University of Nigeria Author Research Publications TEKE, John Nde Title Faculty Social Sciences Department Economics August, 1997 Signature An Empirical Analysis of the Accuracy of Macroeconomic Forecasts in Nigeria (1970-1995) Date PG/M.Sc/93/14327 AN EMPIRICAL ANALYSIS OF THE ACCURACY OF ' MACROECONOMIC FORECASTS IN NIGERIA (1970-95) I'EKE, JOIIN NDE PG/MSc/93/14327 DEPARTMENT OF ECONOMlCS, UNIVERSITY OF NIGERIA, NSUKKA AUGUST 1997 AN EMI'IRICAL, ANA1,YSIS OF' 'THE ACCIJRACY OF' MACROECONOh4IC FORKCASTS IN NIGEMIA (1970-95) nl- C.C. ASU SUPERVISOR F,X'I'ERNRL EXAMINER ---- Prof'. f? .C. I3ob-Th1ru DEAN OF FACII1,TY 01: SClII'IAI, SCIENTES Prof. O.E. Ohimrl IIEAD, 13EP'T. 01:ECONOMIC'S DEDICATION I ACKNOWLEDGEMENT Saying that a work of this nature was execured wilhout making use of the skills, experiences. suggestiom and time of' a myriad of differen1 pcopre, is a fallacy. On the strcngll of his. I sirnxrcly cxpl-css rny hcart-felt appreciation ro all of tllcm. S p c i T ~ a l I y ,I rernai~lilldcbrcd to nly Supc~risors,Prof. A.E. Okorafor and Dr C.C. Agu for their essenrial conlributions throughout the research. 1 also acknowledge in a special manner. Dr C.C. Soludo for his constructive and useful cri~icisms. To the entire staff of the Department of Ilconomics, I am immensely grateful for my general academic up-bringing. To [he TI3KE's family, and to MI-. N.jong pcter Alud, m y cousin, I convey my warrn gratitude for t k i s love. cncouragernent and sacrifices which sustained me through the duration of nly sludics. I a111also thankful to my friends for [heir support. Above all, let us praise God for everything. TEKE, John NDE TARI,F OF' CONTENTS vii 19 31 33 33 37 39 40 42 43 44 LIST OF' FI'GUREITABLES & 1 1.1 Over~~iew of the Study Consciorlsly or not, forecasting occupies us all for much of our lives. It i s accepted as an iutegral and necessary part of pky-making at nlt Icvels of ecmon~ic Inanagenlerlr especially in economic dcveluprnci~cplanning. The household, the hlsincss firm, the local, slate and national governnlents makc decisions ori r~xsourceallocation based on forecasts. The government is expcicfcd to analyze patterns of change, and 10 predict economic trcrds which are ascnriul requiren~e~tts for econ~inicplanning. For instance, for a goverment to makc apprrrpriare decisions on correcfive st-r:~biliza~ion policy actions, it musr h o w what path [Re xonolny follows; it RIUSI have a reason at?!^ accurate forecasts OF future econon~icdevelopmenls, SpcciFbcalEy, consickr a particular case where ;I central bank , in a reserves framework, seeks to stabilize a short-term inkrest rate around a target value. Sellon and Weiner (19?6;5) stress that in this rescrves - framework, errors in forecasting reserve demarid or supply reslilt into volatilily of interest rates around desired levels. The intuition therefore, is that for a government to effectively run or redress an economy, and obtain desired results, setting of targets is a pre-requisite: And setting of such targets reiy essentially on the government's ability ro make reasonably accurate macroeconomic forecasts sibout the tit turo course of the 2 ccoriorlly Rectnt!y, most devclopmg counfries (including Nigeria) art stnrggling 70 no avail with l c l r i ~ i mof economic recowry prugntnn-tes. Even though much research s v ~ r kon the 4 Of course. ib is unrc;cmnal)lc tn expect error-frce forecasts in forwasting si~,u;ttior~s. 'I'liis f'act is rccognisd by ir~cluclin~ nn irregular cnrnpmenl whcn describing a time scrics. Thc prcscncc r>f illis nndoni coniponcnt, which represents uncxplaincd and unprcdict;thlc fluctunlions in tl~cdata, nicans that somc cmmr in kmcasting must bc e x p t ~ r c d . I f Ihc el'fccl oT the 1-;1ridor11facf(~ris sitbstaririal. rhe at)i!ity to .for€cast xcumrcly will hc limit~d.IT~c)\I!cvc~,(IIC effect of the unpredictable conipmenf is small, fnwcxts shcmld track nctuals wirh rnorc accuracy. r i c1101 ~ ~ ~thc only source of crror in Ncvcrthclc.~~.llic irrcgul:~I- c c ~ ~ i ~ p o is forccnsting. Thc accurxy of a forecasting (ecliniquc infiucnccs the ~nagnitudeof crror in frrrecasts. Since thesc crror divisions cannot OL' predicted in a practical situation, f i m . x l crrcrrs rcpr'esclil tlic conlbincd effects of tllc s t ( d m t i c cmpnnenl {unsysfcrnaric crm)r) nrid crrtlrs cri~nrinti~ig fr-urn thc Sorccasting 111ctf~d(sysiemztic error). tlence, signif'icnnt iorecasting errors m y indicate [hat unsysteniatic error is so large that no fowcasring technique will produce accura tc forecasts, or thcy nlay indicate [hat the fill-ccasting nietl~odhcirig used is not cap:~ljlcof' predicting accurntcly - an iriappropriatc rc.c.111liqueis bcing etnployed. C'rwsccpmtly, Illcsc u~irlcrcs~iniationor overestimntiori fc~rccastscould pose sipiil'icant ecollo~iiicins~nl?ilityi l l a n cconon~y. When itilcrr~iedinteforecasts are undcrlxcdicted, tvllich mcans I'orccxts are sliort of actual values, the economy seems to be cnntractcd wllcreas it actually cxparlds witliour nccornpnnyinp plmrwJ ecnnoniic politics. persistent a n d scemirlgly large forecast errors? Why arc Nigerian economic ~ninimuni, ir~spire of thc positive relationship that ex& be[wwn i f ~ n d economic performance? CouItI i t bc hat tIlc random component r l f the forecaslt error (the source o f crror about which ~iotliingcan bc dvnc Lo rcdxe the crror ~lrargin)is substantial? Or, On a more gerieraI noto, tlic primary aim of this invesligahn is ra explore ways to ilxscss tIic potcntinl contril~ulionof Ihrccastin~to policy rornlrllarion. Tlle following spucific ohjcctivcs will guidc the study: 6 'To have an insighl into the nature of fol-ec;w ermrs rhat is observed in nioncy (a) supply. fiscal pap. and nct tlorncsTic credit f o r c c m ~o~f Lltc Nigerian economy sincc 1970. Thc dislrihution ni [lie fcrrccast errors into unsystemlic and y l c r n n t i c errw romponcrlrs wiPl Rc cl'rccrcd 'Thc cvnluation oI' the clircclio~iof rllcsc intcrnic~linlcfo~rcaslsto rtc;lu,rl vaTues (h) since IQ70 \rill I)e carrictl out in o d c r tn ascccrain whether thcse foarcas~sare unclc~cstima~ior~s or nvcrestirnatim~. 1.4 IIypothes~sof the Strl-cly The working h y p ~ h e s c snT this study arc sfatcd as follows: I : There is no s i g n i f i ~ a fr)r;ccast t~~ error in the i n t e r ~ d i a fmecasts t~ observed in the Nigcl ian wonumy since 1970. In otl~crwords. nbscrved tbrccas! errors of the inter~ncdiatcvnriablcs arc tllc hrtrcst minimum rhar nlrlsr m u r (?I,, : ct = Z1 and 0= 1). Shrnrld II,,, Ix: rcjmted. Ilwn two w p a r a l ~lcsts for the respective effect of r k 11m,7n bias 3rd vaciarlcc 6 a c coliymncnls CwT~icliconstiMc the systcmaric crrrvf nf lhc forecast crrvrs arc H,,,: I ~ C C C S S ~ ~ ~They I ~ C ~itrr: ~. Thc expectcd mean of the intertncdiatc forecasts. cqu;~lsthc expected mean ~f aclnnls. That is. thc obscrved systematic liorecafl crmr is ntfl accounted f c by ~ the mean bins [IIn7:E(FJ I : E(h,)l. There is no significant presence of the variance bias in the forecast errors of the intcrrnediate variables of tl~cNigerian economy. ht differently, x l u a l series dc considerably I'lucluatc as I : ITILIC~Ias the forecast series. (I-I,,: 0 = 1). Forecasts of the iritcrn~cdiatevariables o f the Nigcriarl cconomy since 1970 are unclcrestin~atcclas ortcr~n s tlicy are over-eslirnatd [I!,,,: P(F, < A,) = P(F, > f t ~ r c c ~arc s ~ saccurate), r k y allow gnvcnmcnt to act incrcmentalfy on rhc cconorny. That Secor~dly,rn:~crocconomicfol-ccasts can x r v e II monitoring rtmction and provide a basis against which to jr~dgchow economic politics are working mlt. Any serious 8 knowledge of r7 !GI! is p i r l g on currently uughl to he consistent with tTx figures for the ncar fi~lurcand tllc recenl pas[. Tt :ilso i~ivolvcsthc second p i w p k c a u s e if thc ourcorne of cxisting ccoiw~riic policics as rcvcakd in rhc ~nacmccommic forccasls is u n s a M x t n r y . ffcsh rrconrm~icpolicics will hnvc lo be i n t d t w c d . 1'kdissemination of c~fl'icid!~~;~c.l-c~ccc~nrm~ic forecasts tcl invcdvc all governmcrll ofTicials ns ivcll as [lie wlmle cn~n~iitlnity ;I[ I a r p (public and private sectim) prrwidcs C V C ~ ~ O I ~ I Pi ~ n u l t a n ~ with ~~~sty ;t C O I I I I I N ! ~ Iview of ~ ~ ' l l i 1will 1 linllpcrr u~ilcssson~cthingis donc. Each clicrciore. calr then rcxt ktlrnvirly wlinl vicw of' t!ic futrlrc is llclrl c.lsc\vl~crcant1 c;rn cithcs cllallcngc that view or propnse actitm tmed . 1117011i t , . I he foscgoing irnpor-1a11ceattached ( o nlitcri~ccoriomi~forecasts (targets) in ~ of cconoulic pcrt'orrnance is bascd on [he nrssumption that these dicraring t l pace forccrrsls arc r~taair~:ihly xcuratc. Accordingly. inv~stigatinylllc accuracy of targets via tlic dcco~nposirionof tlic discrcpancics hclwccn aclual and target valucs (of major macroeco~~onlic v;lriablcs) is crucial in undcrsta~ldingand explaitling the performance Icvcl ol' tlic Nigcr'inri cccwotn!; since 1970. Ilcncc, thc idcntificalion of forecast errors into bias. random, undcrestirr~alion and ovcrestirilntion errors should illuminate some critical policy issues that coilld hc dcsigncd to improve on the economic pcrfornia~iceof 11ic Nigeri;tn ccononly. Su~nmarily,tllc fincli~igsof this study are intended to assist ccnnnniic analysts/policy -nlakcrs in cvaluaring tlic forecash thcy exploit in order to sat isfiictorily achicvc 11-lc ulility and reliability of the expcctcd uncertainty based on MSE, = ( A 'rvI1~"reE ib tl:e - F = I ) = [E(U)JZ + 0: ..$2.5) value, A , T: ;tnd U are vecrors of I-enli/:~~ions, forecws!:; ;.~rncl CIX~CC[C~ t'orccfist errors, r c s p x r i v c l y . l'lw s q t i a r u - s r w t 01 hlSE, rllci~surtsthc averape ~ L oC f . j'.-0 1.LC.;I>;[ CI'I'OI'. A rigorous a ~ u l y s i s01' rliu MSE, involws i ~ de::onywsirii~!~ s in10~ h r e ec u n i p ~ c ~ i r s of i l ~ csaulws of f o ~ w a s rrr-1.0s: h.1SE, = I 'l'lic Eik;,i cmnprrcllr oil lhc' + I - - ( I -pf,)o: ., . ( 2 . 6 ) right or (2.6) is callccl the 111c;ln(bias) cwltponrnt, the szcond. rlw. slope (variance) colllponent, and tlw thirr!, I! ~e residual (cuv;~ri;~nce) cornpo~ienlof the MSI',. The unar~iin;~blc case of pcl-fectiull occurs u hen E(U) =0, tlritt is, when E(A) -- E(F). C o ~ ~ s i ~ ! c r(2.21, i n g r!ie rcgrcssion s!~)pu 0, ccl~ralsunity cmly when the foremst ci~cjt.IJ,, is ~111,-co1.1.cl:ttcd wilTi 3 1 ~ :f r > o n i ~ ~itlucs, . ~ ~ ~ I:,. 11i:i1 is, when a,' = r~:. 'l'herefore, for perlect forecast (when RT-.coincides wirh LPF), whuse A arid 1' iIli 1hc rlll';lnS d sl ~~)TCC;IS(S and i l ~ t l l ; ! l ~ , variarlcus 01- plcdic~ionsitnd r~eali/;~tions. I, is ~ l l oslopc of' dcler-rni~l;ilionI W I W C L ' ~ I'OI.CC;ISLS and ;ic.luals. ti = 0, and S,, ;ire the esliln;\lerl usti~natei ~ n d;', i s the cocfi'icknt for possible cvalua~icrnas I U illc stmrccs of forcc~slc m r . I:ir~;rlIy, cl~ecwfficiei-~r indicates !he tlegm. c61ch olI1~1'. 'I'lll 5c' 10 whicli eit11t.r 111t acrud or forecast scries fluciuarcs away from p ~ . o p ) ~ . ~ i ((Ib o r ~ s:~rn! U,.] p r w m x 01' systcnt:~riccrrois - 1ll:11 wclrrison~chec:111se ~ F I Lj ifxricarc t l w is, I'cmc;rsls s i p it'ic:~nrlyilil'fcr liunl. crl>scrv;~rio~ls. I.,asrly. ~ l wcovari;tru proporlio~lis .#. ;I rnt.;mirc: of r l ~ eun~ys[en~atic error I ~ Ioccurs . sttx.l~as~ically, a~lclis Itss trouhlijl_~silxc w x m l i ~ ~ 10g1'i~~clycR/I~ul~i1if'eli! (1990; 34 1 j ".., it 1,s unrcasr)n;rble t o c x p x ~predictions [hill iirc perfccrly rlorrcl:trc.d with ;tcru;d outcomes . ..". l'his is s i ~hrciluse r i a ~ i ~ n l l yr, , $ [I. AII i l q ~ m : ~ r ;tcIitilion:d ll crirerio~insed Ibr roreci~srpca-fonl!:l~icc.is INJ:~ su,:i11 :ire rtw iraciclcucc I?!' lurriing p \VIWll b i ~ :~ I l I a c ~ c k r a ~ i errors. o~i A l i ~ r ~ i lhailll ~ l g w r u r i:; observcil ~ A~ Imll - A, # F, - ryl.l . ..(2.19) o = are coun~tI I i ~ ;~ccelcr-:~tio!~ s c n u s~ 011wi 1nc.~suw5or limitccl :qydicxtiun irldudc lllc 1 . ~ ~ 1n1 u n scp~;ircclpel-CCINerror, IIICiIIl p t c ~ f i error, l ~ I ~ cr i .I< i l l I ~OTCC;ISF e t ~ ~rC I K X L ' ~ a ~ dcxpresxcl as ME, = I -1 (1; - A,) 1.1 2.2 Fhpiricnl Lilcr;lt ur.c 1tt.vi~-rv 'I'llc d c i ~ c l a p: ~ I I i~of' ~ ir n u ~ n l of'I:lrgc ~r scale econometric modcls clnfcs :IS the early 19.d's, ;~rld cvcntwtly early U I ~ ; ) I ~ ~ C : I I S~LIL!,.T 011 !II(: i19 far back a ~ c u r a c y of' ~ l ~ x r o e c o n o ~ f'o~w:ists nic follov~A.'I'he~t.ri uclels w r e int?~,c!idto be usefi~1 I U ~ ) ! S irr rlle ;~nalys~s ol rhc economic impacts of tl wide rnnrlc of policy : ~ l ~ ~ ~ - n a r thereby i v c s , m:r'i.ing I-csptuivdy.4hcr rr1rnt', t k s e v;irii~blc.sdc~crior:~rcd ;I-; r l ~ ufor ccast I.rorizon incrc;wA. A cur~~parisor~ 01' scnli-naiw forc.c;~scsgc~i~r:~tctl by AKlMR ~ r u d c l s rcvcal pour r q i s t e r ~ n gI i ~ r g cl<bISFI: or 1 1 .O for norni~ialG N P , 8 , 7 Su, real Gb!P, ~ I N I ~CI-!'OI-I~;IIICC, O ti !'or-[Ire CiNP clcl7;1rrw.I:c~rer.iistsfrom lllcsc rnadcl.;, CIIIis1 noLuc1, could bc i:nprovCJ 1111 tntg!~s o l ~ c zbvc l j~rdgcr~~c~~l. aiulysis in c t m n cc;rntmic.~.111 vicw ill' h i s (lpinion, Ilcz~kl;(1979) ~sx;ru~incs r c s u l ~of' S I . F ~ < ' C I~)I ~ C I C ~ L ' I K I C I I Ico~i\p;~l'iso~~s i ~ \ p u t n. ~ t p u tn~oilclsw i h (less costly :i,d I less so~d~is~icaiccl) :rllcrna~ivcI;rrc.i.as!ing ii.cln~iq~it.s ~ ~ m i l i ~ l tCIui.i~l;: c'J 1115 priucll !'rwi 195 1 rlwouyl~1972 in 11lc l.!~~itcd S1:ttes of A ~ w r i c ; ~l ,~ d y Uri , rain, kurwny, Tn~!l.l,;mil t l NL'IIIC~I;IIICIS. ~ Using wi,qhwcl and ~ ~ n \ v e i ~ percer~l;~gl: ht~d en-ur-s h r cii:~l1urion,111. ul~sel-vcsrh:cl rt-rcrc were !-ton>alrsrwtivc ' l i w c a s i i ~ i gnwl!lc>cls which werl-.coiisisle~~ily ;I., goo~l;IS itqiul ~ I U I ~~ L Im\ ~ k l ls!c. I\IIY-C!'~NX:, rcw!vcd [ h a t f u ~ ~ ~ : : vI'~.L~II\ its i , : ~ ~ r i d i ~ s ! f y 111oclt.t~prop:lg;~lcr i A ; ~ ~ i w lsrnnll y crwn-s i!vcr;~~.ir~g ~ULWCL'R livc ;mi 11ir1epcrcc'nb, a114 \VL'IX ilt BCWI ;IS ;~ccLI~.;\~c. !JICYW I'I.OIII 1'111e~11;tfisc ICC~I~~C~UC'S. f!:.l~iployiriglllu nwan ithsolu~ccr-rol , arid Ihc Chow's 1'-test sti~tistk't;,rcsdts shawcd that 28 ~imdcl~ :vi uppuwd,10 ~ I I : I C C L I Ii ~l l;prc!i1;4i11ary K~ esti[~mtcior ev;ilii;irion mrlutlcd rhc FrrSli, turmi~u; " I K I ~ I ~ ' ' ~urcc:\sl~, 1 ! ! d k t s of crror r:ounr, and acct:lcr;\rion error counr. Furtcnsrs gencrxrcJ fro111IIK h t c h r i u . k ! n1isr.c k)urld t o hu marc ;wur;lle 'rl BII$ l wive ~ rli;111t forccasts. i ~ n dsuperior rrl Norwcgi;in ~oi~ecasts. Si.ns reharks r l w c ~ w - & ,h i t ' h e lo\v h1SFE obscrvcil c ~ u l d!x i i ~ ~ ~ i b uto~ the c d i~npruvc~t~cnls cj'~'ccrcdnn t l ~ cDutclr mn:!. !. Irl I(' validnring n xiinplc t ~ : c l ~ ~ ) ~ l ri~odcl l ~ l l . kdluy ~ l c \ ~ c ' l q )Ibr ~ d the nangl;: ';:dl I II economy, I,:~ck~n;inand You (1979) c v a ! u ; ~ ~1d1 1 ~H C C U ~ K Y 01' lhe n u I e l by coinparing s;rn~plc-csti~nn~ion-peritxl l b f ~ ~ i iwirh : . ~ ~n m a l v a h c s . The mwsurcs of imessrnent I I included clnrts o r :~cruaha n d iorc~aslsvnl~rcs,arid TtMSl$ of G N P ; ~ n dI!:, conlpunznts. 7 111c cny>iric;~l~-cslrlrswcrv gcrlcrally ol~sr.rwd it: IIC I':~irly grv~tl. Il~rwcvcr, poor . pc~.l'un~i:~~lcc was nc&:ci for pricc level irlltl noniin:il ,rrlpo:.l (:\rgds. A I T ~ ~ ~[lie ~ I n~i sn L y nlncru-cconornic eco~imterric rr1ode:s of the hli~eri;~n economy, ~uusrlyh i l l by no~~-Nigcri:tn cconornic c :iperu, Ar1:lrnsun (1978a) !it-lcclcd LIIW lo c o n l p r c rhcir rcspccrivu forccils~ingpel-lbrmanccs wit!^ 11me ol)txi~uIfrom naive ~nc~hcxls adopted by 111c Nir_lerim Centr.;ll lll;rnning Ol'l'icl:: (CI'O). Obscrv;~tiulls O \ ~ C I - the. peri(~I01' slirdy (1977 - 1970) ic!..lj~:;lt~LILLtile ;iwr;LZc ~)CI.CCII[:I~Cci.ross of GNP, invesrnicl~r, corrsir~npriot~, exports and irlil:or!s rccorrfed 50.8, 50.6, 37.2, Y. 1 a1;11 44.9 lbr rnncr~c~-cconame~~~ic ~cchniqws(models), and 30.6, 27.2, 13.3, 417.9 nnil 21.1 for naive ~nerhfils,respectively . A~l;now!cdginglht superiority of the n,;\cro-.ccorrri!ne!ric ~iiodclsw e r naive i~lcrl~ods, Adri~nsun hoi~~cver, ~.n111 ioned ~ ht n 11ln.w:!:i,pl~istic;~red cslinlalc :!.; tw :!!I @ihcf~~ ~ . I A C ~ O ~ ' C O I I N~ IO~T~~~~:C ; ' ~ S!Jc~'~cc!; IS, ; ~ ~ s e ? ; s lit;.: td .. ; ! . i i i ~ I)y computing the frcq~~--:ncy w i ~ hwhich actual outcon-re k!I within rhe r:ingc of IWVlc',: forecasts - u:dIcd 'success raw'. ZTc. observcll that rllc s a c c i s I:OMU Forecasts of a!l rl~;ic~-oeconmuic variabl;:; f t w all ntcs DT rhe r;~ngi: I t c ~ r h m < r.c!a~ivc. to prtli~~iiiiaayi ~ n d revised dal:i wcrc 51; 2nd 49 p m m t s , rcspeclivcly. A .JI' ~Iir? ~OLII--W;LY comparisnn of' FOh4C's l i w c a s ~ s10 CEA, CI30 and privak scctor forccxts, rL-l:ualcri nlixcd 1.c-slilrs. prices, and low levels of eniploymcrir. The slump was roo great rm even $hi: most m!m 321belicva- ot' [he c:q;ir:~lislsystem ro wholly continuc ill .she Lxlirr or !he sysicm. Not orilg the experimcc of' Ihe great depression, but nlso thc work UP Keynes (1930s) w:~sI:igl~ly insin~rwn!al r f i the iwreasing role of government in ccimr)nlyI il11~1s ~ l ~ ~ c l q ~fo ~e~ul !hi: adoplion of econon~icplanning as a strategy of ~namigingtlx ccunrtmy. Keynes S:IW s r m inrtrrucntim "as a nicans of p c l'ccting ~ capilalisrn.". 111 his ~ L ' I I C[II.m~lIy~. Kuyncs foi-mul;tted rr 1ltuure1ic;rl t'ricmcivork in which he cilllccl for inlrncnse govenurroi!. in~ci,Scr~:ik:c in IIIC L ' W I I O I I ~ 111rotig.Ii ~ active moni:tary and fiscal policks. Eking an ;;iPvwaic 01' dci~wcl-sick cu.o~lo~ny.Kcynes ur:cs govurnmcnr to bc m i v e in consunq3iio1l. invos[nicnt, t'rnp!oyiricrx, aid r;tke all neccsx!ry actions tint are p;mmoItnr ill w:ril~\ai~-~ii~y, t ~ i~icrc;\sing r :v!wj~:w~ ~ : I I I : I I :IS I ~ il i s dcsircd. Kcynes"ssug!ptions h:ivc sirlcc lxcnme powct-till tools in grwxnmenr ccunomic policies I w R in Ihe dcvclopaf 3rd 1-crnovcd ~ l i c .wuakrlcss inliewnr ill 111o c;~pit:~lisl sywli. dcidoping cconumits. I ! ~ w ' r w , they Ii;ive n~ As wouIcl be recalled, Keynes Gene~.d'Elwry of'~'crsunly shorr run solutions. In thc ! ( j ~ run, ~ g his suggcstiori.; bccon~c inadquati: so 1ha1cconon~icpl;innirlg 113s cuu~t'Lo he ;~l'i'q)[ed, 2s a n ~ S S C R L ~imd ~ I ! 1:;votal inearls of guiclirr~2nd accelcra[ing cconornic growth cspcci:~llyin developkg countries. I-Ience, I ~ ra[iona!o E uf economic plnnr:irig as a dcvelopmenr tool r w : or1 the funclanwnr:!l uconmk i11rcl ins:itutiorlal argu,rxnts of inarkel failure, resource mobilization and al'lrnlton anlong o~llcrs.I-Ielfgoh and Sc11invo.Carnpo (1970, 11) in n Llnitcd Nations Ir~lusrria!Ikvelopnlrnt Organi~atiurl(UNIDO) pblici![ioll pl-ovidt. rhe- following ex;ilicit nliirkc~Liilul-c r:iliol~alt.tc~rscoiluinic pIarl~.ri~lg in developing counlries: Covcrnnxn~sc:innot, arid shou!d not, h k c ;I i n ~ ~ ~ .passivc cIy rolc in 1hc prwxss uf i~dw!ri;d e q x m i o ~ ~Pi.i~irti~lg . lwi \wxrnc ill1 essenlial and iiltugr;tl 3;n1 of iml~rs~rihl devrlopcucm progrnrimrs, for marker Sorczs by ~ h m s e l v c s cmrur , overcome rhc c1ti.p-sealed structural rigidities in the ecotwmies of developing cou~lrries... 'I'oday, l l x nved tbr sclrnc degror ol' c:cunomic planning is u~~ivcrsally rccugnized.. . In developing, coun~ries,platling is more ?'ca.;ihk ; r r d t)lrll.i' dcsirahlc ~ Tlu y,rc:t[cr ft:asil)ili\y i s a I io Ii t c v c l u l x c l ITSFI!! I ~ ; L I ' ~ c :w[ n ~ n i i r s . 01' [Itc s111:1\lerlri~i!~Ix-r v;irIh!cs riw 1nus1 bc v l k c n irrlo consiclcrnrion, ;lnJ clcsil.,?l)ilirys r m s I'l-oni thr: ~lli: grwrcr fael 1ha1 the autc~rnsricrncc1li1nist-mfbr lowci~ns.This is nchicved ~ l ~ m o g~t Ir K l b t ~ a s error t :lnalysis which r l h ~ t u d ~ ' ' i ~tod o ~ t ~ h cNigerian siru;l~ionwith n Gicw ro irnprming on the ciisiins lewls of lbrccns;ing following iwu swps: Firstly, [he mtim paint (1:) is o n ; l parallel shil't of the rqression liw (KL) to n!~c4 5 " di;igonal (LPF) ;IS mean bias, and reduces llw 1 m 1 n S C I ~ I ; I fbremsf ~C~ 11ic right unti! in Figure 2.1. T h i s climir~ales!Ile error MSfI, IC oU2iis in equation ( 2 . 5 ) . cnincidcs with 11-~i- I.PF ([hat is, 3 = 1). This Ii~rlhrr recluws rhe MSE, LO the scaikr depending on which variable is [reared as thc p c d i ~ l ~and r , wliich lhc pr~:dic~anci. where A, = change in actual I:, = change irl forecast, (M%) crircn-ior~which is a conlbina~ion af ur~bI;~,iudncss :lid m i n i l r ~ u l r l rnriance (1990; 223) says of [lie rni1iitnu1-n MSE crirerinn: and ux~crdcilupon by Mi~~ucr. and Zar~lowiu(1969) prnvidcs tlx b:~sis I'h. which !he (1992; 232) p t s i t , can kt: eslln~ntctl by [lie OIS rnclhod provir!cb the errors in t h i (1086) implies ~ h a cregressions in l c v ~ l s{or log-lewls) may hc of a sl!urims anzure. Tliis: rnenris t l r~cilhsr ~ ~ the parii~iwterc-alinwtes nor ~hcirs!a~~lar'J errors could be reslccl. I n I-laving pr.ocul.ed the pamnwer estilnares, the technique f'oa evaEuation involves of' 111csmcruritl rrdjusrn~enrprasramrncs (SAP). Simple tincar regressions ;I,: cfliicrcd covering 1970 - 95, 1970 - 85 arxl 1986 - 95 period-; for c.wh of [he rnncroccanornic 1986 - 95: A, = 0.79 + 0.5~1'~ - I*' = 0.71 F, = 6.76 (I) (-2.04) d, I-' - 1.01 = 0.23 4.13 I'crccnugc Growth R w of Ncr rloniesric Credit 1980 - 95: A; = 1.00 + O.l31", -1.-7 = 0.77 I:, = 8.01 dL = 1.19 I-' = 0.46 [NDC] statistics are d d i n c d anil expressed in sec~ion2.1 Tablc 4.1: 1'ar'ri:il I n q u a l i r y COL-fl'icizn~s :IS 1)ct'incd by Kou~wyiannis(19771, and I'indyck Rubinl'ctcl (I99 1). NDC IYHU-95 U.97 ILJ8?)-RS 1 - 3 3 0.74 0.42 U.31) 17.02 19.01 16.57 3.73 7 3 1 7 7 1.50 0 2 2 4.53 4 1 5 5 (1969) Ithey txpardcd and dircctlg cxretnled the Il'heil's fbrirmh~ianj~c displayed have a clmr t~r~rlcrstnnifing of the diffennrials of forwasr error. EO O h c r tllan the rcgrcssion of ~norwy supply actuals on L'orccas[s, all' ? of the rcgressions covering ihc whole pcriod under study appear ro be high. SAP period ?'s seen1 to be hcllcr !ha11 those of RED period. Tlhe t-ratios arc reasonably high for all IL. cases. According LO Kou~.soyian~iis (1977:97), a Iiigh ? (7') is believcd to l i ~ v t :a clear rncrit oilly when combincd wilh a high L-ratio( t h a ~is, low stmilard error). Hence, the turning poiut e i n m w c ~ ccommitrcd. 1'tler.c were 19 Lases of ~celeralioncrrurs. 4.2 Statistical Tests ol Signifkance '~'llis section involves ~esllng 11le sign;ficrtnce of uveraall regression and the valucs. III olliel- words, tl!e cocf'fi~ien[of' de~~'rinin;l1i0~1 lxlween acliialu and Sorcc;!us is nor s~g~lil'ici~nr (I-'= O), 4.3 Esm~iluliorlut' Algcbrl~icS i p s of Pr~rameterEsqinrates To :tcctlpr the outcomc of an estim;tlion exercise, the csrinra~csof h e p;zran'lett.rs are required ro be th~v~rc.tically nlcarlingf~~lll m d statislica1I.y satkf;;rciury, Though [he first order rusts st'sta~isricalcl.i[cria arc l-sscl~ti;~!, preference is still given 10 !he ecanomic 'a priori' criteri;~. In support of t h i s view, khis i s whal Koutsoyiannis (4977: 26) says: IL shoiild be rwreil rhar ilie 'ir i k s\atis~JC;lt criteria arc .jeco~daryonly to prioir' rfnomical crirl-ria. '1%~ esrrmates of 111cp;irari~eters should' bc rcjecred in .n_cmr;lliI' they happen ta have ~ l w'wrong' sign (or size) cvcri rhough rhe corrdnlion coefficienl is high, or [tic signi ficarir . Irr wlr cases, the pamrnetcrs, ll~iughslat islically Slwuld ehe estin~aresof alle paramctzrs rurn up with signs or sixes not conf~nniri~r tu economic: rheory, Kmtsc~yiannis (1977: 261 believes they r m y bc attributed to def'iciencics of tIrc empirici~ldata en~ployedfor .the estirna~ionof the model. Also, Rao and Miller (1977: 44) ~ h i n k h a t 'wrung signs are usually warfiings of irlcovecr definitions, spxifiurions or inwprcra~ions.' Nevertheless, Koursoyiannis (1977: 26) suggests i h a l n wrong sign es!ins:itc is acccpretl if and only if thcrc Is gotril reason lo bclicve !hat in rhe parrlcular instance, rhe principles of econollric theory do not hold. I'k hli~~cr-Zarnowitz hypo:hesis posn~lstesrhat a forecast is ~m'biasedand efficient c M n u c l in sec~torl3.3 are sratlrd ..,. . , .. ;IS fulfows: l'k signs of 'a' aml 'I,' arc expetred td be Fur nfl n i w rcgaessiun c1-lses, 'a' and 'b' are positive. This conforms wizfi ' a prior'? (bcc:rux of'hi3h w k i n u c of ahc. csrimates) for the evaluation 01' rlw srarislic:~lsign; l k i > ~ l c c c;~lcl~l;~rccl 411: 11113 a s s u ~ n p i o nof' I ~ o ~ - : ~ L I I L ) c o ~~~. S~I c TU ' IT~~~; ~cI ~I IC;ire C I . :Y ~ I ~IIuwcvcr, c!. il'd- < (1, u r cI, > PI,., her(: is scrious :turocorrclation (positive or wg;ttiv:, rq-xclivcly). Wilh samplz size (n) and K dcgress I:)[' I'rccclo~n,the rnbukired D W d-s~:~rislis limirz are .csr:Misllsrl rzr one or flue percent Ievd of significnnrrc. K i:; h c n ~ m b c ro!' 'I'ablc 4.6; 12esulrs of' Aulocor~clatior-iTest 'I'ablc. 4 . 0 indicates that the regression ecluarions are free of scrious autoco~.rr:larion. 4 . 4 . 2 Test Foi. I-lctcl.osct!dasii~ily , . 1 llc ;tss~rmprionof hornosc.~.~l;~siiciry says rhst [he puliahiliry distrihricrn d ~hc relied u p n for interprewion aild predictive purposes. *. 1 he empirical evidence emanating from [he ;inal.ysis as reucalrd in Table 4.8 is uscd to ev;ilu;~tetllc t~yplhesesof [his study. 1970-95 1970-85 1086-95 -0.06 -2.38' 2.40' 4.55' 4.34' 6. IS' -2.3 1' -2.26' -2.03 = 1 rejects h e joint null lcbcl of significance, 'I'hi:; Talrlc 4.8 s!~owsthar ~lleF-tario tcsr for a = O nrrd /3 hypolllcsis of' unbiascdmss and efficiency m a n s 1ha1 there arc sigriifificanl f'onu5r 211 five pcr'ccnr 0.500 0.500 0.363 errors in tflc intermctliatc furccasts ul~sorvcdi r ~ 111, 58 I rhz Nqwinn cconom) sioclc 1970. inrc~.~lwdiatc vasiablw arc Iri r n w c;i.;es riot In cllller words, observed forecast eciors r;f i h : th'-l biiwst rninirnim~rhat could occur. L-lestfor E(F) = 13(:2) inJicates that the expected n u n s of thr I!IC forecasts :?rc nut equal to i l i t e x p c c ~ dmeans of xrllals. That is, the nbwrved sys~crn;rtic. and/or unsystematic f w x a s t err-ur is parrly ascourrtcd Tor by h e rncm bj:rs. Ilowrvzr, rhe dcgrct. of unbl:mxlr~ess(as disclosed in TahIc .I .5) is relntively minor. Most of [he forecasts app::ar inefi'icient. Tclc regression sbpes (b), bias in h e FOI-ccas~crru:-s of the Nipxiim economy shown in 3s Pur d~fl'cn:t?tly, i!ct~~:\l x~.icsi10 1101 ci;~~sidcritbiy l'lu~ruart:;:s ~ n i ~ cash 111u Sor~.zasrserics. 1.irstlp. the sign t c ' s ~ [13,[s)1 inlp!ics [h:!~ I~ICI-c are a nurllbcr of cascs wl~isrethe ibrwasts ;-\reitmtcrestirnnted as often as they are overes~i!nn[ed. FIowcver, a systematic a1K1 rc-~;~rcdly 0bse:~ve3 property of forccasrs in t h i s s w t y is u~xlcrestimarechanges. I-Ience, tile prirrclpal bias is [hat of L~K' lendency !~rtdti*l:s~ilitarlon. 4-6 t)isctrssii,i\ 'i'lil: empirical cvidencc: from ~ h an;rlysis t uxprevxs tTwt tlx 'TIwil\ coeITicic.slts are low ([II ir~~plic:t~ic)ri uf this is < I ) for t ! l ~rhc ~ a r i a b l c u;!lu h i r l y gnuil. 211 iinequaliiy sels nT forecasts eniployccl in h e sruily. !'orccas~ing pcrl'o:rrl:mce The o r tlicsc monttiiry arxJ fiscal Novcrthclcss, 1!1esc f'lrrec~srsstill Z ~ v zsonx ~1e~rt.t: of rarional cxpcct:r~bnsin ;!nnlyzii~gmacroeconomic policy. According to Papell (19%: 151). 'pcoplc arc rorwrrrd looking, and their expectations of r k fu~urecan !w modc!cLl will1 some accuracy by rtssuming economic iluctuntions a n d r h ~ [hey t use lthcir i ~ ~ i i m ~ i i l r i ~ ,* fortmsr error is ~ririciil in rhc formation of accurate cxprcr~rionswlricli are csscnri;il i n any economic envirorunm. Tlic empirical findings iri rhc preceedi~~g srcfio11 may bz s~ilrecl a!nmarily [hat ~ I I C i11rerlnedi;trcn~ncrocconomil: forccasis of 111eNigerian economy are rclnrivcly biased, inefficiellt. 3. . Tablc 4.5 shows rhnr \tie ccrefil-iccirs of dcrerminmior~s(rZ) r:inge from 62-88 percent - .,.rJ, irnplymg r l u r forccnsrs explained as nluch as die r;lnge of the variations in acrual values of the ;rflLciecl variables. 'Illus, o11e would no^ expect a l# pcnuciil a' because of he existence or olllcr fi!ccor.s txononly whcrms rhr economy ncru:~llycxpmds wirhour any accorupmyiiig plallncd cconornic 1980s [lmugll the early 1990s (Ojn. 1945: 96). This period [lad an a v c n p GDP growrh rare o f 3.5 pescrlrr. Considcrisna, a CDI' growrh raw of' 1.2 pcrcerlr in 1387 againa 3.5 pcrcenr. ojie could think of [he i i ~ s ~ i ~ l lpcriocl ~ l i ~ y ;rs rxpansruwry (CBN, 1989). During the period jusr srated :tbove, [he actual ourconlcs (on rhe avera_sejoI' F:G = 9.2, MS = 32.6 and NDC = 37.5 exceeded forrcascs o f FF. MS nnd N K were undcrfitiorared rrspectivcly, by 5.2. 17.4 arid 26.6. 'l'llcse ~i~di.resrir~~arionu arc psistenr, a d as ruch rRt iorqoing analysis is in conl'ormiry wirh the rc;d 61 situation in the Nigerian economy. In agrccmcnr with the above ernpi~icalfinding, CI3N (1991: 14) rctnarked in A1111ualRcport am1 Staten~mtof Accou~ltsthat 'Overall, the potcnlial L'or impressive ccotiomic perforrnancc in 1991 was only partially realtlixd which could IK attributed partly to tlic cu~i~mnlafivc cffecl ~ f the ' continued (Xivcrgcntebetween actunt macroeconomic outcomcs am! Ilie targuts. espccialiy in respect of government fiscal opcrat ions, rnonctnry a g p g a r e s and tlic exchnngc rate'. The qualifying word 'cunirtlul;~tivc'cmptiasizc.; ~ l l epcrsiq~cntdiscrepnncim hctwecns tlic rolecast and actual valucs of' thc inlcrmcdiatc ~nacrrxcrmomicvariables obscrvcd k f v r e 1901 (CRN, 1990: 12). Fi~rthernmre,thc data presented in 1 1 1 ~appendix show lhesc diuergencca. IIaving diagnosed the forecast errors of Illc variables undcr study. we found the sIopc (variance) cornporienr of Ilx systcrriatic error to hc thc dominant and most problematic error ill thc i~lvestigntion[scc 'Thhks 4,2 and 4.3). 'I'bk is thc crror-type which denmnstrates thc inability of fluctualions in forecasts to [rack fluctuations in actual valucs. Put dif'ferently, the estiniatcd functions oT a formas~ingsplstcni do not track x t u a l trend. IIence, o n thc basis oT thc result of this diagnosis, we recomnlend that the cxisting forecasrisn~tcchniqucs in thc Nigcrian forecasting system i,Itile economy should he rcstmturcd in ordcl- to gct rid of tIic systematic error. In so doing, we shouId bc ahlc to improve on our forccasling accurac? or the Nipcriarl cconomy. and thereby elirni~iate some of the losses I-esulring iron1 uncertainty in the niacroecono~nic rnanagenxnt decision-ma king pmcss of t k economy. . poinu are ~iiissacd.'l'hc sludy onds w i t h thc discussion mcl irnplicnrion of the resul~s:rI1;lt . ilk -- instabili~yobserved in [he Nigerian economy since I:m 1980s thluugh thz carly 1930s Cwgg , J . V . ct a1 ( 1 %7), Mlrrl~~tnntiiul T I - ~ JCIII-VCS: T I ~ An Aid No. 1, Edinburgh, Olivcr Lf;r Boyd L,rd. lo F~JI-wcusriu~, Mnlmgr::ph iloopcr, J . and A . Zcllncr (I96 11, "The Error of Forecasr for kful tivariart Regression Models, " Lc~numerricn,29, 544-555. thc I'ltxlictivt: P~'rfon~lancc of Econrimclric Motlcls. " f~lkni~rrion~rl G m ~ u ~ t rReview, ic 15, 3G6-383. Hc)wrey. 17. el ill ( I1)74j."N:iolcs on 'l'csting L x k m a r ~ C. , J . You (1970), 'An F*oimne[ric hlodel of Bangladesh and Evaluatinn of rhe I-:we-year L)c.ll;clop~-~cr~~ I'Enn, " Economics of P l n m i n g , 15 ( 2 S: 3). 126141. I:ao, P , a n d R. I,. Millt.1. ( 197 I ) , Appiirci Erano~rterrics,USA, Wardsworth. Zuvl-kas, C . ( 1979), I:c'~vio~rlicDewlopmnr: An I t l ~ r c ~ h r f E a ~USA, r, M m n i thn I n r c r ~ n ~ ~ r i Press ~ n a l I-rtl. - (I-',) FUI LT.IS~ 5 I 7 70 Y .3 75 7.1 87 .1 5 Y I Y S 13.1 It1 b l j L 17 j 0.4 16.2 -- -- --