower's behavior that stem from moral hazard or from restrictions in the contract intended to contain moral hazard (for example, restrictive covenants or collateral requirements). According to advocates of the credit channel, monetary policy affects not only the general level of interest rates, but also the size of the external finance premium. This complementary movement in the external finance premium may help explain the strength, timing and composition of monetary policy effects better than is possible by reference to interest rates alone. Two mechanisms have been suggested to explain the link between monetary policy actions and the external finance premium: the balance sheet channel (sometimes called the net worth channel) and the bank lending channel. In describing how these channels work, in the rest of this section, we focus on the behavior of firms. Subsequently, we will discuss how the analysis may be extended to consumer behavior, like spending on housing and consumer durables. The Balance Sheet Channel The balance sheet channel is based on the theoretical prediction that the external finance premium facing a borrower should depend on borrower's financial position. In particular, the greater is the borrower's net worth-defined operationally as the sum of her liquid assets and marketable collateral-the lower the external finance premium should be. Intuitively, a stronger financial position (greater net worth) enables a borrower to reduce her potential conflict of interest with the lender, either by self-financing a greater share of her investment project or purchase or by offering more collateral to guarantee the liabilities she does issue. This basic insight underlies many real-world financial arrangements, such as the requirement that borrowers meet certain financial ratios, that they post collateral and/or that they make down payments. Since borrowers' financial positions affect the external finance premium, and thus the overall terms of credit that they face, fluctuations in the quality of borrowers' balance sheets similarly should affect their investment and spending decisions. An extensive theoretical literature has exploited this idea to argue that endogenous procyclical movements in borrower balance sheets can amplify and propagate business cycles, a phenomenon that has been referred to as the "financial accelerator." This approach has been supported by a wide range of empirical work linking balance sheet and cash flow variables to firms' decisions concerning fixed investment, inventories and other factor demands, and to household purchases of durables and housing. Bernanke, Gertler and Gilchrist (forthcoming) offer a discussion and references