货币政策的分配效应：以马来西亚为证

\Al_P___ __ __/S21r0 POLICY RESEARCH WORKING PAPER 2170 The Distributional Policymakers in Malaysia should weigh the Consequences of distributional consequences M onetary Policy of policy actions. They should also consider measures to allevfate the disproportionate Evidence from Malaysia impact that market imperfections have on small and medium-size industries Ilker Domaq The World Bank East Asia and Pacific Region Poverty Reduction and Economic Management Unit August 1999  | POLICY RESEARCH WORKING PAPER 2170 Summary findings Domac provides a descriptive analysis of credit and given policy shift. monetary policies in Malaysia and investigates the Policymakers should weigh the distributional distributional consequences of monetary policy there by consequences of policy actions and should consider focusing on small and medium-size industries and large measures to alleviate the disproportionate impact that manufacturing firms. market imperfections have on small and medium-size Domac suggests that "payoff" or "default" risk - industries. as captured by the spread between safe and risky debt - Measures to alleviate information asymmetry in credit is still well above its pre-crisis level, underscoring the markets - including the promotion of cooperative or increased agency costs of external finance. The decline in mutual guarantee schemes for small and medium-size lending activity in the first half of 1998 can be attributed enterprises - are one useful option. Groups of firms in to the reduced supply of bank credit relative to demand. Southern Europe have made wide use of mutual Empirical results from vector autoregression analysis guarantee schemes -usually within a specific industry demonstrate that monetary tightening disproportionately - to provide a privately organized "insurance system" affects small and medium-size enterprises. for lending banks that allows the banks to rely less on the Moreover, monetary shocks contribute substantially assets of individual companies within the group in more to small and medium-size firms' variance of making loan decisions. The pooling effects of such a production (71 percent) than to that of large system would reduce the risk to the bank of default and manufacturing firms (30 percent). would give members of the society an incentive to reveal These findings corroborate the notion that small and information to the society that they might hesitate to medium-size industries face greater market give to the bank. imperfections, which in turn magnify the effects of a This paper - a product of the Poverty Reduction and Economic Management Unit, East Asia and Pacific Region - is part of a larger effort in the region to analyze the patterns and consequences of the East Asian crisis, with particular reference to the links between the real and financial sectors. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Armanda Carcani, room 04-015, telephone 202-473-0241, fax 202-522- 2751, Internet address acarcani@worldbank.org. Policy Research Working Papers are also posted on the Web at http:// www.worldbank.org/html/dec/Publications/Workpapers/home.html. The author may be contacted at idomac @worldbank.org. August 1999. (36 pages) The Policy ResearcP Working Paper Series disseminates the findings of work in progress to encourage the exctange of ideas about development issues. An objective of the series is to get the findings out qyuickly, even if the presentations are less than fully polisbed. The papers carry; the names of the auitbors and sbould be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the couintries they represent. Produced by the Policy Research Dissemination Center  The Distributional Consequences of Monetary Policy: Evidence from Malaysia by Ilker Domac* JEL Classification Numbers: E44, E5 1, E52 * llker Domac is an economist in the Europe and Central Asia Region. I would like to thank Cevdet Denizer, Giovanni Ferri, Ejaz Ghani, Arvind Gupta, Kim Leng, Maria Soledad Martinez Peria, Saha Meyanathan, Ngozi Okonjo-lweala, Vivek Suri, and Michael Yap for their helpful comments and contributions.   1. Introduction The initial policy response to the Asian crisis which, inter alia, included tight monetary and fiscal policy along with structural measures, has attracted as much attention as the crisis itself. In particular, many experts questioned the effectiveness of the use of high interest rates to defend the currency, to restore confidence, and to attract capital inflows. It has been argued that high interest rates not only undermined the recovery prospects of the economies in the region, but also that they increased the fragility of the banking sector.' There are, however, studies which dispute this argument and contend that the monetary tightening was neither excessive nor a major factor contributing to the sharp decline in output.2 The different conclusions reached by different studies appears to stem from the fact that those who argue that monetary policy tightening was not extreme and that it was unlikely to have been a major factor behind the output decline focus largely on aggregate consequences of policy actions. By contrast, those who claim that monetary policy was excessively tight and that it contributed to increased bankruptcies, thus undermining prospects for economic recovery, recognize the importance of distributional consequences ofpolicy actions. More specifically, the first view -- often characterized as the money view -- argues that higher default-risk-free interest rates following a monetary contraction decreases desired Feldstein (1998) contends that an increase in interest rates will lead to widespread bankruptcies when: (i)inflation is low; (ii) the roll over of bank loans and the demand for the domestic currency hinges more on confidence than on interest rates; and (iii) the failures will reduce the prospect of loan repayment. Although it is true that the depreciation would increase the risk of bankruptcies among companies with large foreign debt, the overall damnage would be less extensive than the bankruptcies caused by a very high domestic interest rate that would hurt every company. Furthermore, as Drazen and Masson (1994) have demonstrated, if the costs of implementing tight policies are too high, the temporary policy would actually reduce credibility because investors know that the policy could not be sustained. Similarly, to the extent that investors believe that high interest rates are not sustainable in the presence of a fragile banking sector and to the extent they anticipate a government bailout of the financial sector, they would expect that tight monetary policy will eventually be abandoned. As a result, higher interest rates will not be an effective tool to defend the domestic currency. See also Stiglitz (1998), Ding, Domac, and Ferri (1998), and Domav and Ferri (1998) for more on the potential adverse consequences of high interest rates. 2 See, for instanlce, Lane, Ghosh. Hamann, I'hillips, Schulze-Ghattas and Tsikate (1998) as well as Goldfajn and Baig (1998). I  investment by firms and households.3 While desired investment declines, the reduction in business and lhousehold capital affects the least productive projects. Since the most profitable projects continue to be funded, there are no direct efficiency losses associated with the distributional aspects of the policy-induced interest rate increase. On the otler hand, the second view -- often referred to as the credit view -- analyzes the ways in which mnonetary policy can influence the interaction between borrowers and lenders in financial markets and the resulting consequences for economic activity.4 By emphasizing a combinatioti of capital market imperfections and portfolio balance effects based on imperfect asset substituitability, the credit view suggests the possibility that the policy's incidence may differ substantially across agents in the economy. Moreover, the impact of policy shifts has to do with the characteristics of the individuals that are not linked to the inherent creditworthiness of the investment projects. An entrepreneur may be turned down by the lender simply because of a currently low net worth, in spite of the social return to the project. It is, therefore, important to understand whether the decline in investment caused by monetary policy shifts has these consequences. The paper has two main goals. First, it attempts to provide a brief descriptive analysis of the credit and monetary conditions in the aftermath of the crisis in Malaysia, with a specific reference to the credit channel literature. Second, it aims to empirically investigate the distributional consequences of policy actions by focusing on the response of small and mediurn size industries and large manufacturing firms to monetary policy shifts in Malaysia. Since the 3 'Ihe money view relies on four important assumptions: (i) the central bank must control the supply of outside money for which there are imperfect substitutes; (ii) the central bank can influence real and nominal short-term interest rates -- prices do not adjust instantaneously; (iii) policy induced changes in real interest rates have an impact on longer-term interest rates affecting household and business spending decisions; and (iv) changes in interest sensitive spending in response to a monetary policy innovation correspond reasonably well with observed output responses to such innovations. 4 As Bernanke and Gertler (1995) point out, the credit view should not be considered as a distinct, free-standing alternative to the money view, but rather as a set of factors that amplify and propagate conventional interest rate effects. As they 2  existing literature on this topic has largely concentrated on industrialized countries, this paper will contribute to the literature by considering a developing nation -- Malaysia. The results from the descriptive analysis suggest that "payoff' or "default" risk, as captured by the spread between risky and safe debt, is still well above its pre-crisis level, thus highlighting the increased agency costs of external finance. The empirical results from the vector autoregression analysis demonstrate that monetary tightening has a larger impact on small medium size industries (SMIs) than it does on large manufacturing firms (LMFs). The findings also underscore that the effect of monetary shocks on the production of SMIs is more persistent than that of LMFs, and monetary shocks contribute substantially more to the variance of the production of SMIs than that of LMFs. Therefore, the findings of this paper lend support to the notion that SMIs face greater market imperfections -- imperfections which in turn magnify the effects of a given policy slhift. I'ihe remainder of the paper is organized as follows. Section 2 describes the analytical framework for evaluating disruptions to the credit market. Section 3, drawing on the framework, provides a descriptive analysis of the credit and monetary conditions. Section 4 presents the empirical framework and results. Finally, Section 5 concludes the paper. 2. Identifying Credit Market Disruptions: A General Framework According to the money view, reductions in the quantity of outside money increase real rates of return, which in turn leads to a decline in investment. Since this theory is based on the notion that there are no externalities or market imperfections, only the least socially productive projects drop out. Although following a monetary contraction, capital stock is marginally lower, the allocation of the decline across sectors is nonetheless socially efficient. suggest, the tern "credit view' or "credit channel" is something of a misnomer; the credit channel is an enhancement 3  The credit view, on the other hand, argues that the direct effects of monetary policy on interest rates are augmented by endogenous changes in external finance premium -- the difference in cost between funds raised externally (by issuing equity or debt) and funds generated internally (by retaining earnings). The magnitude of the external finance premium reflects imperfections in the credit market -- imperfections that create a wedge between the expected return received by the lenders and the cost faced by potential borrowers. This theory argues that a change in monetary policy will alter the external finance premium in the same direction. As a result of this additional policy effect on the external finance premium, the effect of monetary policy on the cost of borrowing and on real spending, as well as real activity, is amplified. Why should policy shifts induced by the central bank have an impact on the external finance premium in credit markets? The credit view offers two distinct linkages arising from information problems in credit markets: those that operate through effects on firms' and households' balance sheets -- a balance sheet channel -- and those that operate through effects on bank lending -- a bank lending channel.5 Table 1: The Link Between Monetary Policy and Output Types of Monetary Transmission Channels According to the Credit View Bank Lending Channel Balance Sheet Channel Mtznbank depositst=>bank IoansT=*IT=*Y7 MlPet=P>adverse selection;, moral hazard4K>lendingT=lT4=I>YT Cash Flow Channel Unanticipated Price Household Level Channel Liquidity Effects MT=* i4-=>cash flowl'=> MT=1 unanticipated Ptf MT=>PJ=>financial adverse selection4-, moral =>adverse selectionl, assetsT =:>likelihood of hazardJLoendingt=, IT=: Yt moral hazard-=>lendingT financial distress4-> =m rt=> Yt consumer durable and housing expendituret=> Yt MT': expansionary monetaly policy; l: investment spending; Y: GDP; P: price level; Pe: equity prices; i: interest rates mechanism, not a truly independent or parallel channel. See Table 1. which draws on Mishkin (1996), for a schematic presentation of the transmission of monetary policy according to the credit view. 4  The balance sheet channel stems from the fact that monetary policy has non-neutral effects on the balance sheet of firms. A tightening of monetary policy stance leads to a decline in the net worth of firms for two reasons. First, as contractionary monetary policy causes interest rates to rise, the servicing of outstanding debt becomes more expansive and firms experience decreasing cash flows. Second, rising rates are followed by declines in asset prices. As a result, the value of marketable collateral declines, and the wedge between the interest rates at which corporations can borrow and the yields on otherwise analogous risk-free assets rises. These adverse developments lead to more stringent conditions under which external finance becomes available. In practice, the extemal cost of finance rises, which ultimately causes interest- sensitive expenditures such as investment to decline. A subset of the balance sheet channel, referred as the cash flow channel, operates through its effects on cash flow.6 Expansionary monetary policy, which lowers the interest rate, also leads to an improvement in corporate balance sheets since it increases cash flow. The increase in cash flow improves the balance sheet in that it raises the liquidity of the firm (or household) and thus makes it easier for lenders to know whether the firm (or household) will be able to pay its bills. Consequently, the adverse selection and moral hazard problems become less severe, leading to an increase in economic activity. The bank lending channel is based on the notion that banks play a special role in the financial system in that they are particularly well suited to solve asymmetric information problems in the credit market. Due to the special role of banks, certain borrowers will not have access to the credit markets unless they borrow from banks. As long as there is no perfect substitutability of retail bank deposits with other sources of funds, the operation of the bank 6 To save space, unanticipated price level channel and household liquidity effects will not be discussed here. See Mishkin (1996) for a comprehensive discussion of these balance sheet channels. 5  lending channel of monetary transmission can be explained as follows. Relaxation of monetary policy, which increases both bank reserves and bank deposits, raises the quantity of bank loans available. Since many borrowers rely on bank loans to finance their activities, this rise in loans will lead to an increase in investment (and possibly consumer) spending. This implies that monetary policy shifts will have a greater effect on small and medium-sized enterprises (SMEs), that are relatively dependent on bank loans, than it has on large firms, that can directly access the credit market via stock and bond markets without going through banks.7 Box 1. What are the Adverse Consequences of Increases in Interest Rates? A quick review of the literature points out three main categories of the adverse effects of increases in interest rates on both lending and aggregate economic activity. First, as Stiglitz and Weiss (1981) demonstrated, asymmetric information and the resulting adverse selection problem can lead to credit rationing in which some borrowers are denied loans even when they are willing to pay a higher interest rate. Therefore, a higher interest rate leads to even greater adverse selection; because of the resulting increase in adverse selection, banks will want to make fewer loans, possibly producing a steep decline in lending that will lead to substantial decline in investment and aggregate economic activity. Second, increases in interest rates have a significant impact on the balance sheets of both firms and households. As Bernanke and Gertler (1995) pointed out, a rise in interest rates, and therefore in households' and finns' interest payments, decreases firms' cash flow, which causes a deterioration in their balance sheets. This, in turn, makes it harder for lenders to know whether the firm or household will be able to pay its bills. Consequently, adverse selection and moral hazard problems become more severe for potential lenders to these firms and households, leading to a decline in lending and economic activity. Third, increases in interest rates have an important effect on bank balance sheets. Since banks typically borrow short and lend long, a rise in interest rates directly causes a decline in net worth (the interest rate rise lowers the value of assets with longer duration more than it raises the value of liabilities with shorter duration). Therefore, even if the credit quality of banks were to remain unaffected, a rise in interest rates causes a decline in net worth which then leads to a decline in bank lending. An additional reason for SMEs being disproportionately affected by credit channel effects emanates from the possibility that the monetary squeeze leads to a flight to quality in bank lending. More specifically, banks may respond to the monetary restriction, not only by restraining credit generally, but also by adopting more stringent lending policies vis-a-vis 7 See Certler and Gilchrist (1994) for more on this. 6  customers that are perceived to be less credit worthy.8 That is, when a deposit drain squeezes their resources, banks will attempt to select customers who are ex ante more credit-worthy: e.g. those lhaving a more established credit record or those able to post more collateral.9 As underscored by Bemanke, Gertler, and Gilchrist (1996), the flight to quality in bank lending may, in turn, trigger afinancial accelerator effect along the following causal chain: the negative shock pushes the economy into a recession; the recession tightens borrowing constraints; tighter borrowing constraints amplify the recession, and the cycle goes on.'0 The flight to quality in bank lending could also come in the form of banks' increasing their holdings of government securities in relation to their assets, in response to a tightening of monetary conditions. This, combined with the decrease in loans in relative to bank assets, is often interpreted as a sign of a declining bank willingness to extend new loans. Finally, there are two additional factors causing SMEs to be disproportionately penalized by the credit channel effects. First, to the extent that the smaller banks are viewed as less likely to be bailed out by the Government, they may be the ones to suffer most in the deposit flight."1 The fact that SMEs, more than other firms, typically rely on small bank lending implies that an additional credit squeeze may hurt them disproportionately.'2 Second, for a small firm, the cost of administering bankruptcy proceedings is larger, relative to the amount borrowed, than for a 8 Bernanke. Gertler, and Gilchrist (1996) report evidence in line with this hypothesis. A negative bias similar to that regarding SMEs might apply to fast-growing firms, since they have a higher ratio of expected future profits to the current value of physical assets and thus can provide lower collateral. 9 Lenders perceive SMEs to be more risky since they generally have a shorter track record and typically release less --and less structured-- information. 10 See Kyiotaki and Moore (1997) for more on this. Lang and Nakamura (1995) report evidence of a flight to quality in bank lending in the US. II Kashyap and Stein (1994, 1997) contend that small banks, rather than large ones, are more likely to be hit by monetary restrictions. 12 Berger, Kashyap and Scalise (1995) demonstrate a strong correlation between relative size of the lending bank and relative size of the borrowing firm in the US: small firms tend to borrow from small banks and large firms tend to borrow from large banks. 7  large firm. Therefore, a rise in market interest rates will lead banks more than capital market lenders. In a crisis situation, during which the relationship between monetary policy instruments and nominal income changes drastically, assessing the monetary policy stance becomes intricate. Consequently, it may be misleading to focus solely on key indicators of monetary policy when evaluating the monetary policy stance and credit conditions in order to detect "a credit crunch". Evaluation of monetary and credit aggregates must be accompanied with a more detailed investigation of the channels through which firms, banks, and households are affected by changes in monetary policy. Box 2. Illustration of the Relevant Spreads The following expression presents the set of interest rate spreads which capture the credit channel effects of the transmission mechanism of monetary policy: SPRo = LR-TB = SPRI + SPR2 where LR = lending rate; TB = Treasury bond rate; SPR1 = corporate bond rate - Treasury bond rate; SPR2 = lending rate - commercial paper rate; The difference between rates on corporate and on Treasury bonds (SPRI) measures the general risk premium as it is perceived by the market. If the balance sheet effect is at work, we expect that this spread will increase after the monetary restriction, reflecting the fact that private sector debt has become relatively riskier vis-a-vis sovereign debt. The spread between the lending rate and commercial paper (SPR2) is a proxy for the lending chainel effect. The difference between lending rates and commercial paper rates quantifies the premium that bank-dependent borrowers must pay in order to raise external finance relative to those firms able to issue debt on the market. The lending channel effect contends that this spread will increase in the aftennath of the monetary squeeze. Source: Doma,c and Ferri (1998) As widely acknowledged, a decline in either bank loans or a decline in their growth following a monetary tightening is not sufficient to pin down an adverse movement in banks' loan supply. This is because the decline could be induced either by the corporate sector 8  demanding less credit -- because fewer investments are undertaken -- or by bank reluctance to lend. By contrast, if the decline in (the dynamics of) bank loans -- a quantity variable -- is coupled witlh a widening of the spread between bank lending rates and interest rate on non-bank debt (See Box 2) -- a price based variable --, then it can be argued that an adverse shift in the bank supply of loans is curtailing credit. In fact, such a situation is consistent with only two possibilities: either supply has declined whereas demand has not, or supply has declined more than demand has. 3. Descriptive Analysis of the Credit and Monetary Conditions in the Aftermath of the Crisis Many years ago, Irving Fisher (1933) emphasized the interaction of debt and speculation in describing business and credit cycles: Thus, over-investment and over-speculation are often important; but they would have far less seri .ous results were they not conducted with borrowed money. That is, over- indebtedness mnay lend importance to over-investment or to over-speculation. The same is true as to over-confidence. I fancy that over-confidence seldom does any great harm except when, as, and, iJf it beguiles its victims into debt. The excessive credit growth became particularly acute in Malaysia during 1995-97 (Figure 1).13 The Central Bank, in tact, considered the expansion of loans to be excessive but, in light of the potentially adverse consequences on highly leveraged Malaysian corporations, was reluctant to respond by raising the intervention rate. In October of 1997, the Central Bank responded to the excessive credit growth by issuing a directive to limit the growth of credit lines on the grounds that it might have less serious consequences than increasing the interest rates. The adoption of 13 Excessive credit growth (EXCR) is calculated as: EXCRt = A,- Alogy, Ml - where A is the first difference operator. D, M, and y are, respectively, the total domestic credit, the broad money stock, and the real GDP. This measure assumes that money demand has a unitary elasticity with respect to real income. 9  this directive also seems consistent with the fact that Malaysia increased its intervention rate less than other crisis countries did (Figure 2). With the benefit of hindsight, it is quite likely that this directive played -- possibly by causing banks to over-react in the wake of the crisis -- a role, both in curtailing the availability of credit and in increasing lending rates relative to other market rates. Figure 1. Excess Domestic Credit Growth Figure 2. Evolution of Overnight Rates 0 25 - __ _ _ _ _ _ _ _ _ _ _ _ __ _ _ _ _ _ _ _ _ _ _ _ 00 02_.10Q fMlaysia ~ Korea Thailand - .. . Indonesia .a..Philippines 80- 0.20_ 80- 0 15- 40_ 0.10- 20- oos o e1 r ffi R 17lw eLI0g 0!llsl 1990 1991 1992 1993 1994 1995 1996 1997 9d019d049d079 109 9 0499079 109d01 :049 :079:10 In line with the authorities' efforts to invigorate the economy by engineering an increase in domestic demand, interest rates have declined noticeably (Figure 5). Despite the marked decline, the annual growth rate of loans has remained anemic and, indeed, turned negative in real terms in September (Figure 3). Although this observation appears, at first blush, to imply that the slow- down in lending largely emanates from the demand side, several comments are in order. First, the base lending rate is more reflective of administered changes and more representative of the lending rate charged to prime-customers. The average lending rate incorporates the lending rate that is charged to borrowers perceived to be more risky. Accordingly, an increasing spread between the latter and the former implies that banks are still 10  reluctant to extend new loans (Figure 6). 14 In fact, this observation suggests that the adverse selection problem has become more severe in recent months. Figutre 3. Growth Rate of Real Loans (year-on-year) Figure 4. Growth Rate of Real Loans by Type of Banks 40 6 _ Commercial Banks Finance Companies 30 ----- _____Merchant Banks o\ ~~~~~~~~~~40 -_ 201 20- 10- 9 01 9t03 19t:O 01107 O1:09 911: 9E:Ci1 91:03 9:05 98.07 9r099sE11 9t:01`9t:03 9i04 9t:07 9t08 I I' 9}01 3 Figure 5. Evolution of Average Lending rate Figure 6. Average Lending Rate and Base Lending Rate 20 2.4_ 14 .il Spread (left) cnCommermial Banks 2.2 Average tending rate (fight) -13 Finance Companies It Bs. l ( 18- Mechant Banks / .____Base lending ntter(ght) /\1 / . 2.0 ~~~~~~~~~~~~~~~~~~~~~~~12 16 168 1.8-~~~~~~~~~~~~~~~~~~~1 14 1.42_ Second, the increasing share of revolving credit and overdraft in total loans suggests that borrowers are relying more on shorter term and generally pre-committed facilities. This, combined with the sharp decline in lending activity, furthier underscores bank reluctance to 14 The base lending rate (BLR) for commercial banks is calculated as: BLR= [-IR * 80%]+2.25% l- SJ?R% where IR and SRR are the Central Bank intervention rate and statutory reserve requirement. Prior to September 1, 1998 the BLR was computed using average KLIBOR in lieu of the intervention rate and the fixed administrative charge was 2.5% instead of 2.25%. 11  extend new loans (Figure 7). In addition, to the extent that shorter-term and pre-committed facilities are utilized mainly by large firms, as was shown by Ellihausen and Wolken (1990), SMEs without commitments may find themselves served last, if at all. Figure 7. Share of Revolving Credit and Overdraft in Total Figure 8. Share of Securities in Commercial Banks' Total Loans Assestsa 32.S 3.0 32.0 ~~~~~~~~~~~~~~~~~~~~~~2.8- 315 2.6- 31.0 2.40 30 5 2.2- 30.0 ____________________________ 2.0 ___________________________ 97.01 97!0397!0597107'970997.1 1 98.01 98.03 98!0S 98107 98.09 98.11 9 01 99:03 9:05 9:07 9:09 9.11 ' 019d:03 9 .05 9:07 9d:09 9:11 a: Domestic Commercial Banks Third, available evidence also suggests that there has been a flight to quality -- a re- allocation of banks assets towards government securities away from lending to the corporate sector. As can be seen from Figure 8, banks have increased their holdings of government securities in relation to their total assets since January 1998 -- a sign of declining bank willingness to lend as banks devote an increasing share of their assets to securities. Fourth, a reshuffling of deposits across banks took place as depositors enacted aflight to quality. The flight to quality took place from finance companies and merchant banks to commercial banks (Figures 9-11). Perceived by depositors to be less likely to fail, foreign banks also experienced an inflow of deposits from domestic banks between December 1997 and January 1998. The flight from domestic to foreign banks has been less pronounced: the market share of foreign banks in total deposits has increased from 20.9 percent in January 1997 to 21.5 percent in December 1998, reaching its peak of 23.1% in January 1998 (Figure 12). To the 12  extent that some customers were exclusively borrowing from those banks that experienced a deposit flight, at least temporarily, such borrowers may have found it particularly difficult to obtain new credit. In fact, at a time of heightened corporate risk and of tight liquidity, banks are very likely to turn down new credit applicants with whom they do not already have an established customer relationship. Figure 9. Fixed Deposits at Commercial Bankss Figure 10. Fixed Deposits at Merchant Banksa 66- 8.0 ___________________________ 64-75 7.0- 62- 6.5- 60- 6.0- S~~~~~~~~~~~~~~~~~~~~~~~~~~~5 5 9 )01 9903 i9059i9:i0 i9909 ' i1 e1'4i01's40 959607 96109911 6 __ 97.01 97!04' 97a07 97!10' 98101' 98104' '9807' 9ss!i Figure 11. Fixed Deposits at Finance Companies3. Figure 12. Share Foreign Banks in Total Bank Deposits 36 23.5 P*~Od of Rligh toF 23.0- 34- 22.5 22 22.0d 21.5- 30- 21.0 28 " ; - 20.51 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 9 :~ 709:oss7o' 99:0999:1911910 60 60 60 61 17.01'97!03'97405'97!07'97!09'97 I 198!01 98aO3'98!05'98!07'98!0, .T1 a: (% share in the total) Fifth, evolution of the yield curve, particularly beginning with September 1997 and throughout 1998 (see Box 3), suggests that economic agents expect short-term rates to increase (or they expect future monetary tightening) as demonstrated by the steep upward slope of the yield curve. 13  Box 3. Evolution of the Yield Curve Yield Curve as of January 1997 Yield Curve as of March 1997 Yield Curve as of June 1997 634 632 650 6 30 6i N 626~~~~~~~~~~~~~~~~~~~6 3~ ~ ~ ~ ~ ~ ~ ~~ ~~~~~~~~~~~~~~~~~ 3 * 625 I~~~~~~~~~~~~~~~~2 20- 623 2 6 6 'O 14 6 ~~~~~~~20 4 0 1 1 406 6 10 2 1 134N410I 1 26 8 0 1 1 M.hbtlA M66y Short-terim interest rtes expected to fall Short-term interest rates expected tofall Short-term interest rates expected tofall Yield Curve as of September 1997 Yield Curve as of December 1997 Yield Curve as of January 1998 a61s 74. 72M -~~~~~~~~~~~~~~~~~~ ~~~~~~~~68 - ~~~~~~~~72. hor r . ir r 6.6 X70 64 60. ,.~~~~~~~~~~~~~0 / 60 6 2 a l 12 14 66 Ss-__ _ __ _ _ _ _ _66 _ _ _ _ _ __ _ _ _ _ _ 2 4 6U 6 10 2 4 6 a 10 12 14 2 6 8 10 12 14 M"iN M~Ky Mday Short-term interest rates expected to rise Short-term interest rates expected to rise Short-term interest rates expected to rise Yield Curve as (if March 1998 Yield Curve as of June 1998 Yield Curve as of September 1998 a53 / 5.5 64. ~ ~ ~ ~ ~ ~ ~ ~ 1 6.6 6.0~~~~~~~~~ 6 64 ~~~~~~~~~~~~a5, 62. /7D 2 4 6 86 10 12 14 2 .4 6 6 110 12 14 2 6 6 86 2b 12 14 Short-term interest rates expected to rise Short-term interest rates expected to rise Short-term interest rates expected to rise Yield Curve as of December 1998 5 35 536 Short-term interest rates expected to rise 14  The steepening of the yield curve is also consistent with a future decline in economic activity.15 Indeed, the information content reflects the expected effects of monetary policy via interest rates on economic activity. The expectation of future decline in economic activity, in turn, should have an adverse effect on lending activity. Although the relationship between monetary policy actions and the term structure is unlikely to remain invariant over time, and may be affected by other factors such as changes in expected inflation and in the real ex ante long-termn rate, the present situation may be explained in either one of two ways: either agents view current declines as insufficient to revive economic activity or they do not view policy actions as effective. Finally, as was explained previously, a decline in loan growth or in the lending rate by itself is insufficient to determine whether the decline in bank lending arises from a shift in either demand or supply. The evolution of the lending activity should be accompanied with that of the spreads presented in Box 2.16 Although the spread between the average lending rate and the 3- month T-bill (Figure 13), which captures both the general risk premium and the bank dependent borrowers' specific spread, has declined in recent months, it is still above pre-crisis levels.'7 More precisely, in the case of commercial and merchant banks, the spread reached its historic peak during the earlier part of 1998, before declining towards the end of 1998. It still remains, however, well above the pre-crisis period. The spread in the case of finance companies rose in March 1998 and, thereafter, exhibited a downward trend. Overall, the increase in the spread in the case of finance companies is less pronounced compared to that of commercial and merchant banks. i See Estrella and Mishkin (1995) for more on the role of the term structure in the conduct of monetary policy. 16 Indeed, numerous studies demonstrated that increases in the interest rate between risky and safe debt are associated with subsequent downturns in output [see, for example, Stock and Watson (1989) and Friedman and Kuttner (1989)]. Consequently, the behavior of the spread also has implications for economic activity. 17 Put differently, an increase in this spread can also be interpreted as a tightening in the terms of external finance. 15  Figure 13. Spread between Lending Rate and 3-Month T-bill Rate 18 Commercial Banks Merchant Banks 7 12, 6- 10- 5 8 4 8 3- 4- Finance Companies 10D 9- 8- 7- 6- 5- 41 j, 4, '9 ;9 1094 95, 199"T;l 197 ..... 1998 ll The overall analysis combining the evolution of the spread -- the price variable -- and that of the lending activity -- the quantity variable -- along with the other key indicators of the framework suggests the following: (i) evolution of the spread between risky and safe debt implies that "payoff' or "default" risk is still well above its pre-crisis level, thus underscoring the increased agency costs of external finance; (ii) the decline in lending activity between the end of 1997 and first half of 1998 can be explained by the reduction in bank credit supply relative to demand; (iii) it appears that the decline in demand has probably played a more prominent role in I Three-month moving average. 16  contributing to the sharp slow-down in lending activity during the second half of 1998 compared to the earlier period; (iv) the sharp decline in the growth rate of loans by finance companies -- which in part can be attributed to closures -- when combined with relatively moderate increase in the spread, can be interpreted as the decline in demand contributing to the slow-down in lending activity more than in the case of commercial and merchant banks; and (v) customers borrowing exclusively from banks that experienced deposit flight, namely merchant banks, finance companies, antd, for a very short period, domestic banks, are likely to find it more difficult to acquire new credit. Considering that SMEs usually have a customer relationship with only a single bank, while large firms usually have customer relationships with several banks, the latter can typically find a bank that is able to make loans. Thus, it is quite possible that the phenomenon of deposit flight will hurt SMEs disproportionately.'9 4. Empirical Framework and the Main Results 4.1. Background A growing body of research -- mainly on the industrialized countries -- has established the empirical significance of credit market imperfections and has provided evidence on dispropotionate effects of monetary policy changes on bank-dependent borrowers. More spefically, this strand of the literature has demonsted that the informational frictions that add the costs of external finance apply largely to younger firms and firms that are not well collaterlized. On average, these are small firms [Gertler and Gilchrist(1994)]. Smaller firms depend heavily on intermediary credit, while large firms make far greater use of direct credit, including equity, public debt, and commercial paper [see Gertler and Hubbard (1988)]. Morever, beginning with a Beyond the possible discrimination against SMEs, there may be particularly severe discrimination against newly founded firms. In a period when credit is tight, banks are unwilling to take on new loan customers, so that people who want to start new firms often cannot obtain credit. 17  study conducted by Fazzari, Hubbard, and Peterson (1988), a large number of investigations on investment have highlighted that smaller firms are more likely to face liquidity constraints. These findings, in trmn, imply that monetary policy shifts have an important distributional aspect that cannot be tackled within the traditional money view.20 Furthermore, since smaller and faster growing firms are likely to be firms with highly profitable investment opportunities, the fact that they bear a disproportionate share of the burden imposed by a recession has serious consequences for social welfare: not only are recessions associated with both aggregate output and investment declines, but the declines themselves are inefficient. While the largest industrialized countries have been the subject of the above mentioned studies, very little wvork has been conducted to investigate the validity of distributional consequences of monetarv policy shifts for developing countries.21 This section attempts to fill an important void by investigating whether monetary policy shifts have a greater influence on the production of sinall and medium sized manufacturing firms than on that of larger firms in one such developing country, Malaysia. The first step of the empirical investigation is to construct a reasonable proxy for the production of small and medium sized as well as large manufacturing firms. To this end, the paper relies on the information published by the Federeation of Malaysian Manufacturers (FMM) on small and medium sized industries (SMIs) and on industrial surveys conducted by the Department of Statistics. Appendix 1 presents both information on SMIs in Malaysia as well as the compostion of manufacturing component of the industrial production. Available information suggests that the majority of SMIs operate in: (i) food and beverages; (ii) fabricated metal products; (iii) wood and wood products; and (iv) basic metal. Using the corresponding weights 20 See Cecchetti (1995) for more on this. 18  of these components from Table Al of Appendix 1, the paper calculates the production of SMIs, while the production of large industries is calculated by using the remaining components and their corresponding weights. Figures 14 and 15 illustrate the evolution of the constructed series for the production of SMIs and large manufacturing firms (LMFs). It is clear that the downward trend in production is more prominent in the case of SMIs compared to that of LMFs. Indeed, the growth rate in the index averaged over 1997 and 1998 for SMI production is almost -1.4 percent, while that of large industries is around 1.8 percent. Figure 14. Index of SMI Industrial Production Figure 15. Index of Large Industries Industrial Production 160. 193 180- 150L 1 70- 140- 100 130- 150- 120-10 110. ___________________________ 130.___________________________ 9g :01 9d:04 9607 96:1O 9901 97h04 99:07 97:10 9B8&1 bd:04 98:07 98:10 9 :01 9604 9B:07 9s:1o 99:01 99:04 99:07 W9A:1 98:01 98:047 901O Table 2 also presents descriptive statistics of respective index of SMIs and large industries between 1992:01 and 1998:12. It is interesting to note that the variability of SMIs' production is almost twice as much as that of large industries, as demonstrated by the coefficient variation. A possible explanation for this phenomenon is provided by Gertler and Gilchirst (1994) who 21 A recent study by Domac and Ferri (1998) on Korea concluded that SMEs suffer more than other businesses do from the adverse ceffects of the credit channel. 19  suggest that large firms smooth the impact of variation in demand by contracting out to SMIs in good times, but servicing all business internally in economic downturns. Table 2. Descriptive Statistics of Manufacturing Production (1992:01-1998:12)a Large Manufacturing Firms SMIs Mean 0.71 0.51 Median -0.18 -0.25 Maximum 16.75 18.76 Minimum -14.83 -16.39 Standard Deviation 6.28 8.22 Skewness 0.19 0.21 Kurtosis 2.86 2.58 Coefficient Variation 8.85 16.10 Jarque-Bera 0.59 1.23 Observations 83.00 83.00 a: Growth rate of the respective index. 4.2. Responses to Policy Shocks by SMIs and Large Manufacturing Firms: Evidence from Vector Autoregressions Quantifying the effects of monetary shocks has been one of the most significant challenges of empirical monetary economics. Economic research has historically employed several competing approaches to identify policy effects from the observed data. On the one hand, there are techniques based on the Choleski decomposition of vector autoregression (VAR) residuals. Within this approach, the main debated issue pertains to whether an indicator of monetary policy exists which can justifiably be placed within a Wold-causal ordering. On the other hand, the approach proposed by Romer and Romer (1989, 1990) uses episodes viewed as monetary contractions as a means of identifying monetary shocks. Here the key issue causing 20  disagreement is whlether such episodes can be reasonably considered exogenous with respect to other shocks to the economy.22 Since the seminal work of Sims (1980), VAR models have become an increasingly popular tool for empirical studies of the monetary transmission mechanism. 23 A 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 or a disturbance to any other variable in the system. This paper, in line with the general convention in the empirical literature on the transmission of monetary policy, also employs a VAR model. The basic strategy for identifying an exogenous shock to monetary policy is to focus on the disturbance term in the following regression equation: St= ''(Rt) + as5t (1) Where St stands for the policy instrument; 'V is a linear function; Qe summarizes the information set that the Central Bank considers when determining S,; a is a positive number; and finally st is a serially uncorrelated shock that is orthogonal to the elements of 0, and has unit variance. The information set consists of the past history of all variables in the system and the time t realizations of subset of those variables. The dynamic response of a variable to a monetary 22 Evaluation of these competing approaches is beyond the scope of this paper. Interested readers, however, should refer to Chirstiano, Eichenbaum, and Evans (1996a., 1996b), Zha (1997), and Reifschneider, Stockton and Wilcox (1997). 23 Sec Watson (1994) as well as Reifschneider, Stockton and Wilcox (1997) for a comprehensive discussion on the use of VARs and the role of econometric models in the conduct of monetary policy. 21  policy shock can be captured by the coefficients in the regression of the variable on current and lagged values of the fitted residuals in the above equation.24 Following Bernanke and Blinder (1992), Chirstiano, Eichenbaum and Evans (1996a, 1996b), and others, the paper employs the short-term interest rate, namely the overnight money rate (R), which is under the direct influence of the Central bank, as an indicator of the stance of monetary policy.25 Consequently, the disturbances to the overnight money rate equation in the VAR are identified as shocks to monetary policy. The response of other variables in the system to an interest rate shock is interpreted as the structural responses of those variables to an unpredicted shift in monetary policy. More specifically, the VAR system in the case of SMIs (in the case of LMFs) includes the log of SMI production [YSMI] (yL-) along with the log of real effective exchange rate [RER], ovemiight money rate [R], and the spread between the average lending rate and the 3-month T- bill rate [SPR], in that order.26 This ordering, consistent with the Central Bank's behavior in practice, assumes that the nmonetary authority looks at the contemporaneous state of real economic activity (Ysmi or YL) as well as the real effective exchange rate (RER) before deciding on st.27 The spread between the average lending rate and the 3-month T-bill rate (SPR) is 24 See Chirstiano, Eichenbaum, and Evans (I 996b) for a detailed technical exposition of this strategy. 25 In defining the transmission mechanism. exogeneity of monetary instrument is important in two senses: one economic and one econometric. In an economic sense, the monetary policy instrument must, at the margin, be controllable by the Central Bank: the endogenous response of the instrument to developments in the economy must take place merely via the Central Bank's reaction function. Exogeneity, in the econometric sense, entails that enough restrictions be imposed upon the system to allow the identification of the Central Bank's reaction function and the shocks pertaining to it. 26 There is an issue of whether the variables in a VAR need to be stationary. Sims (1980) and others recommend against differencing even if the variables contain unit root. They contend that the goal of a VAR analysis is to determine the interrelationships among the variables, not to determine estimates. The main argument against differencing is that it "throws away" inforrnation concerning the co-movements in the data. Although the results of the Phillips-Perron stationarity tests indicate that variables involved are integrated order of 1, I(l), following the argument put forth above, variables are not differenced. By not imposing co-integrating relations, the estimation avoids a long-run identification problem, which may be in principle difficult to solve, with no loss of information on the long-run properties of the system, incurring some loss due to the reduced efficiency of estimation but at no cost in terms of consistency of estimators. See Sims et al. (1990) for more on this. 27 'he empirical results are qualitatively the same when R is ordered last. 22  included to capture the impact of the changes in the terms of external finance (or the combined impact of the balance sheet and lending channel effects) on economic activity (see Box 2). It should be noted that the variable SPR understates the actual magnitude of this effect since the average lending rate would not entirely capture the increase in cost to the typical bank borrower, as fewer borrowers qualifv for this rate during tight money episodes. The VARs are estimated using monthly data from 1992:01 to 1998:12, including 12 lags of the relevant quantitative variables.28 The estimated VAR coefficients are not in themselves very interesting and thus are not reported here. Instead, the impulse response functions and the variance decompositions are reported. In order to assess the overall quantitative impact of a monetary tightening and the credit channel effects on production of SMIs relative to LMFs, Figures 16 and 17 plot the impulse response functions of ySME and yL with respect to innovations in the interest rate and the spread residuals over a horizon of 36 months (Appendix 2 reports the overall results of this exercise). Standard errors are calculated by the Monte Carlo method, with 1000 repetitions and standard error bands (of