Would "Cold Turkey" Work in Turkey? Author(s): Oya Celasun, R. Gaston Gelos and Alessandro Prati Source: IMF Staff Papers, Vol. 51, No. 3 (2004), pp. 493-509 Published by: Palgrave Macmillan Journals on behalf of the International Monetary Fund Stable URL: http://www.jstor.org/stable/30035959 . Accessed: 28/06/2014 17:55 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

.

Palgrave Macmillan Journals and International Monetary Fund are collaborating with JSTOR to digitize, preserve and extend access to IMF Staff Papers.

http://www.jstor.org

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

IMFStaffPapers Vol. 51, No. 3 c 2004 International MonetaryFund

Would"ColdTurkey" Workin Turkey? OYACELASUN, and ALESSANDRO PRATI* R,GASTONGELOS, Wheninflationrates in a countryare persistentlyhigh, observersoftenbelieve that the inflationprocess has become "inertial,"posing an obstacleto disinflation.Using an innovativeapproach,we assess the empirical validity of this argumentfor the case of Turkey.Wefind that the currentdegree of inflationpersistence in Turkeyis lower than in Brazil and Uruguayprior to their successful stabilizationprograms. More significantly,expectationsof future inflation are more importantthan past inflation in shaping the inflationprocess, providing little evidence of "backwardlooking" behavior Using surveydata, we find that inflationexpectations,in turn, largely depend on the evolution offiscal variables. [JELE31, E52]

B

ringingdown inflationfrom persistentlyhigh levels while avoidingeconomic disruptionremainsone of the most importantchallenges for policymakersin many developing countries.The critical question in a disinflation attemptis how quickly inflationwill respondto a tighteningin monetarypolicy. If agents are predominantlyforwardlooking and the monetarytightening is credible, actual and expected inflation will quickly adjustto the new regime and the outputcosts will be small, even if disinflationis rapid-that is, if a "cold turkey"policy is adopted.1 *Oya Celasunis an Economistin the Research Department,R. GastonGelos is an Economistin the WesternHemisphereDepartment,and AlessandroPrati is a Deputy Division Chief in the Research Department;all are with the InternationalMonetaryFund.The authorswish to thankRichardClarida, MarkGriffiths,JuhaKiihkinen,RatnaSahay,AntonioSpilimbergo,CarlosV6gh,an anonymousreferee, andthe Editorfor helpfulcommentsanddiscussions,andthe ResearchDepartmentof the CentralBankof Turkeyfor useful suggestions. 'The literatureon stabilizationis too largeto be discussedhere.See Sargent(1982) for a studyof the andCalvoandV6gh (1999) for an extensivediscussionof inflationstaprocessof endinghyperinflations bilizationin developingcountries.Fischer,Sahay,andV6gh(2002) providean overviewof modernhyperandhigh inflations,includingdisinflationepisodes.Hamannand Prati(2002) studywhy manyinflation stabilizationssucceedonly temporarily.

493

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

Oya Celasun, R.Gaston Gelos, and Alessandro Prati By contrast,if agents'price-settingbehavioris backwardlooking because of wage and price indexation or adaptiveexpectations, disinflation will need to be more

to be ableto assess gradualto minimizeoutputlosses.Forthisreason,it is important the extentto whichinflationhas an "inertial" component.Methodological questionson how to addressthisproblemhavethereforereceivedsignificantattention in the academicliterature.2 Turkeyis an ideal case study for these issues, because it has experiencedper-

sistentlyhigh inflationsince the 1970s.In contrastto manyotherhigh-inflation countries,whichat somepointexperiencedhyperinflation, Turkey'sinflationhas neverexploded,with annualinflationrateshoveringaroundstablebutgradually increasingplateaus,rangingfrom40 percentto 120 percentoverthe last decade. Thispersistencehas led manyto believethatinflationis largelydrivenby inertia, withentrenched inflationexpectations attempt. posinganobstacleto anydisinflation International the Infact,thefailureof the2000 stabilization by programsupported by manyto the presenceof inflation MonetaryFund(IMF)has been attributed inertia.3 Wefindlittleempiricalsupportfor this argument.First,the univariatepropertiesof inflationdynamicsshowthatthe currentdegreeof inflationpersistence in Turkeyis lowerthanthatof BrazilandUruguaypriorto theirsuccessfulstabilizationprograms.Second, using a frameworksimilarto that in Gall and Gertler(1999), andemployingbothprice and surveydata,we find thatexpectations of futureinflationare more importantthan past inflationin shaping the inflationprocess. Third,an examinationof the determinantsof inflation expectationsshows thatexpectationsdependlargelyon the evolutionof fiscal variables. Contraryto a widelyheldview,thesefindingssuggestthatbackward-looking contractsdo not pose a seriousconstraintto the disinflationprocess.Indeed, authoritiescouldeffectivelycontrolthe speedof the disinflationby adoptingand maintainingpolicies that influenceexpectationsof futureinflation.Moreover, to a largeextentby forward-looking sinceinflationis determined behavior,output costsassociatedwitha rapiddisinflationprogramarelikelyto be relativelylow.A crediblefiscal consolidationis probablythe key to reducinginflation,because inflationexpectationswill declineonly if the publicperceivesthatthe need to monetizefiscaldeficitsor inflateawaythe debtstockhas cometo an end. On methodological grounds,ourpaperinnovatesby makingextensiveuse of we go surveydata.Whilesurveydatahavepreviouslybeenusedin the literature, to explainthe formationof expectations,explicitly one stepfurtherby attempting betweenfiscalandmonetaryfactors.4 distinguishing The mainaimof this paper,however,is to answera policyquestion;for this butpresentthe reason,we do notrestrictourselvesto a singleempiricalapproach, 2See, amongmanyothers,the discussionsin Chadha,Masson,andMeredith(1992), Celasun(2001), Dotsey (2002), andGall andGertler(1999). 3See the conferencereportof the 2001 conferenceof the NationalBureauof EconomicResearch (NBER)on Turkeyunderhttp://www.nber.org/crisis/turkey_report.htmil. 4Foran exampleof the use of surveydatain the estimationof the Phillipscurve,see Roberts(1995).

494

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

WOULD"COLDTURKEY" WORKINTURKEY?

results from various alternativemethods, hoping to assemble a body of convincing evidence supportingour conclusion. I. Macroeconomic

Setting in 1990-2003

High inflation and volatile economic growth have been defining features of the Turkisheconomy over the last few decades (Figures 1 and 2). Against a backgroundof low and declining inflation in most other countries,including many of Figure 1. 12-Month CPI Inflation, Turkey,1991-2003 (In percent) 140 120 100 80 60 40 20 0

Jan-91 Oct-91 Jul-92Jan-91

Jan-94 Oct-94 Jan-91 Jan-91 Jan-91 Jan-91 Jan-91 Jan-91 Jan-91 Jan-91 Jan-91 Jan-91 Jan-91

Figure 2. GDP Growth, Turkey,1991-2002 (Inpercent) 10 8

6 4 2 0 -2 -4 -6 -8 -10 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

495

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

Oya Celasun,R,GastonGelos,and AlessandroPrati the formerlychronic-inflation casesof LatinAmerica,Turkishinflationaveraged 80 percentper yearin 1991-99. While a numberof stabilizationattemptswere madethroughoutthe decade,for most of the 1990s disinflationwas not on the forefrontof the policyagenda.5 Turkeyenteredthe 1990switha newlyliberalizedcapitalaccount,strongcapitalflows,andexpansionary fiscalpoliciesfinancedto a significantextentby monetaryexpansion.Whenthe governmentattemptedto increasethe shareof direct monetaryfinancingto curbinterestpaymentsin early1994,theloss of confidence in theliraandthereversalof capitalinflowswereswift,anda majorcurrencycrisis ensued.An IMF-supported programstemmedthe crisis,butthe programwent off-trackin mid-1995.Nonetheless,capitalflows to emergingmarketsweresubstantialin 1995-97,andtheTurkisheconomyenjoyedstrongeconomicgrowthon thebackof short-term capitalinflows. by incomplete Althoughthe aftermathof the 1994 crisis was characterized structuralreformeffortsand the absenceof a crediblenominalanchorto coordinateinflationexpectations,progresswas madein some areas.The balance on the primaryfiscal accountaverageda surplusof 2.2 percentin 1994-97, as opposedto a deficit of 0.8 percentin 1990-93 (Figure3). CentralBank financingof the fiscal deficitswas graduallyreducedandeliminatedfrom 1997 onward.Nonetheless, the improvementin the primarybalance was partly reversedafter 1996, and the overalldeficit of the public sector-the primary deficitplusinterestpayments-remainedhigh.Thisreflectedalso thehighinterest bill due to the uncertaintyassociatedwith the pathof inflationandincomplete confidencein the fiscal policy adjustments.Monetarypolicy remained accommodating,subordinating any quest of inflationreductionto the goal of stabilizingthe financialmarketstowarda smoothfunctioningof the government domesticborrowingprocess.Inflation,as a result,ratchetedupto around100percent at end-1997. By the end of the decade,Turkey'sincreasinglyexceptionalstatusamong emergingmarketsin experiencinghigh and volatileinflationled to heightened perceptionsof the cost of thesehighinflationrates.Theresultingbroadenedconstabilizationpaved sensuson the desirabilityof disinflationandmacroeconomic the way for the disinflationeffortsstartingin 1998. Thegovernment programin early1998,taradopteda gradualiststabilization getingan inflationrateof 50 percentby end-1998.TheRussianfinancialcrisisin August1998led to severecapitaloutflowsanda massiveincreasein publicsecand torborrowingcosts,butthe monetaryprogramwas by andlargemaintained, inflationwas broughtdownto about70 percentat end-1998. a continuedcompressionof Theyear1999witnessedtwo severeearthquakes, foreigncapital,anda grossdomesticproduct(GDP)declineof 4.7 percent.Against a backgroundof unwieldyborrowingconditions,the governmentadoptedan ambitiousexchangerate-baseddisinflationprogramin December1999,adopting a crawlingpeg of the liraagainsta basketof 1 U.S. dollarand0.77 euro,andtargetinga consumerpriceindex (CPI)inflationrateof 25 percentby the end of SUnlessotherwisenoted,all dataarefromthe CentralBankof Turkey.

496

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

WOULD"COLDTURKEY" WORKINTURKEY?

Figure 3. Overall and Primary Deficit, Turkey,1990-2002 (Inpercent) 20 - PrimaryDeficit/GDP --Overall Deficit/GDP

15 10 5 0 -5 -10 1990

1990

1990

1990

1990

1990

1990

1990

1990

1990

1990

1990

1990

2002. A sizable fiscal retrenchmentwas achieved, but the exchange rate came under severe pressureas a result of a liquidity shortagein the banking sector in November 2000. The peg was broken in February2001 following renewed pressures in the banking sector and a massive outflow of foreign capital. Despite a reboundof inflation following the devaluationand the switch to a floating exchange rate regime in February2001, inflation was on a declining path in 2002-03, and the fiscal stabilizationwas sustained, underpinnedto a sizable extent by the long-postponedstructuralreforms.

II. WhatDo We KnowAbout Wage and Price Indexation? Although there is a perceptionthat a significantshareof wage and price contracts in Turkeyare indexed to past inflation, there is little comprehensivedocumentation of the natureof indexation in various sectors of the economy. The contract length in the small fractionof the privatesector covered by collective bargaining (about 350,000 workers)is typically two years, with six-monthbackwardindexation. The wages of the vast majorityof workersare negotiatedon a firm-by-firm basis, and little is known aboutthe contracts'nature,except thattheirlength is usually six months.The 2000 programagreedon with the IMF includeda shift in the inflation adjustmentof civil service sector wages and pensions from a backwardlooking mechanismto a forward-lookingadjustmentin line with projectedinflation rates,with catch-upclauses. Currently,salariesof civil servantsandblue-collar workersin the public sector are adjustedtwice a year. Little is documentedabout other indexation mechanisms in the economy. Interestingly,Shiller (1997) cites Turkeyas a puzzling example of a countrywith high and variableinflationwithout substantialindexing. Thereis some inertiain public sectorpriceadjustments,possibly contributingto inflationpersistence.The governmentbudgetuses a backward-lookingrevaluation 497

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

Oya Celasun, R.Gaston Gelos, and Alessandro Prati coefficientbased on the averageannualincreasein the wholesaleprice index (WPI) to determinemanytaxes anduserfees. Pricesof utilities,medicalsupplies,andmany food items are controlledby the government;the exact extent to which inflation adjustmentsin these areas are characterizedby systematicinertial(i.e., backwardlooking)behavioris not fully clear. The time-series behavior of CPI subindices suggests that, in importantnontradablesectors of the economy, prices tend to be adjustedwith lags of several month. We find that housing, health, and educationprices appearto be the most sluggish, while prices of food, clothing, andhousewaresare the most flexible.6The nontradablesector price increases seem to be set largely based on past inflation. The left-handpanels in Figure4 show, as we would expect in such a case, thatthe ratio of the index of slowly adjustingprices (housing, health, education) to the overall CPI increasesas inflationdeclines from end-1997 to end-2000, and diminishes when inflation acceleratesin 2001. The right-handpanels in Figure 4 show that the ratios of those price categories that are more flexible than the average move in the same directionas the inflationrate.

Analysis III.Persistence:A Univariate How importantis inflationpersistencein the aggregate?A univariateanalysisof inflationdynamicsindicatesthatinflationpersistencein Turkeyduring1994-2002 was smallin bothrelativeandabsoluteterms.Wefirstestimateda simpleregression of monthlyCPIinflationon its lag anda lineartrendfor the periodJanuary 2002.We thencomputedAndrews's(1993)medianunbiasedesti1994-February matorof the first-orderautoregressiveparameterandobtainedthe half-lifeestimates shown in Figure 5 by replicatingthis procedureon rolling samples of 60 observations.7The estimatedhalf-life of a unit shock to CPI inflation (the length of time needed to halve the magnitudeof the originalshock) is only aboutone month and has been relatively stable over time, remainingbetween 0.8 and 1.3 months over the last three years (top panel of Figure 5).8 CPI inflationin Turkeyis also less persistentthanin Uruguayand Brazil prior to their successful inflation stabilizations.Uruguay,like Turkey,had a history of high but relatively stable inflation before embarkingon a successful stabilization programat the beginning of the 1990s, after inflation reached a peak of 140 percent. The middle panel in Figure 5 shows that in Uruguay,the half-life of a unit shock priorto stabilizationwas similarto or marginallyhigherthanthatof Turkey in February2002. The case of Brazil is also of interestbecause it was a country with extensive backwardwage and price indexationbefore the 1994 stabilization. 6Foodpricescarrythe largestweightin the CPI(31 percent),followedby housing(26 percent),cloth(9.3 percent),housewares(9 percent),health(2.9 percent),entertainment ing (9.8 percent),transportation (2.8 percent),education(1.6 percent),andmiscellaneous(4.5 percent.) (2.9 percent),hotelsandrestaurants 7Stock(2002, p. 379-87), in a commenton Cogley andSargent(2002), notes,"Therearea varietyof estimationmethod ways to measurepersistence,noneperfect."He thengoes on to use a median-unbiased similarto the one employedhere. 8Wealso repeatedthe analysison quarterlydataandobtainedanalogousresults.For all these calculations,we modifieda programthatwas originallywrittenby AntonioSpilimbergo.

498

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

WOULD"COLDTURKEY" WORKINTURKEY?

Figure 4. Inflation Inertia and the Components of the CPI Index InertialComponents

Flexible Components 1.3

Ratio of s.a. HOUSINGprices to s.a. CPt (right axis) "

1.10 Ratio of s.a. FOOD prices to s.a. CPI (right axis)

1.2 1.1

100

1.05 1.00

,

0.95

100

1.0

80 0.90

80. 0.9

60

0.85

60 0.80 40.

40

Year-on-year CPI inflation(left axis)\

20

Year-on-year CPI inflation(left a

20 1995

1996

1997

1998

1999

2000

2001

1995

1996

1997

1998

1999

2000

2001

1.3 Ratio of s.a. HEALTHprice, to s.a. CPI (right axis) ,

1.2 "

1.1 Ratio of s.a. CLOTHINGprices to s.a. CPI (right axis)

100-

1.0

1.1 100-

1.0

80-

0.9

80-

0.9

60-

0.8

60-

40-

0.7

40-

Year-on-year CPI inflation(left axis)\

Year-on-yearCPI inflation(left axis)"

20-

20.

1995 1996 1997 1998 1999 2000 2001

1995 1996 1997 1998 1999 2000 2001 1.6

Ratio of s.a. EDUCATIONprices to s.a. CPI (right axis) t

1.00 ,Ratio of s.a. HOUSEWARE prices tos.a. CPI (right axis) ,

1.4 100-

0.95 0.90

1.2 100-

1.0

0.85

80-

0.80

80-

0.8

60-

0.75

600.70

40-

40-

Year-on-year CPI inflation(left axis)\

Year-on-year CPI inflation(left axis)

20-

20. 1995

1996

1997

1998

1999

2000

2001

1995

1996

1997

1998

1999

2000

2001

Note:The term"s.a."standsfor "seasonallyadjusted."

This is reflectedin half-lifeestimatesbetweenfour and seven months(bottom panelin Figure5).9 Althoughinflationpersistencein Turkeyis small, half-life estimatesare alwayssignificantlydifferentfromzero (top panelin Figure5), suggestingthat furtheranalysisis warranted.Univariatepersistencemay be due to backward9Thebottompanelof Figure5 does not show the upperboundof the confidenceintervalfor the halflife of a shock to inflationin Brazilbecausethe correspondingmedianunbiasedestimatorof the autoas regressiveparameterwas 1, suggestingthatthe BrazilianCPI inflationseriescould be nonstationary, shownin otherstudies(see, for example,Durevall,1999).

499

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

Oya Celasun, R.Gaston Gelos, and Alessandro Prati

Figure 5. Inflation Persistence in Turkey, Uruguay, and Brazil (Median unbiased half-lifeestimates of a unit shock and 90 percent confidence band) 3. 5 Turkey

3.0 2.5.

z

2.0. 1.5.

I

1.0. 0.5 0.0 1999

2000 2001 End of 5-year rollingsample

6 5 Uruguay 4

0 z

3 2

I

1 0 1991 1992 End of 5-year rollingsample

8 Brazil

7 6

z

5 4

I

3 2 1 1991 1992 End of 5-year rollingsample

1993

Note:The solid lines representthe medianunbiasedhalflife estimatesof a unitshock,andthe dottedlines representthe limitsof the90 percentconfidencebands.See footnote9 on the confidencebandsof the estimatesfor Brazil.

500

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

WOULD"COLDTURKEY" WORKINTURKEY? looking price setting,but it is also consistentwith forward-lookingbehavior.Serial correlationin the variables that drive inflation, such as exchange rate depreciation or money growth,would reconcile univariatepersistenceand forward-looking behavior. The next section assesses the relative importance of backward- and forward-lookingbehaviorby estimating a multivariateregression that nests both hypotheses.

IV.The RelativeImportance of BackwardVersusForward-Looking Behavior The distinctionbetween backward-and forward-lookingprice-settingbehaviorsis importantfroma policy pointof view becausethe outputcosts of a rapiddisinflation would tend to be higher with backward-lookingbehavior.10By contrast,forwardlooking price setterswould quickly take into accountthe implicationsfor inflation of a crediblechange in the monetaryregime, thus makingthe costs of disinflation relatively small. In this section, after describingthe empiricalmodel, we use two alternativemeasuresof inflation expectations-actual future inflation and survey inflation forecasts-to assess the degree of forward-lookingpricing behavior in Turkey.We find that,althoughpast inflationcontributesto explainingcurrentinflation dynamics,expectationsof futureinflationplay a much more importantrole.

EmpiricalModel We estimate an empiricalmodel that nests the possibility of both backward-and forward-lookingprice-settingbehavior.Traditionalmodels of price stickinesswith purely forward-lookingagents do not generate inflation persistence.The Calvo (1983) model can be modifiedby assumingthat only a fractionof the firms in the economy are forwardlooking, while the rest arebackwardlooking.11As in the original Calvo (1983) model, only a fractionof firmsareassumedto changepricesevery period, while the rest keep prices constant. Forward-lookingfirms have rational expectations,and backward-lookingprice settersuse a rule of thumb:they update the averagenew price in the most recent round of price adjustmentsby the most recentlyobservedinflationrate.The resultinginflationratein periodt, t, equals + 6E,t,,1 + t, = (1 - 8)7t,_1

x(V, -

P,),

(1)

where correspondsto the logarithmof the price a forward-lookingfirm would Vt select for periodt if it were able to reset its price in each period,and Pt is the logarithm of the aggregateprice level. The firms are assumed to be monopolistically competitive and therefore choose VTas the (logarithm) of a markupover nominal marginal costs, in deviation from steady state. The term VT- Pt then corresponds to the level of realmarginalcost andis often proxiedby the level of excess loSee, for example,the discussionsin Ball (1994), Celasun(2001), or BuiterandGrafe(2001). 11Similarspecificationshavebeen used by Obstfeld(1995), Gall andGertler(1999), Celasun(2001), andGhezzi(2001).

501

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

Oya Celasun, R.Gaston Gelos, and Alessandro Prati demand.12We extend Celasun's (2001) specificationfor the nontradablescomponent of the TurkishCPI to the overall CPI, and we include in the equationterms proxying for marginalcosts for tradablegoods as well as the excess demand for nontradablegoods and services. The degree of inflation inertiais governedby the parameter,1-6. The higher the share of backward-lookingprice setters,the larger is the weight 1-6 on the lagged inflationterm. Similarequationscan be motivated by other models that explain inflation persistence, such as Chadha,Masson, and Meredith (1992), Fuhrer and Moore (1995), Jadresic (2000), and Holden and Driscoll (2001).13

EstimateswithActual FutureInflation We first estimate equations for TurkishCPI inflation, proxying expected future inflation rates with actual futureinflation rates. Given that the measurementerror introducedby using such a proxy is not orthogonalto the actualfuturerealization of the inflation rate, the appropriateestimation technique involves using instrumentalvariablesthatare orthogonalto the expectationalerror.Moreover,the error term in this case follows an MA(1) process and is not necessarilyhomoscedastic, so we use the generalized method of moments (GMM) to take into account the structureof the errorterm. In additionto the lagged and futureinflationterms, we include terms proxying for the marginalcosts and the excess demandof the tradable and nontradablecomponentsof the CPI, respectively-the real wage (rwage) and the real exchange rate (rer)-and as a proxy for nontradablesexcess demand, the relativeprice of tradableswith respect to nontradablesin the CPI index (trntr) and the imports-to-GDPratio (imp).14, 15 We do not impose any restrictionson the coefficients of these terms, letting them be determinedby the data.16Our set of instrumentsinclude three lags of inflation and all the other variablesin the equation, three lags of the nominal interestrate, and a dummy variablefor the second quarterof 1994 to accountfor the effect of the April 1994 crisis. The constantterm in the equationis allowed to vary in the two subsamples, 1990: Ql1-1994: Q1 and 1994: Q2-1998: Q1. The equationestimatedfor 1990: Q1-1998: Q1 is as follows, with standarderrorsin parentheses:17 , = 0.38t,_,

+ 0.44 E1t1,+ (0.11**) rert + (0.09**) 0.34 rwage, + (0.17**) 0.62 trntr, + (0.05**) 0.39 imp,. + 0.62

(0.08**)

(2)

12Galfand Gertler(1999) arguethatproxiesfor marginalcosts shouldbe used insteadof an excess demand/output gapvariable.See also thediscussionin Celasun(2001).Wewill use bothapproachesbelow. Masson,andMeredith(1992) showthatwhen8 is greaterthan0.5, thereis a moneygrowth 13Chadha, ratethatkeepsthe outputgap at zero alongthe disinflationpath. excess demand. 14SeeCelasun(2001) on the motivationfor the measureused to proxynontradables 15Allvariablesare in logarithms,andexcept for the real wage, in deviationfroma lineartrend.For the logarithmof the realwage, the quadratictrendwas also statisticallysignificantandthereforewas used alongwith the linearone in detrending. 16Thepresenceof the realexchangeratetermaccountsfor the possibilitythata shareof firms,most likely in the tradablessector,"index"theirpricesto the currentandexpectedfuturedomesticlevel of foreign prices(i.e., the exchangeratemultipliedby foreignprices).We planto test in futureworkwhether manner. some pricesettersindexto the exchangeratein a backward-looking 17**Denotessignificanceat the 1 percentlevel.

502

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

WOULD"COLDTURKEY" WORKINTURKEY?

Figure6. Turkey:Backward-LookingBehavior in CPI (RecursiveGMMestimates of lagged inflationcoefficient and 95 percent confidence band) 0.6-

ERBS

stabilization

0.50.4C.)

0.30

0.20.1

a) 0.0 -0.1-

1998

1999

2000

2001

End of sample beginning in 1990:1 Note: The shadedareacoversthe periodof exchangeratebased stabilization(ERBS).

The weight on expected futureinflation is significant and positive but below one, and not statisticallydistinctfrom 0.5. All the four drivingvariablesalso enter the equationin a statisticallysignificantmanner,with the expected sign. Hansen's J-statistichas a p-value of 0.378, implying thatthe instrumentsand the model are valid. To examine whether the degree of inflation inertiachanged over time as a result of the recent disinflation efforts, we extend the sample by adding recursively one quarterof observationsat a time, through2001: Q4. Figure 6 presents the estimates of the backward-lookingcoefficient 1-6 for CPI inflation obtainedby graduallyextending the sample period.18The estimates show some evidence of inflation inertia,but the weight on lagged inflation falls from about 40 percent in the 1990: Q1-1998: Q1 sample to less than 15 percent in the full sample, suggesting that price setters'behaviorhas become increasingly forwardlooking. Possibly, this is the result of the disinflationstrategyfollowed in recent years, in particularthe switch to forward-lookingwage setting in the public sector mentionedabove.19

EstimateswithSurveyInflationForecasts We obtain broadly similar results when estimating equations in which survey CPI inflation expectationsfrom ConsensusForecasts are used in lieu of expected 18Theestimatedcoefficienton the relativepriceof tradableswithrespectto nontradables (trntr)is stafromzeroin samplesthatend after1999:Q3, butall othervariablesin theequatisticallyindistinguishable tion remainstatisticallysignificantin drivinginflation. 19Theresultsarebroadlyin line with findingsby Dibooglu(2001) basedon the GDP deflator.

503

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

Oya Celasun, R,Gaston Gelos, and Alessandro Prati inflation.20Surveydataallow us to departfromthe rationalexpectationsassumption made above and bypass the need to use the GMM. For Turkey,Consensus'sforecasts for averageinflationin the currentyear andthe following year are availableon a monthlybasis for the period 1995:1-1998:05 and bimonthlysince then. In addition, since 1998:05,Consensushas been providingexpectationsdatafor the six individual monthsahead,on a bimonthlybasis. We estimateregressionsfor seasonally adjustedmonthlyandbimonthlyinflationfor 1998:05-2002:02and semiannualand annualinflationfor 1995:01-2002:02.21Forthe inflationregressionsfor monthlyor bimonthlyinflation, Consensus data exist only for two years at a bimonthlyfrequency, reducing the numberof observations.For those regressionsthat assume either six-monthor one-yearcontracts,we have more than seven years of data at a As a measureof excess demand,we use a seasonallyadjusted monthlyfrequency.22 series of capacityutilization,which is availableon a monthlybasis. In most specifications, the estimates using survey forecast data reinforce the notion thatinflationpersistencedoes not have its main origin in backwardlooking behavior(Table 1). In the regressionswith forecasthorizonsof six monthsand one year,the coefficient on lagged inflationis statisticallyindistinguishablefrom zero. In the estimates with a forecast horizon of one and two months, the evidence is more mixed. However,these regressionshave small sample size and low explanatory power relative to those with six-monthly and annual inflation, casting doubt on the reliabilityof the estimates.23 tests provideanotherindicationthatprice settersare forward Granger-causality tests are essentiallytests of temporalprecedence.If price looking. Granger-causality settersweremainlybackwardlooking,thenchangesin inflationwouldGranger-cause changes in inflationexpectations,and not vice versa. However,the patternis the opposite:for all horizons,we cannotrejectthe hypothesisthatinflationexpectations Granger-causeactualinflation,butthe hypothesisthatactualinflationGranger-causes inflationexpectationscan be rejectedat the usualconfidencelevels (not shown).

V. WhatDetermines InflationExpectations? As inflationexpectationsappearto play a dominantrole in the price-settingmechanism, we conduct an empirical investigationof their determinants.We run two types of regressions to explain consensus forecasts of average inflation over the 20Weused the meanforecastof 15 differenteconomicforecasters,including13 Turkishand foreign andBusinessmen's privatefinancialinstitutions,IstanbulBilgi University,andthe TurkishIndustrialists' Association. 21Thespecificationremainsconstantacrosstheestimations.Theequationfor,say,n-monthlyinflation includesn-monthslagged(n-monthly)inflation,expectedn-monthaheadinflation,andthe n-monthaverage of capacityutilization,as well as a constantterm.Werestrictthe sumof the coefficientsof laggedand expectedfutureinflationto one, as in the GMMregressions. 22Six-monthand one-yearaverageinflationexpectationsare computedusing a weightedaverageof this year's and next year's expectedinflationrate. In the case of six-monthinflation,we assume that expectedinflationwas equalfor the whole 12-monthhorizon.Giventhatthesedataareavailableonly at a bimonthlyfrequencysince May 1998,we interpolatedatafor the missingmonths. 23Therecursiveestimatesof the coefficientof laggedinflation(not shown)arequitestablesince mid1997.

504

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

WOULD"COLDTURKEY" WORKINTURKEY?

Table 1. CPI Inflation Regressions with Survey Data (Newey-Weststandard errorsin parentheses) Forecast Horizon

One Month

One Month

Two Months

Two Six Six Months Months Months

Lagged inflation

0.529* (0.069)

0.655* (0.129)

0.527* (0.184)

0.868* (0.184)

0.016 -0.240 -0.128 -0.243 (0.166) (0.148) (0.010) (0.136)

Inflation forecast

0.471* (0.069)

0.345* (0.129)

0.473* (0.184)

0.132 (0.184)

0.984* 1.240* 1.128" 1.243*" (0.166) (0.148) (0.099) (0.136)

Capacity utilization (seasonally adjusted)

0.176" (0.046)

0.172" (0.046)

0.003 (0.145)

-0.067 (0.174)

-0.106 0.130 0.057 0.661 (0.305) (0.296) (0.679) (0.831)

Constant

-13.55* (3.330)

-12.831" (3.332)

-0.246 (10.759)

4.962 12.257 -5.144 10.445 -36.961 (13.23) (23.073) (22.25) (53.24) (61.725)

One Year

One Year

Estimation bimonthly bimonthly bimonthly bimonthly monthly monthly monthly monthly frequency Instrumental variables Estimation period Numberof observations R2

no

yes

no

yes

no

5:19982:2002

5:19982:2002

5:19982:2002

5:19982:2002

23

23

23

23

86

75

86

51

0.31

0.24

0.00

0.00

0.55

0.58

0.60

0.70

yes

no

yes

1:1995- 1:1995- 1:1995- 1:19952:2002 2:2002 2:2002 2:2002

Notes: *Denotessignificanceat the 5 percentconfidencelevel. The set of instrumental variables includethe constantterm,inflationforecast,andthe currentvalueandthreelags of the capacityutilizationrate.

following 12 months. In the first set of regressions(first two columns of Table 2), we try to explain inflationexpectationswith a set of standardvariables:monetary growth,the differencebetween the overnightrate and the deposit rate (to measure the stance of monetarypolicy), capacity utilization(to proxy for excess demand), lagged inflation, and exchange rate changes.24,25 In addition, to assess whether inflationexpectationsare drivenby expectationsof monetizationof fiscal deficits, we include the primarybalance and the change in the nominal debt stock (both as 24Wedidnotaddressthe separatequestionof whetherinflationexpectationsarerational.A quickexaminationsuggeststhat surveyforecastsare unbiased,but not efficient.The fact that laggedinflationhelps explainexpectationsis notinconsistentwiththeearlierresults,in whichexpectationsof futureinflationdrive currentinflation.It suggeststhata reduced-form equationcould expresscurrentinflationas a functionof costterm,andothervariablesexplainingexpectationformation. laggedinflation,anexcessdemand/marginal 25Wesubtractthe slow-movingdepositratefromthe overnightrateto controlfor the factthatthe same level of policyratescanreflecta differentstanceof monetarypolicy,dependingon thegenerallevel of nominalinterestrates.Includingtheexchangerateas anexplanatoryvariableis problematic,because,at leastin periodsin whichit is floating,theexchangerateis a jumpvariablethatcapturesinflationexpectations.The qualitativeresultspresentedbelow arenot affectedby the exclusionof the exchangeratevariable.

505

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

O

a

.< O

O

C-) O

CD

IV (1.617) (0.880) (0.260) 1.223" -0.841 -4.484* Specification

(78.752) Yes 0.7123

0.01 bimonthly

-5.775

III (0.693) (0.637) (0.112)

(50.895) No 0.8023

12-months;

0.982* -1.133 -2.687* Specification

0.00 bimonthly

49.213

following the over

II (0.196) (0.515) -

0.452 0.790* Specification Expectations inflation parentheses) in Inflation I average (0.212) CPI of errors (0.095) of Specification 0.486* 0.917"

forecast standard

Determinants Consensus 2. Newey-West mean Table

(0.006)(0.022) (0.093) (0.171) (31.455) (0.011) Yes

0.6982

0.00 monthly

0.017 0.030* -0.085* 0.758* -0.019 33.990

(17.575) (0.061) (0.006)(0.025) (0.165) (0.012) No 0.7982

0.00 monthly

0.005 0.022* -0.104* 0.629* -0.003 -44.127*

level. adjusted, GNP) of adjusted,

confidence

seasonally percent adjusted) percent (p-value) (as 5 seasonally GDP, (lagged) variable: (lagged) test the of inflation? at GDP, rate balance of seasonally change percent lagged (lagged) fiscal(inmonth) correlation deposit rate for percent (Dependent (lagged, one significance (as serial growth stock minus lagged M1 exchangevariables debt observations month) rate frequency *Denotes consolidated utilization (lagged) balance in of one Note: Capacity Change Year-on-year Instrumental Inflation Expected year-on-year, Primary lagged Overnight Year-on-year Constant Estimation R2 Number Breusch-Godfrey

O

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

WOULD"COLDTURKEY" WORKINTURKEY? ratios to GDP) as fiscal variables. In the second set of regressions (last two columns in Table2), we relateinflationexpectationsto expectationsaboutthe size of the fiscal deficit. As data on fiscal expectations are available only since May 1998 at a bimonthly frequency,the numberof observationsin this case is much smaller.Moreover,given that fiscal balanceexpectationsmight be endogenous (at a minimumbecause inflation expectationsaffect interestrates and, thus, expected fiscal balances), we also estimate this specificationwith instrumentalvariables.26 There is evidence that fiscal variablesplay an importantrole in determining inflation expectations.The results suggest that marketparticipantsclosely follow fiscal developmentswhen forming inflationexpectations.In particular,in the first two columns of Table 2, the primarybalance and changes in the debt (lagged one month) do have the expected sign (higher debt and deficits are associated with higher inflation expectations)and are significantat the 5 percentlevel, even after 28The two fiscal variablesalone explain controllingfor a broadset of variables.27, 43 percentof the variationin inflationexpectations(not shown). These resultsare confirmedby the second set of regressions(last two columns in Table 2), where the expectation about fiscal deficits significantly enters the regression, even when instrumented.Moreover, expected fiscal deficits remain significant in regressions explaining survey expectations even when we include expected M2 growth and expected GDP growth over the next 12 months as explanatoryvariables (not shown). The latter finding suggests that fiscal deficits affect inflation expectations, and not only because higher fiscal deficits increase expected money growth over a 12-monthhorizon. In addition, they increase the probabilityof monetizationover a longer horizon,therebyaugmenting,with a sufficiently forward-lookingmoney demand,expected inflation also in the shortrun. VI. Conclusion Backward-lookingbehavior does not seem to pose a major obstacle to reducing inflation in Turkey.Instead,credible fiscal consolidationappearsto be crucial for lasting disinflation. Inflation expectations are likely to decline only if the public perceives that the temptationto monetize fiscal deficits or inflate away the debt stock has come to an end. The fact that the improvementsin the noninterestfiscal balance in 2000-01 have coincided with a markeddrop in inflation expectations lends supportto this view, although the lack of demand pressuresfollowing the February2001 crisis is also likely to have contributedto the disinflation and the downward adjustment of inflation expectations. Concurrently,the authorities could consider policies aimed at eliminating residual forms of backwardindexation in wages and prices (such as housing rents and health and educationprices), 26Weuse pastchangesin the debtstock as instrumentsfor the forecastof the fiscal balance. 27Severalstudieshave analyzedeconometricallythe relationshipbetweenpublic sectordeficits and inflationin Turkey(see, for example,Metin, 1998). See also Lim and Papi (1997) andAlper and Ucer (1998). 280ne mightwonderwhetherthe laggedinflationtermis the regressorthatexplainsmostof the variationin inflationexpectations.This does not seem to be the case. For example,in specificationI, the R2 excludinglaggedinflationis 0.55.

507

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

Oya Celasun, R,Gaston Gelos, and Alessandro Prati which are likely to account for the remainingrole of past inflation found in the econometricanalysis. The approachdeveloped here has a broaderapplicabilitybeyond the specific Turkishcase. In particular,we believe that the use of survey data to assess the importance of backward-lookingbehavior in shaping the inflation process and the examination of the determinantsof these expectations are useful tools to inform policymakers who face the task of reducing inflation from persistently high levels.

REFERENCES Alper,C. Emre,and MuratUcer, 1998, "SomeObservationson TurkishInflation:A Random WalkDown the Past Decade,"paperpresentedat the conferenceon "Stabilizationin an EmergingMarket:The Case of Turkey,"Istanbul,BogaziciUniversity,January. Estimationof First-Order Andrews,Donald,1993, "ExactlyMedian-Unbiased Autoregressive/ Unit Root Models,"Econometrica,Vol. 61, No. 1, pp. 139-65. Ball, Laurence,1994, "WhatDeterminesthe Sacrifice Ratio?"in MonetaryPolicy, ed. by G. Mankiw(Chicago:Universityof ChicagoPress). Buiter,Willem H., and ClemensGrafe,2001, "No Pain, No Gain?The SimpleAnalyticsof EfficientDisinflationin Open Economies,"CEPRDiscussionPaperNo. 3038 (London: Centrefor EconomicPolicy Research). Journalof Calvo, Guillermo,1983, "StaggeredPrices in a Utility MaximizingFramework," MonetaryEconomics,Vol. 12, No. 3, pp. 383-98. and Carlos V6gh, 1999, "InflationStabilizationand BOP Crises in Developing , Countries,"in Handbookof Macroeconomics,ed. by JohnTaylorand MichaelWoodford (Amsterdam:North-Holland), pp. 1531-614. Celasun,Oya, 2001, "PriceSettingandInflationDynamicsin SmallOpenEconomies,"(Ph.D. dissertation;CollegePark,Maryland:Universityof Maryland). Chadha,Bankim,Paul Masson,and Guy Meredith,1992, "Modelsof Inflationand Costs of Disinflation,"StaffPapers,International MonetaryFund,Vol. 39 (June),pp. 395-431. and Thomas J. 2002, Sargent, "EvolvingPost-WorldWarII U.S. Inflation Cogley,Timothy, Dynamics,"in NBER MacroeconomicsAnnual 2001 (Cambridge,Massachusetts:MIT Press). Dibooglu, Selahattin,2001, "InflationaryExpectationsand Costs of Disinflation:A Case for Costless Disinflation in Turkey?"in Inflation and Disinflation in Turkey,ed. by A. Kibritcioglu,L. Rittenberg,and F. Selcuk (Aldershot,England:Ashgate). Pricingin New Dotsey, Michael, 2002, "Pitfallsin InterpretingTests of Backward-Looking Vol. 88 Federal Reserve Bank of Economic Richmond, Quarterly, KeynesianModels," (Winter),pp. 37-50. Durevall,Dick, 1999, "InertialInflation,Indexation,and Price Stickiness:Evidence from Brazil,"Journalof DevelopmentEconomics,Vol. 60, No. 2, pp. 407-21. Fischer,Stanley,RatnaSahay,and CarlosV6gh, 2002, "ModernHyper-and High Inflations," Journalof EconomicLiterature,Vol. 40 (September),pp. 837-80. Fuhrer,Jeffrey, and Gerald Moore, 1995, "InflationPersistence,"QuarterlyJournal of Economics,Vol. 110, No. 1, pp. 127-60. Gall,Jordi,andMarkGertler,1999, "InflationDynamics:A StructuralEconometricAnalysis," Journalof MonetaryEconomics,Vol. 44 (October),pp. 195-222.

508

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions

WOULD"COLDTURKEY" WORKINTURKEY?

Indexation,Credibility,and InflationPersistence," Ghezzi, Piero, 2001, "Backward-Looking Journalof InternationalEconomics,Vol. 53, No. 1, pp. 127--47. Hamann,A. Javier,and AlessandroPrati, 2002, "Why Do Many DisinflationsFail?"IMF WorkingPaper02/228 (Washington:International MonetaryFund). and John C. "A Note on InflationPersistence,"NBERWorking Holden,Steinar, Driscoll,2001, No. 8690 Massachusetts: National Bureauof EconomicResearch). (Cambridge, Paper Jadresic,Esteban,2000, "CanStaggeredPriceSettingExplainShort-RunInflationDynamics?" (unpublished; Washington:International MonetaryFund). Lim, ChengHoon, and LauraPapi, 1997, "AnEconometricAnalysisof the Determinantsof Inflationin Turkey," IMFWorkingPaper97/170(Washington: International MonetaryFund). Metin,Kivilcim,1998, "TheRelationshipBetweenInflationandthe BudgetDeficitin Turkey," Journalof Businessand EconomicStatistics,Vol. 16, No. 4, pp. 412-22. Obstfeld, Maurice, 1995, "InternationalCurrencyExperience:New Lessons and Lessons Relearned," BrookingsPaperson EconomicActivity:1, BrookingsInstitution,pp. 119-96. John Roberts, M., 1995, "New KeynesianEconomics and the Phillips Curve,"Journal of Money,Creditand Banking,Vol. 27, No. 4, pp. 975-84. Sargent,Thomas J., 1982, "The Ends of Four Big Inflations,"in Inflation: Causes and Consequences,ed. by RobertE. Hall (Chicago:Universityof ChicagoPress),pp. 41-97. Shiller, Robert, 1997, "PublicResistanceto Indexation:A Puzzle,"BrookingsPapers on EconomicActivity:1, BrookingsInstitution,pp.159-228. Stock,JamesH., 2002, "Discussionof CogleyandSargent's'EvolvingPost-WorldWarII U.S. InflationDynamics,'" in NBERMacroeconomics Annual2001 (Cambridge, Massachusetts: MITPress),pp. 379-87.

509

This content downloaded from 193.0.146.141 on Sat, 28 Jun 2014 17:55:43 PM All use subject to JSTOR Terms and Conditions