The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
____________ Review____________ Vol. 67, No. 3 March 1985 5 Why the Big Rise in Business Loans at Banks Last Year? 14 Money Demand Dynam ics: Som e New Evidence 24 The Federal Reserve Reaction Function: Does Debt Growth Influence M onetary Policy? The Review is published 10 times per year by the Research and Public Information Department o f the Federal Reserve Bank o f St. Louis. Single-copy subscriptions are available to the public free o f charge. Mail requests fo r subscriptions, back issues, or address changes to: Research and Public Information Department, Federal Reserve Bank o f St. Louis, P.O. Bo}c 442, St. Louis, Missouri 63166. The views expressed are those o f the individual authors and do not necessarily reflect official positions o f the Federal Reserve Bank ofSt. Louis o r the Federal Reserve System. Articles herein may be reprinted provided the source is credited. Please provide the Bank’s Research and Public Information Department with a copy o f reprinted material. Federal Reserve Bank of St. Louis Review M arch 1985 In This Issue . . . In th e first article in th is Review, "W hy th e Big Rise in B u sin ess L oan s at Banks Last Y ear?” R. Alton Gilbert an d M ack Ott exam in e th e rea so n s for th e u n u su al p atte rn o f b u sin ess lo an s at large co m m ercial banks d uring th e cu rre n t e co n o m ic exp an sio n . T h e au th o rs find that, b o th before an d after th e first h alf o f 1984, b u sin ess lo an s w ere grow ing at rates sim ilar to th o se fo r co m p arab le p erio d s in earlier ec o n o m ic ex p an sio n s. A surge o f b u sin ess lo an s at large co m m e rcia l banks d uring th e first h a lf o f 1984, how ever, raised su ch lo an s above th e level th at w ould have b ee n ex p ected on th e basis o f p ast cy clical p attern s. Am ong th e p o ssib le exp lan ation s for th e surge in b u sin e ss lo an s in th e first h alf o f 1984 is th e u nu su ally rapid rise in th e p a ce o f ec o n o m ic activity d urin g th at p eriod ; th is co u ld have in d u ced an u nu su ally large rise in cred it d em an d by b u sin ess firm s. Gilbert an d Ott do n ot find em p irical su p p o rt for th is exp lan ation . In stead , they co n clu d e th at th e m o st im p o rtan t fa cto r co n trib u tin g to th e surge o f b u sin ess lo an s w as ban k fin an cin g o f co rp o ra te m ergers an d leveraged bu youts. Unlike th e o th e r exp lan ation s, b o th th e tim ing an d m agn itu d e o f th e u nu su ally large am o u n t o f m ergers an d bu youts in early 1984 m a tch up w ith th e su rge o f b u sin ess loans. In th e seco n d article in th is Review, “M oney D em and D ynam ics: Som e New E vid en ce,” D aniel L. T h o rn to n briefly reviews th e n atu re o f th e th ree stand ard d ynam ic m o d els o f m o n ey d em an d a d ju stm en t. T h e a u th o r p o in ts out th at th ese sp ecificatio n s are n ot statistically co m p arab le. In stead , h e n o te s th at th ese m od els sh o u ld be co m p ared in term s o f th e ir co n fo rm ity w ith th eo ry an d th e ir stability over tim e. E stim atin g th e se sp ecificatio n s for th ree su b p erio d s b etw een II/1951-II/1984, T h o rn to n finds th at (1) all th ree sp ecifica tio n s are sensitive to the sp ecificatio n o f th e long-run d em an d for m oney, (2 ) n o n e is co n siste n t w ith th eo ry over all th ree su b p erio d s an d (31 n o n e exh ib its tem p o ral stability. He co n clu d e s, how ever, that, sin ce n o n e o f th e se sp ecifica tio n s m ay ad equ ately re p resen t th e sh o rt-ru n d em an d for m o n ey w h en estim a ted u sin g aggregate data, th e resu lts m ay say little abou t th e instability o f m o n ey d em an d . T h e p ro sp ect o f large federal d eficits has ren ew ed co n c e rn s abou t th e Fed eral R eserve’s ability to co n d u ct in d e p en d e n t m o n etary policy. In th e th ird article in th is issue, “T h e Fed eral Reserve R eactio n F u n ctio n : D oes D ebt G row th In flu en ce M onetary P olicy?” R ichard G. S h eeh a n offers so m e evid en ce on th e ex ten t to w h ich th e Fed eral Reserve h as altered m o n etary p o licy in re sp o n se to federal deficits. T h e relatio n sh ip s betw een th e deficit, th e m o n ey stock an d in terest rates d ep en d on th e n atu re o f th e d eficit an d th e targeting p ro ced u res u sed by the Fed eral Reserve. T h e a u th o r u ses a sim p le m od el to sh o w th a t a p o licy -in d u ced or stru ctu ral d eficit m ay lead to h ig h er in terest rates u n d e r a m o n etary aggregate targeting strategy o r to h ig h er m o n ey grow th u n d e r an in tere st rate targeting strategy. In co n trast, a cy clica l deficit, in d u ced by a re cessio n for exam p le, will be a cco m p an ie d by low er in terest rates o r slow er m on ey grow th. Using a reactio n fu n ctio n ap p roach , Sh eeh a n finds evid en ce th at stru ctu ral d eficits led to h ig h er m o n ey grow th before 1971 w h en th e Fed eral Reserve tar- 3 In This Issue . . . Digitized 4 for FRASER geted in tere st rates. Sin ce th en , th e Fed eral Reserve h as, at le a st in part, fo cu sed on a m o n etary aggregate target, an d th ere is n o ev id en ce th at m o n ey grow th h as b ee n in flu en ced by stru ctu ral d eficits. F urther, th ere is n o evid en ce suggesting that h ig h er stru ctu ral d eficits have in crea sed in terest rates over any period . Why the Big Rise in Business Loans at Banks Last Year? /{. Alton Gilbert and Mack Ott B u s in e s s loans at large commercial banks rose sharply in the first half of 1984, after changing little in the first year of the econom ic recovery from the 1 9 8 1 82 recession. Many analysts of business and financial data cite the rate of growth of business loans at large commercial banks as an indicator of the pace of eco nomic activity.1 Given that interpretation, this pattern of business loans would have signaled a sharp acceler ation in econom ic activity in the first half of 1984. Some analysts offer the alternative explanation for the rapid rise in business loans that bank credit was being used on an unusually large scale to finance corporate mergers and leveraged buyouts.2 This article in vestigates the empirical support for these alternative explanations. f t Alton Gilbert is an assistant vice president and Mack Ott is a senior economist at the Federal Reserve Bank of St. Louis. Paul G. Chris topher provided research assistance. ’See Berry (1984), Heinemann (1984), Jasinowski (1984) and Wei ner (1984). In particular, Weiner recounts a banker as emphasizing that ...loan demand has been real_It is being used to finance inventory, plants and equipment — not mergers and acquisitions.... 2See Giordano (1984). Economist Henry Kaufman is quoted by Busi ness Week as crediting merger finance for the bulge in loans (see “Do Mega-Mergers...” (1984)): “Much of the recent credit growth was associated with the merger phe nomenon." He says that merger activity explains why first-quarter busi ness loans at large commercial banks showed an increase of $4.6 billion rather than a decline, as they generally do at this stage of the business cycle. IS THE GROW TH OF BUSINESS LOANS IN THE CURRENT EXPANSION ANY DIFFERENT THAN IN PAST EXPANSIONS? The first step in analyzing the relationship between the pace of econom ic activity and the growth of busi ness loans is to examine the pattern of business loans in the current expansion relative to the patterns in previous expansions. If the pattern of business loans in the current recovery is typical of the pattern in prior expansions, alternative explanations such as bank financing of mergers and buyouts are unnecessary. Business loans by large (weekly reporting) banks tend to stay at about the same level as at the recession trough during the first year of a recovery period, with this sluggish loan behavior giving way to moderate growth in the second year (chart 1). Thus, for the first year of the econom ic expansion following the 1 98 1-82 recession, the pattern of growth in business loans by large commercial banks was similar to the pattern of business expansions since 1960. The growth of business loans at large banks in the first half of 1984, however, was unusually rapid for that stage of an econom ic expansion. By the spring of 1984, the level of business loans, relative to their level at the trough of the preceding recession, was substantially above the average profile during previous expansion 5 FEDERAL RESERVE BANK OF ST. LOUIS C hart MARCH 1985 1 Business Loans at Large Commercial Banks Relative to Levels at Recession Troughs Retession Trough = 1.0 Av er a g e — 1.2 Retession Trough = 1.0 1.2 - — Cur r ent cycl e 1.1 1.1 ____ 1.0 ^____ y v'— .9 - ___ ' ^ ^ 1.0 .9 S ’ .8 .8 1 1 1 I I ...................... -12 -10 -8 -6 -4 -2 1 1 1 1 1 II 0 2 4 6 1 1 1 1 1 ! 1 i II 8 10 12 14 16 1 1 11 18 20 II 22 24 Months to and from Recession Trough NOTE: The shaded area represents the range of values observed in business cycles with reces sion troughs in 1960 through 1975. The profile for the cycle around the 1980 trough was excluded because its succeeding recovery did not last two years. The line labeled “ average” is the average for these prior reference cycles. Chart 2 Cycles in N o m in al GNP Recession Trough = 1.00 1.24 Recession Trough = 1.00 1.24 . 9 2 ________ I_______ I________I_______ ________I_______ I_______ I_______ I_______ I________I________I----------- .92 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 Quarters to and from Recession Trough NOTE: The shaded area represents the range of values observed in business cycles w ith recession troughs in 1960 through 1975. Th e profile fo r the cycle around th e 1980 trough w as excluded because its suc ceeding recovery did not last tw o years. The line labeled “ av era g e” is the average fo r these prior reference cycles. 6 MARCH 1985 FEDERAL RESERVE BANK OF ST. LOUIS Chart 3 Cycles in N onfarm Business Inventories Recession Trough = 1.00 Recession Trough - 1.00 Quarters to and from Recession Trough NOTE: The shaded area represents the range of values observed in business cycles with recession troughs in 1960 through 1975. The profile for the cycle around the 1980 trough was excluded because its suc ceeding recovery did not last two years. The line labeled “ average" is the average for these prior reference cycles. period s. Follow ing th e ir rapid ru n -u p during th e first half o f 1984, th e grow th rate o f b u sin ess lo an s at large banks h as b e e n clo se to th e average grow th rate during co m p arable p erio d s o f exp an sio n . T h is ca n be see n by th e parallel m ovem ent o f b u sin ess lo an s sin ce th e first h a lf o f 1984 and that o f th e cy cle average. C o n se quently, the only asp e ct o f th e grow th in b u sin ess loan s by large co m m ercial banks that disting u ish es the cu rren t ex p an sio n is th e rapid grow th in th e first h alf of 1984. PERHAPS ECONOMIC GROWTH WAS UNUSUALLY RAPID LAST YEAR T his u nu su ally rapid rise in b u sin ess lo an s in the first h alf o f 1984 m ight reflect sim ply an u nu su ally rapid rise in the p ace o f eco n o m ic activity. T h is p o ssi bility is investigated in ch art 2 , w h ich p resen ts n o m i nal GNP for several qu arters befo re an d after recessio n trou ghs in ratio to n om in al GNP in trough qu arters. If th e p ace o f e co n o m ic activity h ad b eco m e unusually rapid by th e first h a lf o f 1984, th e ratio in ch a rt 2 for the cu rren t ex p an sio n w ould b e su b stan tially above th e range for previous ex p an sio n p eriod s. Yet, w hile the ch art show s th at n om in al GNP did rise to th e top o f th e range for previous referen ce cy cles d uring th e first h alf o f 1984, it did n o t rise su b stan tially above the range o f ex p e rie n ce in p rio r ex p an sio n period s. An im portan t link b etw een th e p a ce o f eco n o m ic activity and th e grow th o f b u sin ess lo an s involves the grow th o f b u sin ess inventories. B u sin ess inventories tend to rise w ith th e p ace o f ec o n o m ic activity during ex p an sio n period s, an d b u sin e sse s co m m o n ly fin an ce th eir inventory in v estm en t th rou gh b an k lo a n s .3 Thus, th e u nu su ally rapid grow th o f b u sin ess lo an s last y ea r m ight reflect an u n u su ally rapid grow th o f b u sin ess in v en to ries. C h art 3, how ever, in d ic a te s th a t th e grow th o f b u sin ess in v en tories h as n o t b e e n u nu su ally rapid in th e cu rren t ex p an sio n . In fact, inventory ex p an sion has b een slow er in this reco v eiy th an in any recovery o f th e past 25 y ears. 3See Hicks (1980). 7 FEDERAL RESERVE BANK OF ST. LOUIS MARCH 1985 Chart 4 R atio of Fixed Pius Inven to ry Investm ent to Intern al Funds 1952 54 56 58 60 62 64 66 68 70 Source: Board of Governors of the Federal Reserve System NOTE: Data are for the nonfinancial corporate sector. Shaded areas represent periods of business recessions. C hart 4 provides an ad d itio n al p ersp ectiv e on th e rise in th e d em an d for cred it by b u sin e sse s in th e cu rren t and previous ex p an sio n s. T h is ch art plots the ratio o f th e fixed an d inventory investm en t o f n o n financial b u sin ess co rp o ratio n s to th e ir internally g en erated funds. N onfinancial co rp o ratio n s requ ire ex ter nal fin an cin g o f th e ir fixed an d inventory investm ent w h en ev er th is ratio is above unity, bu t in tern ally g en erated fund s su b stitu te for extern al fin an cin g w h en ever this ratio is below unity. As illu strated in ch art 4, in tern ally g en erated fund s ten d to rise m ore rapidly th a n b u sin ess in vestm ent during th e early stages o f ec o n o m ic exp an sio n s, so th at th e ratio is belo w unity. After a few qu arters o f expan sion , b u sin ess in vestm ent beg ins to ex ceed in ternal funds, an d b u sin esses th e n m u st tu rn to ex ter nal fin an cing. T h e relatio n sh ip betw een internally g en erated funds and b u sin e ss investm en t is o n e o f th e reason s for th e typical lag in th e grow th o f b u sin ess loan s after th e beginn ing o f ec o n o m ic ex p an sio n s il lu strated in ch a rt 1. C hart 4 clearly in d icates th at the relatio n sh ip betw een b u sin ess investm en t an d in ter http://fraser.stlouisfed.org/ 8 Federal Reserve Bank of St. Louis 72 74 76 78 80 82 1984 nal funds in th e cu rren t ex p a n sio n is typical o f prior exp an sio n s. T h u s, ch a rt 4 provides n o exp lan ation for the u nu su al rise in b u sin ess lo an s in th e first half of 1984. Finally, co n sid e r th e co n tra st b etw een th e p attern of b u sin ess lo an s at large an d sm all co m m ercia l banks in th e cu rren t recovery. If b u sin ess lo an s at large banks reflect th e in flu en ce o f th e p a ce o f eco n o m ic activity on b u sin ess cred it d em an d , th e se effects w ould ten d to b e sim ilar for b o th large an d sm all banks. C hart 5 in d icates, how ever, th at th e p attern of b u sin ess lo an s at small banks in th e cu rre n t ex p a n sion is very sim ilar to th e p a ttern s in previous ex p a n sio n s .4 T h u s, th e in flu en ces th at co n trib u te d to th e u nu su ally rapid grow th o f b u sin ess lo an s at large co m m ercial banks in th e first h alf o f 1984 did n ot have sim ilar effects in bo o stin g th e grow th o f b u sin ess loan s at sm all banks. 4The series on business loans of small commercial banks was calcu lated by subtracting business loans of weekly reporting banks from a series on business loans of all commercial banks. FEDERAL RESERVE BANK OF ST. LOUIS Chart MARCH 1985 5 Business Loans at Small Commercial Banks Relative to Levels at Recession Troughs 8 __I__I__I__I__I__I__I__I__I__I__I____I__I__I__I__I__I__I__I__I__I__L_l__I__I__I__I__I__I__I__I__I__I__I__ -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20 22 g 24' Months to and from Recession Trough NOTE: The shaded area represents the range of values observed in business cycles with reces sion troughs in 1960 through 1975. The profile for the cycle around the 1980 trough was excluded because its succeeding recovery did not last two years. The line labeled “ average" is the average for these prior reference cycles. Chart 6 Cycles in Nonfinancial Commercial Paper -4 -3 -2 -1 0 1 2 Quarters to and from Recession Trough 3 4 5 6 7 8 NOTE: The shaded area represents the range of values observed in business cycles with recession troughs in 1960 through 1975. The profile for the cycle around the 1980 trough was excluded because its suc ceeding recovery did not last two years. The line labeled “ average" Is the average for these prior reference cycles. 9 FEDERAL RESERVE BANK OF ST. LOUIS MARCH 1985 Chart 7 Ratio of Funds Raised by Long-Term Borrowing to Total Funds Raised by Borrowing1-1 Source: Board of Governors of the Federal Reserve System LI The ratio of funds raised by issuing bonds and mortgages to total funds raised in the financial markets, net of new equity issues. NOTE: Data are for the nonfinancial corporate sector. Shaded areas represent periods of business recessions. PERHAPS THERE WAS A SHIFT IN THE COM POSITION OF EXTERNAL FINANCING BY BUSINESS FIRMS T h e previous sectio n in d icates th at th e rapid rise in b u sin ess lo an s at large banks last y ea r did n o t reflect an u nu su ally rapid rise in th e d em an d for external fin an cin g by b u sin ess firm s. T h e u nu su al p attern of b u sin ess lo an s at large banks, therefore, m u st reflect an u n u su al ch an g e in th e composition o f extern al financing by b u sin ess firm s. T h is sectio n co n sid ers various p o ssib le ch an g es in th e co m p o sitio n o f b u si n ess fin an ce th at m ight a cc o u n t for th e rise in b u si n ess lo an s at large banks. A Shift to Bank Loans from Other Forms o f Short-Term Credit One p o ssib le shift m ay have involved an unusually large re d u ctio n in sh o rt-term bo rrow in g by b u si n esses from so u rces o th e r th a n co m m ercial banks. T h is p ossibility is investigated by exam in ing th e p at terns o f co m m ercial p ap er o u tstan d in g issu ed by n o n fin an cial firm s in th e cu rren t an d previous eco n o m ic 10 exp an sio n s. C hart 6 in d ica tes that, ra th er th an d eclin ing in th e first h a lf o f 1984, th e grow th o f n o n fin an cial co m m ercial p a p e r d uring th at perio d w as relatively rapid. T h u s, th e u nu su ally rapid grow th o f b u sin ess loan s at large banks in 1984 d oes n o t reflect a shift of b u sin ess cred it from th is type o f sh o rt-term cred it to loan s at large banks. A Shift from Long-Term to ShortTerm Debt A nother p o ssib ility is th at n o n fin an cial b u sin ess co rp o ratio n s re d u ce d th e ir issu a n ce o f lon g-term d ebt to an u nu su ally large d egree in th e first h a lf o f 1984, su bstitu tin g sh o rt-term for lo n g -term d eb t as a so u rce o f funds. C hart 7 p re sen ts th e ratio o f fu n d s raised by n o n fin an cial firm s by issu in g b o n d s an d m ortgages to total fu n d s raised th rou gh borrow ing. T h is ratio has fallen during th e cu rre n t e c o n o m ic exp an sio n , b u t its d eclin e is sim ilar to th e d eclin es in p ast ex p an sio n period s. C on sequen tly, th e u n u su ally rapid grow th of b u sin ess lo an s at large ban k s d o es n o t ap p e a r to b e the result o f an u n u su al ch an g e in th e co m p o sitio n of sh o rt-term an d lon g-term d ebt as so u rces o f fund s for n o n fin an cial b u sin ess co rp o ra tio n s. FEDERAL RESERVE BANK OF ST. LOUIS MARCH 1985 Chart 8 Ratio of N e t N e w Equity Issues to N e t Funds Raised 1952 54 56 58 60 62 64 66 68 70 Source: Board of Governors of the Federal Reserve System NOTE: Data are for the nonfinancial corporate sector. Shaded areas represent periods of business recessions. Reductions in Corporate Equity Accounts Financed Through Short-Term Borrowing O ne featu re o f b u sin ess fin an ce th a t d o es stan d out as u n u su al in th e cu rren t ex p an sio n is th e large n ega tive values for th e n et fund s raised by n o n fin an cial firm s throu gh equity issu es (chart 8 ). A negative value im plies that th e tran sactio n s th at reduced th e equity cap ital a cc o u n ts on th e b alan ce sh e e ts o f so m e p u b licly h eld co rp o ratio n s w ere larger th an th e funds raised by o th e r su ch co rp o ratio n s throu gh n ew equity issues. T h e net red u ctio n s in equity o f the n o n fin a n cial co rp orate se c to r as a result o f fin an cial tra n sa c tion s in ea ch o f th e first th ree q u arters o f 1984 w ere the largest quarterly red u ctio n s ever in th e flow o f funds data series. (T hese qu arterly d ata are available from 1952.) T h e co n tem p o ran eo u s, rapid rise in b u sin ess loan s at large banks su ggests th at n o n fin an cial co rp o rations fin an ced at least part o f th e tran sactio n s in volved in red u cin g th eir equity a cc o u n ts through lo an s from large banks. Types o f financial transactions that reduce co rp o rate equity — O ne type o f tra n sa ctio n th at co n trib u tes 72 74 76 78 80 82 1984 to a negative value for fu n d s raised th rou g h n et co rp o rate equity issu es involves a b u sin ess buying back sh ares o f its ow n co m m o n stock. A n oth er is co rp orate equity red u ctio n s d ue to m ergers in w h ich th e sh a re h o ld ers w ho sell th e ir sh ares receive c a s h .5 T h e re w ere som e large m ergers involving c a sh p aym en t in 1984. In re cen t y ears, leveraged b u yo u ts (LBOs) also have re d u ced co rp o ra te equity cap ital a cc o u n ts. In an LBO, an investor or group o f investors borrow s funds, often from co m m ercial banks, to buy th e sh a res o f publicly traded co m p a n ies an d con vert th em to privately held co m p an ies. T otal p u b licly h eld co rp o ra te equity is re d u ced by th e value o f th e co rp o ra tio n ’s equity w h en it is taken private. A list o f large LBOs (each o f value over $ 10 m illion) tabu lated by an in d u stry so u rce in d icates that investors b orrow ed $13.5 billion to fin an ce LBOs 5For a cash purchase of its own stock, the corporation’s assets and equity decline by the same amount. In the case of a purchase of its own stock financed by borrowings, the debt liability rises by the same amount as equity declines. When a merger is effected by one corporation buying another with cash, the equity of the surviving corporation will be the same as before the merger. Thus, aggregate corporate equity is reduced by mergers effected through cash pay ments to shareholders of the corporations being purchased. 11 MARCH 1985 FEDERAL RESERVE BANK OF ST. LOUIS Chart 9 Stock Retirements and the Change in Business Loans at Large Commercial Banks 1970 71 72 73 74 75 76 77 78 79 80 81 82 83 1984 Shaded areas represent periods of business recessions. in 1984, up from $3.2 billion in 1982 and $2.8 billion in 1983. equity retirem en ts by n o n fin an cial co rp o ra tio n s rose su bstan tially in 1984 relative to th e p ast. T h e negative values for net fund s raised throu gh issuing eq u ities reflect tra n sa ctio n s th at re d u ce p u b licly h eld co rp o rate equity. T h e m ag nitu d e o f th ese tra n sa ctio n s can be m easu red by developing a series o f stock retirem en ts. R etirem en ts are red u ctio n s in co rp o ra te equity claim s held by th e p u blic as a result o f (1 ) re p u rch a ses o f th e ir ow n sh ares by co rp oratio n s, (2) m ergers tra n sa cted as ca sh p u rch a se s an d (3) LBOs. R etirem en ts are co m p u ted as th e d ifference betw een th e total fund s raised by co rp o ratio n s throu gh th e sale o f n ew sh ares m inu s th e n et fund s raised through equity is su e s .6 T h e n et fu n d s raised reflect th e retire m en ts o f equity as a d ed u ctio n . As show n in ch a rt 9, Magnitudes o f Equity Retirements and Changes in Short-Term Debt — T h e tim ing o f th e equity retire 6The series on stock retirements should be considered an approxi mation, rather than a census of retirements at all firms. The retire ments series is negative in some quarters because of differences in the transactions covered in the series on total funds raised by nonfinancial corporations through the sale of new shares and in the series on net funds raised through equity issues. The series on net funds raised includes estimates of dividend reinvestments and con versions of debt to equity, which are not reflected in the series on funds raised through new equity issues. The series on retirements has a downward bias because of these differences in its compo nents. 12 m en ts show n in ch art 9, as w ell as th e ir cum ulative m agnitude, m a tch es up w ith th e a n o m alo u s behavior o f b u sin ess lo an s by large banks sh o w n in ch a rt 1. T h e m agnitude o f th e ru n -u p in th e se lo a n s during th e first th ree qu arters o f 1984 — over $26.5 billion — is about 40 p ercen t o f th e total value o f equity retirem en ts in this period, $65.7 billion. Ind eed , th e largest q u a rter of equity retirem en ts w as th e seco n d quarter, $30.2 bil lion, w h ich w as also the largest loan deviation from th e past p attern — a rise o f $14.5 billion in b u sin ess loan s co m p a red w ith th e average o f $2.5 billion d uring previous ex p a n sio n s .7 'Financing stock retirements is only one reason why businesses borrow from commercial banks. Chart 9 indicates that, from 1970 through about 1976, there was no relationship between the patterns of stock retirements and changes in business loans. The major reason for the pattern of changes in business loans during that period appears to be inventory investment financed through bank loans. Businesses increased their inventories substantially in 1973 and 1974, then gradually reduced them during 1975 and 1976. For an analysis of the determinants of business loan demand during that period, see Gilbert (1976). FEDERAL RESERVE BANK OF ST. LOUIS CONCLUSIONS B u sin ess lo an s o f large co m m ercial banks follow ed th e u su al cy clical p attern d uring th e first y e a r o f the cu rren t ec o n o m ic ex p an sio n . T hey rem ain ed at about th e sam e level as at th e trou gh o f th e prior recessio n . During th e first h alf o f 1984, how ever, b u sin ess loan s of large co m m ercial banks ro se at rates ex ceed in g th o se o f co m p arable p eriod s o f previous e co n o m ic ex p a n sions, before resu m in g norm al grow th in th e seco n d h alf o f 1984. T h e u nu su ally rapid rise in b u sin ess lo an s at large co m m ercial banks in th e first h a lf o f last y ea r d o es not reflect an u nu su al rise in b u sin ess d em an d for ex ter nal financing. T h e rise in th e p ace o f e co n o m ic activity an d th e req u irem en ts o f b u sin ess firm s for extern al fin an cing in th e cu rren t ex p an sio n have b een sim ilar to p attern s in previous exp an sio n s. T h e m ost im p or tant factor that a cc o u n ts for th e u nu su ally rapid grow th o f b u sin ess lo an s in th e first h alf o f 1984 is th e in crease in co rp o rate m ergers an d leveraged buyouts fin an ced w ith lo an s from large banks. T hu s, o u r analy sis o f th e p attern o f b u sin ess lo an s in th e cu rren t MARCH 1985 eco n o m ic ex p an sio n in d ica tes th at th e grow th rate of b u sin ess lo an s ca n b e an u n reliab le in d ica to r o f the p ace o f eco n o m ic activity, sin c e it ca n b e in flu en ced su bstantially bv ch a n g es in th e co m p o sitio n of ex ter nal fin an cin g by b u sin ess firm s. REFERENCES Berry, John M. "Money Supply Figures on Target,” Washington Post, July 13, 1984. “Do Mega-Mergers Drive Up Interest Rates?” 16, 1984), pp. 176-80. Business Week (April Gilbert, R. Alton. “Bank Financing of the Recovery,” this Review (July 1976), pp. 2-9. Giordano, Robert M. “Financial Market Perspectives: Look Before Your Leap,” Economic Research, Goldman Sachs Financial Mar ket Research (July 1984). Heinemann, Sally. “Fixed Income Markets Still Stabilizing,” New York Journal of Commerce (June 18, 1984). Hicks, Sydney Smith. "Commercial Banks and Business Loan Be havior," Journal of Banking and Finance (June 1980), pp. 125-41. Jasinowski, Jerry J. "Less Stockpiling, More Stability,” New York Times, June 17,1984. Weiner, Lisabeth. “Banks Lag Behind as Business Borrowing Booms,” American Banker (August 22, 1984). 13 Money Demand Dynamics: Some New Evidence Daniel L. Thornton J CONSIDERABLE em p irical w ork an d a significant, b u t co n sid erab ly sm aller, volum e o f th eo retical effort has b een devoted to th e q u estio n o f th e short-ru n, d ynam ic a d ju stm en t o f th e d em an d for m oney. M uch o f the im p etu s for th e em p irical w ork cam e from the cla ssic stud y by C how (1966), w ho em ployed th e p ar tial ad ju stm en t m odel to ch a ra cte riz e th e ad ju stm en t o f a ctu al to d esired real m o n ey b alan ces. C how an d th e alternative n o m in al m o n ey an d p rice ad ju stm en t sp ecifica tio n s.- T h e se sp ecifica tio n s have received co n sid era b le atten tio n in th e literatu re, w ith m u ch o f th e em p irical w ork devoted to d eterm in in g w h ich o f th e se sp ecifica tio n s is m o st co n siste n t w ith th e data, for exam p le, G oldfeld (1976), Hafer an d Hein (1980), Ju d d and Scad d in g (1982a), C oats (1983), Milb o u rn e (1983), H etzel (1984) an d M otley (1984). Although th ere w as early co n c e rn over th e e c o n o m ics o f C how 's sp ecificatio n and its relatively slow estim ated sp eed o f ad ju stm en t, th is sp ecificatio n did n o t co m e u n d e r p articu larly clo se scru tin y u ntil the u n an ticip ated rise in velocity in the m id-1970s and the d eclin e in velocity in th e early 1980s.1 As a result, a n u m b er o f alternative d yn am ic a d ju stm en t sp e cifica tion s have b ee n developed. W hile th e se sp ecificatio n s differ in several fu n d am en tal resp ects, they fall into two g en eral categ o ries: th o se th at assu m e th e p rice level ad ju sts to exo g en o u s ch an g es in th e m on ey stock and th o se th at assu m e th e n o m in al m oney stock ad ju sts to exo g en o u s ch an g es in th e p rice level. C o n se quently, th ree fu n d am entally different, sh o rt-ru n dy n am ic a d ju stm en t p ro cesse s have b een co n sid ere d in th e literatu re: th e real a d ju stm en t sp ecificatio n of T h e p u rp o se o f th is p a p e r is th reefold . First, we review th e literatu re on th e se sp ecifica tio n s an d point out th at n o n e o f th em ca n b e th ou gh t o f as re p rese n t ing ad equ ately th e sh o rt-ru n a d ju stm en t o f actual to d esired m o n ey w h en ap p lied to aggregate data. Second , w e d em o n stra te th at n o n e o f th ese th ree sp ecificatio n s are d irectly co m p arab le statistically .3 C onsequently, th e relative p erfo rm an ce o f th e se alter natives can b e a sse ssed only bv th e ir con form ity w ith Daniel L. Thornton is a senior economist at the Federal Reserve Bank of St. Louis. John G. Schulte provided research assistance. The author would like to thank Tom Fomby for useful comments on an earlier draft. ’See, for example, Goldfeld (1976), Carr and Darby (1981), Coats (1982), Laidler (1980, 1982, 1983), Chant (1976), Judd and Scadding (1982a, 1982b), Hetzel (1984) and Motley (1984). Digitized 14 for FRASER 2Nearly all of the specifications that have been suggested in the literature fall into one of these basic categories, at least to the extent that they have the price level, nominal money or real money as the dependent variables. Furthermore, many of the specific alternatives are concerned with how the demand for real money balances ad justs to changes in its arguments and, as such, are consistent with any of the three fundamental adjustment processes considered here. 3lt has been recognized, especially recently, that these alternatives are nonnested, i.e., none can be obtained by placing restrictions on any of the others. Consequently, most studies compare the fore casts of real money [e.g., Hafer and Hein (1980) and Goldfeld (1976)], or the residual sum of squares [e.g., Judd and Scadding (1982a) and Coats] of alternative models. To date, only Motley (1984) has recognized that the nominal and price specifications are not comparable. MARCH 1985 FEDERAL RESERVE BANK OF ST. LOUIS theory and th e ir stability. Finally, w e investigate the p erfo rm an ce o f ea ch sp ecificatio n u sing th e sam e data and th e sam e estim atio n period, II/1951-II/1984. T h e evidence suggests th at n o n e o f th e se sp ecificatio n s have perform ed w ell over th e en tire period and n on e have b een stable. T h e p ecu liarity o f th is p ro cess w as quickly p oin ted out by W alters (1967). He n o ted that, in th e aggregate, m arket equilibriu m req u ires th at th e d em an d for real m o n ey b a la n ces eq u als th e su pply o f real m oney. If the n o m in al m o n ey sto ck is e x o g e n o u s, eq u ilib riu m requires T h e issu e o f tem p o ral stability is particu larly im p o r tant if on e is to rely on sh o rt-ru n m o n ey d em an d in form ulating sh o rt-ru n stabilizatio n policy. If th e shortrun d em an d for m o n ey is u nstable, th e n attem p ts to stabilize ou tp u t an d p rices in th e sh ort run through 131 M /P* = m", m on etary co n tro l will be u n su ccessfu l b ec a u se differ en t levels o f ou tp u t and p rices will be co n siste n t w ith a given stock o f m o n ey at d ifferent p o in ts in tim e. T h is type o f sh o rt-ru n instability, how ever, d oes not rule out the u sefu ln ess o f m o n eta iy co n tro l for achieving e co n o m ic stabilization over th e lo n g er run. DYNAMIC SPECIFICATIONS OF MONEY DEMAND All sh o rt-ru n m o n ey d em an d sp e cifica tio n s are b ased on th e long-run d em an d for m oney, 111 m" = fix, a, u,l = fIZ,), w h ere m d en o tes real m o n ey b alan ces, X is a set of en d o g en o u s and exog en ou s variables w h ich usually in clu d es som e m easu re o f real in co m e o r w ealth and on e o r m ore in terest rates, an d a is a v ecto r o f u n know n param eters. T h e erro r term is d en o ted by u. All variables are in natu ral logs. Chow based h is sh o rt-ru n sp ecificatio n on th e sim ple and co n ven ient partial a d ju stm en t m ech an ism , (21 m , — m,_, = X lm 1' — 141 P, - P, , = XIP,* - P, T h e co m b in atio n o f eq u atio n s 3 an d 4 resu lts in a sp ecificatio n that reflects W alters’ criticism o f the Chow m od el an d exp licitly rep rese n ts th e so-called p rice a d ju stm en t sp ecifica tio n co n sid ere d by G ordon (1984), Laidler (19831, an d Hetzel. Goldfeld (1973, 19761, on th e o th e r hand , argued a g a in st C h o w ’s s p e c ific a tio n o n m ic r o e c o n o m ic ground s. He co n te n d e d th at it is defective b eca u se it im plies th at an individual ad ju sts real m o n ey b a la n ces fully an d in sta n ta n eo u sly to p rice level ch an g es, but only partially to m o n ey d em an d c h a n g e s .11 As an alternative, h e offered th e n o m in al a d ju stm en t sp ecifi cation, 151 M, - M, , = XIM;1 - M,_,) 0 < X .< 1, w here M '1d en o te s th e d esired level o f n om in al m oney. He argued th at eq u atio n 5 m akes m o re sen se than 0 < X < 1. He sp ecified his a d ju stm en t p ro c e ss on th e basis of individual e co n o m ic behavior, arguing that individ uals m ight ad ju st th e ir actu al stock o f real m oney b a la n ces to th e d esired level in m u ch th e sam e way as th ey m ight a d ju st th eir actu al stock o f c o n su m e r d u ra bles to th eir d esired level. T h is sp ecificatio n h as b een rationalized in a m icro eco n o m ic fram ew ork in w h ich th e sp eed o f a d ju stm en t (X) is d eterm in ed by th e co st o f bein g out o f equilibriu m relative to th e co st o f mov ing to equilibrium [for exam ple, M otley (1967) and Feige (1967)]. 40ne could compare the ability of each model to forecast its depen dent variable by, say, comparing percentage forecast errors (e.g., Hetzel). Such an exercise, while interesting, has little to say about the demand for money. Moreover, no objective comparison can be made, because there is no agreement about which variable is most important. w h ere M an d P* d en o te th e n o m in al m o n ey stock and the long-run eq uilibriu m p rice level, respectively. If M is fixed in th e aggregate an d th e p rice level is ad ju stin g to ch an g es in M, eq u atio n 2 ca n b e th ou gh t o f as the price a d ju stm en t eq u atio n : Substituting equation 3 into equation 2 and holding M fixed so M, = M ,, = M, the resulting expression is equation 4. The reader should note that, while we have not changed notation, the interpretation of X in equation 4 is fundamentally different from that of equation 2. The same is true of the interpretation of \ in equation 5 below. 6This is most easily understood by noting that combining equation 2 with equation 1 implies not only that the long-run demand for nomi nal money is unit elastic with respect to price, but that the short-run nominal demand is as well. This aspect of the real adjustment specification is not odd if one believes that money and bonds are close substitutes for each other, i.e., if a strict Keynesian liquidity preference holds. If money and fixed-dollar-denominated financial assets are held in some desired proportion given the interest rate, an unanticipated change in the price level will affect both money and financial assets proportionally so that an individual’s relative holdings of financial assets and money will be unaffected. This will hold in either a pure asset model or in inventory theoretic transactions models. Thus, it is not unrea sonable to assume that an individual’s demand for real money holdings adjusts instantaneously (or at least very quickly) to unantic ipated price level changes if one believes that the only link between the real and financial sectors is the interest rate. 15 FEDERAL RESERVE BANK OF ST. LOUIS eq u ation 2 , a priori, b ec a u se th e a d ju stm en t o f n o m i nal m on ey to a p rice level ch an g e is partial rath er than in sta n ta n eo u s as eq u atio n 2 im p lies .7 G ordon also argues for th e nom inal ad ju stm en t sp ecificatio n on m icro eco n o m ic grounds." He m ain tains th at th ere are no a d ju stm en t co sts asso ciated w ith p rice-in d u ced ch an g es in real m on ey holdings and, co n seq u en tly , th e only co sts involved in a d ju st ing o n e ’s portfolio are th o se a sso ciate d w ith ad ju stin g n om inal m o n ey b a la n c e s .'1 Laidler (1983) notes, how ever, th at w h en eq u atio n 5 is applied to aggregate data, on e co m m its th e fallacy o f co m p ositio n if th e aggregate n om in al m on ey stock is exogenous. Individuals are free to ad ju st th e ir n o m i nal b alan ces, but so ciety as a w h ole is not. M oreover, Hetzel observes th at applying eq u ation 5 to aggregate d ata is tan tam o u n t to assu m in g th at th e p rice level is exogenous to an en d o g en o u s n om in al m oney stock. A ccord ing to th is in terp retation , th e m on etary au th o r ity su p p lies th e n o m in al m o n ey b a la n ces d esired by th e p u b lic w ith a lag. In th is co n tex t, th e n om inal ad ju stm en t m odel is view ed as an eq u atio n re p rese n t ing th e m arket equilibriu m , w h ere X is th e ad ju stm en t p aram eter in th e so -called F ed eral Reserve reactio n fu n ctio n ra th er th an th e sp eed o f ad ju stm en t of m o n ey d em an d . Given th is in terp retatio n , m arket equilibrium requ ires (6) M'VP = m 11, MARCH 1985 w here Mli d en o te s th e aggregate level o f n om in al m oney b a la n ces d esired by th e p u b lic given th e p rice level, P. The Short-Run Specifications T h e above eq u atio n s ca n b e u sed to o b tain th e th ree sh ort-ru n m oney d em an d sp ecifica tio n s. E qu ation s 1 an d 2 can be co m b in ed d irectly to obtain (7) m , = XflZ.I + I t - X l l m , ,1, the real a d ju stm en t sp ecifica tio n o f C how (1966). Likewise, w e ca n co m b in e eq u a tio n s 3 w ith 4, an d 5 w ith 6 , to o b tain w hat Laidler (1983) has term ed quasired u ced -form eq u atio n s: (8 ) m, = XflZ.) + (1 —XHM,_, —P,) and (91 m, = XflZ.I + (1 —X)(M, —P, ,). B ecau se, o sten sibly, all o f th e se eq u atio n s have real m on ey on the left-h and side, it a p p ears th at th ese m od els can b e co m p ared u sin g sta tistica l tech n iq u es. T h is is in co rrect. Note th at th e eq u atio n s 8 an d 9 co u ld ju st as w ell b e specified and estim ated as 18'I M, = XflZ, I + ( 1 - X lM , , + XP, and (9') P, = -X f(Z ,l + (1 -X IP , , + XM,. 7By the same reasoning, one could argue that the nominal adjust ment specification implies that individuals never fully adjust to ex pected inflation — see Carr and Darby. Both of these characteriza tions may be off the mark, however. A more reasonable model might allow both price level and nominal money shocks to affect the demand for money in the short run, but require them to average out to zero in the long run. This has been suggested recently by Gordon. For example, let (M? - P*) = f(Zt) and combine this equation with equations 4 and 5. The result is an equation that can be estimated given a further normalization rule: the residual sum of squares can be minimized in the direction of M, or P,. Unfortunately, the results are extremely sensitive to the normalization rule. In general, if one normalizes in the direction of M„ the results are similar to (and often not statistically distinguishable from) those of the nominal specifica tion. If one normalizes in the direction of P„ the results are similar to the price adjustment specification. These results are available upon request. 8lt is not clear exactly how Gordon means this. Certainly, individuals are free only to adjust their nominal money holdings since price must be taken as given; however, Gordon cites the energy price shock as his only example. He argues that the supply shock reduces real income and, hence, the demand for real money (presumably pro portionally) so that no portfolio disequilibrium occurs. 9Using the standard quadratic adjustment cost approach, it can be shown that the nominal specification results if adjustment costs are associated only with nominal money and if prices are given. See Hwang (1984). 16 Comparing equations 7, 8' and 9' reveals that they all have different dependent variables. Furthermore, no trivial transformation exists that will make these equa tions comparable; that is, regression equations cannot be manipulated algebraically to change the left-handside variable to anything one pleases. Therefore, noth ing can be said about which specification is preferred based on comparisons of these specifications, despite claims to the contraiy. (See the appendix for a more detailed discussion.) Alternatively, on e ca n n o te th at eq u a tio n s 8 ' an d 9' make different a ssu m p tio n s about w h ich variable is exog en ou s (i.e., p rices an d n o m in al m oney, re s p e c tively). Sin ce, at best, only o n e o f th e se a ssu m p tio n s is co rrect, co n siste n t estim a tes o f th e erro rs ca n b e o b tained from only o n e o f th e se eq u atio n s. H ence, any co m p ariso n b ased on the resid u als o f th e se two s p e c i fication s is in ap p ro p riate. F u rth erm o re, th eo ry alon e ca n n o t serve as a guide b eca u se , at th e m icro e c o n o m ic level, th e a ssu m p tio n o f ex o g en o u s p rices seem s m ost relevant, w hile in th e aggregate th e exoge neity o f n om in al m on ey is m o st p lausible. FEDERAL RESERVE BANK OF ST. LOUIS EMPIRICAL RESULTS B ecau se th e se alteratives are n ot statistically co m parable, ea ch sh o u ld be evaluated for its co n siste n cy w ith the theory an d its stability .1" E stim ates o f th e real, n om inal and p rice a d ju stm en t sp e cificatio n s are p re sen ted in th is sectio n . T h e estim ates rep o rted h ere cov er th e period II/1952-II/1984, w h ich h as b ee n d i vided into th ree su b p erio d s: II/1951-IV /1961, 1/1962IV/1973 and I/1974—II/1984. T h is division is som ew hat arbitrary; n evertheless, it h as several a sp e cts w h ich m ake it d esirable. First, th e tw o earlier su bp erio d s co rresp o n d clo sely to p eriod s for w h ich Goldfeld (1973) found th e b asic C how eq u atio n to be stable. H ence, it will be in terestin g to co m p are th e estim ates o f th e nom inal and p rice a d ju stm en t sp ecificatio n s over th e se p erio d s. Second , IV/1973 m arks an observed break in th e n om inal and real a d ju stm en t sp e cifica tio n s." Third, all th ree p eriod s differ rath er signifi can tly w ith re sp ect to th e grow th and variability o f b o th m o n ey and p ric e s .12 Finally, d uring th e first two periods, th e Fed eral Reserve w as relying alm ost ex clu sively on an in terest rate target, w hile, in th e third period, m o re co n sid eratio n w as given to m o n etary aggregate targets. H ence, w e m ight ex p e ct to see som e d eterio ration in th e p erfo rm an ce o f th e nom in al s p e c ification over th e third period. T h e real (R), nom inal (N) and p rice (P) ad ju stm en t eq u atio n s are estim ated w ith ord inary least squ ares (OLS) to facilitate co m p ariso n s a cro ss tim e period s. D urbin's h -statistic is rep o rted to illu strate how the erro r stru ctu re h as varied am on g sp ecificatio n s and th rou gh tim e .13 Fu rth erm o re, all th e eq u atio n s w ere estim ated w ith real m o n ey b a la n ces on th e left-hand '““Consistency with the theory" means that the coefficients should be statistically significant, correctly signed, and the adjustment coeffi cient should obey its restriction. Thus, these equations are inter preted (as they have been in the literature) as money demand equations. It should be noted, however, that since equations 8' and 9' are really quasi-reduced forms, neither is a particularly likely equation for explaining the nominal money stock or price level, respectively. I am indebted to Tom Fomby for this observation. He noticed that equation 9' did not capture the monetarist notion of a long lag from money to prices. "Hafer and Hein (1982) mark the break at IV/1973, while Lin and Oh (1984) record it at 11/1974. ,2The variances (x 100) of M and P, respectively, are (0.3449, 0.4599), (3.2210, 1.8107) and (4.6474, 4.7701) for the three peri ods. The simple correlations between M, and P, over these periods are 0.9601, 0.9947 and 0.9911. ,3The equations also were estimated adjusting for first-order serial correlation using a maximum likelihood, grid-search procedure to estimate the coefficient of autocorrelation directly. In all instances, the qualitative conclusions were unaffected by the serial correlation correction. MARCH 1985 side, so th at th e signs o f th e co efficien ts are th e sam e for all sp e cifica tio n s .14 Also, sin ce th e n om in al and p rice sp e c ific a tio n s re p re se n t o v er-id en tified , red uced -form eq u atio n s, th e rep o rted F -sta tistic is for a test o f th e over-identifying restrictio n s; th e resu lts re p orted are for eq u atio n s w ith th e restrictio n im p o sed .” Finally, th e eq u atio n s w ere estim ated using real in co m e (y), th e co m m ercia l p a p er rate (CPR) and th e passb ook savings rate (PBR) as in d ep en d en t varia bles. T h is sp ecifica tio n o f lon g-run m oney d em and rep resen ts a fairly stan d ard version, follow ing Gold feld (1973). T h e eq u atio n s are estim ated w ith and w ithout th e PBR b ec a u se n u m ero u s stu d ies have found that sim ilar variables have n ot b een statistically significant over la ter p eriod s, for exam ple, Hafer and H ein (1980), M ilb o u rn e an d Ju d d a n d S ca d d in g (1982al. The Three Adjustment Equations E stim ates o f th e th ree a d ju stm en t sp ecifica tio n s for th e th ree p eriod s ap p e a r in table 1. N eith er th e real n o r th e n o m in al sp ecifica tio n s perform s w ell in the early perio d u n less th e PBR is in clu d ed . Real in co m e is insignificant in b o th eq u atio n s an d the over-identifying restrictio n is re je cte d at a very low sign ifican ce level in th e n o m in al sp ecifica tio n w h en th e savings rate is ex c lu d e d .16 Furth erm o re, b o th th e real and n om in al sp ecifica tio n s p ro d u ce sim ilar estim a tes of th e co efficien ts over th is period . T h e only striking dif feren ce is th e ap p aren t first-o rd er serial co rrelatio n in the n om in al sp ecificatio n , n ot p resen t in th e real equation. B oth th e real an d n om in al sp ecifica tio n s perform well in th e last two p erio d s in th at all th e param eters (save th e co n sta n ts) are sign ifican t and have th e a n tici pated sign if th e PBR is exclu d ed . In clu d in g th e PBR for th e I/1962-IV /1973 period, how ever, ten d s to in crea se th e estim a ted co efficien t on real in co m e m ark edly, w hile it re d u ce s it in th e 1/1974—II/1984 period. Indeed, real in co m e is in sign ifican t in th e real sp ecifi catio n in th e last period if th e PBR is in clu d ed . In co n tra st w ith th e real an d n om in al sp ecificatio n s, real in co m e is n ot significant in th e p rice a d ju stm en t sp ecificatio n in th e first period . Fu rth erm o re, it is sig- ,4The adjusted R2s are calculated for their respective dependent variable, however. '5The over-identifying restriction for the nominal specification 8' is that the coefficients on M ,, and P, sum to one. The over-identifying restriction for this price specification is that the coefficients on P,_, and M, sum to one. l6The t-tests are one-tailed if the coefficient has an anticipated sign, two-tailed otherwise. 17 MARCH 1985 FEDERAL RESERVE BANK OF ST. LOUIS Table 1 Estimates of the Three Basic Adjustment Equations Equation Constant y CPR PBR Mw - P,., M,_, - P, M ,-P _ i h R2 SEx 102 F 11/1951—IV/1961 R N P .854* (3.72) .001 (0.06) -.0 1 5 * (4.62) .674* (3.19) .129* (3.20) -.0 1 7 * (5.83) .610* (2.79) .001 (0.05) -.0 1 2 * (4.06) .445* (2.33) .132* (3.78) -.0 1 5 * (5.66) .444* (2.44) .000 (0.04) -.0 0 4 * (1.77) .424* (2.31) .034 (0.93) -.0 0 5 * (2.00) -.034 * (3.31) .842* (22.02) 1.52 .9418 .4828 — .724* (14.66) 0.96 .9537 .4310 — .890* (24.12) 2.99* .9943 .4434 14.29* .764* (17.04) 2.55’ .9959 .3783 0.50 -.035 * (3.95) -.0 0 9 (0.95) .917* (29.87) -0 .4 3 .9971 .3621 0.01 .880* (17.87) -0 .3 8 .9971 .3626 2.04 1/1962-1V/1973 R N P .407 (1.21) .094* (2.54) -.019 * (4.50) .730* (2.20) .188* (3.98) -.018 * (4.79) .051 (0.15) .081* (2.41) -.0 1 4 * (3.77) .379 (1.09) .159* (3.42) -.0 1 4 * (4.08) .385* (2.84) .020 (1.10) -.007 * (3.25) .491* (3.50) .060* (2.29) -.007 * (3.70) -.0 7 3 * (2.89) .812* (7.58) 4.26* .9879 .5242 — .651* (5.72) 4.16* .9896 .4853 — .894* (8.74) 3.90* .9993 .4813 5.53* .749* (6.45) 4.45* .9993 .4591 1.20 .903* (19.11) 0.09 .9996 .2610 4.07 .841* (15.45) 0.00 .9997 .2517 0.84 .894* (16.11) -0 .1 4 .8795 1.0080 — .913* (15.92) -0 .5 0 .8809 1.0020 — .941* (19.93) -0 .7 4 .9985 .8332 0.04 .939* (19.30) -0 .7 5 .9985 .8437 0.37 -.0 5 7 * (2.32) -.028 * (2.08) 1/1974-11/1984 R N P -.0 6 9 (0.20) .096* (4.13) -.0 2 3 * (4.20) -.0 6 0 (0.18) .057 (1.44) -.024 * (4.36) -.0 0 5 (0.02) .051* (2.61) -.0 1 1 * (2.28) - .0 1 3 (0.05) .057* (1.71) -.011 * (2.20) -.1 4 6 (0.87) .050* (4.29) -.014 * (4.77) -.1 3 3 (0.91) -.001 (0.08) -.015 * (5.84) .107 (1.21) -.0 1 8 (0.24) .139* (3.55) Absolute value of the t-statistic in parentheses. 'Significant at 5 percent level. 18 .962* (33.83) 3.15* .9995 .5055 19.61* .988* (38.11) 1.73 .9996 .4424 5.50* MARCH 1985 FEDERAL RESERVE BANK OF ST. LOUIS F urth erm ore, m u ch o f th e a p p aren t instability in th ese sp ecificatio n s is a sso cia te d w ith th e scale varia ble, th e co n sta n t term , th e PBR an d th e stan d ard error itself, ra th er th an w ith th e CPR o r th e ad ju stm en t coefficient. Table 2 Likelihood Ratio Test Results Tests of Equality of Parameters Periods 1 and 2 Specification PBR No PBR R N P 9.61 5.55 6.70 20.90* 13.63* 2.73 Periods 2 and 3 PBR 22.82* 10.30 42.28* No PBR 22.48* 9.43 12.25* Tests of Equality of Variances Periods 1 and 2 Periods 2 and 3 Specification PBR No PBR PBR R N P 0.10 1.84 5.77* 0.01 0.49 4.67* 26.52* 20.07* 13.48* No PBR 23.77* 17.67* 11.97* ‘Significant at the 5 percent level. Critical values: x20S(1) = 3.84, X205 (4) = 9.49, X2m (5) = 11.07. n ificant in th e se c o n d period, but only if the passbook rate is in clu d ed , and, in th e third period, only if the PBR is exclu d ed . In this in stan ce, th e PBR en ters w ith th e w rong sign. Finally, th e over-identifying restrictio n ca n n o t be re je cte d at th e 5 p e rce n t level d uring the first two p eriod s in th e p rice sp ecificatio n , bu t is re jecte d for th e I/1974-II/1984 period. It is in terestin g to n o te that, although real in co m e is n ot significant in th e p rice eq u atio n in th e first period (or in th e seco n d if th e PBR is exclu ded ), th e stand ard error from th is sp ecificatio n is low er th an th at o f ei th er th e real or n om inal sp ecificatio n s. If on e thought that all th ese eq u atio n s had th e sam e d ep en d en t vari able, on e w ould co n clu d e in co rrectly that the price equation is th e p referred sp e cifica tio n .17 M oreover, the results are in c o n siste n t w ith L aid ler’s (1983) c o n je c ture th at th e se eq u atio n s are so sim ilar that, if eith e r the real o r n om inal sp ecificatio n s perform s well, then so will the p rice sp ecification.'" 17Hence, it is not surprising that Coats and Judd and Scadding (1982a) concluded that these specifications are preferred. 16ln fairness to Laidler, he goes on to argue that none of these specifi cations is likely to be stable over time, a conjecture that our empirical results support. Formal Tests o f Stability In o rd er to test the stability o f th ese sp ecificatio n s through tim e, likelih ood ratio tests w ere perform ed on general sp ecifica tio n s th at allow ed for d ifferences in th e v arian ces o f th e eq u atio n an d th e co efficien t of au to co rrelatio n as w ell as th e stru ctu ral p a ra m e te rs .19 T h e resu lts of te sts o f th e equality o f th e co efficien ts an d v arian ces are p resen ted in table 2. T h e results suggest that G oldfeld’s (1973) co n c lu sio n abou t the stability o f th e real sp ecifica tio n over th e first two p eri od s is critically d ep en d en t u p o n th e sp ecifica tio n of th e long-run d em an d for m oney. If th e PBR is in clud ed , the n ull h yp o th esis th at th e stru ctu ral p aram eters are stable ca n n o t be re je cted . If it is exclu d ed , the h yp oth esis is re je cte d . Fu rth erm o re, th e h yp o th esis of stru ctu ral stability b etw een th e se co n d an d third p eri od s is re je cte d for th e real sp ecifica tio n regard less o f w h eth er th e PBR is in clu d ed . T h e p rice a d ju stm en t sp ecifica tio n d oes n ot fare m u ch better. W hile th e null h y p oth esis o f th e equality o f th e stru ctu ral p aram eters ca n n o t be re je cte d for the first two period s, th e in sig n ifican ce o f real in co m e in eith e r period m akes th e resu lt o f little in terest. M ore over, th e h y p oth esis is re je cte d decisively in a co m p a r ison o f th e last tw o period s. T h e results for th e n om in al sp ecifica tio n are m ore encouraging. T h e null h yp o th esis is re je cte d during the first period only if th e PBB is exclu d ed . M ore im portantly, th e h yp o th esis is n o t re je c te d at the 5 p e r mit is well known that the standard F-test for structural stability is sensitive to heteroscedasticity. See Toyoda (1974) and Schmidt and Sickles (1977). Thus, likelihood ratio tests were constructed to allow for heteroscedasticity. This procedure is complicated by the presence of statistically significant serial correlation across some of the partitions. This was handled by obtaining maximum likelihood estimates of the coefficient of autocorrelation over each partition. The tests were conducted with the model transformed appropriately to adjust serial correlation. If there was no statistically significant autocorrelation in a subperiod, the untransformed data were used. If there was prior evidence of serial correlation, the Prais-Winsten transformation was used. If there was no evidence of prior serial correlation, the initial observation was included unweighted (see Fomby, Hill and Johnson (1984), p. 213, and Thornton (1984)]. Maximum likelihood estimates of the restricted model were obtained using an iterative procedure. The resulting likelihood ratio statistics are asymptotically distributed x2(J). where J is the number of restrictions. 19 FEDERAL RESERVE BANK OF ST. LOUIS MARCH 1985 Table 3 Estimates for the 1/1974-11/1984 Period With Dummy Variables for Credit Controls Specification Variable R N P Constant -.1 4 1 (0.52) -.1 5 5 (0.54) -.1 0 2 (0.46) -.0 9 9 (0.45) -.0 8 1 (0.59) -.1 0 0 (0.61) y .047 (1.50) .095* (5.02) .050’ (193) .048* (3.17) -.0 0 3 (0.18) .052* (4.58) CPR -.0 2 3 * (5.16) - .022* (4.77) - .009* (2.41) - .009* (2.47) -.0 1 5 * (6.35) -.0 1 3 * (4.92) PBR .133 (1.89) Mm - P mm m, - - m .933* (20.18) -.0 0 6 (0.11) .148 (4.10) .910* (19.71) .961* (24.91) p, .961* (25.64) pm .978* (40.71) .952* (34.25) Dummy 11/1980 - .038* (4.66) - .036* (4.35) - .028* (4.27) -.0 2 8 * (4.35) -.0 1 0 * (2.50) -.0 1 0 (1.91) Dummy 111/1980 .011 (1.37) .012 (1.40) .019* (2.79) .019* (2.83) -.0 0 7 (1.78) -.0 0 6 (1.23) SE x 102 .7935 .8214 .6518 .6427 .4057 R2 .9253 .9200 .9991 .9991 .9997 h .267 .811 .400 .414 .528 Coefficient equality 27.44* 24.59* Variance equality 12.44* 11.73* 10.19 7.04* .4868 .9995 2.713* 9.30 48.45* 13.87* 5.56* 8.61* 10.62* Absolute value of the t-statistics in parentheses. 'Significant at 5 percent level. ce n t level during th e latter period, regard less o f the sp ecificatio n . T h e test statistic is bord erline, however, esp ecially w hen th e PBR is exclu d ed . Fu rth erm ore, th ere is a significant in crease in th e variance o f th e sp ecificatio n as well as a m arked ch an g e in th e serial co rrelatio n o f th e erro r stru ctu re. (T hese resu lts are co n siste n t w ith re cen t findings o f Lin and Oh.I T h u s, it ap p ears th at th is sp ecificatio n h as ch an g ed in som e fun d am ental way during th e last period. variance o f b o th th e real an d n o m in al sp ecifica tio n s may be due to cred it co n tro ls. Given th e im p o rta n ce of h etero sced a sticity in te sts o f p a ra m e te r stability, it is im portan t that this p ossibility b e a cc o u n te d for. T h u s, cred it co n tro l dum m y variables for II/1980 an d III/1980 w ere in clu d ed in all sp ecifica tio n s. (They w ere in clu d ed in th e p rice sp ecifica tio n ou t o f cu riosity, sin ce a priori it is difficult to d eterm in e th e ir effect on the p rice level.) The Impact o f Credit Controls OLS estim ates o f th e se eq u atio n s fo r th e th ird p e riod ap p ear in table 3. T h e likelih ood ratio sta tistics for tests o f th e equality o f th e p aram eters an d v arian ces over th e last tw o p eriod s also ap pear. In clu d in g th e cred it co n tro l dum m y variables su b stan tially low ered the estim ated stan d ard errors for th e real an d n om in al sp ecificatio n s, as an ticip a ted ; how ever, th e red u ctio n for th e p rice sp ecifica tio n (not surprisingly) is n o t as large. B oth cred it co n tro l dum m y variables are signifi can t in th e n o m in al sp ecificatio n , roughly eq u al in All th ree sp ecificatio n s in d icate a significant in crea se in th e v ariance o f th e eq u atio n during th e latter period. Hein (1982I h as p resen ted so m e evid ence that this ch an g e m ay be d ue in part to th e cred it co n tro ls of 1980; m ore recently, G ordon and Hafer and T h o rn to n (1985) have show n that th e cred it co n tro ls had a sta tis tically significant im p act on co n v en tion al m o n ey d e m an d eq u ations. H ence, th is m arked in cre a se in th e 20 MARCH 1985 FEDERAL RESERVE BANK OF ST. LOUIS Table 4 Estimates of the Nominal Specification With M1 Net of OCD 1/1974-11/1984 1/1974-1V/1980 Constant -.1 8 8 (0.48) -.3 2 2 (0.59) y 0.27 (0.58) 0.26 (1.18) CPR - .020* (3.12) - .007* (1.80) PBR -.0 0 2 (0.01) -.0 4 2 (0.34) 1.010 (25.70) 1.018 (13.27) Dummy 11/1980 -.0 2 1 (1.65) -.0 2 9 (4.94) Dummy 111/1980 0.20 (1.65) .014 (2.23) M ,,- P , 1.1691 .5194 R2 .9963 .9983 h .159 SE x 102 -.3 3 3 Absolute value of the t-statistics in parentheses. 'Significant at 5 percent level. m agnitude and o p p o site in sign. O nly th e first dum m y variable is sign ifican t in eith e r th e real o r p rice sp ecifi catio n s; its co efficien t in th e real m o n ey sp ecificatio n is approxim ately equal to th e su m o f th e co efficien ts o f th e n o m in al m o n ey an d p rice a d ju stm en t sp e c i fications. D espite th e obvious im p o rtan ce o f th e cred it c o n trols to th ese sp ecificatio n s, esp ecially th e nom inalone, th e co n clu sio n s o f th e stability tests are not differ en t from th o se rep o rted in table 2. C on sequen tly, the cred it co n tro ls had no effect on th e o u tco m e o f tests for stru ctu ral stability. Additional Evidence on the PBR T h e p erfo rm an ce o f th ese eq u atio n s is greatly af fected by th e p re se n c e o r a b se n c e o f th e PBR. In p ar ticular, it bears greatly on th e te sts o f th e stability of the stru ctu ral co efficien ts o f th e n om in al an d real sp ecificatio n s. T h e h yp o th esis o f stability is re je cted over th e first two p erio d s and is bord erlin e over the last two p erio d s if th is variable is exclu ded . F u rth er m ore, th e sw itch o f th e PBR itself from statistical sig n ifican ce to in sign ifican ce m ight be co n sid ered evi d en ce o f instability. T h e sensitivity o f th e se sp e cifica tion s to th e PBR co u ld have a so u n d e c o n o m ic basis or b e a m ere sta tistica l artifact. If th e la tte r is co rrect, it w ould ap p ear th at th e se sp ecifica tio n s have b ee n c o n siderably le ss stab le tem p o rally th a n is gen erally su p posed . C on sequen tly, th e role o f th is variable deserves ad d itional a tten tio n . Over m o st o fth e estim atio n period co n sid ere d here, M l w as co m p o se d prim arily o f n o n -in terest-b earin g d em an d d ep o sits an d cu rren cy . C on sequen tly, on e cou ld argue th at th e PBR co n stitu ted an im portan t op p ortu n ity co st variable — esp ecially over th e first two p eriod s — an d th at th e eq u atio n s are seriously m issp ecified if th is variable is exclu d ed . In th e last period, how ever, th e PBR m ight b e co n sid ere d a proxy for th e ow n rate, as in terest-b earin g tra n sa ctio n a c co u n ts (paying ex p licit rates clo se to th e PBR) m ad e up a large part o f M l .20 In o rd er to test th is exp lan ation , M l less o th e r ch eck ab le d ep o sits (OCD) w as u sed in p la ce o f M l in th e n o m in al sp e cifica tio n over th e last period . T his m easure co rresp o n d s clo sely to th e old cu rren cv p lu s-d em a n d -d ep o sits d efin ition o f m on ey. If th e above co n je c tu re is co rrect, th is sp ecifica tio n sh ould perform w ell in th e se n se th at b o th real in co m e and th e PBR sh o u ld e n te r significantly. If th e p erfo rm an ce is poor, eith e r th ere h as b een an u nd erlying shift in m on ey d em an d in th e m o st re ce n t period o r m on ey d em an d has never b een stable. T h is a p p ro a ch is lim ited by th e fact that th e p ro p o r tion o f d em an d d ep o sits h eld by individuals d eclin ed after th e n ation w id e in tro d u ctio n o f NOW a cc o u n ts in 1981. T his co u ld b ias th e resu lts for estim a tes over the en tire I/1974-II/1984 period . T h u s, th e ad ju sted M l m easu re w as estim a ted for th e en tire th ird p eriod and for th e su b p erio d 1/1974—IV/1980.21 T h e results, re p orted in table 4, sh o w th at n e ith e r th e PBR n o r real in co m e e n te r sign ifican tly in th is eq u atio n for eith e r tim e period. F u rth erm o re, th e a d ju stm en t coefficien t is negative, in d icatin g an u n sta b le d yn am ic sp e cifica tion. T h e resu lts are n ot co n siste n t w ith the c o n je c ture th at th e PBR re p rese n ts a critica l variable in the long-run d em an d for m on ey. T h u s, th e co n clu sio n th at n o n e o f th e sh o rt-ru n m o n ey d em an d sp e cifica tion s have b ee n stab le is m o re attractive. “ The PBR might not be a good proxy for the own rate on NOW accounts over this period because it does not account for service charges associated with these deposits. 21The real and price specifications were estimated but not reported. Also, the equations using adjusted M1 were estimated over the first two periods but are not reported because they differ little from those reported in table 1. 21 FEDERAL RESERVE BANK OF ST. LOUIS SUMMARY AND CONCLUSIONS T h is article h as d ealt w ith alternative sp ecificatio n s o f th e sh o rt-ru n d em an d for m oney. It h as p o in ted out that, although th ree b a sic form s o f th e d ynam ic ad ju stm e n t o f m o n ey d em an d have b ee n co m p ared in th e literatu re, they are n o t strictly am enab le to statisti cal testing. T h e sp ecificatio n s w ere estim ated for th ree su b p eriod s over th e perio d II/1951-II/1984. It w as fou nd th at (1 ) all th ree sp ecificatio n s are very sensitive to w h eth er th e passb ook savings rate is inclu d ed , (2 ) n o n e p ro d u ce resu lts co n siste n t w ith ec o n o m ic th e ory for all th ree p erio d s an d (3) n o n e exh ibit tem poral stability. W hile, strictly speaking, th e h yp o th esis o f tem p oral stability co u ld not b e re je cte d at th e 5 p e r ce n t level for G oldfeld’s n o m in al m o n ey a d ju stm en t sp ecificatio n for th e last tw o period s, it cou ld be re je c te d at a slightly h ig h er sig n ifican ce level. F u rth er m ore, th e variance o f th is sp ecificatio n an d th e serial co rrelatio n o f th e erro r stru ctu re ch an g ed signifi can tly in th e last period. MARCH 1985 Chant, John F. “Dynamic Adjustments in Simple Models of the Transactions Demand for Money,” Journal of Monetary Economics (July 1976), pp. 351-66. Chow, Gregory C. “On the Long-Run and Short-Run Demand For Money,” Journal of Political Economy (April 1966), pp. 111-31. -------------- - "A Comparison of Alternative Estimators for Simultane ous Equations,” Econometrica (October 1964), pp. 532-53. Coats, Warren L., Jr. “Modeling the Short-Run Demand for Money With Exogenous Supply," Economic Inquiry (April 1982), pp. 22239. Feige, Edgar L. “Expectations and Adjustments in the Monetary Sector,” American Economic Review (May 1967), pp. 462-73. Fomby, Thomas B., R. Carter Hill and Stanley R. Johnson. vanced Econometric Methods, (Springer-Verlag, 1984). Ad Goldfeld, Stephen M. “The Demand for Money Revisited,” Brook ings Papers on Economic Activity (1973:3), pp. 577-638. ________ _ “The Case of the Missing Money,” Brookings Papers on Economic Activity (1976:3), pp. 683-730. Gordon, Robert J. "The Short-Run Demand for Money: A Reconsid eration,” Journal of Money, Credit and Banking, (November 1984), pp. 403-34. Hafer, R. W., and Scott E. Hein. “The Dynamics and Estimation of Short-Run Money Demand,” this Review (March 1980), pp. 26-35. M oreover, th e stability test re su lts for b o th th e n o m inal and real a d ju stm en t sp ecificatio n s over th e first two p eriod s d ep en d critically on in clu d in g th e p a ss book savings rate in th e sp ecificatio n o f long-run m o n ey d em an d . Su b seq u en t investigation p ro d u ced resu lts th at raise q u estio n s abo u t th e role th e p a ss book rate h as played in m o n ey d em an d . If th e perfor m a n ce o f th e p assbo o k rate in th e first tw o su bp eriod s is m erely a statistical quirk, then, co n trary to co m m o n belief, n e ith e r o f th e se sp ecificatio n s is stable over th ese period s. ________ _ "The Shift in Money Demand: What Really Hap pened?" this Review (February 1982), pp. 11-16. T h e instability o f th e se p articu lar sp ecificatio n s is n o t too su rprising w h en it is reco g n ized th a t they rep rese n t red u ced form s o f th e d yn am ic ad ju stm en t o f m oney and p rices, ra th er th a n stru ctu ral m on ey d em an d eq u atio n s. C onsequently, w hile th e se sp ecifi ca tio n s are stan d ard in th e literatu re, th e ir instability m ay say little abou t th e in stability o f m o n ey dem and . Thu s, o u r resu lts cast d oubt on th e u sefu ln ess o f th ese sp ecificatio n s for sh o rt-ru n m o n etary co n tro l, w ithou t in d ictin g m o n ey d em an d in g en eral o r u sefu ln ess o f m on etary co n tro l for sh o rt-ru n ec o n o m ic stabiliza tion. In any event, th e instability o f th e se eq u atio n s certain ly d oes n o t p re clu d e th e u sefu ln ess o f m o n e tary grow th targets in achieving lo n g er-ru n eco n o m ic stability. Judd, John P., and John L. Scadding. “Dynamic Adjustment in the Demand for Money: Tests of Alternative Hypotheses,” Federal Reserve Bank of San Francisco Economic Review (Fall 1982a), pp. 19-30. Hafer, R. W., and Daniel L. Thornton. “Price Expectations and the Demand for Money: A Comment," Review o f Economics and Sta tistics (forthcoming 1985). Hein, Scott E. "Short-Run Money Growth Volatility: Evidence of Misbehaving Money Demand?” this Review (June/July 1982), pp. 27-36. Hetzel, Robert L. “Estimating Money Demand Functions,” Journal of Money, Credit and Banking (May 1984), pp. 185-93. Hwang, Hae-shin. “Test of the Adjustment Process and Linear Homogeneity in a Stock Adjustment Model of Money Demand,” processed (1984). ________ _ “The Search For A Stable Money Demand Function: A Survey of the Post-1973 Literature,” Journal of Economic Literature (September 1982b), pp. 993-1023. Laidler, David E. W. Press, 1982.) Monetarist Perspectives. (Harvard University ________ _ "The 1981-82 Velocity Decline: A Structural Shift in Income or Money Demand?” Monetary Targeting and Velocity, Proceedings of a Conference held at the Federal Reserve Bank of San Francisco (December 1983), pp. 100-03. ________ _ “The Demand for Money in the United States — Yet Again," in Karl Brunner and Allan M. Meltzer, eds., On the State of Macro-economics, Carnegie-Rochester Conference Series on Public Policy (North-Holland, Spring 1980), pp. 219-71. REFERENCES Lin, Kuan-Pin and John S. Oh. “Stability of the U.S. Short-Run Money Demand Function, 1959-81," Journal of Finance (Decem ber 1984), pp. 1383-96. Carr, Jack, and Michael R. Darby. “The Role of Money Supply Shocks in the Short-Run Demand For Money," Journal of Monetary Economics (September 1981), pp. 183-99. Milbourne, Ross D. “Price Expectations and the Demand for Money: Resolution of a Paradox,” Review of Economics and Statis tics (November 1983), pp. 1287-305. 22 FEDERAL RESERVE BANK OF ST. LOUIS MARCH 1985 Motley, Brian. “Dynamic Adjustment in Money Demand,” Federal Reserve Bank of San Francisco Economic Review (Winter 1984), pp. 22-26. Thornton, Daniel L. "On the Treatment of the Initial Observation in the AR(1) Regression Model," Federal Reserve Bank of St. Louis Working Paper No. 84-003 (1984). _________ “The Demand-for-Money Function for the Household Sector — Some Preliminary Findings," Journal of Finance (Sep tember 1967), pp. 405-18. Toyoda, Toshihisa. “Use of the Chow Test Under Heteroscedastic ity," Econometrica (May 1974), pp. 601-08. Schmidt, Peter, and Robin Sickles. “Some Further Evidence on the Use of the Chow Test Under Heteroscedasticity,” Econometrica (July 1977), pp. 1293-98. Walters, A. A. “The Demand For Money — The Dynamic Properties of the Multiplier,” Journal of Political Economy (June 1967), pp. 293-98. APPENDIX T h e p u rp o se o f th is ap p en d ix is to show that eq u a tion s 7, 8 and 9 m inim ize th e resid ual sum o f squ ares in different d irectio n s and, b ec a u se o f this, th e re sid uals from th ese sp ecificatio n s are not statistically com p arable. T h is ap p en d ix draw s heavily on the work o f C how (1964). C on sid er th e stand ard regression m odel Y = X (3 + u, w here Y is a T by 1 v ecto r o f th e d ep en d en t variable, X is a T by k m atrix o f in d e p en d e n t variables, (3 is a k by 1 v ecto r o f u nknow n p aram eters an d u is a T by 1 vecto r of rand om errors. It is now co m m o n ly u nd ersto o d th at th e least sq u ares estim ate o f th e v ecto r (3 is g eom etrically th e p articu lar lin ea r co m b in atio n o f th e regresso r variables th at m inim izes th e squ ared d is ta n ce b etw een th e v ecto r Y and th e sp ace sp an n ed by th e co lu m n s o f X. It is less well know n that this e sti m ate is obtain ed by im p o sin g a p articu lar d irectio n an d sca le n o rm alizatio n rule. To see this, co n sid e r th e m ore general m od el (3,Y, + [3,Y2 — + U, w here Y„ Y,, X, and X, are T by 1 vectors w ith scalar param eters (3,, |32, jjl, and (x2. C how n o tes that estim ates o f th e p aram eters o f th is m od el cou ld be o btain ed by least squ ares by p ro jectin g the lin ear co m b in atio n of th e Y’s (that is, P,Y, + (3,YJ on th e sp ace sp an n ed by X, an d X2. In th is case, least sq u ares estim ates w ould be ob tain ed by p ro jectin g th e v ecto r (3,Y, + (3,Y, in the d irectio n o f (X,XJ. T h is w ould esta b lish th e d irectio n n orm alization . O n ce th is is a cco m p lish ed , th e scale ca n b e o b tain ed by ch o o sin g an y scale-n o rm alizatio n (for exam ple, (3„ (3,, |x, or |x, = 1). In th is case, d irectio n n orm alization an d scale-n o rm alizatio n are sep arate. C how p o in ts out, how ever, th at if th e re strictio n (3, = 1 w ere im p o sed b efo re th e m inim ization , th e v ecto r Y, alon e is p ro jecte d o n th e sp a ce sp a n n ed by (Y2 X, X J. Th at is, th e analyst is assertin g th a t th e v ecto r Y„ h as a m ean v ecto r in th e sp a ce (Y, X, X J an d an additive rand om erro r orth o g o nal to th e sp a ce sp a n n ed by (Y, X, X2). Alternatively, if th e re strictio n (3, = 1 w ere im posed , th e least sq u a res estim a tes w ould b e o b tain ed by p ro jectin g th e v ecto r Y, on th e sp a ce sp a n n e d by (Y, X, X J. T his w ould im ply th a t th e analyst view ed Y, as having a m ean v ecto r in th e sp a ce (Y, X, X J an d an additive ran d om erro r v ecto r orth o g o n al to (Y, X, X J. Clearly, th e resid u al vecto rs ob tain ed from th ese different orth ogonal p ro je c tio n s are in gen eral differ en t ran d om variables an d are, th erefo re, n o t co m p a ra ble. T h e sam e is tru e o f erro r v ecto rs from eq u atio n s 7, 8 an d 9. We ca n estab lish th is by n oting that m inim iza tion is o b tain ed after im p o sin g d ifferent restrictio n s (norm alization rules). F o r exam ple, th e im plicit co ef ficient on (M-P), is set equal to o n e in eq u ation 7. Likewise, th e co efficien ts on M, an d P„ respectively, are set equal to o n e in eq u atio n s 8 an d 9. 23 The Federal Reserve Reaction Function: Does Debt Growth Influence Monetary Policy? Richard G. Sheehan T M . HE p ro sp e ct o f fed eral governm ent d eficits to ta l ing $907 billion b etw een 1985 an d 1990 h as renew ed d oubts abou t th e Fed eral R eserve’s ability to co n d u ct in d ep en d en t m o n etary p olicy. O ften im plicitly u n d e r lying th e se d o u bts is th e fear th at in crea ses in federal debt will drive up in terest rates an d slow eco n o m ic grow th in th e a b se n c e o f ex p an sio n ary m on etary p o l icy. Given th e m agnitu d e o f p ro jecte d federal deficits, m any an alysts are co n c e rn e d th at th e Fed eral Reserve m ay feel obliged to in crea se th e m o n ey stock faster than it oth erw ise w ould to keep in terest rates from rising.' It is th e p u rp o se o f this p ap er to offer som e evidence on th e ex ten t to w h ich th e F ed eral Reserve h as altered m o n etary p o licy in re sp o n se to fed eral d eficits .2 T h e focus h ere is to d eterm in e if m onetary policy h as re a cted to fed eral d eficits in a co n siste n t m a n n er over tim e. T h e sensitivity o f m o n etary actio n s to debt grow th is co n sid ered over different tim e p eriod s and Richard G. Sheehan is an economist at the Federal Reserve Bank of St. Louis. Larry J. DiMariano provided research assistance. 'Sargent and Wallace (1981) have gone so far as to argue that the Federal Reserve has only a choice between increasing the money stock sooner or later. While Darby (1984) has disputed this conten tion, the issue apparently remains unresolved. See Miller and Sargent (1984). 2The process of a debt increase directly leading to expansionary monetary policy is often labeled "monetizing the debt." Given the ambiguities surrounding that phrase, it is not used here. See Thorn ton (1984) for a detailed explanation of alternate definitions of the phrase. 24 u n d er alternative m easu res o f m on etary a ctio n s and debt. MONETARY POLICY AND DEBT T h e textbook view o f th e relatio n sh ip betw een m o n etary policy an d fed eral d eb t ca n b e d em o n stra ted in the co n tex t o f a sim p le co m p arative sta tic m o n ey m ar ket m odel, w h ich is su m m arized in figure 1. Let us assu m e th at m o n ey d em an d (MD) is a fu n ctio n o f the in terest rate an d th e level o f in co m e an d th at th e Fed eral Reserve ca n effectively fix th e m on ey su pply (MS). W ith so m e initial level o f in co m e, m o n ey d e m and an d su pply fu n ctio n s m ay b e rep resen ted by MD„ an d MS„, respectively. Given a stru ctu ral (or exog en o u s o r active) ch a n g e in fiscal policy, say, an ex p a n sionary a ctio n in creasin g th e deficit, in co m e will rise in th e sh ort ru n .3 T h is in crea se in in co m e, in turn, will lead to an in crea se in m on ey d em an d , shifting th e m on ey d em an d curve from MD„ to MD, in figure 1 an d driving u p in terest rates. If th e F ed eral Reserve is o peratin g w ith a m o n etary aggregate target, m o n etary policy will not resp o n d to th e deficit. T h e stru ctu ral 3A change in fiscal policy, that is, a change in the behavior of fiscal policymakers, is considered structural, exogenous or active. Thus, a fiscal-policy-induced change in the deficit, as one measure of fiscal policy, also is considered exogenous. It is assumed that the fiscal policy change and resulting deficit change are not prompted by a change in the business cycle. A change in the deficit resulting from a change in, say, real GNP is considered cyclical, endogenous or passive. See Tatom (1984) for a more extensive discussion of the distinction between active and passive deficits. FEDERAL RESERVE BANK OF ST. LOUIS F ig u r e 1 Comparative Static M o n e y M arket M odel MARCH 1985 eral fund s rate as its target, th e in cre a s e in th e stru c tural d eficit an d th e resu lting in crea se in m o n ey d e m an d w ill p ro m p t it to re sp o n d d ifferently. T h e in crea se in in tere st rates as m o n ey d em an d in crea ses from MD„ to MD, w ould lead th e F ed eral Reserve to in crea se th e m on ey su pply (from MS„ to MS,) suf ficiently to drive in tere st rates in g en eral an d th e fed eral fund s rate in p a rticu la r b ack to th e ir original lev e ls :’ W ith an in terest rate target, th e exo g en o u s d eficit in crea se w ould n o t in flu en ce th e in tere st rate but w ould in crea se the m o n ey stock. If th e Fed eral Reserve h as n o t follow ed a p u re in ter est rate o r m on etary aggregate target bu t in stea d has follow ed a m ixed strategy u sin g both , a stru ctu ral defi cit w ould still shift th e m o n ey d em a n d curve o u t as before, bu t th e m o n ey su p p ly curve w ou ld shift out only partially, say, from MS„ to M S ,.8 T h u s, th e stru c tural debt in crea se w ould lead to b o th h ig h er in terest rates an d h ig h er m o n ey grow th. d eficit will n ot alter th e m on ey stock bu t will in crease the in terest rate from r„ to r,.J W ith cy clical (or en d o g en o u s or passive) fiscal p o l icy ch an g es, how ever, th e im p act o f ch an g es in th e stru ctu ral d eficit is qu ite different. A ssum e th e e c o n om y en ters a re cessio n as a result o f a n o n -p o licy sh o ck to th e system . T h e au to m atic stabilizing p ro p e r ties o f federal taxes and ex p en d itu res will lead to a cy clical in crea se in th e d eficit as in co m e d eclin es. Further, th e d eclin e in in co m e w ill re d u ce th e d e m an d for m oney, shifting th e m o n ey d em an d sc h e d ule, say, from MD, to MD„ in figure 1. Again, if the Fed eral Reserve is u sing a m o n etary aggregate as its target, th e m o n ey stock will rem ain co n sta n t. An in crea se in th e cy clical d eficit will now be acco m p an ied , however, by a red u ctio n in th e in terest rate from r, to r„. W ith a m o n eta iy aggregate target, th is m od el im plies that stru ctu ral d eficits will lead to in crea ses in th e in terest rate, w hile cy clical d eficits will be a cc o m pan ied by d ecrease s in th e in terest rate. In co n trast, if th e Fed eral Reserve is using th e fed- 4This discussion assumes loanable funds demand is not completely elastic. It further assumes the Federal Reserve is focusing on a monetary aggregate and will not change its desired value of that aggregate in the face of temporary fluctuations in income. W ith a federal fun d s target an d an in crea se in the cy clical d eficit lead in g to a d ecrea se in m o n ey d e m an d from MD, to MD„, th e Fed eral Reserve w ould d ecrea se th e m o n ey stock from MS, to MS,, to keep the in terest rate u n ch an g ed . W ith a m ixed targeting stra t egy an d an in crea se in th e cy clica l deficit, th e m o n ey su pply w ould b e ex p e cted to shift partially dow nw ard from MS, to MS,. T h u s, th e in crea sed d eficit w ould be acco m p a n ied by a lo w er in terest rate an d a low er m on ey supply. W h eth er an in crea se in th e d eficit is a cco m p a n ie d by in crea ses o r d e crea se s in th e m o n ey stock and in terest rates d ep en d s on th e so u rce o f th e d eficit and on th e m a n n er in w h ich th e Fed eral Reserve is c o n d u ctin g policy. T h e alternatives are su m m arized in table 1 . It sh o u ld b e n o ted th at a given d eficit m ay co m b in e stru ctu ral an d cy clica l elem en ts. In that case, the im p act o f th e d eficit o n th e in tere st rate is am bigu ou s if th e Fed eral Reserve targets on a m o n etary aggregate; its im p act on th e m o n ey su p p ly is am bigu ou s if the Fed targets on in terest rates. B oth im p acts w ould be am bigu ou s w ith a m ixed targeting p ro ced u re. Further, th ere is n o g u aran tee th at th e Fed eral Reserve has follow ed (or will follow) a co n siste n t p attern o f target- 5lf the Federal Reserve is operating with an interest rate target, it is also necessary to assume that the Federal Reserve believes that money changes can alter interest rates — as they do in this simple model — and that the Fed has a willingness to alter the money stock based on that belief. 6Lombra and Moran (1980) cite evidence suggesting this is typical of Federal Reserve behavior. 25 FEDERAL RESERVE BANK OF ST. LOUIS MARCH 1985 Table 1 Expected Coefficient on Debt Term in Measures of Monetary Policy Structural debt increase Cyclical debt increase Combination debt increase Monetary aggregate target M r 0 + M r 0 - M r 0 ? Interest rate target M r + 0 M r - M r 0 Mixed targeting procedure M r + + M r - M r ? ? 0 — ? ing on either. T h u s, th e d ebt co efficien t n eed not be stable over tim e. From 1958 to 1984, th e Fed eral Reserve in term ed iate policy targets ap p aren tly u nd erw en t su bstan tial revi sion. F o r exam ple, th rou gh th e 1960s, it is generally assu m ed th at th e Fed eral R eserve’s prim ary co n c e rn was co n tro llin g in terest ra te s .7 M onetary aggregates began to receive m ore atten tio n in th e early 1970s. From O cto b er 1979 to O cto b er 1982, th e re w as an em ph asis on m o n etary aggregate targeting; sin c e th en aggregate targeting h as b e c o m e m ore flexible w ith less p ro m in en ce given to M l.” T h u s, at least fou r regim es can b e id entified : (1) from 1958 to approxim ately 1970, ch a ra cte rized by in terest rate targeting, (21 from the early 1970s to O cto b er 1979, a m ixed targeting strategy, (3) from O cto b er 1979 to O cto b er 1982, a m onetaryaggregate target, an d (4) from O cto b er 1982 to the p resen t, again a m ixed targeting strategy. W hile it w ould prove fruitful to ex am in e “reactio n fu n c tio n s” estim ated sep arately over e a c h o f th e se period s, the sh ort tim e fram es o f th e la tte r tw o p eriod s p reclu d e that op tion. T hu s, th e sam p le is divided into two subperiods, th e first p rio r to 1971 ch a ra cte riz ed by in ter est rate targeting an d th e se c o n d from 1971 w ith a greater focu s on m onetary aggregates. THE “REACTION FUNCTION” APPROACH T h ere have b ee n a n u m b er o f previous stu d ies that have exam in ed th e relatio n sh ip b etw een m onetary 7See Lombra and Moran (1980) and Wallich and Keir (1979). 8See Thornton (1983) and the sources cited there. 26 policy an d fed eral d eficits. M ost o f th e se stu d ies fall u n d er th e gen eral h ead in g o f estim atin g a “rea ctio n fu n c tio n ” for th e Fed eral Reserve." T h e re a ctio n fu n c tion a p p ro a ch assu m es th at th e Fed eral R eserve’s p o l icy a ctio n s are b a sed on its goals, its m o d el o f th e eco n o m y an d th e co n stra in ts th at th e m od el im plies. Thus, th e estim ated rea ctio n fu n ctio n is b ased im p lic itly — o r exp licitly in th e ca s e o f M cM illin an d B eard (1980) — on o u tp u t an d fin an cial m arket m od els, to geth er w ith a ru le (that is, an a ssu m p tio n abou t h ow the Fed will rea ct to d istu rb a n ces to rea ch its goals) for d eterm in in g Fed eral Reserve behavior. C om bining the behavioral assu m p tio n s o f th e p o licy rule w ith th e output an d fin an cial m arket m o d els p red icts h o w th e Fed eral Reserve will re a ct to d istu rb a n ces to th e e c o n o m ic system — h en ce, a “re a ctio n fu n c tio n .” Previously estim ated re a ctio n fu n ctio n s have dif fered w ith re sp e c t to th e c h o ic e o f d ep e n d en t and in d ep en d en t variables, th e fu n ctio n al form em ployed, the tim e period u sed for estim atio n an d the c o n c lu sio n s b a se d on th a t e stim a tio n . T h e y also have reach ed different co n c lu sio n s abou t th e stability o f the estim ated re a ctio n fu n ctio n . T h u s, it is useful to briefly survey previously estim ated rea ctio n fu n ctio n s. T h ree variables co m m o n ly have b ee n em p loy ed as th e d ep e n d en t variable, th at is, as th e m e a su re o f m o n etary policy. N iskanen (1978) an d B arro (1977) am on g oth ers u se a m e a su re o f th e m o n ey stock, M l, a ssu m ing that th e m o n ey sto ck is th e b est in d ica to r o f m o n e tary policy d uring th e p eriod o f estim atio n . Froyen (1974), Levy (1981), an d B arth, Sickles, an d W iest (1982) u se th e m on etary b a se in stead , co n ten d in g th at the b ase co rresp o n d s m o re clo sely to o p en m arket o p era tion s and is a good m easu re o f ex o g en o u s m onetary policy a ctio n s. T h e th ird alternative, u sed by Abram s, Froyen, an d W aud (1980), D eR osa an d Stern (1977), and Havrilesky, Sapp, an d Sch w eitzer (1975), is th e federal fund s rate. T h ey argue th a t th is variable is a m ore ap p rop riate m easu re o f m o n etary p o licy in p erio d s in w h ich th e F ed eral Reserve is targeting on in terest rates. T h ey fu rth er co n te n d th at th e F ed eral Reserve, in fact, h as targeted in tere st rates d urin g m o st o f th e post-W orld W ar II period. Previously estim ated re a ctio n fu n ctio n estim ates also have u sed a w id e range o f in d e p en d e n t variables an d have a ssu m ed altern a te goals o f th e F ed eral Re- 9For example, see Allen and Smith (1983), Barth, Sickles, and Wiest (1982), Froyen (1974), Hamburger and Zwick (1981, 1982), Levy (1981), McMillin and Beard (1980, 1982). Two studies that do not use the reaction function approach are Dwyer (1982) and Thornton (1984). For a detailed statement of the deficit problem, see Tatom (1984). MARCH 1985 FEDERAL RESERVE BANK OF ST. LOUIS serve (e.g., p rice stability, low u nem p loy m en t, high real grow th rates and fin an cial m arket stability). M ost previous stu d ies have u sed ord inary least squ ares (OLS) estim atio n te ch n iq u e s, and in d ep en d en t varia b les generally are in clu d ed w ith n o m o re th an one lag."' T h e estim atio n resu lts have b een in co n sisten t in a n u m b er o f resp ects. For exam ple, u sing th e m onetary b ase as th e policy m easu re, Allen an d Sm ith (1983) found that th e u nem p loy m en t rate w as significant, w hile Levy (1981) fou nd it in significan t. On th e im p act o f th e debt, in clu d ed as a m easu re o f fin an cial m arket stability, Levy co n clu d e d th at d ebt grow th influ en ced m on etary policy, w hile H am burger and Zwick (1981) re a ch ed exactly th e o p p o site co n clu sio n . On th e sta bility o f th e estim ated reactio n fu n ctio n , Allen and Sm ith (1983) argued in favor o f a stable relatio n sh ip ; Abram s, Froyen, an d W aud (1980) rep o rted findings of instability. It is u n c le a r to w hat ex ten t th e se differ en c e s are due to d ifferent sam p le periods, th e ch o ice o f in d ep en d en t variables, th e sp ecificatio n o f th e m o n etary policy variable o r th e u se o f d ifferent fu nctional form s." N either o f th e se tw o m easu res is a p erfect in d ica to r o f th e p ressu re on th e F ed eral Reserve to a lter policy in re sp o n se to ch a n g es in fed eral d ebt. NFD is p o te n tially in flu en ced by m a cro e co n o m ic sh o ck s, w h ich m ay also have an im p act on (or be th e result of) m o n e tary policy. T h u s, NFD in clu d es b o th stru ctu ral and cy clical co m p o n e n ts. NFD d o es have th e advantage of in clu d in g off-budget item s, an d th e re ce n t grow th in off-budget item s m ay re p resen t su b stan tial ad d itional p ressu re on m on etary p o licy m ak ers .'4 T h e HEBD m ea su re is a d ju sted for real in co m e c h a n g e s .13T h u s, it may be co n sid ered a m easu re o f stru ctu ral policy ch an g es. HEBD, how ever, d o es n ot in clu d e off-budget item s. T h e eq u atio n s are estim ated over th e interval from 1/1958 to III/1984 (excep t w h ere noted) as w ell as over th e su b p eriod s from 1/1958 to IV/1970 an d from 1/1971 to III/1984. T h e en tire perio d is b est ch a ra cte riz ed in term s o f table 1 as a m ixed targeting p ro ced u re. T he early su b p erio d is b asically a tim e o f in tere st rate tar geting, w hile th e latter co n fo rm s m ost clo sely to a m on etary aggregate targeting p ro ced u re. T h e estim ated eq u atio n s are o f th e form p resen ted below, a sp ecifica tio n sim ilar to th at o f F royen (1974): EMPIRICAL M ETHODOLOGY T h e b asic reactio n fu n ctio n ap p ro ach is also em ployed here. Tw o alternative m onetary policy m ea su res are u sed as d ep en d en t variables: th e m oney stock (M l) and th e federal funds rate (iPF), given th at the Fed eral Reserve h as alternately fo cu se d on in terest rates and the m oney s to c k .12 T o fu rth er allow co m p ari son o f th e estim atio n resu lts w ith th e p o ten tial rela tio n sh ip s b etw een m o n etary p o licy an d d eficits as p resen ted in table 1 , w e em ploy tw o m easu res o f debt grow th in th e follow ing em p irical analysis: th e net federal debt (NFD) an d the high em p loy m ent deficit (HEBD ).'3 ’“Levy (1981) used instrumental variables and Abrams, Froyen, and Waud (1980) used 3SLS. Froyen (1974) and studies using Barro s (1977) basic specification used more than one lag. "See Barth, Sickles, and Wiest (1982) or McMillin and Beard (1981) for a more extensive review of the reaction function literature. Thornton (1984) uses a different framework focusing on the “causal” relationships between monetary policy and debt rather than using a reaction function approach. His results are consistent with the findings of the reaction function literature. There apparently exists a relationship between monetary policy and federal debt, but this finding is sensitive to the period of analysis chosen as well as the precise measure used for debt. 12The monetary base is not used as a measure of monetary policy since Thornton (1984) has shown the linkage between debt growth and the monetary base is influenced by a number of other factors. ,3Previous reaction functions have generally used either NFD or HEBD although Froyen (1974) used both in the same equation. I (II X, = a„ + J K 2 a, X,, + 2 P, Zw + 2 7 , D,_t i= 1 j= 0 k= 0 w here X = a m easu re o f m o n etary policy; Z = a v ecto r o f m easu res o f th e goals an d c o n strain ts o f th e Fed eral Reserve; D = a m easu re o f debt; and a, (3, an d 7 are th e estim ated p aram eters. T h e righ t-h an d -sid e variables in clu d e lags of the d ep en d en t variables as w ell as cu rren t an d lagged values o f th e stab ilizatio n o b jectiv es or goals u sed by th e Fed eral R eserve .'6 In clu d ed in th e sp ecifica tio n are th e gen eral p rice level (P), th e u n em p loy m en t rate (UR), an d alternately e a c h o f th e two m easu res o f fed eral debt. Follow ing th e previous rea ctio n fu n ctio n 14For example, off-budget items totaled $17.3 billion in fiscal year 1982. 15See deLeeuw and Holloway (1982). 16Froyen has noted that the estimated reaction function actually rep resents a joint test of the influence of the chosen stabilization goals and constraints together with the appropriateness of the chosen dependent variable. Lags of the dependent and independent varia bles are included (1) to allow gradual adjustment to goals so that monetary policy is not a source of instability and (2) to capture the effect on monetary policy of variables omitted from the model. 27 FEDERAL RESERVE BANK OF ST. LOUIS literature, in terest rate term s are in clu d ed in the m o n ey eq u ation , w hile m o n ey term s are in clu d ed in the in terest rate eq u ation. All variables w ere in clu d ed in log d ifference form excep t for HEBD, w h ich is in clu d ed in level form . M ax im um lag len gth s w ere arbitrarily re stricted to 12 lags on th e d ep e n d en t variables an d six lags for th e o th er righ t-h and -sid e variables. T h e c h o ic e o f approp riate lag length w as th e n d eterm in ed by Akiake’s final p re d iction erro r (FPE) c rite rio n .17W hen th e FPE sea rch for th e p referred lag sp ecificatio n in d icated th at no values o f a righ t-h an d -sid e variable im proved th e sp e cifica tion, th at variable w as d ro p p ed from th e b a sic eq u a tion . E xcep t w h en n oted, a variable w as in clu d ed in th e estim ated eq u atio n only w h en an F -test on its join t coefficien ts in d icated it w as significant at th e 10 p e r ce n t level. Tw o-stage le ast sq u ares w as u sed as th e estim atio n te ch n iq u e to avoid problem s o f sim u l tan eity .18 ESTIMATION RESULTS MARCH 1985 Table 2 Basic Results Using NFD: 1/1958-111/1984 M1 *FF c .003 (2.01) -.0 1 1 ( - .3 3 ) SM .833 (6.76) 3.316 (93) 2UR -.0 4 3 (-2 .6 0 ) -1 .2 5 6 (-2 .9 7 ) ZiFF -.0 2 1 (-2 .3 4 ) -.1 4 4 ( - .8 6 ) ZP 2.746 (.77) 2NFD R2 .39 RMSE Q(20)' T h e re actio n fu n ctio n resu lts estim ated over the 1 9 5 8 -8 4 perio d are p re sen ted in tables 2 and 3. T ables 4 and 5 in clu d e th e resu lts o f eq u atio n s estim ated from 1958 to 1970, w hile tables 6 an d 7 p re sen t results o f eq u atio n s estim ated over th e 1 9 7 1 -8 4 interval. T h e focu s o f th e follow ing d iscu ssio n is o n th e d ebt varia ble an d th e ex ten t to w h ich fed eral d eficits have in flu en ced m o n e ta iy policy. T h e d ebt co efficien ts are in terp reted in light o f th e p re d icted co efficien t signs from table 1 . Full Period Results -2 .5 5 5 (-1 .9 7 ) .0065 9.09 .54 .122 17.90 Significance levels2 M .0001 .0102 0) (1) UR .0110 (0) .0001 (2) Iff .0056 (2) .0214 (2) P .0001 (4) NFD .1175 (1) T able 2 p re se n ts th e eq u atio n s estim ated initially w ith NFD as an in d e p en d e n t variable. T h e top part o f th e table p re sen ts th e co efficien t su m s and th e tstatistics on w h eth er th at su m is significantly different from zero. At th e bo tto m o f table 2, th e significan ce 'The Q-statistic tests for autocorrelation in the presence of lagged dependent variables. It follows a chi-square distribution and is calculated for 20 degrees of freedom. The critical value at the 95 percent level is 31.41. 2Given the varying degrees of freedom, the significance levels of the joint F-statistics are presented. The number of lags are included in parentheses. 17See Batten and Thornton (1984). In one instance below, the FPE chose the maximum lag length allowed. In that case, the maximum lag length was increased but further lags were insignificant. values are p re sen ted for th e jo in t h y p oth esis th at all the co efficien ts for a p a rticu la r variable are equal to zero. T h e se sig n ifican ce levels are p re sen ted sin ce th e lag len gth s an d co rresp o n d in g d egrees o f freedom vary from o n e sp e cifica tio n to an o th e r. T h e lag len gth s are in clu d ed in p a re n th e ses. Zero in d ica tes th at only th e co n tem p o ra n e o u s variable is in clu d ed . t8Only one equation is estimated, and this period’s inflation, unem ployment rate, etc., may be influenced by this period’s monetary policy. In the first stage, each of the dependent variables was re gressed on 10 lags of itself and four lags of all other variables in the model. The maximum lag lengths were arbitrarily restricted. The second stage, reported in the text, replaces the current values of the independent variables with the first stage estimates. If HEBD were an exogenous policy tool, the use of an instrument for HEBD would be unnecessary. There is no reason, however, to assume that current fiscal policy is independent of, say, current monetary policy actions. 28 Sin ce n et federal debt, on average, h ad n o significant im pact on m o n ey d uring th e 1 9 5 8 -8 4 period, it w as om itted from th e M l eq u atio n . NFD is in clu d ed in the MARCH 1985 FEDERAL RESERVE BANK OF ST. LOUIS the h yp oth esized m ixed targeting p ro c e d u r e .13 Table 3 Basic Results Using HEBD: 1/1958—111/19831 M1 Iff c .007 (3.74) .003 (.10) SM .399 (2.50) 1.279 (.39) SUR -.0 2 9 (-1 .9 4 ) -1 .3 8 9 (-3 .1 2 ) SiFF -.0 1 2 (-1 .6 8 ) -.1 4 9 (-■85) IP .214 (.06) 2HEBD .00026 (4.08) R2 .45 .53 RMSE .0063 .125 Q(20) 14.47 23.14 .0392 (5) (1) .0554 (0) .0001 (2) .0379 .0965 (2) Significance levels M UR Iff (1) P HEBD .0058 .0001 (4) .0003 (1) NOTE: See footnotes to table 2. 'Equations including HEBD are estimated only through 111/1983 since the data series has been discontinued. federal funds rate eq u atio n sin c e th e sum o f its co e f ficien ts is significant at th e 10 p e rce n t level. A 1 p e r ce n t in crease in NFD low ers th e fed eral fund s rate by an estim ated 2.56 p e rce n t. Sin ce NFD co n ta in s both stru ctu ral an d cy clical co m p o n e n ts, b ased on table 1 , it ap p ears that th e cy clical co m p o n e n t o f NFD d o m i n ates th e stru ctu ral co m p o n e n t in th e fed eral fund s rate eq u ation . F urther, sin c e NFD significantly en ters the federal funds rate eq u ation, th e Fed eral Reserve app aren tly did not follow a pure in terest rate strategy over th e 1 9 5 8 -8 4 period . T h is result is co n siste n t w ith T h e HEBD resu lts p resen ted in table 3 apparen tly yield co n clu sio n s at od d s w ith th e se resu lts. W ith the HEBD m easure, th e d eficit has a significant positive im pact on th e m o n ey stock bu t no im p act on the federal fund s rate; co n seq u en tly , it w as o m itted from th e final estim ated federal fun d s eq u atio n . Given HEBD as a m easu re o f th e stru ctu ra l deficit, th e im p act o f HEBD on M l an d iKf is co n siste n t w ith th e Fed eral Reserve, on average, p u rsu in g an in tere st rate target ing strategy d uring th e 1 9 5 8 -8 4 period. T h e co n d itio n s p re sen ted in tab le 1, how ever, re p re sen t only su fficien t co n d itio n s for th e stru ctu ral defi cit to have no im p a ct on th e fed eral fu n d s rate. In o th e r w ords, it is n o t n ecessa ry fo r th e F ed eral Reserve to b e targeting in terest rates in o rd er to g en erate th e result that HEBD d oes n ot in flu en ce iFK. For exam ple, if HEBD is sm all relative to th e lo an ab le fu n d s m arket or if th e su pply o f lo an ab le fu n d s is in tere st-ela stic, th e n HEBD w ould have little in flu en ce on itKeven w ith, say, a m ixed targeting strategy. Further, th ere is evid en ce to suggest th at th e stru c tural d eficit re p rese n ts a relatively sm all fractio n o f th e total d em an d for lo an ab le funds. F o r exam ple, in 1982, HEBD averaged $32.6 b illion w h ile n et cred it m arket borrow ing by n o n fin an cial se c to rs w as $404.1 billion. Thu s, th e HEBD co m p o n e n t o f fed eral b orrow ing w as only 8.1 p ercen t o f fund s borrow ed. In co n trast, on average from 1975 to 1981, sim ilar figures in d icate HEBD w as only 4.6 p e rce n t o f n et fund s borrow ed. HEBD m ay have little o r no im p act on in terest rates n o t b eca u se o f th e p articu la r targeting p ro ced u re u sed by th e Fed eral Reserve, b u t ra th er b e c a u se o f th e sm all relative size o f th e stru ctu ral deficit. Given this in terp retation , th e resu lts in table 3 are also co n siste n t w ith a m ixed targeting strategy. 19The coefficients on the non-debt terms in table 2 deserve comment. Inflation does not significantly enter the M1 equation and unemploy ment enters with a negative coefficient. While the negative coefficient on the unemployment rate is signifi cant in all equations, its economic impact is minor. For example, a reduction in the unemployment rate from 7.5 percent to 7.0 percent would increase the growth rate of money by only 0.2 percent. The procyclical response of monetary policy to the unemployment rate is certainly not intuitive; it is, however, consistent with the findings of Abrams, Froyen and Waud (1980). Although the sum of the coefficients on the inflation term in the federal funds rate equation is not significant, the joint impact is significant. The short-run impacts are large in magnitude although approximately offsetting over a year. Similarly, the sum of the coeffi cients on money growth in the federal funds rate equation are not significantly different from zero. Again, it is the result of offsetting individual coefficients. 29 FEDERAL RESERVE BANK OF ST. LOUIS MARCH 1985 Table 4 Table 5 Interest Rate Target Using NFD: 1/1958-IV/1970 Interest Rate Target Using HEBD: 1/1958-I V/1970 M1 c Iff c M1 Iff .005 (3.02) —.019 (- .5 4 ) .001 (.27) -.0 0 8 (- .2 3 ) 2UR -.0 6 4 (-2 .8 6 ) -1 .8 6 0 (-5 .6 0 ) 2UR -.0 3 1 (-1 .4 2 ) -2 .0 1 7 (-5 .7 2 ) ZiFF -.0 6 5 (-5 .1 5 ) -.1 5 9 (-1 .2 9 ) ZiFF -.0 5 3 (-4 .0 3 ) -.0 4 4 (-3 5 ) IP 1.164 (6.83) 6.861 (1.58) IP .789 (3.60) 6.757 (1.42) 2NFD .281 (2.29) -.8 4 2 (-3 2 ) SHEBD .00029 (3.09) R2 .67 .67 R2 .69 .59 RMSE .0038 .110 RMSE .0037 .121 Q(20) 23.05 11.41 Significance levels Q(20) 30.54 17.69 Significance levels UR .0001 (4) .0001 (0) UR .0001 (4) .0001 (0) Iff .0001 (2) .0001 (2) Iff .0002 (2) .0031 (2) P .0001 (1) .0024 (1) P .0001 (1) .0136 (1) NFD .0275 (0) (1) .0045 NOTE: See footnotes to table 2. Subperiod Results T h e re actio n fu n ctio n resu lts estim ated over the interval from 1/1958 to IV/1970, w h ich co rresp o n d s to w hat is com m only th o u g h t to be a perio d o f in terest rate targeting, are p re sen ted in tab les 4 an d 5. T h e estim atio n p ro ced u res are id en tical to th o se em ployed for th e en tire period resu lts above. W hen NFD is u sed as th e d ebt m easu re, its co efficien ts are jointly significant in b o th th e m o n ey sto ck an d fed eral funds rate eq u atio n s. A 1 p e rce n t in crea se in NFD w ould result in a p erm an en t 0.28 p e rce n t in crea se in the m on ey stock. In co n trast, in th e federal fund s rate eq uation, w hile th e NFD co efficien ts are jo in tly signifi cant, th e ir sum is n o t significantly d ifferent from zero. An in crease in NFD th is period w ill be a cco m p an ie d 30 HEBD .0132 (1) NOTE: See footnotes to table 2. by low er in terest rates th is period , b u t th at d eclin e in th e fed eral fund s rate w ill b e offset n ex t period , w ith th e fund s rate retu rn in g to its previous level. T h u s, the NFD resu lts are co n siste n t w ith in terest rate targeting, assu m in g a o n e-q u a rter lag b efore th e F ed eral Reserve ca n effectively offset in terest rate ch an g es. T h e HEBD resu lts in tab le 5 gen erally are co n siste n t w ith th e NFD resu lts. HEBD is sign ifican t in th e m on ey eq u ation bu t in sign ifican t in th e fed eral fu n d s rate eq u ation . An in cre a se in th e stru ctu ra l d eficit lead s to an in crea se in th e m o n ey sto ck d uring th e early period bu t h as n o effect on th e fed eral fu n d s rate. T h is HEBD result is also c o n siste n t w ith in tere st rate targeting. T h e NFD and HEBD resu lts differ only in th e ir tim ing. NFD has a slightly faster im p act on th e m o n ey stock FEDERAL RESERVE BANK OF ST. LOUIS MARCH 1985 Table 6 Table 7 Mixed Targeting Procedure Using NFD: 1/ 1971- 111/1984 Mixed Targeting Procedure Using HEBD: 1/ 1971- 111/1983 M1 c .020 (4.69) 2M -.1 4 6 (-1 .0 8 ) 8.246 (1.26) HJR Sifp 2P R2 RMSE .020 (3.87) 2M -.0 0 6 (- .7 7 ) -.5 4 7 (-2 .0 6 ) SiFF -.3 5 0 8.549 (1.59) .087 (1.14) .0072 9.81 -3 .6 1 3 (-2 .2 8 ) -1 .5 6 2 (-3 .2 3 ) 2P -.0 0 5 (- .6 4 ) - .4 2 7 (-1 .5 9 ) -.2 9 8 (-1 .1 7 ) 5.611 (1-06) 2HEBD .00014 (1.64) R2 .25 .57 .111 RMSE .072 .116 11.34 Q(20) 9.06 6.38 Significance levels M .0097 (6) UR .0052 M .0089 (6) .0047 UR (1) (1) .0202 (1) .0228 (2) P .1047 (0) .0967 (3) NFD .2616 (0) (1) .0882 NOTE: See footnotes to table 2. and a tem porary effect on th e federal fund s rate. HEBD takes on e q u arter lo n g er in reach in g its full im p act on m o n ey an d h as n o effect on th e federal fund s rate.2" E qu ation s estim ated only over th e 1 9 7 1 -8 4 period, w h ich co rresp o n d s to a period o f g reater relian ce on a m onetary aggregate target, are p resen ted in tables 6 and 7. T h e NFD and HEBD eq u atio n s b o th im ply that debt grow th did n ot in flu en ce th e m oney stock over this period. Again, w h en th e federal funds rate equa- “ With respect to the non-debt terms, there are some interesting differences between the early period and the full period results, in particular for the money stock equation. The money stock continues to respond countercyclically to unemployment, but it also responds countercyclically to inflation in the early period. Also, lagged money terms are insignificant. -.2 1 0 (-1 .3 5 ) .58 Significance levels Iff •f f 9.589 (1.22) 2UR .21 Q(20) c -1 .3 7 2 (-3 .0 1 ) ( -1 .6 5 ) 2NFD M1 Iff (1) .0551 (2) P .2474 (0) .0663 (3) HEBD .1083 (0) If f .0542 NOTE: See footnotes to table 2. tion is estim ated w ith NFD, that variable is significant; w h en it is estim ated w ith HEBD, th e d eficit m easu re is in significant. As in th e d iscu ssio n o f th e co m p lete p e riod results, th e find ing th at HEBD h as n ot in flu en ced th e fed eral fund s rate m ay b e d ue m o re to th e sm all size o f th e stru ctu ral d eficit vs. total cred it d em an d th an it is to th e targeting p ro ced u res o f th e Fed eral R eserve .’1 Thu s, th e later perio d estim atio n resu lts are 21Even for the first three quarters of 1983, the last period for which HEBD is available, the structural deficit increases to only 10.4 per cent of net credit market borrowing. Of course, the insignificance of HEBD could also be a result of other causes. For example, believers in currency substitution would argue that any increase in HEBD leading to increased real interest rates would also lead to foreign capital inflows that could drive interest rates back to approximately their original levels. 31 FEDERAL RESERVE BANK OF ST. LOUIS co n siste n t w ith th e Fed eral Reserve follow ing a m o n e tary aggregate target. T h e federal fund s rate eq u atio n s estim ated over th e later p eriod w ere sim ilar to th o se for th e early an d th e full periods. In co n trast, the m o n ey sto ck eq u atio n s w ere su bstan tially different in th e la ter period . T h e m on ey stock eq u atio n s c h o se n by Akiake’s FPE an d Ftests co n siste n tly im ply th at virtually all variables e n tered, w ith th e p o ssible ex cep tio n o f th e federal funds rate and th e inflation rate, are insignificant. From th e p ersp ectiv e o f estim ating a reactio n fu n c tion that “e x p la in s” m u ch o f th e variation in the m on ey stock, th e 1 9 7 1 -8 4 resu lts leave m u ch to be d esired. T h ey are, how ever, co n siste n t w ith tw o very different th eo ries o f Fed eral Reserve behavior. First, it is p o ssible th at over th is period th e goals of th e F ed eral Reserve o r th e w eights on th o se goals w ere ch a n g ing frequently, p erh ap s d ue to shifts in m o n ey d e m and, deregu lation o r fin an cial innovations. If true, it w ould be im p o ssible to estim ate a co n siste n t relatio n ship betw een goals an d th e m o n ey stock. In th e ex trem e, th e m o n ey stock after d etren d in g w ould b e a rand om walk. Alternately, th e Fed eral Reserve, on av erage, m ay have follow ed a co n sta n t m o n ey grow th rate rule. In th is case, th e m oney stock after d etren d ing w ould also be a rand om walk. E ith er o f th ese hy p o th eses w ould be co n siste n t w ith a poorly perform ing sh o rt-ru n reactio n fu n ctio n for th e m on ey stock. SUMMARY AND CONCLUSIONS T h is p ap e r h as exam ined w h eth er federal debt grow th h as in flu en ced alternate m easu res o f m o n e tary policy. It w as d em o n strated th at a stru ctu ral defi cit w ould have very d ifferent im p licatio n s th an a cy cli cal deficit. A stru ctu ral deficit in th e static m odel p re sen ted h ere co u ld lead to an in crease in m oney grow th an d /or in terest rates. In co n trast, a cy clical d eficit co u ld be a cco m p an ie d by a d ecrea se in m on ey grow th an d /or in terest rates. W h eth er debt alters m on ey grow th or in terest rates d ep en d s on th e n atu re o f th e targeting strategy u sed by th e Fed eral Reserve. T h e resu lts o f a reactio n fu n ctio n , developed and estim ated over altern ate intervals, suggest that p rior to 1971 d ebt grow th did lead to m o n ey grow th but did not in flu en ce in terest rates. Sin ce th en , d ebt grow th h as not altered m o n ey grow th bu t m ay have b een a sso ciated w ith in terest rate ch an g es. Net federal debt grow th, w h ich co m b in es bo th stru ctu ral and cyclical debt ch an g es, is acco m p an ie d by a low er federal funds rate for th e 1 9 7 1 -8 4 period. T h is resu lt su ggests that cy clical debt ch an g es d o m in ate stru ctu ral in NFD’s 32 MARCH 1985 effect on in terest rates. In co n trast, th e h igh-em ploym ent budget deficit, a m easu re o f stru ctu ral debt ch an g es only, has h ad no im p act on th e fed eral funds rate over any tim e period . T h is resu lt m ay b e d ue to H EBD’s sm all size in co m p a riso n w ith total cred it d e m ands. T h e resu lts p re sen ted h ere are co n siste n t w ith m on etary policy bein g in d e p en d e n t o f federal deficits even thou gh m on ey m arket variables do apparen tly resp o n d to th o se d eficits. D uring th e p eriod w h en the Fed eral Reserve w as targeting in tere st rates, th e as su m ed policy m easu re, th e fed eral fund s rate, w as unaffected by federal d eficits. W hile th e m o n ey stock d oes resp o n d to d eficits in th e early tim e period, 1 9 5 8 70, th e m on ey stock w as n ot b ein g u sed as a policy target in th at interval. Conversely, in th e later period, 1971-84, th e Fed eral Reserve paid m ore atten tio n to th e m oney sto ck an d less to in terest rates. In that interval, th e prim ary p o licy variable, th e m o n ey stock, was again u naffected by fed eral d eficits w h ile th o se d eficits m ay have h ad an im p act o n in tere st rates ~ REFERENCES Abrams, Richard K., Richard Froyen and Roger N. Waud. "Mone tary Policy Reaction Functions, Consistent Expectations, and The Burns Era,” Journal of Money, Credit and Banking (February 1980), pp. 30-42. Allen, Stuart D., and Michael D. Smith. “Government Borrowing and Monetary Accommodation,” Journal of Monetary Economics (No vember 1983), pp. 605-16. Barro, Robert J. “Unanticipated Money Growth and Unemployment in the United States,” American Economic Review (March 1977), pp. 101-15. Barth, James R., Robin Sickles and Philip Wiest. “Assessing the Impact of Varying Economic Conditions on Federal Reserve Be havior,” Journal of Macroeconomics (Winter 1982), pp. 47-70. Batten, Dallas S., and Daniel L. Thornton. “How Robust Are the Policy Conclusions of the St. Louis Equation?: Some Further Evi dence,” this Review (June/July 1984), pp. 26-32. Darby, Michael R. “Some Pleasant Monetarist Arithmetic,” Federal Reserve Bank of Minneapolis Quarterly Review (Spring 1984), pp. 15-20. deLeeuw, Frank, and Thomas M. Holloway. “The High-Employment Budget: Revised Estimates and Automatic Inflation Effects,” Survey of Current Business (April 1982), pp. 21-33. DeRosa, Paul, and Gary H. Stern. “Monetary Control and the Fed eral Funds Rate,” Journal of Monetary Economics (April 1977), pp. 217-30. Dwyer, Gerald P., Jr. “Inflation and Government Deficits,” Eco nomic Inquiry (July 1982), pp. 315-29. 2'AOne hypothesis not tested here is that the Federal Reserve has shifted targets, say, from monetary aggregates to interest rates, in response to federal deficits. This change would represent another way in which debt growth could influence monetary policy. FEDERAL RESERVE BANK OF ST. LOUIS MARCH 1985 Eisner, Robert, and Paul J. Pieper. “A New View of the Federal Debt and Budget Deficits," American Economic Review (March 1984), pp. 11-29. McMillin, W. Douglas, and Thomas R. Beard. “Deficits, Money and Inflation: Comment,” Journal of Monetary Economics (September 1982), pp. 273-77. Froyen, Richard T. "A Test of the Endogeneity of Monetary Policy," Journal of Econometrics (July 1974), pp. 175-88. ________ _ “The Impact of Fiscal Policy on the Money Supply in the U.S.: Theory and Empirical Evidence,” Rivista Internationale di Scienze Economiche e Commerciali (1981), pp. 941-57. Hamburger, Michael J., and Burton Zwick. “Deficits, Money and Inflation,” Journal of Monetary Economics (January 1981), pp. 141-50. ________ _ “Deficits, Money and Inflation: Reply,” Journal of Mone tary Economics (September 1982), pp. 279-83. Havrilesky, Thomas, Robert H. Sapp and Robert L. Schweitzer. “Tests of the Federal Reserve’s Reaction to the State of the Econ omy,” Social Science Quarterly (March 1975), pp. 835-52. Hoffman, Dennis L., Stuart A. Low and Hubert H. Reineberg. “Re cent Evidence on the Relationship between Money Growth and Budget Deficits,” Journal of Macroeconomics (Spring 1983), pp. 223-31. Levy, Mickey D. “Factors Affecting Monetary Policy in an Era of Inflation,” Journal of Monetary Economics (November 1981), pp. 351-73. Lombra, Raymond E., and Herbert M. Kaufman. “The Money Sup ply Process: Identification, Stability, and Estimation,” Southern Economic Journal (April 1984), pp. 1147-59. ________ _ “The Short Run Impact of Fiscal Policy on the Money Supply,” Southern Economic Journal (July 1980), pp. 122-35. Miller, Preston J., and Thomas J. Sargent. “A Reply to Darby,” Federal Reserve Bank of Minneapolis Quarterly Review (Spring 1984), pp. 21-26. Niskanen, William A. "Deficits, Government Spending, and Infla tion: What is the Evidence?" Journal of Monetary Economics (Au gust 1978), pp. 591-602. Plosser, Charles I. “Government Financing Decisions and Asset Returns,” Journal of Monetary Economics (May 1982), pp. 325-52. Potts, Glenn T., and Dudley G. Luckett. “Policy Objectives of the Federal Reserve System,” Quarterly Journal of Economics (August 1978), pp. 525-34. Sargent, Thomas J., and Neil Wallace. “Some Unpleasant Mone tarist Arithmetic,” Federal Reserve Bank of Minneapolis Quarterly Review (Fall 1981), pp. 1-17. Tatom, John A. "A Perspective on the Federal Deficit Problem,” this Review (June/July 1984), pp. 5-17. Lombra, Raymond and Michael Moran. “Policy Advice and Policy making at the Federal Reserve," in Karl Brunner and Allan H. Meltzer, eds., Monetary Institutions and the Policy Process, Carnegie-Rochester Conference Series on Public Policy (Autumn 1980), pp. 9-68. ________ _ “The FOMC in 1982: De-Emphasizing M1,” this Re view (June/July 1983), pp. 26-35. Makin, John H. “Real Interest, Money Surprises, Anticipated Infla tion and Fiscal Deficits,” Review of Economics and Statistics (Au gust 1983), pp. 374-84. Wallich, Henry C., and Peter M. Keir. “The Role of Operating Guides in U.S. Monetary Policy: A Historical Review,” Federal Reserve Bulletin (September 1979), pp. 679-90. Thornton, Daniel L. "Monetizing the Debt,” this Review (December 1984), pp. 30-43. 33