The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
CO o GO O c= 03 February 1982 Vol. 64, No. 2 QQ CD > CD CO CD c r ~cu CD “O CD 3 The Three-Year Experience with the Community Reinvestment Act 11 The Shift in Money Demand: What Really Happened? The R eview is publ ished 10 tim es per year by the Research D epartm ent o f the Federal Reserve Bank o f St. Louis. Single-copy subscriptions are available to the public f r e e o f charge. Mail requests f o r subscriptions, back issues, or address changes to: Research D epartm ent, Federal R eserve Bank o f St. Louis, P.O. Box 442, St. Louis, M issouri 63166. A rticles herein may be reprinted provided the source is credited. Please provide the Bank’s Research D epartm en t with a cop y o f reprinted material. The Three-Year Experience with the Community Reinvestment Act NORMAN N. BOWSHER c \ ^ < R E D I T is a scarce co m m o d ity . As lenders allocate available funds on the basis of a variety o f considerations, in clu din g price (interest rates), abil ity to repay, maturity o f the loan and costs o f servic ing, it is generally difficult for an outside observer to determ ine why one loan application is refused w hile an apparently similar one is accepted. D uring the 1970s, banks and thrift institutions w ere charged with “ redlining” in allocating credit. Many charged that lenders, in essence, drew a line (presum ably red) around certain areas on a map and deliberately redu ced the supply o f mortgage and other credit to residents o f those areas. R edlining was cre d ite d w ith both un fairly discrim in ating among those seeking credit and hastening the e c o nom ic declin e o f the affected areas. Lenders, the oretically, did this because they w ere shortsighted, bigoted or insensitive to the needs o f these indi viduals and com m unities.1 In response to such charges, Congress passed the C om m u n ity R ein vestm en t A ct (C R A ), e ffe ctiv e N ovem ber 6, 1978, to encourage financial institu tions to m eet the credit needs o f their local com m u nities. T his a rticle d iscu sse s r e d lin in g and examines the CRA experien ce during the three years o f its existence. Since a study by the C ouncil o f State Planning A gencies has recom m en ded enactment o f lTake the Money and Run! Redlining in Brooklyn (New York Public Interest Group, Inc., 1976), pp. 13-16; and Redlined: A National Survey by National Peoples Action o f Mortgage Lend ing Policies in the United States, October 1976 (U.S. Senate Hearings, November 23, 1976, pp. 154-87). a law similar to the CRA, but aimed at increasing credit to small businesses in the bank’ s com m unity, this is an appropriate tim e to rev iew the C R A experience.2 M ORTGAGE MARKETS AND REDLINING E con om ics o f M ortga ge L en d in g T he purchase o f a hom e is typically the largest financial outlay that an individual makes in his life time, usually am ounting to tw o or more years o f a buyer’ s incom e. H om e purchasers generally rely on substantial credit to facilitate their purchases since they do not have sufficient savings readily available to buy the hom e outright. By m id -1981 total mortgage debt in the country am ounted to $1.5 trillion, more than 50 percent greater than the total federal debt. This mortgage credit was granted by a vast num ber o f diverse lenders. Savings and loan associations h eld 34 per cent o f the debt, com m ercial banks had 18 percent, life insurance com panies carried 9 percent, and federal and related agencies held 8 percent. The remaining 31 percent o f the mortgage debt was dis tributed am ong mutual savings banks, mortgage p ools or trusts, relatives and other individuals, 2Beth K. Vogt, “ Small Business Loan Act Urged,” American Banker, December 24, 1981. 3 FEDERAL RESERVE BANK OF ST. LOUIS mortgage com panies, state and local credit agencies, credit unions and pension funds. Because of the unique features o f each property, the lim ited kn ow ledge about borrowers outside the comm unity, and legal restrictions on som e lenders, most mortgage loans are granted by lenders located in the area of the property to be financed. N ever theless, there is, in essence, a national mortgage market, and terms on mortgages vary only slightly betw een regions. The broader market reflects the fact that some lenders, such as insurance com panies, in searching for the most profitable opportunities, len d in various sections of the country. Also, mortgage bankers fre quently resell mortgages to institutions situated in other parts o f the country to enable them to make additional loans locally. FH A insurance and the secondary markets further im prove the acceptance of mortgages outside the local com m unity. In addi tion, savings tend to flow from areas o f relatively low interest rates to areas o f relatively higher interest rates. As a result, there is, in reality, a national mort gage market, bringing com petition for mortgages into virtually every locality. Since lenders are in business to m axim ize their wealth, it is natural for them to seek the most profit able loans available. It is rational, therefore, in determ ining w hether to grant a loan, for lenders to consider such e con om ic factors as the present and future value o f the collateral, the incom e, wealth and other measures o f the creditworthiness o f the bor rower, and the probable collection costs, in addition to the interest rate charged. On the other hand, it is irrational for lenders to refuse to len d for reasons unrelated to the likely profitability o f the loan. T h eoretical O b jec tio n s to the E xisten ce o f “ Irrational ” R edlining T o forego profitable opportunities by discrim i nating against potential borrowers on the basis o f non-econom ic criteria is generally con sidered irra tional behavior on the part o f lenders. Im posing less favorable terms in one area than another, or refusing to lend altogether, w hen not justified by differences in risk or cost, is inconsistent with the self-interest o f lenders or borrowers. I f private lenders are profit maximizers, non-profitable redlining w ou ld be o f lim ited duration.3 Although som e lenders, at times, 3See Jack M. Guttentag and Susan M. Wachter, Redlining and Pul)lic Policy, Monograph Series In Finance and Economics, Monograph 1980-1 (New York University), p. 5. 4 FEBRUARY 1982 may derive satisfaction from denying certain loans for non-econom ic reasons, com petition from other lenders w ho seek such profitable loans assures that such actions are neither com m on nor w idespread. D espite its practical drawbacks, many b eliev e that such redlining is com m on and that laws are n eeded to correct this abuse.4 These observers b eliev e that many financial institutions exercise local m onopoly pow er; thus the potential com petition to reduce unprofitable redlining is severely lim ited. H ence, lenders allegedly have sufficient market pow er to indulge their lending prejudices for a considerable time. Although a lender with sufficient m onopoly pow er can b ecom e lax or biased if he chooses, h ow ever, m ost m o n o p o list len ders have p ecu n ia ry incentives to make the most profitable loans, in cen tives that are r e in fo rce d w h en m an agem ent is accountable to stockholders. An exception, w here prejudicial discrimination may be practiced without pecuniary cost, is a m on opoly lender already so profitable that it fears pu blic p olicy actions may be forthcom ing if it becom es even more profitable.5 This does not appear to be a problem for mortgage lenders. Is There E vid en ce that R edlining E xists? The principal m ethod of demonstrating the exis tence o f redlining is to count mortgages made by certain lenders in an inner-city, low -in com e area and in a suburban, high-incom e area for about a year and com pare the tw o figures.6 Such arguments w ere supported by data supplied by financial institutions under the H om e Mortgage D isclosure Act. T hese data, com bin ed with census information on housing, incom e and population, indicate that low -incom e areas receive proportionately less credit than other neighborhoods. H o w e v e r , th ese stu d ies have seriou s sh ort comings. M ost careful analyses have generally been consistent with the im plications cited above for 4George J. Benston, “ The Persistent Myth of Redlining,” Fortune (March 13, 1978), pp. 66-69. 5See Armen A. Alchian and Reuben A. Kessel, “ Competition, Monopoly, and the Pursuit of Money,” Aspects o f Labor Eco nomics (National Bureau of Economic Research, 1962), pp. 15783. Also, Alfred Nicols, “ Stock Versus Mutual Savings and Loan Associations: Some Evidence of Differences in Behavior,” American Economic Review (May 1967), pp. 337-46. 6See Michael Agelasto II and David Listokin, “ Redlining in Perspective: An Evaluation of Approaches to the Urban Mort gage Dilemma,” in Donald Phares, ed., A Decent Home and Environment: Housing Urban America (Ballinger Publishing Company, 1977). FEDERAL RESERVE BANK OF ST. LOUIS profit-maximizing lenders. T w o studies — for Sac ramento, California and L ou isville, K entucky — demonstrated that arguments advanced to show red lining om itted many important sources o f credit such as mortgage bankers and private funds.7 In an anal ysis o f data provided by T oled o, O hio, savings and loan associations, the dem and for mortgages was also found to be an important om itted factor.8 A study o f FH A insured mortgage foreclosures in six major cities w hich focu sed on the risks (costs) o f lending on properties in alleged redlining areas found that dif ferences in loan terms w ere based on econ om ic rather than prejudicial factors.9 Allegations by com munity groups that properties in low -in com e areas w ere systematically underappraised w ere not sup ported in a study o f savings and loan data for Miami, San Antonio and T o le d o .10 On the other hand, an examination of denials o f m ortgage applications based on a survey o f 176 banks by the Com ptroller and the Federal D ep osit Insurance C orporation (F D IC ) found slight e v id en ce of n on-econom ic dis crim ination.11 An in-depth study o f the Rochester, N ew York, metropolitan area found insufficient e v id e n ce to conclude that redlining was a serious prob lem .12 Its authors noted that previous inquiries claim ing to find significant amounts of redlining made incom p lete surveys o f len d ers (particularly m ortgage bankers), ignored the effect o f rent controls, used time periods too short for meaningful generaliza tions, ignored the collateral offered and the credit worthiness o f borrow ers, or did not com pare d e mands for credit by areas. 7Dennis Dingemans, “ Redlining and Mortgage Lending in Sacra mento,” Annals o f the Association o f American Geographers (June 1979), pp. 225-39; and Theodore Koebel, Housing in Louisville: The Problems o f Disinvestment (Urban Studies Center, University ol Louisville, 1978). 8James R. Ostas, J. David Reed, and Peter M. Hutchinson, “ An Intertemporal Comparison of Urban Mortgage Lending Patterns in the Toledo, Ohio SMSA: 1977 vs. 1975,” Unpublished paper (Bowling Green State University, 1979). 9Richard G. Marcis and Everson W. Hull, “ Analysis of the SocioEconomic Determinants of Foreclosures on 2219(d)(2) and 235 Mortgages,” Federal National Association Meeting, 1975. 10A. Thomas King,“An Analysis of Mortgage Lending in Three SMSA’ s,” Office of Economic Research, Federal Home Loan Bank Board (preliminary report, 1979). "H arold Black, Robert L. Schweitzer and Lewis Mandell, “ Dis crimination in Mortgage Lending,” American Economic Re view (May 1978), pp. 186-91. 12George J. Benston, Dan Horsky, and H. Martin Weingartner, An Empirical Study o f Mortgage Redlining, Monograph Series In Finance and Economics, Monograph 1978-5 (New York Uni versity). FEBRUARY 1982 Another extensive em pirical analysis was made of virtually all hom e mortgage and hom e im provem ent loans granted in C uyahoga C ounty, the central county o f the C leveland area, from 1977 through 1979.13 A fter co n tro llin g for in com e and other dem ographic variables, the study con clu d ed that n eighborhood racial com position had little impact on either the total num ber o f deed transfers financed by mortgage loans or total housing-related financing. H ow ever, it also appeared that the portion o f con ventional mortgage financing provided by banks and savings and loans was significantly low er in inte grated and all-black neighborhoods than in all-white neighborhoods. On the other hand, black and racial ly m ixed areas w ere significantly more likely to be served by mortgage bankers offering FH A and VA financing. Also, banks and savings and loans were m uch more likely to make hom e im provem ent loans in these areas. One can obtain additional evidence that irrational redlining does not exist by looking at the operating history o f new banks established primarily to lend in low -in com e areas. Twenty-six black-ow ned banks, for exam ple, w ere established to serve this demand in low -incom e areas in the last 10 years. O f these new banks, five have failed, and at least a dozen others w ere near collapse before other organizations bailed them out.14 Although minority banks came into existence to deal with specific minority credit problem s, their lack o f success suggests that most creditworthy demands w ere already bein g accom modated, although other factors such as management and capitalization may also have played a role. Further tests of banks’ lending behavior support the profit-maximization m odel. One recent study, using data on 30,000 com m ercial bank consum er loans, tested whether sex discrimination existed in credit allocation by banks.15 The study found no systematic pattern o f sex discrim ination — even before the Equal Credit Opportunity Act was passed. Instead, banks typically beh aved as profit maxi mizers, making loans on equivalent terms to equally risky customers, regardless of their sex. 13Robert B. Avery and Thomas M. Buynak, “ Mortgage Redlining: Some New Evidence,” Economic Review, Federal Reserve Bank of Cleveland (Summer 1981), pp. 18-32. 14Michael L. King, “ Black-Owned Banks Facing Doubts About Their Continued Usefulness,” The Wall Street Journal, August 10, 1981. 15Richard L. Peterson, “ An Investigation of Sex Discrimination in Commercial Bank Direct Consumer Lending,” The Bell Journal o f Economics (Autumn 1981), pp. 547-61. 5 FEDERAL RESERVE BANK OF ST. LOUIS THE C O M M U N IT Y REINVESTMENT ACT D espite theoretical objections and the lack oi e v i d en ce that such redlining actually existed, Congress passed the CRA. The congressional action generally reflected the p u b lic’ s sympathy with the anecdotal arguments o f those living in blighted areas. The suc cess of com m unity groups in con vin cin g the press and pu blic that lenders w ere not serving older urban areas was primarily the result o f skillful publicity rather than substantial confirm ing ev id e n ce .16 The act was intended to elim inate irrational redlining in detexmining w hether a loan application is accepted; lenders w ere still perm itted to evaluate applications on rational econ om ic grounds. The act specifically requires financial institutions to . . demonstrate that their deposit facilities serve the con ven ien ce and needs o f the com m unities in w hich they are chartered to do business.” It directs four regulatory agencies — the Board o f Governors o f the Federal R eserve System, the Com ptroller o f the C urrency, the F ederal H om e Loan Bank Board (FH LBB), and the F D IC — to encourage each insti tution under their jurisdiction to help m eet the credit needs of its entire local community. Under the act, a financial institution is required to adopt a CRA statement, maintain pu blic CRA files and display a CRA notice, w hich includes informa tion on the availability o f the institution’s CRA statem ent. T h e C R A statem ent m ust in c lu d e a delineation of the area that com prises the institu tio n ’ s com m u n ity and a list o f prin cip a l types of credit that the institution is prepared to extend. T he pu blic files must contain any signed com m ents receiv e d from the p u b lic about the institution’s record of serving the credit needs o f its community. In a ddition , the C R A requires the regulatory agencies to assess regularly each institution’ s record of m eeting the credit needs o f its com m unity, in clu din g low -to-m oderate in com e neigh borh oods, consistent with the safe and sound operation o f the institution. These assessments are taken into ac count in rulings on merger, holding com pany and other applications by the institution. Also, the CRA offers anyone the opportu n ity to ch a llen g e any merger, holding com pany or branching expansion 16See George J. Benston, “ Mortgage Redlining Research: A Re view and Critical Analysis Discussion,” The Regulation o f Financial Institutions, Conference Series No. 21 (Federal Re serve Bank of’ Boston and the National Science Foundation, 1979), pp. 114-95. 6 FEBRUARY 1982 plans o f financial institutions that are considered unresponsive to the credit demands o f their co m munity. Protests U nder the C R A T o date, about 100 protests in opposition to the applications o f banks and savings and loans on CRA grounds have b een filed with the regulatory agen cies. M ost have been lod g ed by com m unity orga nizations, a few have com e from the press or indi viduals, and approxim ately one-third have b een lodged by com petitors. M ost protests have b een against institutions located in or near low -in com e areas o f major cities. At first, com m unity groups w ere hesitant about using the CRA to challenge applications, perhaps because o f their unfamiliarity with the operations o f financial institutions and regulatory agencies and because most creditw orthy dem ands w ere bein g accom m odated. Over time, these organizations have gained experience and b ecom e more active. A num ber o f protests have been supported by considerable inform ation; these have frequently b een instru mental in gaining concessions from financial insti tutions.17 Although there are several technical requirements in the CRA, such as displaying a CRA notice in the lobby, protesters have had little com plaint c o n cerning them. The ch ief issue raised in most protests is the failure o f the financial institution to serve adequately the housing-related credit requirements o f lo w -in c o m e n e ig h b orh ood s, e sp e cia lly those com posed o f minorities or those with a shifting racial balance. T hese complaints typically contend that the lending institution systematically refuses or severe ly limits credit to certain neighborhoods becau se o f location, age of property or general conditions in the area, with little regard to the creditworthiness o f the applicant. Protests also have b e e n re g istered on other grounds. These include: the institutions’ failure to advertise the availability o f credit in the low er in com e neighborhoods; a low level o f involvem ent with governm ent programs, particularly the Small Business Administration and the student loan pro grams; excessively restrictive mortgage loan p o l icies, such as larger d ow n paym ents than other l7See Thomas M. Buynak, “ The Community Reinvestment Act: Early Experience and Problems,” Economic Commentary, Federal Reserve Bank o f Cleveland, April 20, 1981. FEDERAL RESERVE BANK OF ST. LOUIS lenders in the com m unity require; pre-screening o f potential loan applicants; and inadequate efforts to ascertain “ com m unity credit n eeds.” FEBRUARY 1982 Table 1 CRA Examination Ratings of Member R egula tory R esponse to the C R A Congress provided little specific guidance in the act as to what is satisfactory or unsatisfactory per form ance in regard to com m unity reinvestment. The act does not explain h ow a financial institution’ s com m unity should b e selected, h ow credit needs are to b e d e te r m in e d , w h at co n s titu te s lo w - and m oderate-incom e neighborhoods, or to what extent the act was to be interpreted by considering the costs, liquidity, safety and profitability o fth e lender. Since the provisions o fth e act are vague, regulatory agencies have had to both enforce the act and inter pret it as w ell. T he regulatory agencies invited the pu blic to sug gest h ow to interpret and im plem ent the act in a series o f hearings h eld in cities across the nation.18 T o provide a focus for the hearings, a num ber o f questions that the statute raised w ere included with the p u blic notice o f the hearings.19 C onsequently, the im plem entation o f the act began m odestly and cautiously and has b een gradually form ulated over time. T h e agencies, evolvin g their ow n standards on a case-by-case basis, have exam ined a variety o f e v id e n c e in evalu a tin g a le n d e r ’ s C R A p e rfo r m ance.20 Under the CRA, regulatory agencies have a num ber o f responsibilities. T h ey must regularly assess each lending institution’ s record o f perform ance in helping to m eet its com m unity credit needs. This assessment or inspection covers both the technical com pliance with regulations and a qualitative eval uation o f the institution’ s overall perform ance in serving the credit requirements o f its com m unity. The regulatory agencies have agreed on a uniform rating system to provide a consistent application o f the act. H ow ever, they assigned no explicit weights 18See statement by Philip C. Jackson Jr., Federal Reserve Bulle tin (August 1978), pp. 631-33. 19See “ Community Reinvestment Act of 1977 to Be Imple mented,” Federal Reserve Bank o f Dallas Voice (March 1978), p. 12, for questions posed. Also, see “Community Reinvestment Act Heaiing Held at Fed,” Federal Reserve Bank o f Dallas, Voice (May 1978), pp. 22-24, for a sampling of the mixed com ments received at the public hearings. 20Glenn Canner and Joe M. Cleaver, “The Community Reinvest ment Act: A Progress Report,” Federal Reserve Bulletin (February 1980), pp. 87-96. Banks During 1980 Rating 1 - Outstanding Number of banks 31 Percentage of banks 3.5% 2 - Good 328 36.7 3 - Satisfactory 507 56.7 4 - Needs im provem ent 5 - Unsatisfactory TOTAL 26 2.9 2 0.2 894 100.0 to the assessment factors, since they b e lie v e d that any such w eighting w ou ld constrain an institution’ s responses to local credit demands. A significant aspect o f the C RA inspection is an overall judgm ental evaluation o f a lender’ s perfor mance in m eeting the credit demands o f its com munity. C R A inspections o f a given bank occu r roughly every 12 to 18 months and, b y and large, have revealed that the banks served the credit needs o f their com m unities (table l).21 Even in cases o f un satisfactory perform ance, most potential borrowers w ere protected by other com petitive institutions in the area. Regulatory agencies take into account both the CR A assessments and actions taken b y the lender to bring about future im provem ent w hen they evaluate an institution’ s application for a charter, branch, office relocation, deposit insurance, m erger or acqui sition. T he agencies w ill deny any application if they ju d ge that the bank or savings and loan has not com plied with the substantive provisions o f the CRA. Also, since the pu blic may challenge a financial institution’ s application on CR A grounds, the agen cies must evaluate the merits o f CRA protests and objections w hen considering an institution’ s appli cation. T o aid protestors, the Federal Reserve Sys tem issued Regulation BR, w hich lists the criteria the Board o f Governors considers in evaluating the CRA record o f a bank (see insert). In addition, each Reserve Bank has appointed a Com m unity Affairs O fficer w h ose resp on sib ilities in clu d e advising 21Glenn Canner, “ The Community Reinvestment Act: A Second Progress Report,” Federal Reserve Bulletin (November 1981), pp. 813-18. 7 FEDERAL RESERVE BANK OF ST. LOUIS com m u n ity groups and banks on p roced u res to follow in C RA disputes. T o date, only four applications have b een denied on C R A grounds. T h ree rejection s w ere b y the F D IC , two involving branch applications and one a merger. The fourth denial was by the Com ptroller o f the Currency. T he three F D IC cases follow ed pro tests by com m unity groups; in the other cases, there was no protest. Three o f these cases w ere subse quently approved after specific actions by the insti tutions — such as hiring a full-tim e com m unity rela tions officer, im proving its marketing programs and FEBRUARY 1982 com m itting specific amounts o f funds for mortgage and hom e im provem ent loans in low-to-m oderate incom e neighborhoods — im proved their C R A per formance. T he relatively few denials under the CRA, h ow ever, is not a reliable measure o f the effect o f the act on bank lending practices. Recause o f the act, a num ber o f institutions have changed certain len din g practices, and many approvals o f applications have been based on com m itm ents to im prove C R A per formance. O f the 19 protested cases approved b y the Assessing the Record of Performance Regulation BB, Section 228.7 In connection with its examination o f a State m em ber bank, the Roard shall assess the record of perform ance o f the bank in helping to m eet the credit needs o f its entire com m unity, including low - and m oderate-incom e neighborhoods, con sistent with safe and sound operation o f the bank. T he Board w ill review the bank’s C R A State m e n ts) and any signed, written com m ents re tained by the State m em ber bank or the Federal Reserve Rank. In addition, the Board w ill con sider the follow in g factors in assessing a bank’s record o f perform ance: (e) T he geographic distribution o f the State m em ber bank’s credit extensions, credit applica tions, and credit denials; (f) E vid en ce o f prohibited discriminatory or other illegal credit practices; (g) T h e State m em ber bank’ s record of open ing and closin g offices and p rov id in g services at offices; (h) The State m em ber bank’s participation, in cluding investments, in local com m unity d e v e l opm ent and redevelopm ent projects or programs; (a) Activities con du cted by the State m em ber bank to ascertain the credit needs o f its com munity, including the extent of the bank’s efforts to com m unicate with m em bers o f its com m unity regarding the credit services bein g provided by the bank; (i) T h e State m em ber bank’s origination o f residential mortgage loans, housing rehabilitation loans, hom e im provem ent loans, and small busi ness or small farm loans within its com m unity, or the purchase o f such loans originated in its com munity; (b) T h e extent o f the State m em ber bank’s marketing and special credit-related programs to make m em bers o f the com m unity aware o f the credit services offered by the bank; (j) The State m em ber bank’ s participation in governm entally-insured, guaranteed, or subsi d ized loan programs for housing, small b u si nesses, or small farms; (c) T he extent o f participation b y the State m em ber bank’ s board o f directors in formulating the bank’ s policies and review ing its perform ance with respect to the purposes o f the Community' Reinvestm ent Act; (d) Any practices intended to discourage ap plications for types o f credit set forth in the State m em ber bank’s CRA Statement(s); (k) The State m em ber bank’ s ability to m eet various com m unity credit n eeds based on its financial condition and size, and legal im pedi ments, local econ om ic conditions, and other fac tors; and (1) Other factors that, in the Hoard’s judgm ent, reasonably bear upon the extent to w hich a State m em ber bank is helping to m eet the credit needs o f its entire community. FEDERAL RESERVE BANK OF ST. LOUIS FEBRUARY 1982 Board o f Governors, com m itm ents have played a role in seven.22 some reallocation o f credit and increased the costs o f financial intermediation. It is the explicit p olicy o f the regulatory agencies to encourage discussion b etw een applicants and protestants to help resolve or narrow their differ ences. A num ber o f such discussions have been held, and, in six protested cases d e cid e d by the Board o f Governors, a privately negotiated agreem ent b e tween the parties was a factor in resolving the prob lem. These discussions resulted in commitments by financial institutions to change lending practices and other policies. At times, in private agreements b e tw een the parties, lenders have gone much further than the act requires, for exam ple, by allocating specific amounts of mortgage credit in certain d e pressed areas at interest rates of V2 percentage point b elow the prevailing market rate.23 The philosophy incorporated in the CRA o f re quiring each financial institution to give preference in its lending to those in their local com m unity can b e questioned. Financial institutions, their stock holders and their depositors are clearly better o ff if funds are loaned w here the interest rates are higher, regardless o f location. Similarly, the prospects for increased investment and production are enhanced w hen credit is allocated efficiently. C om petition am ong len d ers, the b o rro w e rs’ b est p rotection against prejudiced lenders, is strengthened when financial institutions seek to make the best loans possible. EVALUATION OF THE CRA The CRA has b een controversial from its in cep tion. Prior to its passage, com m unity groups claim ed that irrational redlining was com m on, w hile finan cial institutions asserted that they w ere m eeting neighborhood credit demands consistent with pru dent lending practices. There is little agreement, how ever, on the extent o f the problem , though most careful studies have found little discrim ination in lending. C onsequently, the desirability o f the act and the role, if any, that it should play in credit markets is still in doubt after three years. The Am er ican Bankers Association has requested Congress to repeal the C RA.24 Even if som e managers o f financial intermediaries w ere w illing to forego profits to satisfy their prej udices, the sizable num bers o f lenders operating in most local markets, especially in the major cities where redlining is alleged to be greatest, makes it unlikely that many actual cases o f credit unavail ability on com petitive terms w ould be observed. The experience o f the last three years has been con sistent with many econom ists’ view s that the lack of credit availability to borrowers caused by irrational redlining is uncom m on. Yet, also as expected, the act has placed a burden on lenders, w hich has caused 22lbid. 23See the order in the Landmark case, News Release, Federal Reserve Bank of St. Louis, November 30, 1979. 24“ ABA Calls For Repeal of CRA; High Costs are Cited,” Ameri can Banker, December 11, 1981. The ABA contends that costs far exceed the benefits of CRA, and the act merely requires what good sense dictates anyway — serving the communities. Some analysts regarded the passage o f the act as a major step toward governmental allocation of credit. Such concern still exists, even though the regulatory agencies deny that the act and its enforcem ent allo cates credit.25 In fact, the Federal Reserve has stated that it w ill not endorse any agreements to allocate cred it.26 Yet, w h en financial institutions desire favorable rulings on applications, and, as part o f this process, must demonstrate that the credit needs o f low-to-m oderate incom e areas are being adequately served, credit w ill be allocated differently, if for no other reason than to increase the probability o f a favorable ruling. Thus, some financial institutions have adjusted their lending policies to grant more credit in low -to-m oderate in com e areas in their communities. Another result o f the act is that financial institu tions have undertaken large projects in w hich finan cial risks can be shared. One example o f such efforts was in Springfield, Massachusetts, w here 11 local com m ercial and savings banks and two insurance com panies com bin ed developm ent efforts to revi25The Federal Reserve has stated, “Although CRA is directed at the problem of meeting sound community credit needs, it was not intended to establish a regulatory influence on the allocation of credits. In implementing the Act, the Board has acted on the belief that banks are in the best position to assess the credit needs of their own local communities . . Federal Reserve System, Community Reinvestment Act, Information Statement, January 3, 1980, p. 1. 26(The Board) . . is aware that many banks have on their own initiative adopted special purpose credit programs, or pilot programs to test new credit offerings. The Board does not wish to discourage these efforts. However, the Board will closely scrutinize any agreements to ascertain that they are not incon sistent with the safety and soundness of the bank involved, and do not establish a preference for credit extensions inconsistent with evenhanded treatment o f borrowers . . Federal Reserve System, Community Reinvestment Act, Information Statement, January 3, 1980, p. 3. 9 FEDERAL RESERVE BANK OF ST. LOUIS talize the dow ntow n area.27 The longer-run anti com petitive im plications of creating what are es sentially lending cartels, how ever, may be unfavor able for borrowers. An evaluation o f the net impact on the com m unity of C R A -ty p e agreem en ts and com m itm en ts by financial institutions to com m unity groups is diffi cult. Perhaps borrowers in low-to-m oderate incom e areas have received som ewhat more credit than they w ould have otherw ise received. The costs to dep os itors, stockholders and other potential borrowers, how ever, are largely hidden. T o the extent that con tacts with com m unity groups im prove the banker’s k now ledge o f loan opportunities and risks, and gen erates new sources o f sound loans at current market rates, such activities im prove the financial system w hile rem oving som e inequities. T o the extent that bankers engage in these activities m erely to create harmonious pu blic relations, they m erely increase the costs of financial intermediation. The costs im posed by the CRA on financial inter m ediation have run into many m illions o f dollars. T he expense o f adopting formal p o licy statements, 27See “ Investing in the Future of America’s Cities: The Banker’s Bole,” Six Case Studies, prepared by the National Council on Urban Economic Development for the Office of the Comptroller of the Currency, Community Development Division. Digitized for10 FRASER FEBRUARY 1982 appointing com m unity relations officers, familiar izing em ployees with the legal requirements, h old ing meetings with com m unity groups, record k eep ing and reporting must be financed by each lending institution. The administrative costs o f the regu latory a g e n cie s in p e r io d ic a lly assessin g each financial institution’ s CRA perform ance and in ap plying CRA standards in the review o f each appli cation is a burden on taxpayers. A protested CRA application generates the additional costs o f pre paring a defense and often delays for six months or more the outcom e o f the application. Som e have e x p re sse d c o n c e r n that the C R A eventually w ill reduce the supply o f credit in low incom e neighborhoods. A study prepared for the FHLBB found that the act shifted housing-related credit into certain central-city areas, but only in the short run.28 The regulations also raise costs more sharply for lenders serving these localities, w hich could eventually result in a reduction in the supply o f such credit. With less credit available, it becom es more expensive, adversely affecting the low -incom e areas. Also, w hen allegation o f C RA violations com e from com petitive institutions seeking to prevent or delay a market entrant, the flow o f credit to the local area is im peded, an outcom e presum ably opposite to the act’ s intent. 28Guttentag and Wachter, “ Bedlining and Public Policy.” The Shift in Money Demand: What Really Happened? R. W . HAFER AND SCOTT E. HEIN T A HE m oney dem and function is a key relation ship in conventional m acroeconom ic m odels. W hen it appeared that during the mid-1970s the con v en tional specification had undergone an unforeseen shift, analysts devoted considerable ingenuity and research effort to testing alternative explanatory variables that w ou ld account for the change.1 Some specifications have p rod u ced marginally superior forecasting results. N one, how ever, has been suc cessful in explain in g the post-1974 b eh a v ior of m oney demand. Discussions o f the temporal stability o f parameters in econom etric m odels generally differentiate b e tween tw o distinct types o f shift. O ne type of shift is an intercept, or level, shift, in w hich the estimated relationship sim ply undergoes a parallel change that leaves all marginal (slope) coefficients unaf fected. T he other type o f shift occurs w hen at least 1See, for example, Michael J. Hamburger, “ Behavior of the Money Stock: Is There a Puzzle?’’ Journal o f Monetary Eco nomics Quly 1977), pp. 265-88; Gillian Garcia and Simon Pak, “ Some Clues in the Case of the Missing M o n ey," American Eco nomic Review (May 1979), pp. 330-34, and “The Ratio of Cur rency to Demand Deposits in the United States,” Journal o f Finance (June 1979), pp. 703-15; Richard D. Porter, Thomas D. Simpson and Eileen Mauskopf, “ Financial Innovation and the Monetary Aggregates,” Brookings Papers on Economic Activity (1:1979), pp. 213-29; H. Robert Heller and Mohsin S. Khan, “The Demand for Money and the Term Structure of Interest Rates,” Journal o f Political Economy (February 1979), pp. 109-29; David J. Bennett, Flint Brayton, Eileen Mauskopf, Edward K. Offenbacher and Richard D. Porter, “ Econometric Properties o f the Redefined Monetary Aggregates,” Board o f Governors o f the Federal Reserve System, Division of Research and Statistics (February 1980), processed; G. S. Laumas and David E. Spencer, “ The Stability of the Demand for Money: Evidence from the Post-1973 Period,” Review o f Economics and Statistics (August 1980), pp. 455-59; and Thomas D. Simpson and Richard D. Porter, “ Some Issues Involving the Definition and Interpreta tion of the Monetary Aggregates,” in Controlling the Monetary Aggregates III, Conference Series No. 23, (Federal Reserve Bank o f Boston, 1980), pp. 161-234. one o f the relative slope coefficients changes. Sur p risin g ly , p rev iou s exam inations o f the m on ey dem and p u zzle have not explicitly investigated this basic distinction. The approach used in most previous work has b een to presum e that the change was not necessarily parametric, but due to the exclu sion o f an important variable. H ence, most studies focu sed on searching for the “ correct” scale or opportunity cost measures to be used in the rela tionship.2 G iven the unsuccessful nature o f this approach, w e consider a different tack. The purpose o f this article is to study explicitly the nature o f the shift in m oney demand. The evid en ce suggests that the con ventional m oney dem and specification was subject to a once-and-for-all level shift during the mid-1970s. Our results further suggest that the econ om ic rela tionship underlying the estimated slope coefficients of the conventional equation remained remarkably stable throughout the turbulent 1960-79 period. This result conflicts directly with much previous research. The format o f the paper is as follow s: First, the apparent deterioration in the standard specification for M l during the I/1960-IV/1979 period is d ocu m ented.3 Then, a procedure to determ ine likely point(s) o f intercept change(s) in the m oney dem and function is suggested and im plem ented. Finally, the im plications o f our findings are presented. 2For a critical analysis of attempts to repair the conventional specification, see R. W . Hafer and Scott E. Hein, “ Evidence on the Temporal Stability of the Demand for Money Relationship in the United States,” this Review (December 1979), pp. 3-14. 3The 1960-79 period is used to focus attention explicitly on the problems associated with money demand estimations through the mid-1970s. Estimation of the function through 1980 and 1981 would necessitate allowances for the possible effects o f the credit control program and the change in Federal Reserve operating procedures. Such analysis would divert attention from the previously unresolved issue. 11 FEDERAL RESERVE BANK OF ST. LOUIS FEBRUARY 1982 Table 1 Regression Results for Equation 1 Sum mary statistics2 C o efficie nt1 Period RCPt RCBt (M/P)m R2 SE(x10-3) h rho 0.125 (2.69) -0 .0 1 6 (3.02) -0 .0 3 2 (2.08) 0.778 (6.02) 0.967 3.96 1.78 0.31 0.057 (2.51) -0 .0 1 9 (3.45) -0 .0 3 9 (1.79) 0.962 (13.55) 0.874 5.27 0.87 0.55 Constant yt 1/1960-1V/1973 -0 .6 1 0 (2 .8 2 ) 1/1960- IV/1979 -0 .2 7 5 (2.35) 1All variables enter logarithm ically. The log-level equation is estimated using Hatanaka’s procedure. The num bers in parentheses are absolute values of t-statistics. 2lR 2is the coe fficien t of determ ination corrected fo r degrees of freedom , SE is the standard error of the estimated equation, h is the Durbin h-statistic and rho is the Hatanaka estim ate of the autocorrelation coefficient. TH E SHIFT IN THE M O N EY D E M A N D FU N CTIO N : A REVIEW OF THE PROBLEM The conventional m oney dem and specification is (1) In ( M /P ) t = « o + j8i In yt + p-2 In R C P t + /3a In R C B t + (3-1 In (M /P )t.i + et, where M represents the narrow definition of m oney (new M l),4 P is the im plicit GNP deflator (1972= 100)', y is real GNP (1972 dollars), RCP is the com m ercial paper rate and RCB is a w eighted average o f the com m ercial bank passbook rate. W hile many differ ent m oney dem and equations have b een estimated, equation 1 is generally the standard used for com parison. Initial estimates o f equation 1 revealed a signifi cant degree o f first-order serial correlation in the error process. Previous estimates o f equation 1 gen4In response to a changing financial environment, the monetary aggregates were redefined. Thus, checkable deposits can now take the form of negotiable orders of withdrawal (NOW), automatic transfer system (ATS) and credit union share draft accounts. The old M l measure has been augmented by the introduction of these deposits. To the extent that this empirical redefinition ofthe “ transactions” measure of money is induced by the advent o f near-money substitutes, the use of old M l may reveal unstable relationships. Whether other financial innova tions, such as money market mutual funds, repurchase agree ments, overnight Eurodollars and the like, impinge upon the estimation of equation 1 is an empirical matter to be addressed below. Indeed, this line of reasoning has been used to explain the poor post-1973 performance of equation 1. See Garcia and Pak, “ Some Clues in the Case ofthe Missing Money” and “ The Ratio of Currency to Demand Deposits,” and Porter, Simpson and Mauskopf, “ Financial Innovation,” for examples of such arguments. 12 erally have corrected this problem through the use o f the Cochrane-Orcutt iterative procedure. This ap proach, how ever, yields inefficient coefficien t esti mates in the presence o f a lagged dependent vari a b le .5 T h e r e fo r e , to ob ta in estim ates that are (asymptotically) efficient and consistent, Hatanaka’ s residual adjusted Aitken estimation procedure is used in this study.6 Table 1 presents estimates o f equation 1 for the I /1 9 6 0 - I V /1 9 7 3 a n d I /1 9 6 0 - I V /1 9 7 9 s a m p le periods. T he estimates for the earlier sample period are quite similar to those o f other studies. These estimates suggest that real m oney balances adjust toward their equilibrium levels at the rate of about 22 percent per quarter, ceteris paribus. T he esti 5Roger Betancourt and Harry Kelejian, “ Lagged Endogenous Variables and the Cochrane-Orcutt Procedure,” Econometrica (July 1981), pp. 1073-78; and Michio Hatanaka, “ An Efficient Two-Step Estimator for the Dynamic Adjustment Model with Autoregressive Errors,” Journal o f Econometrics (September 1974), pp. 199-220. It has been shown also that the CochraneOrcutt procedure may not iterate to a global minimum of the regression standard error. See R. W . Hafer and Scott E. Hein, “ The Dynamics and Estimation o f Short-Run Money Demand,” this Review (March 1980), pp. 26-35. 6Previous money demand studies using the Hatanaka procedure include Charles Lieberman, “ The Long-Run and Short-Run Demand for Money, Revisited,” Journal o f Money, Credit and Banking (February 1980), pp. 43-57; Laumas and Spencer, “ Stability ofthe Demand for M oney;” and Stuart D. Allen and R. W . Hafer, “ Money Demand and the Term Structure o f Inter est Rates: Some Consistent Estimates,” Journal o f Monetary Economics (forthcoming). For an examination o f the Hatanaka procedure vis-a-vis Cochrane-Orcutt, Hildreth-Lu and maximum-likelihood estima tion techniques, see Edward K. Offenbacher, “ A Comparison of Alternative Estimators of a ‘ Standard’ Money Demand Equa tion,” Special Studies Paper No. 157 (Board of Governors ofthe Federal Reserve System, July 1981). FEBRUARY 1982 FEDERAL RESERVE BANK OF ST. LOUIS mated elasticities also are similar to other estimates. For example, the estimated long-run incom e elastic ity is 0.56, a value that roughly coin cides with the theoretical value given by a sim ple transactions d e mand framework.7 Finally, the summary statistics indicate that a large amount of the variation in real m oney balances is captured by the right-hand vari ables, and the error process appears w ell-behaved. The regression results for the 1/1960-1V/1979 period are quite unlike those of the I/1960-IV/1973 period. The estimated short-run incom e elasticity is halved, w hile the coefficien t on the lagged dep en dent variable increases markedly. T he estimated speed of adjustment (0.04 percent) from the 1/1960IV/1979 results indicates that the mean adjustment lag exceeds 26 quarters, considerably different from that for the pre-1974 period (4.5 quarters). M oreover, the estimated long-run incom e elasticity for the full period is now 1.50, three times the estimate o b tained from the earlier sample period. The I/1960-IV /1979 estimates seem to support the claim that the m oney dem and relationship has b een altered. T he regression evid en ce presented in table 1 suggests that the estimated coefficients have shifted dramatically. M oreover, a standard F-test for structural stability allows one to reject the hypothe sis o f stable regression coefficients across the com m on ly h y p o th e siz e d IV /1973 break p oin t: T h e calculated F-statistie of 4.51 exceeds the 5 percent critical value o f 2.23.8 Table 2 Post-Sample Static Simulation Results: 1/1974-I V/1979 Log-level sp e cifica tio n 1 Year and Q uarter Forecast e rro r (x10~2) 1974 I II III IV 0.02 -1 .2 8 -1 .2 9 -1 .6 0 1975 I II III IV -2 .2 8 -0 .6 8 -1 .3 2 -2.51 1976 I II III IV -1 .3 2 -1 .3 0 -2 .2 2 -1 .7 5 1977 I II III IV -1 .4 8 -2 .2 5 -1 .8 5 -1 .3 6 1978 I II III IV -1 .4 9 -2.41 -1 .8 6 -2 .2 6 1979 I II III IV -2 .6 6 -1.21 -1 .4 9 -2.31 Summary statistics2 Further ev id e n ce of the breakdow n is d em on strated by an analysis o f the equation’ s forecasting ability. Post-sample static forecasts for the natural log o f real m oney balances are presented in table 2.9 These forecasts are based on the coefficient esti mates from the I/1960-IV /1973 regression . The results in table 2 indicate a continual overprediction o f real m oney balances. T he T h eil bias coefficient 7See William J. Baumol, “ The Transactions Demand for Cash: An Inventory Theoretic Approach,” Quarterly Journal of Eco nomics (November 1952), pp. 545-56; and Robert J. Barro, “ In tegral Constraints and Aggregation in an Inventory Model ol Money Demand,” Journal o f Finance (March 1976), pp. 77-78. 8This test is complicated by the presence ol first-order serial correlation. In the results reported, the serial correlation coeffi cient (p) is allowed to vary across subperiods. 9Those familiar with the recent money demand literature may find it surprising that static rather than dynamic forecasts are em ployed. The latter technique yields an exaggerated picture of the shift in a relationship without proper interpretation. Conse quently, the more widely understood static forecasting pro cedure is employed in this paper. See Scott E. Hein, “ Dynamic Forecasting and the Demand for Money,” this Review (June/ July 1980), pp. 13-23. RMSE UM US UC 1.782 (x10-2) 0.882 0.010 0.109 'T he forecast errors (actual less predicted) are logs of M1 (billions of 1972 dollars). They are obtained from sim ulating equation 1 and are based on coefficient estimates in table 1. 2RMSE is th e ro o t-m e an -squa red e rro r in term s o f real money balances (billion s of 1972 dollars) for the log-level specification. U M is the Theil bias coefficient, Us the variance coe fficien t and Uc the covariance coefficient. For an expla nation of these statistics, see Theil, A pp lied E conom ic Fore casting. (U M) indicates that almost 90 percent o f the forecast error is attributable to bias, that is, one-sided pred ic tion errors.10 M oreover, the root-mean-squared error (RM SE) o f 0.01782 is more than four times the insample standard error. l0For a complete description of the derivation and interpretation of the Theil coefficients, see Henri Theil, Applied Economic Forecasting (Amsterdam; North Holland Publishing Co., 1966), pp. 27-32. 13 FEBRUARY 1982 FEDERAL RESERVE BANK OF ST. LOUIS CH O O SING B E TW E EN INTERCEPT AND SLOPE SHIFTS The precedin g evid en ce suggests that the param eter estim ates o f eq u a tion 1 from the 1960-79 period no longer represent a viable em pirical m odel o f short-run m oney demand. O bviously, “ som e sort o f shift has occu rred.” 11 The question is, “ What type o f shift has occu rred?” I f the estimated slope coeffi cients have changed, this im plies changes in the underlying econ om ic relationship (i.e., betw een real m oney balances and real incom e or interest rates). W hile the estimates o f the slope coefficients show marked change over the two periods in table 1, the true s lo p e c o e ffic ie n ts m ay not h ave a ctu a lly changed. If, instead, an intercept shift occurred during the m id-1970s, then em pirical estimates o f equation 1 for the I/1960-IV/1979 sample period may be seriously biased because of the failure to account for the level shift in the relationship, which produces a “ missing variable” problem .12 C on se quently, i f the slope coefficien t estimates are biased, they cou ld lead a researcher to falsely reject the hypothesis o f slope coefficient stability. The major difficulty with an analysis o f intercept shifts is in pinpointing exactly w h en the shift(s) occu rred . A u seful p ro ce d u re to determ in e the likely points o f an intercept or slope shift is to reestimate equation 1 in first-difference form .13 Firstdifferencing equation 1 yields (2) A in (M/P), = I3i A in y, + fi2 A in KCP, + /3s Ain RCB, + A in (\1/P),.i + Aet, w here A is the first-difference operator. Equation 2 provides useful diagnostic information in the event o f an intercept shift in the level equa tion. For exam ple, a once-and-for-all intercept shift in equation 1 w ill appear as a one-tim e increm ent in the disturbance pattern o f the first-difference specification.14 M oreover, changes, if any, in the slope coefficients in equation 1 also w ill appeal' in equation 2. If, as many have argued, the marginal relationships em bod ied in equation 1 have changed, the first-difference specification also w ill exhibit similar changes in the coefficient estimates. Thus, the first-difference specification serves a dual pur pose: It can locate the most likely points of an inter cept shift, and it provides evid en ce on whether the slope coefficients have changed. T o locate and test for potential intercept shifts, the fo llo w in g p ro ce d u re was a d op ted : T h e 1/1960IV/1979 first-difference specification (equation 2) was estimated using ordinary least squares, the re siduals w ere plotted over time and the large residual “ outliers” w ere selected.15 Based on this procedure, three points w ere identified and selected as candi dates for points o f intercept shift: 11/1974, IW 1975 and 11/1979. T he first two residuals w ere negative, suggesting downshifts in the log-level m oney d e mand equation. The 11/1979 residual was positive, suggesting an upshift. Equation 1 was estimated (again using the Hatanaka procedure) assuming on e tim e shifts at those points using (0,1) intercept dummy variables: D l = l for I/1960-I/1974, 0 other w ise; D 2 = l for II/1974-111/1975, 0 otherw ise; and D 3 = 1 for IV /1975-1V /1979, 0 otherwise. Preliminary significance tests revealed that only the 11/1974 intercept shift term was statistically significant at the 5 percent level. C onsequently, we report the version o f equation 1 that incorporates 14This increment will be noticeable if the intercept shift is “ suf ficiently large” relative to the variance ol the disturbances. Thus, the residuals of equation 2 are examined to determine the likely point at which “ large” shifts occurred. 15The focus of this article concerns the possible intercept shift in the log-level money demand equation. Consequently, the reader is referred to Hafer and Hein, “ Investigating the Shift in Money Demand,” for a more detailed analysis of the firstdifference estimation results. To give the reader some idea of the outcome, the OLS estimates of equation 2 for the 1/1960IV/1979 period are (absolute value of t-statistics in parentheses) A In (M/P)t = 0.190 A lnyt — 0.017 A In RCPt (3.51) - “ Stephen M. Goldfeld, “ The Case o f the Missing Money,” Brook ings Papers on Economic Activity (3:1976), p. 726. 12Excluding a relevant variable, in this case the intercept shift term, may bias not only the coefficient estimates, but also the estimate o f the residual variance. On this point, see G. S. Maddala, Econometrics (McGraw-Hill, 1977), pp. 155-57. 13A more complete discussion o f this derivation appears in R. W. Hafer and Scott E. Hein, “ Investigating the Shift in Money Demand: An Econometric Analysis,” in Empirical Studies o f Money Demand: Proceedings o f a Conference Held at the Center fo r the Study o f American Business, Working Paper No. 70 (Center for the Study of American Business, Washington University, August 1981), pp. 1-28. Digitized for 14 FRASER 1 {2 = 0 .4 4 8 (2.94) 0.038 A In RCB, 4- 0.562 A In (M/P)t-i (1.68) (5.70) SE = 0 .0 0 5 h = 0 .4 7 Not only do the coefficient estimates appear reasonably close to the pre-1974 estimates, but ex post forecasts indicate a substantial improvement in the pattern. The resultant RMSE is well within two standard errors of the equation’s in-sample standard error and the Theil decomposition statistics indicate that only 7 percent of the forecast error is attributable to bias. Moreover, an F-test for structural change at IV/1973 yields an F-value of 0.06. For a discussion of these results, see Edward K. Offenbach er, “ Discussion of the Hafer and Hein, Smirlock and Webster Papers,” in Empirical Studies o f M oney Demand, pp. 88-106. FEBRUARY 1982 FEDERAL RESERVE BANK OF ST. LOUIS only the 11/1974 intercept shift variable (D l). The resulting coefficien t estimates are (absolute value o f t-statistics in parentheses)16 (3) In (M/P), = - - dum m y variables is (absolute value o f t-statistics in parentheses)17 In (M/P), = - 0.482 - 0.008 D l + 0.099 D l In y, (2.76) (0.53) (2.61) 0.406 + 0.013 D l + 0.076 In yt (3.95) (2.88) (3.83) + 0.124 D2 In y, - 0.018 D l In RCP, (3.69) (3.41) 0.021 In RCPt - 0.020 In RCB, - 0.013 D2 In RCPt (1.76) - 0.019 D l In RCRt (1.39) - 0.015 D2 In RCB, (0.15) + 0.832 D l In ( M / P ) n (7.45) (4.84) (1.28) + 0.917 In (M/P),.i (16.09) R2 = 0.960 SE = 0.0048 h = -0 .0 5 j> = 0.24 T hese results support the contention that the mar ginal relationships in the short-run m oney demand equation w ere not altered as much as previous evi dence suggests. T he evid en ce, how ever, points to the existence o f a significant, once-and-for-all d ow n ward level shift in the relationship in 11/1974. T h e regression results indicate that the constant term in the log-level specification decreased from —0.406 for the I/1960-I/1974 period to —0.419 in 11/1974. This change (0.013) is small relative to the standard error o f the coefficien t estimate. It is, h ow ever, almost three times as large as the standard error o f the regression (0.0048) for the I/1960-IV/197.3 period. Thus, its exclusion significantly affects the full-sample, level estimation. + 0.560 D2 In (M/P),.i (2.77) R2 = 0.969 SE = 0.0044 D W = 1.90 Standard t-tests w ere used to test the hypothesis that each slope coefficient had rem ained stable once the dow nw ard level shift in 11/1974 had been taken into account. The resulting t-statistics indicate that each coefficient had not changed statistically over the full-sample period. T he variables and the t-statisties for their coefficients are In y (0.43), In RCP (0.19), In RCB (0.03) and ln (M /P )t-i (1.35). This evid en ce supports the view that m oney demand was subject to a level not a slope shift during the m id-1970s.18 A superficial com parison o f the shift-adjusted, log-level estimates with those for the I/1960-IV/ 1973 sample period in table 1 suggests that the slope coefficients may have changed across the period tested. T he question to be addressed n ow is, once the dow nw ard displacem ent o f the constant term has been accounted for, have the slope coefficients changed statistically ? T o formally test this hypothe sis, equation 3 was re-estimated for the full sample period with the individual slope coefficients allow ed to take on different values in the two separate sub periods. Z ero-on e dum m y variables again w ere used to delineate the relevant subsamples (1/1960-1/ 1974 and II/1974-IV/1979): the dum m y variables are D l = l in I/1960-I/1974, 0 otherw ise; and D2 = l in II/1974-IV/1979, 0 otherwise. The esti mated equation using both the intercept and slope T he purpose o f this article has been to investigate the nature of the shift in the conventional m oney dem and specification that occurred during the mid1970s b y determ ining whether it was an intercept or slope shift. The em pirical results presented in this article indicate that the conventional equation was subject to a level, and not a slope, shift in early 1974. Our analysis of the first-difference results and the 16The use of the dummy variable for the I/1960-I/1974 period and the constant term is interpreted in the following manner: The true constant term for the I/1960-I/1974 period is obtained by adding the estimated constant and the estimate on the dummy variable. The constant for the II/1974-IV/1979 period is represented by the estimate of the constant term reported in the text. 18If the preceding evidence were not sufficient to sway the skeptical reader, more support conies from the shift-adjusted, log-level equation’s ex post forecasting record: The RMSE for the shift-adjusted equation for the period II/1974-IV/1979 is 0.67 (xlO'2). This value is well within two standard errors of the estimating equation’s in-sample standard error, and is less than half the RMSE reported in table 2 (1.782 xlO*2). CO NCLUSIO N 17Since the Hatanaka procedure requires estimation o f the resid ual error in the last-stage equation, it, too, was constrained in the above manner. A test revealed that neither the error process nor p had changed. This procedure, in conjunction with the dummy variable test, precludes obtaining a direct estimate of p. 15 properly specified log-level equation suggests that 11/1974 is the most likely point o f the significant downward shift in the m oney dem and function. An important im plication o f this study is that the econ om ic relationships inherent in the conventional m oney dem and function are more stable than previ ous investigations have suggested. Changes in m oney dem and since 11/1974 can be explained by changes in the exogenous variables without relying on tenu ous assertions that the underlying econ om ic rela tionships have degen erated. A lthough previou s analyses have suggested that there has been a con Digitized for 16 FRASER tinuous, unexplained deterioration o f the m oney dem and function after 1973, our analysis suggests that the marginal relationships have remained stable over the I/1960-IV/1979 period, providing useful information in estimating the level o f m oney dem and.19 Thus, claims that the short-run m oney dem and function is highly unstable and is responsi ble for the erratic behavior o f m on ey growth during this period must b e reconsidered. 19For example, see Stephen H. Axilrod and David E. Lindsey, “ Federal Reserve System Implementation of Monetary Policy: Analytical Foundations o f the New Approach,” American Eco nomic Review (May 1981) pp. 246-52.