Tài liệu technology diffusion and postwar growth potx

Because ‘commercial paper and bankers acceptances ae short-lived, they are more timely measures ofthe availabilty of short-term credit than are bank or business Failures and the level o

đồ

Federal Reserve Bank of Dallas

Preventing a Repeat of the Money Market Meltdown of the Early 1930s John V, Duca’

Vice President and Senior Policy Advisor Research Department, Federal Reserve Bank of Dallas P.O Box 638906, Dallas, TX 75265 (214) 922-5154, john v.duewdi Phong and Southern Methodist University, Dallas, TX ‘April 2009 (revised November 2009)

‘This paper analyzes the meltdown of the commercial paper market during the Great Depression, and relates those findings to the recent financial criss, Theoretical models of

financial frictions and information problems imply that lenders will make Fewer non-

collateralized loans or investments and relatively more extensions of collateralized finance in times of high isk premiums This study investigates the relevance of such theories tothe Great Depression by analyzing whether the imereased use of a collateralized form of business lending, (bankers acceptances) relative to that of non-collteralized commercial paper ean be

ceconometically attributable to measures of corporate ereditfinaacal risk premiums Because

‘commercial paper and bankers acceptances ae short-lived, they are more timely measures ofthe availabilty of short-term credit than are bank or business Failures and the level or growth rate of | the stock of bank loans, whose matuttes were often longer and were renegotable- In this way, the study adds to the literature on financial market Irctions during the Great Depression, which aside from analyzing securities prices, typically investigates the behavior ofcredit-elated

variables that lag curent conditions, such as bank failures, bankruptcies, the stock of money, oF

‘outstanding bank foans In particular, the real level of bankers acceptances and their use relative to non-

collateralized commercial paper were strongly and positively related to spreads between corporate and treasury bond yields Also significant were short-run events, such a the October

1929 stock market rash and the 1933 bank holiday episode that sparked flights to quality inthe bbond market and a flight to collateral (As) in the money market and pethaps away from the loan market Furthermore, these shifts in the composition of extemal finance were large

‘supporting the view that financial fictions and reduced credit availability may have played an important rae in depressing the U.S economy’ during the 1930s, The paper also relates these Findings to the current financial crisis by examining how the relative use of commercial paper reacted to yield spreads during the current crisis, taking into account Federal Reserve actions to improve liquidity condition inthe money markets, Results suggest that these efforts may have, at least sofa, helped prevent the commercial paper market from melting down tothe extent seen during the early 1930s,

JEL, Classification Codes: B43, B50,

“I thank Danielle DiMartino and Gustavo Suarez for suggestions, David Luttrell and Jessica Renier for research assistance, and seminar participants at Oxford University, Johns Hopkins University, the Bank for International Setlements, the 2009 Bank of Canada - Simon Fraser University Conference on Financial Market Stability, the 2008 North American Economies and Finance Association the 2006 Western Economic Association Meeting, the 2007 Swiss Society for Financial Market Research Meeting, and the 2007 Financial Management Assocation

“Meeting for comments and suggestions on an eaie,pre-2008 Financial crisis version of this paper The views expressed are those of the author's and are not necessarily those ofthe Federal Reserve Bank of Dallas or the Federal Reserve System,

‘This paper analyzes the meltdown ofthe commersal paper market during the Great Depression and relates those Findings tothe recent financial eis, particulary’ with respect 0 Federal Reserve and Treasury efots to improve liuidity conditions in the money markets

“Theoretical models of financial feetions imply tha ereit extensions wil shift tom sky to safer

borrowers if economic Factors increase default risk or increase the cost of loanable funds wia

uid csk premiums (eg, Bernanke and Blinder, 1988; Bernanke and

Gener: 1989; Bemanke, Getler, and Gils, 1996; Jaffee and Russ, 1976: Keeton, 1979 Lang and Nakamura, 1995; and Súgliz and Weiss, 1981) Based on these implications, post

‘Weld War Il data onthe composition of business credit has bean used to assess the eeevance of suc theories, dating back Wat leas Jae and Movin (1969), who assess the composition cofbunk business Lending, and extending to reeet ste, such as Kashyap, Wileos, and Stein (1983), who analyze the reatve use of commercial paper and bank loans This iterate

especialy the Migi-to quality model of Lang and Nskamata (1995) and the financial aocelerator approach of Berar, Gener, and Gilerist (19%), implies that lenders wll make fewer nom secured loans and extend elaively more collateralized fiance in times of high default risk

Swiis ofthe curent isis maybe hampered by what some may misperceve aa lack of bistrial precedent, but the current study provides an analysis of he elative use of commecial

peper daring the Great Depression ta shed light on how the money markets have heen feted

by the current crisis also provides evidence tht Fed Treasury efforts may have (G0 f4r)

prevented the commercial paper market fom imploding onthe eal that itd in 1932, In general, empires studies ofthe impact of Financial Fetons and monetary Factors daring the

1930s have focused on examining the links between bank filurestloans and economic asivity

(Field, 1984; and Wheelock, 1990), the ol of unanticipated deflation (Hamilton, 1987), of inks between default isk spreads and economic growth (Bernanke, 1983), Although the credit and financial fetons literature, particulary the financial accelerator work of Bernanke and Genter (1989) and Berane, Geter, and Gilchủs (1996) is pardy motivated bythe Great Depression, there are no published studies that empirically analyze the composition of business ret

extensions during the ora, pal reflecting a lack of data on imely measures of external Finance

‘The current paper fills pat ofthis gap by investigating whether the increased use of a

collateralized form of business lending (bankers acceptances, B.As)elative to that of non- collateralized commercial paper during the Great Depression (Figure 1) can he eonometicdly attibutable toindieatrs of coeporte credit sk and the isk of bank uns, For robustness, these Financial factors ate assessed in models ofthe levels and ist differences of rel bankers

acceptances In addition to the default and liquidity risk premiums in eorporate-Treasury vield spreads, the risk oF bank runs may also have encouraged the use of BAS, The main reason is that

Figure 1: Domestic Bankers Acceptances Rise Relative to Commercial Paper ‘Shortly Ate Risk Spreads Rise During the Depression

in rss, banks had the option of borrowing aginst o selling BAS othe Federal Reserve,

‘Which committed ise? to rediscounting BAS and commercial paper held by hanks and whose silty to rediscount nk assets was limited 0 these types of paper until the Banking Act of

1038, Asa esl, BAS were mare liquid than other bank asses But the Fed's conduct oF pen rarket operation in BAS and discounting BAs lowered the liquidity risk of investing in BAS,

[By implication, the relative use of BAs asa source of exteral finance should he increas

int) icasures of eomporate-Tw suy ri spross, which blond daft se and liquidity sn, and (2) possibly measres ofthe sak of bank rans, uch asthe curreney-to-dopot ratio Nevertheless, these are reasons hy higher hank sun isk may not induce greater use of BAS This papers sclted to Bernanke's (1983) empirical Findings that two proses Por the destruction of Fnancil intermediation during the Depression, deposits of suspended banks and labiliis of filing businesses, were signiicant in explaining changes in industriel production in te presence of

‘money growth, Bemanke also referenced finding that a pros’ for default sk, the spread

bbeween yields on Baa and Ans-rated corporate bonds, was also statistically sig

‘or industil production, but was als highly collinear with oer indicators oF financial distess Although Bemanke discussed how deflation destroyed collateral and thereby plausibly reduced the availability of credit, the literature has not direct assessed whether the

‘composition of business finance during tae Great Depression was affected by rising isk

pvenloms and bank run isk in ways consistent with theories of financial fitions

“The sunent study tess whether an empirical implication ofthe ere inancal fictions channel is casistent with Bernanke's finding that Financial fietions were macro-conomically

‘nxportant, This sty also uses mare imely data on basiness nance than hes been often used in studying the Great Depression, Because commercial paper and bankers acceptances are short lived, they ate more timely indicators ofthe availability of extemal finance than are bank ot

business failures and the level or growth of the stock of bank loans, which Ig the economy and

‘hich have been analyzed in much ofthe htestre on the Great Depression A particular problem with analyzing outstanding bank fons is hat bankers may have given voobled

borrowers mor ime to epay loans or delayed wens ofoans, Asa result, growth in she stack cof bank loans may net give a consistent nor timely indication ofthe availability of new eri

Consistent witha financial fietions channel, bankers aecepances—real or retative to commercial paper—ae increasing inthe spreads between investmentgrade corporte snd

‘Treasury bond yields, Such spreads are used to gaus cris premiums, without breaking them

ow ito expected default losses and changes in the price of deal nd iguiity snk [Terni (1976) argues that wider corporate quality spreads reflects heighened isk of ines failures rather than ight iguidity cantons | Because the mosey market and corporate spread variables exhiit dit and nonstationary using the Greet Depression a, cointegration lenis ae

‘sed, Given the short matures of bankers acceptances and commercial paper, ests imply tha viskpreniums persistelly afected the composition and availabilty of external finance

To relate these iin othe cutrnt isis, his study also analyzes the cureat isis Owing to the deepening of financial and credit markets, and the Fed's shift toward conducting

‘pen market operations in Treasury and agency deb, bankers acceptances have become a tivial

source of domestic finance Instead, the paper-bank loan mix (Kashyap, Wileoy, and Stein,

1992) is model

to see how the same Baa-Treasury spread usc inthe Greet Depression sample

is related tothe composition of shorter husines ince in the current crisis Renlts sugee

‘hat ations taken since Oetober improve liquidity i the money markets may have helped prevent the commercial paper market from imploding by as much asit did in the erly 1930s

“This study is opanized as follows, Seetion 2 provides dtl about bankers acceptances aid hypotheses about their use daring the Great Depression, Secion 3 presents the data and

section 4, the empirical results for this era Section S examines whether new Federal Reserve

Fiquidity programs may have prevented she commercial paper market from collapsing The conclusion relates the findings forthe Great Depression with those onthe curentctsis

Det and Hypotheses Regarding Bankers Acceptances and Commercial Paper

What Are Bankers Acceptances and What Purposes Do They Serve?

‘A bankers acceptance is atime drat drain on a bank to finance the shipment or storage

‘of goods A draft is “aovepted’ when a bank guarantes payment to the holdes This makes the

[BA tradeabe in securities markets bectuse the issuing bank unconditionally promises payment (othe investor at a specified date regandless of whether the yoods-buying firm repays the bank and because investors usually have better information about the bank than the goods purchaser

“The goods-buying im essentially receives funding to pay the seller from a bank, whet Funds

the extension by selling the bankers acceptance From the perspecive of banks guaranteeing 'BAs, the goods purchased or held in inventory colatealize the BAS Thos, extending BAS ails less default sisk than making sn unsecured lan to fem of similar ered quality

Why Risk Premiums May Affect The Use of Bankers Acceptances

hooves of credit and financial ection imply that lenders wold pressure horrowers for

‘more collateral, hich s relevent for the BA market during ant era when fins had fewer

financi ‘options Goods sllers, for instance, may be reluctant to uive trade credit for an extended period toa x

In addition, i he event that the goods buying fim doesnot repay, the Buyer's bank may

bbe more able ro repossess the pools financed with 2 BA using collection processes than the

goods-seling fim However, the exta pape an legal work in creating aBA versus e standard fon suagess that many borrowers would be better off with foam unless the bank requires collateral thatthe bortower could not provide absent a BA or unless the value of ellateral

induces banks to offer loan rates on unsecured loans above those on BAS, These fetorsimly thatthe relative use of BAS increases with macro default isk, while that of commercial paper

‘would, consistent ith the ratios plotted in Figure | and real levels shows in Figure 2!

Why Liquidity Rsk May Affect The Use of Bankers Acceptances

Another atuactive aspect of BAs was hanks’ ability sell BAs, which made BAs easier

‘of han business loans for which deposits were needed (Duield and Summers 1981) The

BA market was very liquid the 1920s and 1930s because the Federal Reserve condaced open snark operations in it (Federer, 2003) and Small and Clouse, 2006) For this reason, BA

issuance ued the impact ofthe risk of bank deposit ans on te banks’ bility to meet

customer eredit needs An adiional funding advantage was tha banks Were requied to hold reserves against deposits needed to fund leans, but ao against BA that they issued and sod

Furthermore, even if hanks held BAs they issued, BAs were more liquid than loans Prior

eas collateral for discount

+9 1932, only commercial paper and BAs held by banks were eli

founs; and it was not until late 1935 that the Federal Reserve allowed some types af loans 0 serve as collateral for discount loans In 1932, Congress granted the Federal Reserve limited

temporary authority w rediscount promissory notes secure to the satisfaction ofthe Pederal Reserve under the Glass-Steagall Act The Federal Reserve interpreted this provision as giving it limited authority to accept government securities as well as commercial paper and BAS as

collateral for discount loans (The New Vouk Clearinghouse Associaton, 1953, p76)? The Zac baked commeral paperiselavey new Unlike BAs which wv olaraiad ty me poo mab

Likely reflecting ths lmded chenge al the Fed, dun variable conteliag Cor this ited and temporary

suo was tenia reins at show,

Figure 2: During Depression Era, Real Levels of Domestic Bankers Acceptances ose With Risk Spreads While Commercial Paper Tends to Fall „„,

“These considerations imply that the relative use of BAS may also rise with the risk of

bank runs as proxied by the curreney-to-deposit ratio, Nevertheless, there may be counter= vailing factors One appealing aspect of owning BAs to investors is that the bank issuing the BA

‘quaroates sepayent of principal and implied interest epardess of whether the goods-buying, firm repays the issuing bank That sid, confidence in the value af ths gurantee may be Low

‘when she risk of bank failures is high, Finally testing or the impact of hank rn rik would be appropriate ifthe analysis focused on otal BAS, Hoviever, in this study detecting any efeet may

be cficuleosause the analysis focuses on Bs tat finance domestic commerce—which were not the bulls of total BAs—to Focus on financial fictions effects on domestic borrowing

Why Liquidity and Bank Rum Risk May Also Affect The Use of Commercial Paper

‘The impact of defaulea liguidty isk on BA use may have been relevant to the Great

Diamond, 1991), Firms issuing paper are wpically required by the market ta have back-up ines fbank credi that eliminate most liquidity risk associated with their abit to pay on time

For vo reasons, hoseever, higher default ea igi risk daring the Grest Depresion also plausibly made BAs a more feasible source f external Finance than commercial pape

First, even highly ded paper issuers pose default risk in that unusual era which preceded

she asset-backed commercial paper market and during which the creitratings of many ighly rated firms had been eut by ratings agencies (Wigsnore, 1985) Secon it was not unl the Ite 1970s, when large banks had acess to wellaeveloped, large tine deposit and otber funding snakes tha paper-issung fms could easily and normally obtain ines of bank ereit tobeck up thei issuance of paper Thus, when soacems about quit risk rose dramatically, they may

have affected investor perceptions about the liquidity risk of commercial paper during the 1930s

U1 Data and Variables

Bankers Acceptances and Commercial Paper

Federal Reserve data are used to construct the ratio of bankers aecepranees to commercial

"

peper (ACP), Both components were seasonally adjusted using a multiplictiv

procedure, which yielded a rato that was similar had the ratio ater than the components besa seasonally adjusted The advantage of separately adjusting each component is hat he same resulting BA series is used to constact areal BA series thats also modeled The BA series that

is analyzed excludes BAs used to finance intemational trade, which ar ess reetive of

onset activity and more reflective of swings in international trade that were butted by chhanges in trade bartiers such asthe Smoot-Hlawley Tariff A big advantage in using a ratios that, in principle, it largely eiminates the nee to sale fr the level of business activity Another sxivantage is that it wnerlly avoid the need to deflate nominal level, which entails choosing»

ce index from the very Himited set of available pre- WWM price indexes However the overall prwieerprige index is used to deflate Bs and commercial paper to see how each component behaved and to test For any impact of financial iietions onthe real level of BA use A dectine in real commercial paper and risen eal Bas were behind the mp in HAC during the eaety 1030s (Figure 2) ‘This is consistent wth fll inthe availability of non-collatralized Finance (es, commercial paper) and a surge toward collateralized Finance either fom former paper issuers or traonal bank loan borrowers who lost access to unsecured bak loans

‘Monthy Federal Reserve data on BAs and commercial paper are available between December 1924 and December 1941, In 1940-11, sere ws a large jump in commercial pape

“which plausibiy reflecied a suge in US production of defense goods purchased hy the US and

UK during 1940 and 1941, Because many firms with defense contuets could be viewed as having both safe revenue sources and even iplicit government-backing on thir deb obligations

(like commercial paper, the data used inthe analysis i restricted 10 December 1924 0

December 1938 10 prevent the sample rom being contaminated withthe omitted variable bias associated with the ramping up of defense spending inthe early 1940s,

Default and Ligiudiy Rsk

‘Default and liquidity sk combined are tracked by spreads here yields on investment ride corporat and longer U S goverment bonds based on the view that yield spreads between risky and data ree bonds generally reflect cyclical swings i def riek and dale risk premiums (aff, 1975) One such spread is that between yields on A-rated corporate

bonds (Moods) and long term U.S government bonds, denoted as 47K (source: Federal

Reserve (5M), also available fom NBER's Maceohistory Database) ATR was sed instead of the spread between Baa-tated corporate and Treasury bonds (BLLATE, whic, unlike and the BA variables, doesnot have a uit root ona quartelyfrequeney (upper panel, Table S) and exhibits only marginal evidence ofa unit rao ona monthly basis (upper panel, Table 1)

Also usd isthe spread between average inesment ara compote bond yields and longster U S government honds (OTR) Under more “normal” market condition, there is 8 preference inthe empirical literature to lok a spreads between yields on dierent ratings

classes of bonds, a ssh spreads may arguably capture swings i the overall default isk

However, thre wore many downgrides of corporate honds during the Depression in response to the worsening economy (Wigmore, 1985), This factor suguests that a quality sprcad based om

ve average corporate bond yield miuht beter track changes in deft sk than a spread based

ca parcularratings clas of bonds, Indeed, th gread between on average investment grade corporate and longterm Treasury bonds (OTR) somewhat eutpetons the spread between Sields on A-rated corporate an long-term US government bonds (47%) Although they tend to

‘ove closely together, (he quality spread based on average corpora yields moved up alte less

than the 4T dosing the 1931-1932 portion ofthe Depression exs (Figure 3),

Figure 3: Bond Quality Spreads and Currency-Deposit Ratios

Because ATR and COTR are spreads of corporate yields over long-term Treasury yields, they arguably track both default risk and liquidity tsk in contrat to spreads between yields on low-rated and high-rated investment wrade bonds ATR can be decomposed into the spread

between Aaa- and A-rated corporate bonds, plus the spread between Aaa-rated corporate and long-term Treasury bonds The Aaa-Treasuty spread likely includes a premium refleting both the extra liquidity of Treasury bonds and less prepayment risk—the latter because the Treasury has never prepaid and then refinanced outstanding bonds (Duca, 1999), Hence, in principle, the two spreads (Aaa-A and Aaa-Treasury) could be used instead of a single spread (A-Treasury),

However, in practice, several considerations favor using a single spread, First, in nusually risky periods like the Depression, even Aaa-rated firms posed some default risk

‘Second, suc risk is also reflected in the unusually high incidence of corporate downgrades during the Great Depression, Asa result, during this era even the Aaa-Treasury spread reflects

‘movements in default risk, ‘Third, this factor is even more important because logs of variables

‘are used, which raises problems arising from Jenson’s inequality associated with decomposing, spreads like 47 and COT into Tog subcomponents, Fourth, there is a high deuree of

correlation between the Asa-Treasury and A-Aaa spreads during ths ea, raising problems of

‘nulti-collinearity-—particularly using cointegration techriques on & sample spanaing, 18 years Likely reflecting these considerations, oer results (not shown inthe tables) which used

components like the logs ofthe Asa-Treasury and A-Treasury spreads, gave unusual results

For thre reasons, the analysis uses bond quality spreads rather than interest rate spreads between BAS and commercial paper Firs, the latter may introduce simultaneity bies Second, switch othe BA-commercal paper rate spread dd not materially fect the hey results

“Third, imerpreting bond spreads easier since the BA-comsmercil paper rate spread may aso

‘eft elatve money market conditions from factors oer than dealt o guy isk

ratio of currency’ to demand plus time deposits (CZDI47), Both are plosted in Figure

Other Financia Criss or Regulatory Variables

(Changes in bank regulation plausibly affected the we of Bs Tivo monthly dummies are used to conto forthe impact ofthe Banking Act of 1935 RANKACT339 equals eto belore September 1935 (and one thereafter), when the Basking Aet oily took effet, The othe, BANKACTSS1, equals one since Janay 1935, when money market participants may have acted

{in anticipation of the Acts passage, These long.-term dummy variables did not perfonn well,

‘This was parintarly ue with a quartety dummy, B4NKAC'TS5, which equated | from 1035:Q3 on, ‘To conserve space these rns are na shown inthe table, Nevertheless, there are some outsized swings in BACP and the rea eve of BAs jus before an jus afer he Banking Act was enacted As check to sce whether hese short-run effects tered the shot: and long- coeticionsin quarterly vctor-eror correction models, a shortsn dummy RACTDD) was included in soove models, which equaled In 1935-93, nauntive Tin 1935.94, and O otherwise

In ation, some egresions sso inchaded (+! vale to account fr any temparty, coutsize, impact of he banking crisis of March 1933 The monthly version ofthis variable, BANKHOLID, equals one in Mare 1933, when fears of a banking collapse srippa the US, inducing President Roosevelt to order bank holiday Daring that shutdown, regulators

examined the books of banks and reopened only the ones dened solvent ones that mon

Reflecting that the holiday oecurted late Match sd that the public's confidence in hanks was sil shaken in April the monthly version of BAKO also equals one i Api 1933 and the quarety version ofthis variable equals In 1933:g2 and 0 otherwise) The large jump in relative

BA use without a large jump in default risk spreads during these months stronely suswests that

tion tests to assess the relationship bebwcen the relative use

‘of BAS an he two measures of risk on a monthly basis "Thes,shortnun findings from vector erorcorrection models ae reviewed Quartry results are provided asa robustness check Finally the absfate level of tel Bs are analyzed ta Se if Bindings regarding the elative use

oF BAS (e., BAC?) ate not mecly an atic of comparing BA use with commercial paper

41 Monithy Cointegration Results

Cointepration analysis shoud he used to detet torn tlationships anion

nonstationary vacables (Engle and Granger, 1987), preferably in Togs Since the Lops of monthiy

nd quately BA use (denoted wit an L before ACP) have unit oats (lloing Fr possible {ime tends) as do the deal sk spreads LCOTH and ATR, cointegration tests are used to test for ona-rn relationships among the relative use of BAS and one default isk spend (LA or COIR) Bach ofthese variables has a unit roo, being nonstationary in fevels and stationary in fis ierences, according to augmented-Dickey-Fallr sais that are insignificant forthe levels and significant forthe fest differences ofeach variable (upper panel Table 1) Somewhat stronger results were fund using LCOTH, whose results are presente first in Table 1

“Tests found only one cointeurating vestr for LEACH and LOTR for samples including and excluding observations after the Banking Act of [935 (vectors and 8) based on vn

standard cointegration tes statistics (Table 1) One's the trace statistic, which rejected only the

absence of one evinteuraing vector in each case atthe S pervent level using Johansen-luseius's (1980) rank significance eviterion The other statistic is the maximus eigenvalue statisti, which

also only rejected the absence of one cointegrating vector in each case a the S percent level In cach case, yetors minimizing the Akaike information statisti favored a lag length of | month,

‘Combined withthe insignificant statistics for the existence of mare thant cointegrating vector

(not shown), these findinas suppor the hypothesis that one long-ron (conte

lationship exists among each group of variables, For the unique estimated vectors I and 2, the quality spread is highly significant withthe hypothesized sign, since pp

thecosfiden signs inthe coinegrating vector (ew, LBACP, = 1 798746 | 3 705096*LCOIR, from vest 1) implies that eulbvium use of BA's relative to commercial paper is inereasng inthe spread between yields con investment grade corporate and Tong-tesm Treaty vel Similar results that suppor the Financial fictions hypothesis were obtained in vector eror-comtection models (VECMS which jointly estimate long: and shorttem eatonships) that included BANKHOLIDAY’ asa short-run

‘vatable (vectors 2 and 8, Table 1) Also encouraging is that etal log levels of B4CP are

‘socked well and are slighty le by the equilibrium leves implied by vectors I and 2 (Figure),

Figure 4: Tracks and

Slightly Leads Actual Use During the Depression Era

‘vectors could not be ejected Further, even if one significant vector were assumed (vectors §

and 6), the curreney-to-deposit ratio was insignificant, while the eoetTicients and significance of

COT were unaffected Thus, tere appears wo belt ole forthe currency to-<deposit ratios

As another robusness check, «dummy forthe Banking Act of 1935 was added to vectors

1 and 2 in vectors Sand 6 Inthe absence ofthe Bank Holiday dummy, a nique coimesating

‘vector (vector 5) was found in which the significance and magnitude ofthe bond spread variable

‘were basically unaffected, whereas the Banking Act dummy was insignificant Inthe presence ofthe hank holiday dummy, evidence of cointegration was weaker, Even assuming that a unique

‘vectors found (vector 6), the banking st variable emained insignificant while the significance

and magnitude ofthe bond spread variable were basically unaffected, Combined with evidence

eacluding te 1935.39 Ibpedadyidlđed neelt (vetr37 and 8) that were similar to thse af

‘vectors 1 and 2, he financial fietions hypothesis isnot oversees by controling for the eects

of the Banking Act of 1935, which appear tobe minor and laraely insignificant

[Al ofthe above pastors of results are also abained when the average corporte-

Treasury spread (COTR) was replaced by the A-Treasory hond yield spread (A7R), as ean be seen by comparing modes in Table 2 with corresponding mets in Table 1

42 Monthly Veto

Error Correction Rests

‘Mogel of changes vsibles whose levels have unit root and ae cintegrated, should

‘not omit information about long-run reatonships to avoid misspecification (Engle and Grange 1987), In sdition to seeing how controling fo shorten efests may alter estimates of long-ron relationships (see Tables 1 and 2, estimating those vetoreror correction models (VECM'S)

‘whic jointly estimate long-run and short-run elationships, is help in seeing wheter the long-

an celationships help explain shor-san movements In estimating short-run movements (ist itferencesoflogclevels), the VECM’s regres the First ference i each longer variable oa

lagued frst differences of each long-run factor and include a lagged eror-corection (EC) tem

cult ata minus the estimated equilibrium values of the long-run variable modeled Table 3 presents the sorta models ofthe veltive use of BAs, withthe erv-corection lems based on the corespondingly numbered soinesrating vectors in Tale I hat use the spreads bese the average investment grade corporate bond and fong-tem U.S, government bond yield ‘Thus, the even umbered models each include te hank holiday dummy, while ther models der inthe constation oftheir eror-corestion term and whether they incude an estat ag ist

itference term fom adding any aditional long-run variable to the cinteratng veto:

Several intresting pattems aise, Fis, the error-eorretion rm is hiahly ssntican

‘with the amcipated negative sign, Since the EC term equals actual minus equilibrium use, the

negative ceticien's imply thatthe change in acts! use will decline if atl use exeeeded its estimated euiibsium level in the prior period Since the bond spread variables significant with the amicipated sign in the long-run relationships, the short-nn potions ofthe VECM's also support the financial frictions view that high default risk wil induce a greater relative use of collateralized finance, Second the sizes ofthe EC coefficients ae reasonable, implying that 3:0

4 percent of dsequbria ase cliinated om average each month or roughly 30-10 perent per year, A thied pattern is that dhe hank holiday variable is highly significant, implying some ational light to collateral effet beyond that captured hy the bond risk spree, either

inopliciy either through the eroronretion othe short-run ALCOTR term Foust, consistent

‘withthe comparisons of wectons Ian 2 with vectors Tan 8 in Table | (and 2), models 1 and 2 have similar oeticcnts when the sample excludes the 1935-39 subperid, asin models 7 and 8

‘which ate placed next 0 models | and 2 in Table 3 This implies that evidence i suppot ofthe financial ietons hyposhesis is unaested by the Banking Act of 1938 Fifth, although the banking at variable is statistically insignificant in the long-run relationships fn vectors 5 and 6

in Table 1 its rst dtference is significant in models Sand 6, suggesting thatthe Banking Act

‘of 1935 might have had some short-run effects onthe pattems of Finance Sith, only models

2, 7,and 8, al of which focus solely ona longe-sun relationship between relative BA use

(LBACE an he corporste-Treasury vidd sprcad (C119, ha resins that wore well

‘have in contrast to oer models that tried! to include other (ttsicallyisigificont) longs term factors Finally, any evidence fr short-run effects captured inthe cutrency-to-eposit aio

‘is obtained in model 5, but disappears in the presence of the bank holiday variable ede! 6)

Asa robustness check, similar short-run moves from VECMS tat replace the average corporate bond Treasty yield spread (LCOTR) withthe corporate A-Treasuny (LATE were estimated Comparing modes in Table with corresponding ones in Table 3 indicates thatthe

results are qualitatively and quantitatively similar The only difesence is that the models using

"

the averse corpoate-Treastry yield spread have slightly beter fis and move statistically

Significant error-corretion tems than models using the corporate A-Treasry spread

4.3 Beogencity

‘Anata question is whether the bond yield spreads may he driven by the compsiton of the demand for business Finance, which would greatly complicate the interpretation ofthe above findings, In vector-eror conection systems modeling monthly relative BA use quarecly

relative BA vse (analyzed in seston 44), and quarterly real use oF BA's, the err correction terms are highly sigificant in modeling BA use but ae highly insignificant in modeling the comporte-Trensury bond yield spreads, indicating that these yield spreads are weakly exogenous

{OBA use, These results point to an asymmetry to how the vector components adjust to

isequiibi, with BA use making the Significant adjustments Thas, consistent with theory, the guilibdam compostion of short-term, esteral business eis drives by the combination af etalt and lguidity isk measured i the spreads of corporate over Treasury ond yields

44 Quarterly Models of Relative BA Use

Because monthly data are noisy and might abscure long-run patterns, the approach was

repeated using quarterly averages of monthly data, Table Sand 6 summarize findings that

correspond to the monthly results in Tables 1, with three differences First, Tables § and 6

omit esl that included a Longa damn forthe Banking Act of 1935 (1 fom 1935 g3 1030.0), which wer similar to monthly results but were exch to conserve space

[Nevertheless some madels were tests to cant for any short-run eects by including a short

‘un dammy, BACTDD, which equals {in 1935ig3 and -1 in 1935iq4 Second, because the unsber of degrees of eeedon is wraty smaller, sample excluding 193-1939 was not

resented ong oa very short-sample Third, a quately du variable forthe stock market rash of 1929 (STCRASHDD, equals 1 in 1929.44, -1 in 1930 aH, and O otherwise) was included

in some models a5 a robustness check By construction, S7CRASHDD imposes that any

short-tam boosto he uc oŸ BA's rdlaiYe to contneial papetìn 1929:g† is ungound completeiy in the First quarter of 1930, Likely reflecting an unwinding thats aster than the norms! ero comecton process, this variable outperformed a dummy that equaled 1 in 192994 and,

«athens in earessions not shown In monthly models not reported, a monthly dummy forthe (October 1929 crash was very insignificant, in contrast othe significa quarry coefficient found in every model, "This iference may reflect thatthe nosines of vontly data sight scuresomte shortun effects, sfc, vetly lhe patton seen in he monthly nests were obtained inthe quarterly models The notable exception s that the inclusion of controls forthe short-run effets ofthe bank holiday, the October 1929 stock market crash and the Banking Act

(of 1935 more substantially improves the fit ofthe short-run models Nevertheless,

comesponding monthly nd quacely models that omit such short variables had similar fis

“The non stationarity of corporate spreads use over the sarmple may concern some senders

‘who are uncomfortable with the notion of ni arin corporate spreads (aot that he currency

«epasit ratios clearly had long-lasting changes in thei levels) "Tall this conser, one-stage versions of models 1-3 and 5-7 were nun replacing errr-correction terms (dated t-1) from

ration tests with thet lag levels of elative BA use and corporate spreads, Two quarterly lags ofthe First ference of relative BA use and one quately lag ofthe st difference in corporate spreads are inchaded to contol rote shortrun dynamic effects Ae shown in Append Tale At, the qualitative results ae Similar: In these model, the igificant negative

coefficients on the lag low-level of BA use imply dat the wap between the implied long

equilbvium and actual BA use is eliminated at quarterly speeds of adjustment of 11-16 percent

45 Quarterly Models of Real BA Use

‘To see whether these basic results are robust to using absolute rather than relative levels

(of BA use, the [oy fevel of veal BAs (RAFET")is analyzed with sniar variables with afew adjustments Fist, BAs are deflated using the overall wholes price index (source: NBER.

.Maeiohistory Database) Second, electing the need for some control forthe impact of overall conan atvity on the demand fr finance, the NBER index of business setivity (HUSACT) is included in addition ro the ong-nun variables use earlier As indicated inthe upper-panel of Table 7, the log levels ofboth variables (denoted with L's atthe stato level variable names) have unit roots according to augmented Dickey-Faller statistics Cointgration tests were frst conducted on vectors numbers 1 and Sin Table 7) that inelided just ALE, one spread

‘stable COIR or LAR), and LBUSACT, Thìn also tested wore oer vectors (numbers 2 and 6, Table 7 tha ed the bank holiday variable, as well as vectors (numbers Sand 7) that adied all hrc short rn variables, AANKHOLIDAY, BACEDD, and STCRASH These are sila o variables usin aalyaing quately, relative BA use (HAC?) except thatthe October

1028 stock crash dummy equals ia 1929-4 and O otherwise This impli assumes less oF an vnsinding len 1930 g1 than inthe stock crash dummy ithe BACP models, which also equaled evinus 1 in 1930:q1 Consistent wit the diferences inthe constsction ofthese stock

«rash variables was a sharp plunge of commercial paper in 192094 followed bya sharp recavery

n 1930 g1, hích would affect JBACP more than LBALEV, Finally, corresponding 0 vectors 3 and 7 which include al thee short-run Financial variables, vectors 7 and 8 also inchue helo of the cursency-todemsand deposit ratio, LCDRATIO, which ha a8 statics close to exhibiting anit oot than the curreney-to-total deposits aia (LCZDRATIO), (in ober models

presented 9 conserve space—a long-run dummy For the Banking Act of 1935 was insignificant)

Unique cointgrating relationships were found in each cas, with ly lengths of 4 forall butthe two VECMs (vectors | and 5) that exchided RANKMOLIDAY, RACED, and SICRASH,

im which ease S las were need fo Find # unique vector Several pater emerge fiom these

‘vectors in Table 7 and the shortnn models in Table 8 Fis, the quality spread and business setivity variables are statistalysiaican, wit the hypothesized signs, reflecting tha BA use

{is incveasing in the levels of default risk and economic activity Second, the eurteney-to-deposit

rao is gaisdealy sigifican, vì aien ndicadng thất high curency-to-deposit ratios ace associated with lower BA use, ceteris paribus This result suggests that thế nggnveeies oŸ investor concerns about the value of bank guarantees on BA that may have been reflected inthe cureney-depasit ratio likely outweighed the postive effects of increased incentives fr banks to issue BAS than make traditional loans, ‘Third, the inclusion of the eurreney-deposit ratio

increases the magnitude ofthe coefficient on the quality spread, while reducing that ofthe

business activity index Nevertheles, the qualitative results regarding the effects of

liquidity/defaul risk and business activity on BA use are unaffected by the presence of currency: deposit ratios Also noteworthy is how the inclusion ofthe short-run banking/financal variables

very noticeably improves theft ofthe corresponding models, implying thatthe banking act of

1935 had temporary depressing effects on the real level of BA use, while the stock market erash and bank holiday episodes induced more use of BA financing, consistent with a Might to

collateral during flights to quality inthe corporate debt markets Nevertheless, in the presence of

Figure 5: Estimated Equilbrium Real BA Levels Tracks and

Slightly Leads Actual Levels During the Depression Era,

autre inl by "esters (abo 7),

Aeeeetee

these short-run event contos, the implied long-run equilibrium relationship from quality spreads and the busines atvity index track an shy lead actual log levels of real Bs well (Figure 5), Overall, the findings forthe level of real BA use are consistent with those for the use of BAS relative to commercial paper

Y Bond Yield Spreads and Business Credit Composition

sakes the 10 of BAs commercial pape lest informative, Also, the deepen The

commercial paper sakes from the ise of asset-backed paper implies ha he lnks between commervial peper an veal activity are more complicated then during the 1930s Accordingly instead of BACP or te real levels of BAs a commercial paper, we model commercial paper as a share of commercial paper plas business loans (CP BEAM, se Kashyap, Wileos, and Sei, 1997), CPBLAMINs total commercial paper oustanding (consistently defined since 2001) as a shar of commercial paper and commercial and industrial loans tll commercial banks It itters fom the mix variable of Kashyap, etal (1997) in neludig ll commercial paper, rather than ost paper issued by nonfinancial corporations, The reason forthe diference isthe

increasing we duns

he carly pat ofthe decade of erat at many nonfnancial companies who Yorowel financial entities who funded the coi by issuing asset-backed paper Paper ised inclly by nonfinancial fins declined i relative terms during much ofthe decade, as did bank loans, beth of which lost marketshare to asse-backe paper funded exe

“To make the analysis moe comparable with that forthe Gret Depression, CPREMENS

‘odeled as # fneton ofthe spread berween Baa corporate and 10-year Treasury bond yields

(B4A101R).` Rises inthis spread reflect that investors demand higher premiums for corporate default and liquidity risk Such rises are less ofa threat othe funding of bank loans, as banks hhad greater access to insured deposits and the Fed's liquidity facilities, particularly before mid October 2008, As.a result, the price and non-prie terms of commercial paper would likely rise relative to those of bank loans, implying a negative relationship between the commercial paper share of short-term business credit (CPBLX) and the corporate yield spread (Figure 6a and 6b), Interestingly, spreads between Baa and Aas-rated corporate bond yields rose almost as high

in late 2008 as during the worst of the Great Depression (Figure 7), Being the largest component

of the Baa-10-year Treasury yield spread, this spread suguests that default and liquidity risk between corporate and Treasury bonds likely rose to levels not een since late 1933,

‘igure 6a: Relative Use of Commercial Paper Usually Vares eset Inversely withthe Corgorate Bond Yield Spread