Full text of Review (Federal Reserve Bank of Dallas) : March 1990
The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
FEDERAL RESERVE BANK OF DALLAS March 1990 • ConOInlC • eVlew Monetary Aggregates and the Rate ofInflation Joseph H. Haslag Oil Prices and Manufacturing Growth: Their Contribution to Houston}s Economic Recovery Robert W. Gilmer This publication was digitized and made available by the Federal Reserve Bank of Dallas' Historical Library (FedHistory@dal.frb.org) Economic Review Federal Reserve Bank of Dallas Robert H. Boykin President and Chief Executive Officer William H. Wallace First Vico President alld Chlof Operating Officer Harvey Rosenblum Semor Vice President and Oirector of Research Gerald P. O'Driscoll, Jr. Vice President and Associate Director of Research W. Michael Cox Vice President end Economic Advisor Stephen P. A. Brown ASSistant Vice President and Senior Economist Economists National and International John K. Hill Robert T. Clair Evan F. Koenig Cara S. lown Kenneth M. Emery Joseph H. Haslag Linda C. Hunter Mark A. Wynne Regional and Energy Robert W. Gilmer William C. Gruben Mine K. YOcel Keith R. Phillips Lori L. Taylor Fiona D. Sigalla Editors Rhonda Harris Diana W. Palmer Virginia M. Rogers The Economic Review is published by the Federal Reserve Bank of Dallas. The views expressed are those of the authors and do not necessarily reflect the pOSitions of the Federal Reserve Bank of Dallas or the Federal Reserve System. Subscriptions are available free of charge. Please send requests for single-copy and multiple-copy subscriptions, back issues, and address changes to the Public Affairs Department, Federal Reserve Bank of Dallas, Station K, Dallas, Texas 75222, (2141651 -6289. Articles may be reprinted on the condition that the source IS credited and the Research Department is prOVided with a copy of the publication containing the reprinted material. b Contents Pag 1 Money Aggregates and the Rate ofInflation Joseph H. Haslag ome economi ts advocal focu ing les mphasis on moneta lY aggrega l s b aus th I' lalio n hip of mo n ta lY aggregates l th ulLimat gals f mo n lalY p Ii y is less reliabl now than in the past. But the stabi lity of lh relationship b lwe n m ney gr wth and inflalion is a lestable hypothes is. J se ph H as lag inv stigates whelh I' mo n y growth i urreml y as useful a pI' dicLO r of inflation a it previous ly has bee n. H e l st thr e differ nl measur s f mo ney growth: lhe mo n lalY l ase, Ml , and M2. Ha lag find lhat th r latio n hip r main table ov r Lime. Haslag al find lh at bOlh th mon lary I a and M2 are usefu l in predi ting th behavior o f th infl ati n rat. Informalion that i unique to Ml, how v r, make a ·tati ti ca lly in ignifi anl o ntributi n to predi cting infl atio n. Pag 13 Oil Prices and Manufacturing Growth: Their Contribution to Houston 's Economic Recovery Robert W. Gilmer Th HUlon conomy w nt from b m to bust to r ov IY during th 1980s. Exp tati ns o f o il pri e at 50 per barr I and hig h r in th lat '70s and ea rly '80 timulated hundr cis of oi l-r lat d I roj clS in th are l. An o il-IXi cle lin , how v 1', I cI lh e H ouston onomy into a sharp r ce i n lhat la l d from 1982 to 1986. The size of H USl n's w rk for e shrunk I y m I' than 12 per nt wilh lh loss of more than 200,000 job . Th lat '80s br ught r n w cI e onomic expansi n, and Hou ton rega in d n arl y 120,000 job. R bert W. Gi lm r xamin s th Huston n mi r ove lY and draw ix co nclu ions an ut lh city' onomy with re pect to lh busin ss ycl , th cI li ar, and th price of oi l. H find s thaL w hil e th city's I' new d gr wth omes primarily in th s rvi tor, oi l and ga induslries wi ll continu e LO play an imp rLant I' I in Houston 's futur . .. Joseph H. Haslag Economist Federal Reserve Bank of Dallas Monetary Aggregates and the Rate of Inflation f mo n y is t have a ro l in the Feder·tJ R serv 's p licy process, two fund amental questi ns must b answ red. First, a a pr dictor of infl atio n, i mon y growth still as us ful now as in the past? eco nd , insofar a m ney remains a us ful predict r o f infl ati o n, w hich measure of mon y sh ulcl receive the m st 'lttenti n? The Oep sito lY Institutio n D reg ul atio n and Mon talY o ntro l Act f 1980 (OTDM CA) pres rib d r gul atolY c1Ylng s and permitted fin ancial innova ti o ns. Each change potenti all y affe ted the relati nship I twe n money growth and econ mi a tivity.1 Rec ntl y, some ec nomi lS began t advo at that monetary aggrega t r eiv less mphas is in th po li y pro ess b au e th yare less r liab ly related to th e ultimate goa ls of poli y n w than in the past. Benjamin Fri dman (1988, 440), ~ r xam pl e, claims "mo n y growth has simply l e n ilT I vant LO any outc me that matters fo r m netalY p Ii y ."2 Friedman also id ntifies the tability of the rel ationship betw en mo ney gr wth and ' onomic a tivity as th c ntral is u in the api ropriare o ndu ct of mon tary p Ii y. The apparent instability in thi r lation hip, Friedman believ s, justifies dismissing a rol e for m n y in th p li cy proce s. In 1987 the Federal Op n Market ommitt (FOMC), citing "un rtainties al o ut it und rl ying r lati n hip to the I ehavi r f th eco nomy and it se nsitivity LO a va ri ety of o n mic and fin ancial cir umstance " (Boa rd of Governors f th e Fed ral Reserve ystem 1987, 21 , d ided not to stablish a sp cific ta rg t range for Ml growth .3 This articl e fo uses o n the relatio nship betw en money gr wth and th inflati n rat. Pres ntly, achi ving pri c stability onc rn s policymak rs.4 Th re ar two main conclusio n in this arti Ie pertaining to th r lation hip b tween th 1110n talY aggregates and inflati n. First, th vi- I Economic Review - March 1990 d n pr S nted here suggests Lhat mo n y gr wth as a predi LOr o f inflation i CIS usejiJ/ now a it was befor fin ancial d r gul atio n; thaL is, the r lationship b Lwe n m n y growth and infl aLi n is stable ove r time. c nd , Lil ev id nc uggest that both th m neta lY I ase and M2 ar useful in pr di Ling Lh behavior of Lh infl ati n raL . Ml, how v r, makes a tati sti ally in ignifica nL ntribution .s imilarl y, r suits from o ut-of-sa mpl ~ recaSLS show that the highesL for casting accuracy is obs IV d wh n th p ifi ati n inclu I either The author wishes to thank W. Michael Cox, Thomas F. Fomby, Scott E. Hein , Evan Koenig, and Cara S. Lown for helpful comments. Of course, any remaining errors are solely my responsibility. , See Tatom (1983) and Thornton (1983) for a discussion of the potential effects of financial deregulation on the relationship between money and economic activity. 2 The apparent breakdown in the relationship between money growth and inflation is also discussed in Kilborn (1986) and Hill and Robinson (1988). 3 MI targets were deemphasized in 1982, relegated to a monitored status and rebased from the previous four quarters during 1983, reestablished as a primary target in 1984, subject to rebasing in 1985, and targeted in 1986. For a discussion of these episodes, see Thornton ( 1983), Hafer (1985. 1986), Nuetzel(1987), and Hafer and Haslag (1988). , Alan Greenspan (1989) states, "That objective remains to maximize sustainable economic growth, which in turn requires the achievement of price stability over time. " 5 In related work. Michael Darby, Angelo Mascaro, and Michael Marlow (1989) compared the predictive performance of MIA (MI less other checkable deposits), MI , and M2. Consistent with the findings presented here, Darby, Mascaro, and Marlow find that other measures of money are more useful than MI in forecasting inflation. 1 the mo n tary base alo ne or the base and M2. Overall, the vidence uggests that appropriately mea ur d mo n y ca n continu to pl aya useful role in th policy process a a pred icto r o f inflatio n. Chart 1 Monetary Base Growth and Inflation Inflation-rate behavior: a theoretical overview W hat is the ratio nale ro r a I' lati o nship between mo ney g rowth and inflati o n? A simpl e fram ework or money d mand and mon y supply-the quantity th o ly-can chara teri z infl atio n-rate behavio r as dep n lent o n changes in mo ney g rowth and change in no nm o n talY factors. The ent rpiece o f the qu anti ty th Oly is the equatio n o r xchange. Form all y, this id nti ty is rep resent d as (1) MV= Py, w her M i a measure of th mo n y up pl y; V i velocity, o r th numb r of tim that mo n y turns ov l' in a given p ri od; Pith pric I vel; and y rea l o utpu t. In growth rates, equati n 1 may be rewritten a (2) B 7 2 SOURCE : Federal Reserve Bank of St. Louis Switching from levels to growth rates primarily reflects technical factors that arise in the empirical analysis. Money and prices are nonstationary time series; regressions that use nonstationary series risk spurious correlation. To mitigate this potential problem, growth rates of money and prices are used in the regression results that are reported later in this article. Under the null hypothesis that the growth rate of each series has a unit root, results from Dickey- Fuller specification indicate that the null hypothesis is rejected. This finding holds regardless of whether or not the specification includes a time-trend variable . The data plotted in Chart 1 are eight-quarter moving averages of both monetary-base growth and the GNP fixedweight deflator. This particular moving-average representation was chosen because the regression analysis indicates that a change in monetary-base growth affects the inflation rate for eight quarters; in other words, the interim multipliers are significantly different from zero for eight quarters following a permanent change in money growth. Equatio n 2 shows that inflatio n an b expl ained by comparing the growth rat of thr e ractors: mo ney, velocity r mo ney, and rea l gross national produ t (G p ) .6 Th I a ic intuition b bind this r laLi onship i th at inflatio n ris s, for instanc , wh en gr wth in th f40 'ective mo ney suppl y o utstrips growth in mo ney d mand. The effective m ney supply grows ith r w h n mo n y grow o r w hen the tran a ti ns fri ci n y (known a v loc ity) o f each unit o f mo n y grows. Mon y d mand growth is largely du to growth in r al income in th lo ng run. In additi n to th ff cts of r al G P growth on inflatio n, th l' no nmonetalY racto rs aff ct the infl ati o n rat thro ugh v locity growth . C nsider th effects on transactio ns ffi ci n y I y financial innovati n . P o pl r duc d th ir transactio n balan sa n w high-yield financial in trum nts such a mo ney market depo it account (MMDAs) and mo ney market mutual funds (MMMF ) w ere cr ated . Fin ancial innova tio ns, ther fo r , increased th eff ti ve mo n y uppl y b ca use eacb do ll ar was u ed mo l' effici ntl y. Tog th r Va nd l' fl ct the effects o f th o e no nm netary facto rs th at influ nce th e inflation rat . Chart 1 pl ots th growth rate o f the moneta lY base and price from 1961 to 1988.7 A general upwa rd trend in mo n taly-ba e growth and th infl atio n rate exi t until about 1980. Since y Federal Reserve Bank of Dallas 1980, however, the moneta lY base doe not have a discernibl e trend , while inflatio n cI arl y has be n d c1ining. Thus, bas d on the evidence before 1980, Chart 1 ugge t that a pos itiv as ocia tion exist d between mon y growth and inflation . Aft r 1980, the existence of this relati o nship is I ss clear. Overa ll, the vid nc in Chart 1 indicate that a mark d change occurred in th r lati onship betwe n mon y growth and inflatio n after 1980. As equation 2 states, the apparent I reakdown in the relationship b tw en mo n y growth and infl ation must refl ect shift in v I city growth or output growth-shift that may r ult fr m movements in nonmo n talY factors. To capture some of th m ovements in infl ation attributabl e to nonmo n tary factors, the growth rate of the relative pric of nergy i inc1ud d . The notion i that o il -pri c ho ksthrough their ff cts on output grow th-are among the principal nonmoneta lY factor that recently have influenced th inflatio n rate. Chart 2 plots th differ nce of mon tary-ba v locity growth and output growth ( that is, Vand the growth rate of relative oi l pri s from 1961 to 1989. 8 A Chart 2 shows, th r are two ca e where th growth rate of relativ n rgy pric s in reas d dramatically; th se in ra e correspond to th oil-price shocks of th 1970 . In both epiod th differ nce betwe n velocity growth and o utput growth rose sharpl y after a hort d lay. Th is ugg sts that th growth rat of relative enrgy pric may s rv as a u ful pr d icto r of the inflation rat . I s the growth rate of r lativ n rgy pric as u ful a pr dictor o f the inflation rat now as it was b fore 1980? In the mid-19 Os, th gr wth rate of r lativ energy pric s decl ined sharply. But, as Chart 2 shows, this decline in oil prices did not pr ede a similar d c1in in th difference b tween velOCity growth and rea l GNP gr wth. Thus, hart 2 sugge ts that a mark d change in the relationship between relativ oi l pri sa nd ( V- 0 curr d in the 1980 . Relative o il prices do not account f r mov ments in velocity growth. One explanation of the appar nt br akdown in th r lation hip between relativ n rgy prices and ( P-M) depi t d in Chart 2 is that the b havior of VI' f1 ected the ff t of finan ial innovations. 9 As not d above, finan ial innovatio n may hav incr a d the y) y) Economic Review - March 1990 Chart 2 Difference Between Velocity and Real GNP Growth and the Growth in Relative Energy Prices Relative Energy Prices (Percent) Velocity Less Output (Percent) 40 8 30 6 Relative Energy Prices 20 4 10 2 o o -10 -2 -20 -4 -30 +: '6~1~~ ' 6"!'5~~'~6~9~~''=7'T''3.....--.-.'=77.,.....,..--.--..~81~~.-85~.....--..8--+9 -6 SOURCE : U.S. Department of Commerce tran actio n effi ciency of mon y balanc s. Alternatively, movem ent in ve locity simply may be due to change in the gen ral I v I of interest rat s. To ca pture the effects of predictable chang in velOC ity growth, th m d I includes lagged valu of infl atio n and mo n y gr w th to xplain curr nt inflation b havior. The quantity theolY provides a ba ic framework to analyz inflation-rat behavior as a function f mon ta lY and nonm n tary fa tors. Th k Y qu stio n i whether the relationship b tween mon y and infl ation is stabl over tim -once it ontroll d for the influence f nonmo n talY hock , u h a hang s in il pric s, that are unrelat d to financial innovation . If , th n B Hafer (1983) defines relative energy prices as the ratio of fuels and related products and power component of the producer price index to the business sector deflator. • John A. Tatom (1988) offers an alternative explanation for the behavior of inflation to the oil-price decrease. He provides evidence suggesting that oil-price shocks have asymmetric effects; that is. the marginal effect of an oil-price shock on economic activity is larger when the relative price of oil is rising than when it is falling. 3 Fri dman's argument that mo ney is irrelev'lnt to po li y is not substanti ated . Model specifications The r latio nships betw en th infl ati o n rate and the growth rates of the mo netary base, Ml , and M2 'Ire exa min ed u ing general reduced-fo rm sp cifi atio ns that ca n b w ritten as II, (3) "l 11\ '-I , I 7C, = a u+ Lf3,7C + L 8,MI-I + L r ,J!.P H '-I H + v, , ilJ w her n repres nt the rate o f infl atio n; denot s the growth rate of th e mo ney m asur ; l 'p is the growth rat o f th relative price of en rgy; v is th e ITOI' te rm ; a, {3, and 8 are the p'lrameters to be estimated ; and 11" 112, and nj are the numb I' of lagged va lues includ d fo r the inflatio n rate, mo ney growth , and relative energy price gr wth , r spectively. Equatio n 3 indicates that movements in the inflatio n rate d pend o n lagged va lues o f the growth in mo n y, infl ati o n, and relati v energy pri c s. This spe ifica ti on is essentiall y th sa me as o ne e lim at d by R. W . H ar r (J 983). B af r's interests, h w v r, w er limited t assess ing the impli ca ti o ns th at chang s in Ml growth wo ul I have for the infl atio n rate. Lag-length electi on is an impo rtant i sue in estimating equati o n 3. Daniel L. Th o rnto n and Dall as . Batten (985) provided evid nce that po licy o n lu i ns ca n b nsitive to the lag struclLIre specified . To address this probl em, th ji'l1al predictioll error (FPE) rit ri o n d velo ped by II. A kaik (969) se lects th lag I ngth f th e va riabl es. lo The FPE cri t ri o n suggests that eac h speC ifica tio n include seve n lagged va lu s of the inflatio n ["lte. With the growth rates of mo ney and relative energy price, however, the number o f 10 Cheng Hsiao (1981) outlines a procedure tor choosing the optimal lag length in a multivariate regression setting. Up to twelve lagged values of the inflation ra te and money growth were allowed to enter each specification. /I See Godfrey (1978) for a description of this test for serial correlation. This test uses four lagged values of the residuals. 4 lagged va lues depends o n the mo ney measure considered. With monetary bas and Ml , the model includ . o nl y o n lagged va lu f the growth rate o f mo n y. The monetary-base model includes seven lagg d va lu es o f the relativ enrgy pri e measure, w hereas th model w ith Ml includ s nin e lagged va lu s. sing M2, the FPE in lica tes th at th m od I sh uld includ e the first and eighth lagg d va lues of mo n y and nine lagged va lu es of the r lative n rgy price. Th e first goa l o f this empiri ca l investigatio n is to determin e w h th r the relati n hip between money growth and infl ati o n is stabl e. This first invo lves tablishing th x i ten e of a I' lationship betw en mo ney g ro wth and infl ati o n, and next determinin g if this relatio nship h'IS undergone ith I' a stru ctural shift or a r du ti o n in its predi tive str ngth . The econd goa l i to d termin w hich f the mo ney m as ur s that pass the stability test i th best predicto r of inflatio n. Compari ns ar based o n w ithin- amp l perform ance and o n uto f-sa mpl fo recasting experim ents. The empirical results Tabl 1 displays the r suits of estimating equ ati o n 3 fo r each of the monetary va ri' tl les fro m th fir t qu arter o f 1959 to the first quarter o f 1989. T he Br us h-Godfrey test is used to d termin w heth r erro rs are se ri all y co rrelated . In each cas , th Br usch-Godfrey tatistics are less than th riti al va lu e o f 2.37; th r fo re, th vidence is consist nt w ith the hypoth sis that the ITOrs are seri ally uncorrelated." Are chang in la gg d valu es o f the infl ati o n rat , mo ney grow th , and th gr wth rate o f relati ve energy pri es systemati ca ll y related to changes in the urI' nt infl atio n rate? As Ta bl e 1 shows, the coe fficient o n the lagged value f mo netary I a e 'lnd Ml growth are Signifi ca ntl y differ nt fr m z roo Tn th e m del w ith M2 as the mo ney measure, the f statistic fo r a test of w h ther the coe ffi cients on lagged va lu e o f mo ney growth ar jOintl y qual to z 1'0 is equ al to 7.79, w hich is grea ter than its critical value of 2.2 1. Th e coe ffi ci nt on lagged va lues of th growth rates o f the mo netary bas, Ml , and M2 are significa ntl y different fro m ze ro; therefo re, the data suggest that changes in mo ney growth are Federal Reserve Bank of Dallas . Table 1 Coefficient Estimates for the Inflation-Rate Model with Various Money Measures, 1959-1989 n1 General Model : n2 n3 n l = a o + Lf3ln l - 1 + LDIMI_I + L ;=1 ;=1 A YI E~_I + VI ;=1 Money Measures Estimated Coefficient Monetary Base M2 .1947 (.45) -.9868" (4.33) .7868(21 .87) .8165" (24.64) .8464" (26 .31) .2048(14.92) .0945(9.49) .2002' (7.79) .0114(2.90) .0182(3.96) -.003(2.98) - .4689 (1 .69) ' I; 0, M1 Regression Diagnostics: Breusch-Godfrey (with 4 lags) F=: S.E.E. .31 .04 .11 .83 .82 .83 1.085 1.092 1.065 'Statistics in parentheses are F statistics. The null hypothesis is that the coefficients are joinlly equal to zero. " Indicates that the test is rejected at the 5-percent level. ystemati ally related LO changes in inOaLio n . Th vi I n e als indica t s thaL hanges in th e growLh rate o f relaLi v energy pri ces 'lI1d lagged inO aLi o n are useful as pI' dicL rs of Lh infi aLion rate. nd er th null hypoth sis LhaL the effi ienLS n the lagged valu e o f are jOintl y equal L ze r , Lh e F sLaListics fr m th e Lhree spec ifi ca tions are all gr aLer Lh an the criti al value of 2.17. imilariy , under Lh null hypOlh sis Lhat th c effi cients on lagg d inOaLio n raLes ar jOintl y qual t z r , Lh F staListi cs fo r Lh thr e different EP Economic neview - March 1990 specifi ca ti o ns ar all grea t r than 21. Thus, lagg d va lues o f the inOaLi n raL and lagg d growth raLes f relaLiv en rgy pri s ar SYSL maLically r lat d LO the inflatio n rat Test for stability Thi analysi will d nnin wh nifica nt shift oc urr d in the b havior Li n rat , and , spec ifi ally, wh th I' a change oc un'ed in the I' lation hip b Lh r a igof the inflaignifica nt tw en 5 mon y growth and inflation. An important i sue in co ndu cLi ng stability t t i th timing of th point at which the potentially significant change in stimat d parameter OCCU lT d. W. Michael Cox and Harvey Rosenblum (1988) find eviden e suggesting that the fo urth quart r f 1982 is the app ropriat break point in t sts to determine if th ave rage leve l of M2 v locity experienced a significa nt shift during the 1980s. This date correspo nds to the inrroductio n of MMDA and up r-NOWa ounts . The break point h re for stability tests is th ~ urth quarter of 1982.12 The first test looks for cha ng s in th coeffici nts that characteri ze the marginal effe t of each explanatolY variable on th infl atio n rate. When testing for significant changes in the esLimated coefficients (a lso known as a Chow test), the regressio n include a seri es of inter'lctive terms, each of which is th product of a dummy variable and an !2 John U. Farley, Melvin Hinich , and Timothy W. McGuire (1975) outline a procedure for conducting stability tests when the break point is unknown. The Farley- Hinich procedure was employed to test for structural change from 1980: 1 to 1982:4. The procedure indicates that a structural change most likely occurred at either 1982:4 or 1980:3. Results do not change substantially if the break point is 1980:3, rather than 1982:4. !3 !4 David J. Stockton and Charles S. Struckmeyer (1989) also find evidence indicating that the estimated coefficients in an inflation-rate model are temporally stable. Ml is used as the measure of money. Note that the Chow test results assume that the variance is constant over the two subsamples. Thus, the stability of the coefficients is conditional on finding the errors not heteroskedastic. In the paper by Darby, Mascaro, and Marlow (1989), the authors find evidence of a structural change in the relationship between Ml and inflation. The differing results may be due to the failure of the authors to include lagged inflation rates in their estimated equations. 15 A. Steven Holland (1984) asked if the variance of errors in predicted inflation were related to the inflation rate. If so, the decelerating inflation that occurred during the 1980s would correspond to a lower variance in prediction errors. Holland provided mixed evidence about the relationship between the level of inflation rate and the variance of prediction errors of the inflation. '6 6 See Fomby, Hill, and Johnson (1984) for a description of the test procedure of heteroskedasticity. xplanato lY variable (including th int I' pt). The dummy variabl has a valu of ze ro for each period before the fourth quarter f 1982, and is given a valu of o n for each peri d thereafter. Th null hypothes is is that the coeffic ients on the intera tiv terms are jOintly eq ual to z 1'0 . Under this null hypothes is, the Fstatistic are 0.97,0.91, an I 0.91 for the models with mo netary I a e, Ml, and M2, r spectively. Becaus the valu s are below th 5-pe rce nt critica l va lu e of 2.1, th null hypothe is is not reject d. Th I' for , the evidence suggests that financial d regulation did not hav a significanr impa ct on th relationship between infhtion and its determinants.I .~ Th Chow test exam in s wh ther th I' is vid n e of a general stru ctural change in the inflaLion-ra t specification. Was th ere a signifi ca nt change in the sp cific marginal ef~ ct that money growlb has on the rate of pric inc I' ase? To test this hypoth si ,on ca n use th co fficients on the inL ractive t rms associated w ith lagg d money growth after the fourth quarter of 1982. U nder the null hypoth sis that coefficients on the int ractiv terms are jointly eq ual to z 1'0, the Fstatisti sa l' 0.64,0.75, and 1.1 for th e models with m netary base, Ml, and M2, respective ly, ,compa r d to a 5-pe rcent riLi ca l va lue of 2.1. Thus, the marginal ffect of a hange in money growth on th inflation rate also s ms stable over tim .14 Anoth r test for tability determines whether there is a ignifica nt change in the accuracy of the infl ation quations over tim . To determine whether a signifi ca nt sh ift in the accura y of th eq uations occurred, the fo ll owing sp cificatio n is stim ated: IS (4) where V i th re idual obta in d from estimating eq uation 3, D is a dummy variable with the value z ro befor 1982:4 and the value ne over the p riod 1982:4-1989:1, u is the rror term, and a and f3 ar the estimated coeffic ients. 16 Equation 4 in licates that the accuracy of the inflation eq uation hifted after 1982 if f3 is signifi cantly d iff r nt from zero. The t st statistics are -1.66, -1.53, and -1.33 for the mod Is with mon tary base, Ml, and M2, respectively. In each case, the abso lute value of the t stati tic i be low its Federal Reserve Bank of Dallas critica l va lu f 1.96; th r fo re, the evidence suggests that the accura y f the infl atio n equ ati o ns did not chang after 1982. The r suits of th sta bility t ts indicate that the current b havior of the inflation rate is not significa ntly different fro m b fo r fin ancial del' gul atio n. More sp cifica ll y, th t st sugge t tll'lt the marginal effe t o f a change in mo n y gr wth i th ame now as b for fin anciaJ del' gulation . The evid nc , th refo re, do s not substantiate th claims o f Friedman and others that th e relatio n hip betw n mo ney growth and inflatio n has waken d . Table 2 Coefficient Estimates for the InflatlonRate Model Comparing Monetary Base and the Broader Monetary Aggregates, 1959-1989 n1 General Model : n2 fCt =a o + L.8;fCt_1+ L 0;Bt-I ;=1 ;=1 n2 n3 1=1 ;=1 L 111Mt_l +L Y;EP,_;+Vt + Money Measures Within-sample comparisons of the monetary variables Estimated Coefficient Now that eac h m n y measure pass d th e stabiJity test, w hi ch mo ney m a ul' is th be t predi to r o f th e inflation rate? Comparing th mo n y m asur s hea d-to-h ad in a I' gres io n answers this questi o n. F rmaliy , the mpari son is condu t d by stimating the foll owing regres io n mode l: "I (5) fc , = "2 "3 a o + L.81fC '_1+ L 0 113,_1+ L 1= 1 II, 1= 1 't'l EPt-I L Tit M'_I + V, , M2 aD -.3283 (.76) - 1.110' (5.86) Ef3i .7764' (22.51) .7342' (15.16) LO, .1611' (4.53) .1695' (7.78) E TJ , .0291 (.45) .1288' (4.69) Er, .0186' (2.81) .0207* (3.17) 1=1 + M1 1= 1 wh re 13 denot s monetalY ba e growth and Mi the growth rat f M1 I' M2. Th pecifica tio n in clud s on lagged va lu e of bas growth. On lagged va lu is inclu I d wh r M1 is th e mon y m asure conjoined with the mo n talY base, whereas the first and ighth lagg d valu s of M2 growth are includ ed in th e pe ifi ca ti o n. 17 What d th results fro m stimating quati n 5 suggest? Suppose that th e coe ffi cients fo r lagged values of both mon ey measures ar Signifi ca ntly diffi rent from zero. In thi s ca ,both mo ney m as ur s are important w hen predi cting infl ati n. Cons id r a s cond scenario where th coe ffic ient for lagged va lu es of o ne mo ney m asur ar statisti ally Significant, but th co ffi cients for th th I' mon ey m asur are not. 18 The ev i I n e th n suggest that th info rmati o n fro m o ne mo ney meas ure i imp rtant in pI' di ting infl atio n wh il e th e other m as ure is not. Thus, th e mo ney measur that is systematica ll y I' lated to Economic Review - March 1990 • Indicates that the test is rejected at the 5-percent level. infl ati n in this encompassing model is superio r to th oth r meas ur in predi cting infl ation . Tabl 2 prese nts th results from quation 5 estimat d se parately w ith M1 and the monetary base and th n w ith M2 and th e m n talY base. " Note that the encompassing model conjoins the model specification with the monetary base to the specification with MI or M2. .. There is actually a third scenario. The coefficients on both money measures are not statistically significant. In light of the empirical findings reported above - when money measures are specified individually- multicollinearity is the most likely cause of this occurrence. With multicollinearity we would simply be unable to draw inferences about which money measure is a better predictor of inflation. 7 The oeffi cient o n lagged va lu s o f mo n tarybas growth are ignifica ntly differ nl from zero in bOlh models. Tn th M l - m neta ry base combinati n , th e 0 ffi ient on lh lagged va lu o f Ml gr Mil is not Significa ntly differ nt from zero . Th refo re, the evidence i consistent w ith th e hypothesis that th e moneta ry ba e i sup ri o r to M l as a p red icto r of infl atio n. U nd r the null hypoth sis lhat th 0 ffi cie ms o n lagg d va lu s o f M2 gr w th are qual to z r , lh F ta tistic is 4.69, w hich is greater than th 5-p I' ent cri tica l va lu of 2.04. The c fficient o n the lagg d va lue o f th mo n lary bas is also sta tisli ca ll y signi fica nt. Thu , the ev id nce uggesls lh at both ba money an I M 2 contribute u eful info rmatio n w hen predi ting inflatio n. Tn sho rt, info rmati on in th mon tary base is imp rlant w b n predicting infl atio n. Info rm alio n uniq ue lO M l , how v r, is not systemali ally related l in fl atio n , but the vid nc ugg sts that in fo rm ati o n uniqu e to M2 is us ful w h n predi cting infl atio n. Out-of-sample forecasting performance Out-of-sa mpl fo recasling ca n indi cat how w II th specificatio ns x pl ain inflatio n-rat b of hav ior. Four p cifi atio ns ar exa min Ii thr th s incl ude th thl' e mo n y mea ure alo ne. Ba d o n the w ithin-sa mpl e compari sons, a comb inati n f ba mo n y and M2 is co nsid r d as ,. The forecasts are generated with actual lagged values of the monetary variables. Thus, the forecasts are conditional on the actual history of monetary behavior. 20 21 22 8 This period does not entirely exclude supply shocks. The forecasting period includes the summer drought of 1988. An alternative method to evaluate forecasts is the mean error criterion. A biased forecast may be less preferred than an unbiased forecast. The mean forecast errors were calculated, and in each case the evidence suggests that forecasts are unbiased. Suppose that the forecas ts are unbiased, then the expected values of the random variables P and 0 may be written as: E[(e, + e,)(e,-e, )] . This expected value equals (cr,2_cr ?). Thus , the difference between the mean squared error is equal to the covariance between P and O. The correlation coefficient is simply the covariance divided by the fourth specifica tio n. Tn lhis articl e, th fo recasting x perim ent us s a rolling regressio n l generale one-step-a h ad fo r caSls ove r lhree di fferent ho ri zons. T h term rollil1g regressiol1 re~ rs to forecasls generated by a model th al i re stimal d each quarter lO in lude additi o nal rea li zaLio ns f th data. 19 B ca use fo r cast r ult may di ffer by the sa mpl e peri od chos n, lhr alt m ati ve fo recast p ri ods we re u I. The span 1981:3-19 9: 1 enmpas e the peri od since the lasl busin ss cycle p ak and ther fore cove rs an mire busin S5 cycl e. Th pan 1982:4-1989: 1 r p r se nts the peri od sin e th last I u in s cycle lr ugh and thu s cov rs the currenl expansio n. The span 19 6:21989: 1 cove rs the peri od since the dramatic o il pri c de lin e. To th extent thal inflati n-I"l te b havio r during th o il -pri ce sho k did not repr nt mo n ta ry influ nc , this x perim el1l all mpts l focus o n a p ri od Wilh ut eXlrao rd inary factors. 2o Th fo r ca ting ca pabilities of ac h m del are va lual d by minimi zing th e rool-mea n-sq uare rTo r (RM E). 21 The implicit a umpli n is that COSlS rise al an in creas ing rate w ilh fo recasl erro rs. Tabl 3 presents th results o f th o n -sl epah ad f I' cast fo r the thr e for ast peri ods. Am o ng the models w ith the mo n y mea ures includ d al n , th RM E is alway mall est w hen th m n w ry bas is the mo n y measure. Indeed, thi RM E is small st amo ng all four sp ifica ti o ns over th 1981:3-1989:1 fo recasling ho ri zo n. Ta bl e 3 also shows that the RM E fo r lh mode l o mbining bas mo n y and M2 is even lowe r lh'lI1 th mod I Wilh base mo n y alo n fo r the fo r casling ho ri zons f 1982:4-1989:1 and 1986:2-1989:1. Th questi on lhat naturall y fo ll ows is w h ther lh differ nces in the RM Es are statisti ca ll y signif'i ant. C. W . ] . Granger and Paul ewb Id (987) o utlin a pr edur to tesl fo r dif~ renc s in th RM E. The proc c1ur defin es Pt a th Lllll of two f re a t err r in peri od I, w h I' as Qt is the differ nc bel w n the fo r cast erro rs in peri od I. If th simple correlatio n oeffi i l1l b tween Pt and Qt i ignifica ml y different fr 111 Z 1'0, th n th difference betwee n l wO HM E timates is also stati slicall y signifi ant. 22 Appl ying III Granger-N w bo ld proced ure lO lh four inflati n-rale models ind icat s lhat no n of lhe RM Es is slatistica ll y diff r nt fro m lh others. Federal Reserve Bank of Dallas Table 3 Results from the One-Step-Ahead Forecasting Experiment RMSE Forecast Period (Starting Date) BaseM2 Monetary Base M1 M2 1981 :3 1.185 1.378 1.571 1.33 1982:4 1.384 1.428 1.51 1.334 1986:2 1.948 2.417 2.092 1.80 Th m del with th low t RM E depend o n the ~ I' asting hori zon. O v rth lo ng st ho ri zon , the model w ith th mo netary ba e alone actuall y ha the low t RM E; but the model that combin es I a money and M 2 ha a lower RM E over the two short I' fo r casting h ri zon . The fact Lhat the RM Es are 11 t ignifica ntly differ nt, howver , sugg sts that the deLeri oratio n in fo recasting a cura y i not large if the broader aggr gate i used al ne, in tead f th I as ,as th predictor of infl ati o n. Summary Th aim of th e analysis is to d termin w heth I' a significa nt d L ri o ratio n occurred in the rel ati o nship b tween mo ney growth and infl ation and to det rmine w hi h mo ney m asur has the trong st I' latio nship to inflatio n. Th mo netary l ase, Ml , and M2 are th mo ney m asur s considered . A simple th o retica l framework characteri z inflatio n-rate behavior as a fun ctio n of changes in mo netalY and nonmo n Lary facto rs. Th vidence suggest Lhat ach infl atio n-rate model is stable. Moreov 1', the evidenc indi cat that th stimated marginal ffect f a chang in mon y grow th n inflatio n has not chang d ignifi antl y over tim . Thi latter findin g directly contradicts claims that d uring th 1980 a br akdown 0 Economic Review - March 1990 curred in the relatio nship between mo n y growth and infl ati o n. Th mod I were compared to d termin e w heth r o n mo ney m a ul' is b tt r than the others at x plaining va ri ati o n in th infl ation rat Th e resul ts suggest that informatio n from th mon talY I a and M 2 is impo rtant w h n predi ting inflatio n, w hil info rmatio n uniqu to Ml i less impo rta nt. Out o f ampl e, th e vid nce show that a model w ith the mo netalY bas ha th lowest forecast H O I'S in each f th e thr e horizons on ider d . The differenc s in the foreca ting accuracy, how v r, are n t tatistica ll y ignifica nt. Thu , the evid nc favor ither the mon tary ba e alo n r the mon talY ba and M 2 together as predictors of the inflatio n rate. Th I' ul ts presented in this articl sugg t that mo n y r mains u eful as a predicto r of inflati o n. Furtherm o re, the vidence indica tes that either the mo netalY ba o r a combinati o n of bas money and M2 should rec ive the most attention w h n pr di ting inflatio n. the square root of the product 0 ; and 0 0 2. So, if the correlation coefficient is significantly different from zero, then the difference between the RMSEs is significantly greater than zero. 9 Appendix p. Versus a General M2 Model Jeffrey J. Hallman , Richard D. Porter, and David H. Small (1989) recently proposed an alternative indicator of inflation-rate behaviorthat measures the long-run equilibrium price level consistent with the level of M2 balances . This measure , called P*, is derived from the equation of exchange under the assumption that M2 velocity has a constant trend value. Hallman , Porter, and Small then specify a simple inflation-rate regression with five lagged values of the inflation rate and one lagged value of the price gap, which is the difference between P* and the actual price level. Formally, the inflation-rate specification in Hallman, Porter, and Small is given by 5 (A 1) lCt = a o + Lf3;lCH + A.(P *(-1 - P'-1)+ Il t, ;=1 where P denotes the values of the fixedweight GNP deflator, and A. is the estimated coefficient of a change in the price gap. In constrast, the M2 growth specification introduced in this article is 7 (A2) lCt = a o + Lf3,lCt-; +OlM2 (-1 +02M2 t_s '=1 9 + L y,Ep,_,+ V" ~ A '=1 The question arises concerning which specification is better at explaining inflation-rate The calculation ofthe price-gap term resulted in equation A 1 not being a nested version of equation A2. Consequently, comparing the alternative specification involves nonnested test procedures . Russell Davidson and James G. MacKinnon (1981) outline a J test procedure designed to make comparisons of competing models. To implement the procedure, initially allow the price-gap speci fication to be the null hypothesis (that is, the P* variable is presumed to be the correct way to capture the influence of a monetary variable on inflation) , and take the M2 growth specification as the alternative. The price-gap specification includes the fitted values from the M2 growth model as a separate explanatory variable. The J test statistic is the tstatistic on the coefficient on the fitted value from the M2 growth equation . As Table A shows , the J test equals 7.6 when the price-gap model is the null hypothesis. Switch the null hypothesis so that the specification with M2 growth is the true model. Under this null hypothesis, the t statistic on the fitted value from the price-gap model is 1.8. The critical value at the 5percent level is 1.96. The evidence suggests that the model with M2 growth used here is significantly better at explaining inflation-rate behavior than the specification with the price gap developed by Hallman, Porter, and Small. movements. Table A Results of Nonnested Specification Tests Comparing the Models with M2 Growth and p. Null Hypothesis J test statistic 10 M2 Growth p. 1.8 7.6 Federal Reserve Bank of Dallas References 1 I Aka ike, H. (969), "Fitting Autoregressive Models for Prediction, " Annalsfol' the In titute of tatistics and Mathematics, 243-47. Green pan, Alan (989), "Statements to Congress," reprint d in the Federal Reserve Bulletin 75 (Septembe r): 614-24. Bel ley, D. A., E. Kuh, and R. E. Wei h (980), Regression Diagnostics: Ident(fying In:!luential Data and ou rce of Collinearity (New York: John Wiley & Sons). Haf r, 11.. W. (983), "Inflation: ASing Its Recent Behavior and Future Prospects," F deral Rerv Bank of t. Loui Review (August): 36-4l. Board of Governo rs of the Federal Reserve Syst m (987), "MonetalY Policy Report to the Congr ," Federal Reseme Bulletin 73 (April): 239-54. Cox, W. Mi hael, and Harvey Ro enblum (988), "Money and Infl ati n in a Deregulated Environm nt: An Overvi w ," Fed ralR serve Bank of Dalla Economic Review (May): 1-19. Dar! y, lI;1 ichael 11.., Ang 10 11.. Mascaro, and Micha I L. Marlow (989), "Th Empirica l Reli abi lity of Mon tary Aggregates as Indicators: 1983-87," Economic Inquiry ( eptember): 555-85. Davidson, Russell, and Jame G. MacKinnon (981), " evera l Tests fo r Mode l pecification in the Presence of Alternative Hypothesis," Econometrica 49 (May): 781-93. Farl ey, John u., Melvin Hinich, and Timothy W. McGuire (975), "om Comparisons of T sts for a hift in the Slopes of a Multivariate Linear Time Serie Model," j ou1"/1,al of Econometrics (August): 297-318. , l Fomby, Thomas B., R. Carter Hill, and Stanl ey 11.. John on (984), Advanced Econometric Methods (New York: Sp ringer-V r1ag). Friedma n, Benjamin M. (988), "Monetary Policy with ut Quantity Variables," American Economic Review 78 (May): 440-45. Godfrey, L. G. (978), "Testing for Higher Order S rial Correlati n in Regression Equations w h n the Regr ssor include Lagged Depend nt Variables," Econometrica 46 (November): 1303-10. Granger, C. W. J., and Paul N wbold (987), F01'ecasting Economic Time Series (Orlando: Academic Pr s , Inc.). Economic Review - March 1990 _ _ (985), "The FOMC in 1983-84: etting Policy in an Un erta in World, " Fed ral Reserve Bank of St. Louis Review (April): 15-37. _ _ (986), "The FOMC in 1985: Reacting to Declining M1 VelOCity," F deral Reserve Bank of t. Louis Review (F bruaIY): 5-2l. _ _ , and Joseph H. Haslag (988), "The FOMC in 1987: The Effects of a Falling Dollar and th tock Market Collapse," Fed ral Re erv Bank of t. Louis Review (March/ April): 3-16. Hallman, Jeffrey J., Richard D. Porter, and David H. mall (989), "M2 per unit of Potential GNP as an Anchor for the Price L vel," taff Stud ies n . 157 (Wa hington, D.C.: Board of Gov rnors of the Federal Reserve Sy tem). Hill, John K., and K nneth J. Robinson (988), "Money, Wage, and Factor carcity as Pr dictor of Inflation," Fed ral Reselve Bank of Dallas Economic Review (May): 21-29. Holland, A. Steven (984), "Does Higher Inflation Lead to Mor Unc rtain Inflation?" Federal Re elve Bank of t. Loui Review (February): 15-26. Hsiao, Ch ng (981), "Autoregressiv Modeling and Money-Income Causality D tection," j ournal ofMonetalY Economics (January): 85-106. Kilborn, P ter T. (986), "Moneta rism Falls From Grace, " New Y01'k Times, 3 July. Nuetzel, Philip A. (987), "The FOMC in 1986: FI xible Policy for Uncertain Time ," Federal Reserve Bank of t. Loui Review (FebruaIY): 15-29. tockton, David J., and Charles . Struckmeyer (989), "Tests of the Specificatio n and Predictive Accuracy of Nonnested Models of 11 Inflation, " Review oj Economic and talistics (May): 275-83. Ta lom,] hl1 A. (1983), "Wa th 1982 V I city Decline nusual?" Fed ral Res rve Bank of Sr. Loui Review (Augu t/ ept mb r) 5-15. _ _ 1988), "Are the Macro conomic Eff ts of Oil -Pric hock Symm tric?" Carnegie-Rocbester erie' 011 Public Policy ( pring): 325-68. Thornto n, Dani I L. (1983), "Th FOMC in 1982: De-emphasizing Ml, " Federal Reserv Bank of r. LOlli Review Oun /]uly): 26-35. ___ (1983), "Why Does V locity MattetT' F d ral R serve Bank of r. Louis Review (Dec mber): 5-13. ___ , and Dallas . Batt n (1985), "Lag-L ngth I tion and Test of Grang r Ca usa li ty Between Money and Incom ," Journal q( Money, Credit, and Banking (May): 164-78. 12 Federal Reserve Bank o f Dallas Robert W. Gilmer Senior Economist Federal Reserve Bank of Dallas Houston Branch Oil Prices and Manufacturing Growth: Their Contribution to Houston's Economic Recovery o usto n is the self-pro laim d iI apital o f the natio n, and ind ed th ere are ve ly str ng argum nts fo r th a CLIr'I CY f this titl e. H o usto n i a c nt I' for virtu all y every plYIS of th e il and gas busin ss, much of it o perating o n a glo bal s ale. H o usto n mploys 66,000 worker in th mining s cto r, almost all in iI and gas explo rati n and d velo pment. ea rl y o ne-third o f the metropo litan area 's manufacturing jobs ar dire Li y in o il fi eld machinery, pet I' chemi ca ls, o r o il refin ing. H u to n is ho me to m st f th worl I' large o il compani s o r th ir op erating divisio ns, with Exx n A , T nn e 0, P nn zoil , and Mit hell En rgy amo ng th il firms h adqu arter d in Husto n. Majo r publicl y- w ned , H o usto n-I as d pipeline ompani es includ e th oa w i o rpo rati n, Enr n, Panhandle E'lstern , and Trans o. The T xas ulf Coast regio n, w ith H o usto n at its ntel', ontains o ne-fo urth of the nali on's refinin g capacity and half o f it ba ic petro hemica l ca pacity. M.W. K 1I0gg, Flu o r Daniel, and Brown and H ot are amo ng th giant constructi n and ngin rin g firms lh at. draw o n th e I' g io n's d ep te hnical experti e. Husto n has shar d g d Limes and I ad w ith the o il industry. M ost f the 1980s were a time o f relren hmenL fo r I th ity and industly. In 1982, H o u l n found itself ho me t hundr d f il-relat d I roject I a ed n x pectati o ns of o il pric at 50 per barrel o r mo re. Wh n it beca me clea r lhat the e pri es we re ho peless ly unrea listic, H ousto n' o il boom f th e 1970s turned into an o il bust, and th ci ty slid into 'I sharp reces io n fro m 1982 lO 1986. Betw en Janu ary 1982 an I January 1987, th CilY lost mo l' th an 100,000 wage and alalY jobs in energy industries. The city H Economic Review - March 1990 lost sli ghtl y mo l' th an 200,000 j bs in all e to rs, whi h I' pI' S nts 12. p I' enl of the pre-bust work fo rce . 13 g inning in arl y 1987, H o ust n's con mic fortunes rebound d w ith robust and st ady econo mi c x pansio n. By lo l er 1989, n arl y 120,000 jol s had b en added back to H ouston's LOlaI wag and ablY empl ymenL. As Chart 1 indi ates, most of H o usto n's n w job ar in th servi e secto r, and in many way this fa l refl e ts a hea lthy div rsifi ca ti n of the H o ust n economic Chart 1 New Jobs Added in Houston Recovery (Thousands) 70 60 50 40 30 20 10 0 M ining Private Services SOURCE: Texas Employment Commission NOTE : Abbreviations are as follows : Const =Construction Mfg = Manufacturing Trade = Wholesale and retail trade Gov't = Government Private Services = Private Service-Producing Industries 13 Chart 2 New Jobs Added in Private Service-Producing Industries January 1987-0ctober 1989 (Thousands) 50 40 30 20 10 O~-------' ____ r-- - 10~~~~~~~~~~~ TCPU FIRE PERS __~~----~ BUSS PROF SOURCE: Texas Employment Commission NOTE : Abbreviations are as follows: TCPU = Transportation. Communications. and Public Utilities FIRE = Finance. Insurance. and Real Estate PERS = Personal Services BUSS = Business Services PROF = Professional Services Personal Services include Standard Industrial Classification (SIC) Codes 70-79. excluding SIC 73. Business Services. Professional Services include SIC 80-89. base away fro m o il. Chait 2 show furth er d ta il about s rvi e ; th bar mark d PROF (pro~ s inal service) o n i t o f job in medica l, I ga l, educatio nal, and other groupings. Abo ut half of this prof ssional incr a i in m di ca l rvices (14,500 job ) and n tructi n-e ng in rin g 1'vice (4,000 j b ) . Th ntributio n o f o il and ga t thi I' covelY ha b en light 0 far. Chart 1 hows thal mining co ntributed o nl y 4,400 new jobs, and o nl y 6,700 n w jobs we re add d in machineri , I' fin ing, and chemica l (om 3.5 percent of the total , For the purposes of this article. the definition of the economic base is deliberately limited. Several service -sector industries could and should be added if, for example. this article were intended as an economic base study. See Leistritz and Murdock (1981). Our purpose here is to cast Houston as a one-industry town. and to explicitly see how this one industry performed in Houston in the recent past. 14 new m anufacturing jo bs) . Th purpos of this articl e i to ass ss th role o f o il and gas in H o usto n's futllr in light o f ilS recent lackluster perfo rmance. Altho ugh di v 1'sifi ca ti o n is a pos iti v facto r fo r H ou tn's conomy and I ng-t I'm futur , o il and gas w ill remain dominant for yea rs to com . Why ha it rec nt ontributio n been 0 wea k? H as it hi to ri ca l I' lati nship to th H o uston cono my b en altered by th five y aI'S of re essio n? A a m th d logica l and o rgani zati o nal cI vice, w timat d la ti citi o f o il-relat d mpi ym nt in a numl I' of industry s cto t' in I til Housto n and Texas and computed th elasticiti es w ith I' p ct to o il pri es, th .5. busin cycl, and the doll ar. Th ese qu atio n are d scrib d in f these tim ates the next sectio n. On the ba i we draw six o ncl usions about H o uston 'S econom y (Tabl e 1) . The o n lu i n do not (a n I ca nnot) a Id up to a fo recast of H o usto n' e 0nomic future, but th y provid in ight into how H o usto n will rea t to outsid fo r s from the busin s ycl ,CUlT ncy market , and th wo rld o il market. Def"initions and Methodology Manufacturing and mining make up Ho usto n'S cono mic ba fo r purpos o f this articl e.' W e furth er divid manufacturing into durabl e good and nondurable goods, and defin e a s parat gr up f o il -relat d manufacturing industries as refining, ch mi ca ls, an I no n I lrica l mahinelY . Tabl e 2 ummari z s the distribution o f this empl oyment in b th H o usto n and T xas. Our m thodo logy i b ITowed dir ctly fro l11 W .H . Branson and James P. Love (1987). B'IS d on th ir wo rk , one red u d-fo rm equatio n is appli d to to tal mining and manufacturing and to vari o us di saggr ga ted secto rs. Th d p nd nt va ri able is th natural loga rithm of mpl oym nt. Th ri ght-hand independent va ri ables are: a co nsta nt, th e .. un mploym nt rate to aptur ycli cal change, a trend term to ca ptur s ul ar hang s in demand , the rea l price o f o il , and the rea l exchange value o f the do ll ar. The unempl yment rat ( urI' nt and fo ur lag ), the I rice o f o il ( urr nt and four lag ), and the va lu o f th do ll ar CUlT nt and six lag) are all transform ed to logarithm s. Federal Reserve Bank of Dallas Table 1 Six Conclusions 1. Houston's economic base is no less sensitive than the rest of Texas to the national business cycle , other things being equal. 2. Houston's economic base is more sensitive than the Texas economic base to changes in the price of oil. The difference is the large share of oil-related manufacturing in Houston. 3. The response of mining (upstream oil) jobs in Houston to changes in the price of oil is now 40-percent smaller than before the Houston recovery began in 1987. The response of Houston manufacturing remains unchanged . 4. Houston's economic base is much more sensitive than the state's economic base to the international value of the dollar. 5. Thus far, Houston's recovery has been led by a more robust U.S. economy and stronger world-commodity prices. However, the decline in the dollar since 1987 exerted roughly equal influence on the Houston economy. Oil-price changes slowed Houston 's recovery. 6. Oil maintains a significant presence in Houston. Although Houston's oil-related economic base is now less responsive to oil-price increases, more stable and higher oil prices still promise big employment gains for the Houston area. Table 2 Distribution of Mining and Manufacturing In Houston and Texas, 1988 Base Mining Manufacturing Durables Nondurables Petroleum-Related Chemicals Petroleum Refining Nonelectrical Machinery Economic Review - Ma rch 1990 Houston (Percent) Texas (Percent) 100.0 30 .0 70.0 37.3 32.7 40.9 15.4 5.6 19.9 100.0 16.2 83.8 48.5 35.4 22 .3 8.0 3.1 11.0 , 15 Table 3 Regression Coefficients for Branson-Love Equations, Estimated for the Houston and Texas Economic Bases Base Mining Manufacturing Durables Nondurables Trend Houston Texas .0033 .0038 .0119" .0046 .0025" .0031 .0044" .0048 .0032 .0008 Unemployment Rate Houston Texas -. 324 -.297 -. 130 -. 241 - .448 -.316 -.749 - .475 -.057*" -.114 .424 .292 .570 .580 .404 .224 .629 .316 .128 .115 -1.109 -.481 -1.111 -. 337 - 1.270 -.418 - 1.791 -.620 -.561 -.220 Oil Price Houston Texas Dollar Houston Texas Dummy Houston Texas .164" .180 .394" .558 .109" .080" .115" .154" .097'· -.120" -. 121'· -.070 - .149'· -.121· -.087*' .001" Delta Houston Texas -.132" -.114 -.242 -.267* NOTE: The lack of an asterisk indicates that a coefficient is significantly different from zero at 95-percent confidence level. A (0) indicates that the coefficient is significantly different from zero at a 90-percent confidence level. A ('0) indicates that the coefficient is not significantly different from zero at a 90-percent confidence level. In additi on, we add d paramet rs to test fo r the exi tence of stru ctural change in the relati nship betwe n oil prices and base empl oyment after 1987 as Ho uston's recove ry got und rway. A dumm y variabl e assuming the v'due o ne in 1987 :1 and after was in Iud d, as we ll as a term to test fo r a shift in the o il-price elasti city after 1987: l. Empl oyment data are the no nagri ultural wage and sa lG IlY fi gures provid ed by th Texas Empl oym nt ommissio n; the rea l pri ce o f o il is the refin er-acqui siti o n cost o f crud (bas d o n the ove rall pri c lev I preva iling in 1982) denated by the co nsumer price ind x; th e I' al va lu e f the d li ar is th index computed by the R s ar h D epartment o f the F deral R serv Bank o f D all as (Cox 1987). Data cover 1975: 1 to 1989:2. W e ap16 pl y o rdin ary I ast square w ith an adjustm III fo r first-o rd r autoco rrelati o n. Elasti iti s o f employment w ith respect to the unemployment rate, oil pri ces, and the do ll ar are co mput d a the. LlI1l of currelll and lagged co ffi cients. H ypothesis t t I' quire that the equatio ns be I' estimat d under th e restri cti o n that the LlIll o f the current and lagged valu es is zero fo r a ingle lasti ciry, and an F test is co ndu cted to test th e I' strictio n. Tabl es 3 and 4 summari ze th results. Th tabl es pres nt regr ssio n coe ffi cients fo r th tr n I t I'm , th un mpl oyment rate, o il pri ces, and th do llar. Th e tabl es al show the variabl used to test fo r stru ctural chang in th impact of o il pri ces aft I' 1987. A n asteri k markin g a 0 ffi ci nt d no tes a 90-perce nt to 95-p rc nt probability Federal Reserve Bank of Dallas Table 4 Regression Coefficients for Branson-Love Equations, Estimated for Houston and Texas Petroleum-Related Industry Petroleum Trend Houston Texas Unemployment Rate Houston Texas Oil Price Houston Texas Dollar Houston Texas Dummy Houston Texas Delta Houston Texas .0056" .0005*' Machinery -.0 101 " .0003 Chemicals Refining .0038" .0010' -.0019*' - .0051 -.487 -. 313 -.773 - .624 - .159 -. 151 .506 .341 .760 .584 .1 17 .102 - 1.471 -.784 -1.287 -1 .072 - .724 -.309 .189*' .148*' -.17 1** - .1 13*' .174**' .138*' .131 .172 - .746 - .307" -.468" .219** -.334** -. 068** .263** .197** .214*** - .155 - .01 8** .017" -.241** - .123" NOTE : The lack of an asterisk indicates thai a coeffici ent is significantly different from zero at a 95indicates that the coe fficient is different from zero at the 90percent confidence level. A percent confidence level ; ("") indicates that the coefficient is not significantly different from zero at a gO-percent level , and ("') indicates that the coefficient is significant and carries the wrong sign . n that th c effi ci nt dif~ rs rrom zero; two asteri sks incli ate th at there is a less than 90-percem chane it clirrers from zero; no asteri sk indi ca tes a probability above 95 perc nt that th c erricient i dirrer nl rro m zero. T he economi c interpretatio n o r the coeffici nts is straightfolwa rd . Th lrend t rm is the qUGl rLerl y growth rate of empl oym nt in thaL seclo r, ho lding fixed the ther rfects o f the ri ghthan I va ri 'd les. The coefri cicnts o r the ther variabl are th elasticity or mpl oyment w ith res pe l l each ind pend nt variabl e. Thi s repr se nlS th p rc ntag change in mpl oyment ca used by a I-perc nt change in th unempl oyment rates, o il pri s, I' th e do ll ar. Estim ates wer mad ro r the H ousto n m troEconomic Review - March 1990 po li wn area and for lh state o f T xas .2 Tab l 3 shows the resulls ro r the br adest industry lefini ti o ns-t tal ba se, minin g, manufacturing, dural les, and nondurabl es. Tabl 4 u es th e am format, but shows mo re d tail of the petro leum- 2 Branson and Love also provide estimates for the state of Texas. Their coefficients are estimated for 1970- 87; the coefficients measure real energy prices using the ratio of the energy consumer price index (CP!) to overall CPI, and measure the real value of the dollar using the International Monetary Fund (/MF) index of relative unit labor costs. Their estimated elasticities are: - 0.24 for the unemployment rate. 0.43 for real energy prices. and - 0.34 for the dollar. All estimates are highly significant. For comparison, our estimates are, respectively. - 0.30, 0.30, and 0.48. 17 based industries. Six Conclusions Table 1 pre ents six conclusio ns that might be drawn from this econom tric x r i . W will discuss ach of these concl u ions and add whatever qua lifications are nec ssa ry. Houston 's economic base is no less sensitive than the rest of Texa to the national business cycle, other things being equal. Both Hou ton and T xa hav nearly idential elasticities of base employm nt with resp t to the U.. unemployment rate , -0.32 for Houston and -0.30 for Texas. This i surprising to th extent that Houston econom ic myth logy typica lly holds that the city i immune to the U. . busin ss cycle, and Houston mi sed u.s. rece ions everal tim in its hi tory. A b tter analogy than immunity, how v 1', i that an antidote to U.S. bu iness condition i om tim ava ilable to Houston in th form of rising oil and gas prices. W will s the effe ts of oil pric s more lea rly below, but there is no question that g od times in th oi l industry an carry Hou ton right through U. . re e ion. By the am log ic, Houston missed mu h of the current U. . xpansion from 1982 to 1986 as oil pric D II. Holding oil price and the xchange rate as n utral factors, how vel', Hou ton and the re t of Texas can expect similar stimulus from national bu in condition. Houston 's economic base is mOl-e sensitive than the Texas economic base to changes in the p'-ice of oil. The difference is in the large sbare of oil-I-elated mam~racturing in Houston. Th re pon e of Hou ton and Texa mrnrng to changes in oi l prices are Virtually id ntical, with an elasticity nea r 0.6. However, Houston's manufacturing sector carries an la ticity f 0.4, and that of Texa i only 0.2. Within the manufacturing s tor, we s simil ar e la ticities for o il-related indu try in both Houston and Texa . The machinry indu try i much more sensitive than l' fining or chemicals, and Houston 's shar of the vo lati le machin rie group is much larger than that of 18 Texa . Ab ut half of the machin ry is o il-service products haded dir ctly for o il and gas xp lo ration and d velopment, and thes industries ar ensiti ve to oil pric j chem icals and r fining , howev r, I' spond more to co n umpti o n than to sho rt-run price changes. The response of mining (up tream oil) jobs in Houston to cbanges in the pI-ice of oil is now 40percent smaller tban before the Houston recovery began in 1987. The response of Houston manufactwing remains unchanged. The dummy and coefficient shi ft (delta) variabl s ud to te t for chang in th oil-pric elasticity since 1987 how no indicati n of large o r ignificant change in Houston or T xas manufacturing. Thi is true for individual petroI um-r lat d sector as well a for th enti re manufacturing sector. In the mining ector, however, the oil-price la ticity shrunk more than 40 perc nt in both Houston and Texa . A reduc d elasticity is ertainly cons i tent with the con mics of xploration and d velopment of o il and gas fi I Is si nc 1987. Aft r o il price began to fall in 1982, oil and gas exp lo ration companies that gea red up for o il at $50 p I' barrel found their employm nt levels and c t tructure greatly out f lin j th layoffs, re tructuring, and reorganization of exploration and dev lopment divisions in Hou ton-based oil compa ni s ontinu d through 1987, and indeed continu today. On major qualification to th results stems fr m the re e nt shift in exploration away from oil and toward natural gas. By mid-1989 a cou nt of il and gas rigs in op ratio n indicated that mol' d m stic exploration wa g are I to gas than o il. Continu ed natural ga urplu sand depressed natural gas prices-depr d even mor than o il--could explain this reduc d elasticity a w have measured it. By including only an o il-pric variable, we could ov I' tat the ext nt of indu try r structuring, and w wou ld xp ct to see a r turn to tronger m a ured elasticiti if, as is frequ ntly for a t, natural gas returns to selling at a pI' mium to th price of oi l. Houston 's economic base is much more ensitive than the tate 's economic base to the international Federal Reserve Bank of Dallas value oj the dollar. The lasti ity of base mployment in Texas with res pe t to the va lu e o f the do llar o n int rn ational markets is - 0. 8; lh h sLi city for H o uston i mo re than lw ice thi va lu e al - 1.1 1.3 The patt rn ho lds in mining and manuracLUring, b tween dural I s and nondurabl es, and ro r pelr I umr laled industly. H ou LOn ' stro ng inlern atio nal ri enta tio n is indi pUlabl . Bas d n f reign l nnag , the Po rt o f Hut n is the busie t po rl in the U nit d Stat , and mu ch or th impo rted l nnag is fo reign o il fo r th I' fin ri sa nd petrochemi ,t! plant thal lin Lh Hut n hip hann I. Agri ulLural pr du LS, hemica ls, and fu I o il ar Hulo n' to p thre x p ItS. H owever, as a world p Lro l um c nter, H o u Lo n's int rn atio nal o nn e Li n x L nd bey nd lh po rL and ship channel. In H ousto n in 1989, ther we re 574 foreign rinn s h adquartered in rifty-on nations. The city ranks third amo ng U. . iti es in the number o f f I' ign trade ffi s, fifth in fo reign consulaLes, and SiXLh in intern ational air pa ng r . TL i an inl rn ati nal financi al nter for th outh and Southwest with fifty-Lhr e ror ign banks (Gr aL r H o u ton Partnerhip 19 9). Thi slrong dep nd enc n th do llar al so provide o n mo r I' a on fo r H o uston ' tough tim b tw n 1982 and 19 6. This wa the am period that th doll ar soa r d n int rnationalmarkets, appr ci ating I y 38 p [ cent according to th ind x u d in thi articl . Thi appr ciati n certainl y add d co n id ral Ie mom ntum to the downturn air ady prec il iLal d I y th e d cline in oil pri s.(A tu ally, th do llar peak d in arly 1985, but its high relativ valu in 1985 and 1986, combin d WiLh adjustm nt lags, prolo ng I the impact r th overvalu d do llar n Huston .) ThusJar, H01lston 's recovelY bas been led by a more rob1lst U. . eC0110my and stronger worldcommodity prices. However, tbe decline in the dollal· since .7987 exerted m ughly equal influen ce on the Houston economy. Oil-price changes slowed Houston 's recovelY. hart 3 sho ws th d li ar's path , Lh U.S. unemployment rate, and th r al pric of oil from Economic Review - March 1990 Chart 3 The Dollar, U.S. Unemployment, and Oil Prices: Their Path from 1987 to 1989 120 110 100 Do ll ar 90 80 Unemployment Rate 70 60~----------~----------~----------~ 1987 1988 1989 INDEX = 100 for all variables in the first quarter of 1987 SOURCES: United States Energy Information Administration Federal Reserve Bank of Dallas United States Bureau of Labor Statistics 1987 LO mid -1989, with ach vari abl set at a valu r 100 as H o uston 's econ mi r e v ry b ga n in 1987. The actual valu e r th do ll ar index m v d fro m nea r 100 in 1987 to n ar 90 in ea rly and mid-1988; by 1989: 2 iL had appr ciated to 95.2. Th .. unemployment raL dropped teadil y from 6.6 perc ill L 5.3 per ent. R al o il I ri ces w re n ar $15 p I' barr I aL Lh beginning and nd of th p ri od, bUL f II as Iowa $10.50 in lat 1988. On w ay to quantify the impo rLanc o f these vari abl is t perform Lil f llowing ount rfactu al xp rim ent. What if, instead f following the palh hown in Chart 3, th e . . un mploym nt raLe had rem ain d at 6.6 p rc nt while th oth I' vari able follow d their a tual ur e? Tabl 5 3 For comparison, Branson and Love ( 1987) obtained an elasticity of - 0.34 for Texas, using the IMF index of relative labor costs as the value of the dollar on international markets. Their largest coefficients across all tifty states were: North Dakota (- 0. 73), Nevada (- 0.55), Alaska (- 0.54), and Wyoming (-0.54). 19 Table 5 Hypothetical Forecasts of Employment in 1989:2 if Each Variable is Held at 1987:2 Levels Actual Houston Texas Unemployment Rate Houston Texas Oil Price Houston Texas Dollar Houston Texas Base Mining Manufacturing Durables 224.2 1147.8 66 .1 177.2 158.1 970.6 85.5 558.5 72.2 412.4 210 .7 1081.9 64.4 169.1 145.5 911 .1 74.2 510.8 71 .6 401 .8 242.9 1205.5 75 .0 199.9 168.3 1005.3 92.9 587.3 75 .0 419.8 206.1 1107.2 61.4 173.3 143.8 936.0 75.7 534.2 69 .2 404.1 pre nts forecas ts of th ese hYI Olheti al levels of em ploymenl. imilarl y, the other va ri abl s were each held in LUrn at 1987: J lev Is, 'Ind the hypothetica l experi ment repea ted . The to p of the tab le shows aCLUal va lu es of the va ri ables in 1989:2. T he most signiri ca nt effect o f th ese three variabl s comes fro m o il prices. H o uSLO n's mining and manufacturing would have had an th r 18,700 j bs if il pri ces had simpl y sLayed at th eir 1987 level instead of plunging by 30 percent and then reLUrning to the initial leve l. I Th e employment ga in fr m this x perim enr (or th al pare11l loss s fro m the a tu al o il-pri e decl ine) w re split 45-55 betwe n mining and manu fa LUring. The do ll ar has th second most Signifi ca nt effect, and if the dollar had not been allowed to d prec iate as it d id aga inst other CUlT ncies, H o uston would hav had 18,100 fewer jobs in 1989. Most o f th e hypoth ti ca l losses (75.2 perent) w ukl have been in manufa turing. H o lding the unempl oyment rat at 6.6 p 1'c nt woul d have ost H o usto n 13,500 j bs, w ith 88. 1 pe r ent of the job los s conce ntrated in manu fac tu ri ng. • These estimates use the reduced post- 1gB? oil-price elasticity. 20 Nondurables Oil ma inta in a sig llificant p resence in Houston . Altbougb HOllston 's oil-related economic base is /l OW less responsive to oil-price increa es, more stable a nd bigIJer oil p rices still p rom ise big em p loyment gains/ or tIJe Hall ton a rea. Suppose the refin er-a quisitio n price o f rud e oil rose in current d li ars to $22.50 per ba rrel in 1989:3 and remained at that leve l through 1990:2. uppos that d uring this peri od the d li ar rema ined at its 1987 index va lu e o f 95 and the unem pl oyme nt rat stayed at the full employment 5.5 percent of 1989:2. Th result would b 19,800 new jobs in H o usto n and 55,400 new jo bs in T xas. H ust n's n w jobs would b o ri ented aim st 2-to-1 in favor o f the mining sector, w hile Texas jo bs woul d split ev nl y betwee n mining and manu fa turing. H o u ton woul d ca ptu re 35.7 perce nt o f th sta tew id e job ga ins in the LOta l base, I' presenting 9.3 p rce nt o f th e ga ins in mining, but o nl y 27.7 p rc nt of the ga ins in man u factu ri ng. Th ese e timates assume that both H ousto n and T xas now add il-r lat d jobs at a rat 40p r ent bel w the pre-1987 lasti ity. The incr a e t $22.50 p I' barr I in th refin era q uisitio n pri ce of o il is a 30-perc nt increase over $17.50 per barrel, the price in mid-1989. u h an increase remains poss ibl and pl aUSibl e (p rh aps four qu art rs o f o il -pri ce stability is the Federal Reserve Bank of Dallas less pl ausibl part f the s nari o). Tf achi v d , ould still bring big employment thi in I' a gains to the Housto n area. It d not I' turn Hut n to the day o f the o il b om . Prices at $22.50 per barr I would I' turn mining empl oyment to on ly 69 p rcent of th 1982 p ak I vels; manufacturing would return LO 94.7 p rc nt o f it 19 1 pea k mployment. Implications for Houston's Future The H o usto n econo my share w ith the re t of T xa it sen itivity to bu iness-cycl e o nditi ons. At the sa me tim e, w find it i al 0 far mo re en itive to volatile w o rl I o il mark t and the valu of th doll ar on w o rl d mark t . In th e past, this combinati on led Housto n through remarkable highs and lows in conomi c p rformance. The imm di aL futur looks more s daLe. Ass uming Lh do llar remains tabl e rd e lines lightl y, and Lh at OPE I' mains unabl to I' 'IS en ontro l of world o il pri ,gr Wlh in H o u LOn and the nited States w ill b imilar during 1990. Over the lo ng I' run , il pri c s may o nce more begin LO ri se in rea l terms, as many analysLs expe L. s Tr so, H o usLo n w ill benefit grea tl y. O il surely w ill move H USL n ' gr Wlh rat ahea d o f th veraII p a e. Add to this H ousto n's positio n in medi ca l s rvic s, its potential edge in mergi ng nvir nm ntal rvi e , and its d pLh in t hni al kill deri ved from A A and Lh large engin rin g firm , and th city ee ms po is d t move fo rwa rd in way that may slow ly diminish its image as a o n -indu stry town .6 il and gas, how ey 1', w ill r main aL th e e no mi c h arL o f H ouston fo r many yea r t om . • See Brown and Phillips (1989). • For a careful look at oil in Texas, recent diversification from oil, and its effect on the state 's performance, see Fomby and Hirschberg (1989). Economic Review-March 1990 21 References Bran on, W.H. , a nd l a me P. L v (1987) , "The Real E change Rate and Employme nt in .S. Manufacturing: tate and R giona l Results ," W rking Pap r No. 2435 (Camb ri dg , Mass.: Nati na l Bur au of Econom ic R s a r 11) . Fomby, Tho ma B. , and]o ep h G . HiI' chb rg (1989), "Texas in Transitio n: 0 pend neon O il and th ational Econo my, " Fed ral R serve Bank of Da ll as Economic Review (Ja nu a ry): 11-27. Br wn , t phen P.A. , a nd Keith R. Phillips (1989), "O il Demand a nd Pric s in th 1990 ," Federal Reserv Bank of Dalla Economic Review (January): 1-8. Greater Houston Partnership (1989), Houston Facts (Hou t n: Greater Hut n ha mbe r of omme rc ), 1- 12. Cox, W . Mi hael (1987) , "A Compreh nsive ew Real 0 lia r Exchange Rate Ind ex," Federa l R serve Ba nk of Dallas Economic Review (March): 1-14. 22 Leistritz, F. Larry , a nd Steve n H. Murdock (981), Tbe ocioeconomic Impact 0/ Resou rce Developmel/t: Metbods/or Assessment (Bould r, Colo.: We tvi w Pres ). Federal Reserve Bank of Dallas