DEMANDANDSUPPLYOFREALESTATEMARKETINTURKEY:

ACOINTEGRATIONANALYSIS

AMastersThesis

by

ZEYNEPBURCUBULUT

Departmentof

BilkentUniversity

Ankara

January2009

ToMyHusbandandMyFamily

DEMANDANDSUPPLYOFREALESTATEMARKETINTURKEY:

ACOINTEGRATIONANALYSIS

TheInstituteofEconomicsandSocialSciencesof

BilkentUniversity

by

ZEYNEPBURCUBULUT

InPartialFulfilmentoftheRequirementsfortheDegreeofMASTEROFARTS

in

THEDEPARTMENTOFECONOMICSBLKENTUNIVERSITY

ANKARA

January1999

I certify that I have read this thesis and have found that it is fullyadequate,inscopeandinquality,asathesisforthedegreeofMasterofArtsinEconomics.

Assoc.Prof.alaktenSupervisor

I certify that I have read this thesis and have found that it is fullyadequate,inscopeandinquality,asathesisforthedegreeofMasterofArtsinEconomics.

Asst.Prof.mitzlaleExaminingCommitteeMember

I certify that I have read this thesis and have found that it is fullyadequate,inscopeandinquality,asathesisforthedegreeofMasterofArtsinEconomics.

Assoc.Prof.ZeynepnderExaminingCommitteeMember

ApprovaloftheInstituteofEconomicsandSocialSciencesProf.ErdalErelDirector

iii

ABSTRACT

DEMANDANDSUPPLYOFREALESTATEMARKETIN

TURKEY:

ACOINTEGRATIONANALYSIS

Bulut,ZeynepBurcu

M.A.,DepartmentofEconomics

Supervisor:Assoc.Prof.alakten

January2009

Since inacountrythehousingmarket isa leading indicatorfor

the whole economy, the determinants, that are affecting aggregate

housingsupplyanddemand,arewidelysearched.Inthisstudy,wetry

tofindthevariableswhichareaffectingthedemandandsupplyofreal

estatemarket inTurkeybetween theyears 1970 to 2007.We cannot

specialize on the housing market and rather study the real estate

market in the aggregatenumber of dwellings is our quantity

measuredue todata limitations.WechoseTopelandRosens (1988)

demandandsupplymodelsthatarebasicallybasedondifferentshort

andlongrunelasticity.Asdemandsideindependentvariables,interest

rate,valuevariable, incomeandpopulationarechosenandassupply

side independent variables, value, interest rate and costs are chosen.

iv

Value isusedasaproxysince themarketpricedatadoesnotexist in

Turkey.Value isakindofcost that is taken from thebuilderwithout

interested inwhat thematerialsareandhowmuch the laborcosts to

the builder. Also, the annual data is used because of the data

limitations.Due to the fact that all these variables are I(1), Johansen

Cointegration and VECM are preferred. According to the empirical

findings, the signs of all the variables are as expected and are

significantinthelongrun.However,intheshortrun,onlyinterestrate

andcostvariablesaresignificantin90%confidencelevel.Furthermore,

thepriceelasticityofsupplyis1.5inthelongrunwhileitis0.13inthe

shortrun. This shows us that the adjustment costs for a change in

Turkey issignificantlyhigh.Moreover, the longrunpriceelasticityof

demandis4.97.

Keywords: Housing supply, housing demand, cointegration,

vectorerrorcorrection

v

ZET

TRKYEDEGAYRMENKULPYASASIARZVETALEP

DENGES:

EBTNLEMEANALZ

Bulut,ZeynepBurcu

YksekLisans,ktisatBlm

TezYneticisi:DoDr.alakten

Ocak2009

Bir lkede konut piyasas, genel ekonomi asndan gsterge

nitelii tadndan dolay, konut piyasas toplam arz ve talep

bileenleriyaygnbirekildearatrlmtr.Bualmada,1970ve2007

yllar arasnda Trkiye gayrimenkul piyasas toplam arz ve talebi

oluturan deikenler bulunmaya allmtr. Veri eksikliinden

dolay, zel olarak konut piyasas incelenememiti. Birbirinden farkl

uzun ve ksadnem fiyat elastikiyetleri esasnadayal olanTopel ve

Rosen(1988)konutarzvetalepmodelitercihedilmitir.Konutpiyasas

yaznnda ska kullanlan talep/arz deikenleri esas alnarak, bu

almada telep deikenleri olarak, nfus, faiz oran, gelir ve deer

deikenleri;arzdengedeikenleriolarak,deer,faizoranvemaliyet

endeksi deikenleri kullanlmtr. Deer datas, Trkiyede evlerin

piyasafiyatlarbulunmadndandolay,fiyatdeikeninevekilolarak

vi

kullanlmtr.Ayrca, bina says verisi yllk olarak toplanmasndan

dolay, bu alma yllk veri ile gerekletirilmitir. Btn

deikenlerin birinci farklar duraan olduundan dolay, Johansen

EbtnlemeveHataDzeltmeModelitercihedilmitir.Bualmann

ampirik sonularna gore, uzun dnemde sz konusu arz/talep

deikenlerianlamlkmtrvebeklenen iaretlergrlmtr.Buna

karn, ksa dnemde faiz oranlar ve maliyet deikenleri dnda

btndeikenler%90 gven seviyesinde anlamsz kmtr.Ayrca,

uzundnem arz fiyat esneklii 1.50 olarak karken, ksadnem arz

fiyatesneklii0.13olarakkmtr.Szkonusuesnekliksaylarbize,

konut piyasasnda olan bir deiikliin ksa dnemde gerekleme

maliyetininokyksekolduunugstermektedir.Ayrca,uzundnem

talepfiyatesneklii4.97olarakkmtr.

Anahtar kelimeler: Konut Talebi, konut arz, Ebtnleme,

VektrHataDzeltme

vii

ACKNOWLEDGEMENTS

Ifeelmostfortunatetohavebeenguidedandsupervisedbymyadvisor,Assoc.Prof.alaktenandwouldliketoexpressmydeepestgratitude to her for her valuable recommendations, patience andguidancewhichhelpedmefinishthisstudy.IwouldalsoliketothankAsst. Prof. mit zlale and Assoc. Prof. Zeynep nder for theirvaluable critique and comments on my thesis. Without theirsuggestions, I would not have been able to improve the academicqualityofmythesis.

My thanks shouldgoalso tomyhusband,myparentsandmybrotherfortheircontinuoussupport,encouragementandmotivationinthereallyhardtimesIlivedthrough.

IamgratefultomyfriendsatBilkentUniversityfortheirusefulcomments,moral support and close friendship.Without their help, Iwouldneverbeabletocompletethisstudy.

I alsowant to thank toVakfbank for the support duringmygraduate study. Especially I owe thanks to my colleagues and mymanager,CemErolu,inEconomicResearchDepartmentatVakfbankfor theirgreatunderstandingandenforcingme to finishmygraduatestudy.

My thanks also go to Tbitak for their financial support.

viii

TABLEOFCONTENTS

ABSTRACT... iii

ZET v

ACKNOWLEDGEMENTS... vii

TABLEOFCONTENTS. viii

LISTOFTABLES x

LISTOFFIGURES xii

CHAPTER1:INTRODUCTION.........

1

CHAPTER2:LITERATUREREVIEW...........................................

5

CHAPTER3:HOUSINGINVESTMENTTHEORY...................

10

3.1HousingSupply.. 12

3.2HousingDemand 19

3.3ImplicationsofTheory 25

CHAPTER4:HOUSINGMARKETINTURKEY..

29

ix

CHAPTER5:ECONOMETRICMETHODOLOGYANDDATA.......................................................................

35

5.1Methodology..... 35

5.1.1PhillipsPerronUnitRootTest.. 36

5.1.2JohansenCointegrationTest......... 38

5.1.3VectorErrorCorrectionModel 42

5.2Data........

44

5.3EconometricModel.. 48

CHAPTER6:ESTIMATIONRESULTS...

52

6.1EmpiricalResultsofHousingSupplyandDemand..

53

6.1.1LevelDataAnalysis. 54

6.1.2LogarithmicFormAnalysis 63

6.2LimitationsofResults....

67

CHAPTER7:CONCLUSION......................................................... 69

BIBLIOGRAPHY............................................................................ 72

APPENDICES

A.TURKEYBUILDINGCOSTINDEX.................................. 78

B.DESCRIPTIVESTATISTICS................................................ 80

C.TABLESOFESTIMATIONRESULTS............................... 81

x

LISTOFTABLES

1. ExpectedSignsinDemandandSupply.....................................

51

2.ExponentialRegressionResult......................................................

79

3.DescriptiveStatististicsofRealLevelData................................. 80

4.DescriptiveStatisticsofLogarithmicData..................................

80

5.PhillipsPerronUnitRootTestStatisticResults1....................

81

6.ResultsofPhillipsPerronUnitRootTestStatistics2..............

82

7.TestsoftheCointegrationRankforTurkeyCostIndex...........

82

8.Chisquare(2)statisticsfortherestrictionsunderHo:restrictionsareappropriateTurkeyCostIndex................

83

9.LongRunEquilibriumResults1..................................................

83

10.VectorErrorCorrectionResults1................................................

84

11.TestsoftheCointegrationRank2.................................................

85

12.Chisquare(2)statisticsfortherestrictionsunderHo:restrictionsareappropriate2...............................................

85

13.LongRunEquilibriumResults2.................................................

86

14.VectorErrorCorrectionResults2................................................

87

xi

15.TestsoftheCointegrationRank3..............................................

88

16.Chisquare(2)statisticsfortherestrictionsunderHo:restrictionsareappropriate3............................................

88

17.LongRunEquilibrium3.............................................................. 89

18.VectorErrorCorrectionResults3.............................................. 90

xii

LISTOFFIGURES

1.ShortrunEquilibrium.............................................................

11

2.LongrunEquilibrium.............................................................

11

3.TheShareofHousingInvestmentinGrossFixedInvestments1998CurrentPrices.........................................

32

4.WholeBuildingCostIndex(19912007)andIstanbulConstructionMaterialsIndex(19702007)................

78

1

CHAPTER1

INTRODUCTION

Thehousingmarketisdifferentfrommostoftheothermarkets

goodsandservices.Onereasonforthisisthedualfunction;itisbotha

commodity by yielding a flow of consumer services and also an

investmentassetbybeinga largeportionofhouseholdnetworth.So,

alltheanalysisofthehousingmarketincludesbothproperties.Dueto

not only including these properties but also having different other

features, theanalysisof thehousing is furthercomplicated.According

to Palmquist (1983), the housing market is a kind of differentiated

productduetotheheterogeneousstructure,i.e.ithasastructurebased

on the characteristics of houses like the structures of house or the

location.Also,accordingtoQuigley(1992),therearefourbasicfeatures

that differentiate housing from other goods and services. These are,

high cost of supply because it takes long time to build, durability,

2

heterogeneity no two houses are identical in every respect and

location fixity. These features of housing, in particular its durability,

heterogeneity and location fixity together imply that the housing

marketisacollectionofconnectedbutsegmentedmarkets.

Accordingtotherealestatefinanciersandeconomists,becauseof

the relationbetween themacroeconomicvariables andhousing such

as, the relation between employment and housing construction

housing investment,madebyboth thebuildersand theconsumers in

ordertoincreasetheirworth,isaleadingindicatorofeconomicactivity

(Smith and Tesarek 1991;Wheeler andChowdhury 1993).Holly and

Jones(1997)alsoagreewiththisopinion;duetothefactthathousingis

anelementofpersonalwealth,itsoperationmaybesignificantlylinked

toeconomicconditionsofthatcountry.Theincreasedindemandinreal

estatemarketresultsincapitalgainininvestmentforrealestate.Inthis

environment, households observe two effects depending onwhether

they are the owners of real estate orplanning to acquire one. In the

former group, the rise in asset prices alongwith the decline in the

interestratesasaresultofcontinuinggoodeconomicenvironmentlead

to the so calledwealtheffect.Apositive shock tohouseholds total

wealthleadstoanincreaseintheircurrentandfutureconsumption.In

3

thelattergroup,wherehouseholdsareonthebuyersideofthemarket,

thedecline in interest ratesgenerates an income effect thatmotivates

households to purchase houseswhereas the increase in house prices

leads them to substitute away. The resultant impact depends on

whichever force is greater. (Binay and Salman, 2008) These types of

effects bring about the housing market to be too important and

interesting.

Inaddition,governmentpolicycanhaveaprofound impacton

the operation of the housing market. The vouchers or subsidies to

homeowners in the form of themortgage interestdeduction increase

demandforhousingservices.Thelongrunimpactonpricedependson

the supply response determined by the price elasticity of supply.

Government policy has also impacted the supply side of themarket

directly through the construction of public housing and tax policy

designed toencourage theprivateconstructionofnewhousing.These

interventionsraisean importantpolicyquestionconcerning theextent

towhich thesepolicies result innetadditions to thehousing stockor

simplycrowdoutprivateactivity.

4

Economistshaveusedthefactthatthehousingpriceisanatural

outcomeof thedemand forhousing,equatingwith itssupply.So, the

demand and supply for housing interact to determine the price of

housingrelativetoothergoodsandservices.Basedonthisfact,which

basicallydependson the idea that theprice is formedby supplyand

demandmarketmakers simultaneously, I try to estimate the supply

anddemandequationsforthehousingmarketactivity.Inthefirstpart

of this study, Iwill give some information about the literature about

housingmarketstudies.Inthesecondpart,Iwill introduceTopeland

Rosenhousing investment theory that isconsistentwithmyempirical

researchandwiththestructureoftheTurkishhousingmarket.Then,I

will briefly explain the housingmarket structure in Turkey and the

studiesaboutTurkishhousingmarket.Inthefourthpartofmystudy,I

willexplainmymethodselectionfortheestimationaswellasthedata

and theory underlying the estimationmethod with the econometric

model.Inthelastpart,theestimationresultswillbedisplayed.

5

CHAPTER2

LITERATUREREVIEW

In the literature, while modeling housing market, various

methods are used. Poterba (1984) takes an assetmarket approach to

modeling the housing market. His model of the housing market

examines the impact of a shock to the steady state,mapping out the

adjustmentprocesstoanewsteadystate.Ashocksuchasadeclinein

usercostresults initially inan increase in realhousingpricesince the

housing stock is fixed. Themarket then adjustswith growth in the

housingstockandadeclineinrealpricetoanewsteadystate.

Urban spatial theory, which provides equilibrium models in

which the stock of housing always equals the urban population, is

anotherwayofmodelinghousingmarket.Inthesemodels,there isno

6

supply theorydealingwithconstruction flows sincenewconstruction

or the flow of housing simply equals the growth in population.

DipasqualeandWheaton (1994)use this theoryeffectively inorder to

disproveoneoftheassumptionsaboutthehousingmarketwhichtells

housingmarketclearsquickly.Theyquestion thisbyusingstockflow

approachandshowthehousingmarketsinabilitytorapidlyclear,and

alsoshowtheinefficiencyofhousingmarket.Inordertogetridofthe

problemofslowmarketclearing,theyusepriceadjustmentmechanism

andannex it todemandsupplyequations.Theyestimate theirmodels

by using two quite different approaches in the way of forming

consumersexpectationsabout futurehouseprices,and they find that

the gradual price adjustment statistically holds strongly both when

consumersdevelopexpectationsby lookingbackwardathistoricprice

movementsandwhenhousingdemandisbaseduponrationalforward

lookingforecasts.Moreover,theyuselandfactor,whichdependsonthe

stock of housing not the level of building activity, in defining the

supplyequationofthehousingmarket.

SomeresearcherssuchasPalmquist(1983)thinkthathousingisa

goodexampleforadifferentiatedproduct.So,Palmquistestimatesthe

demand for the characteristic of housing by using hedonic demand

7

theory. He chooses this because previous studies about hedonic

regression could not find any weakness of this theory and also

nonlinear hedonic equation with the data of seven standard

metropolitan areas provides elimination of identification and

endogeonity ofmarginal prices problems. In his paper, he assumed

there is nomarket segmentationwithin an urban area since there is

mobilityamonghousingtypesandlocationsandlittleevidenceofprice

discrimination.Alsoheassumes thatdifferences in consumerswithin

andbetweencitiesaremeasurableandcanbecontrolled.

UnlikePalmquist(1983),Reichert(1990)thinksthattherearebig

differences in housing demand or supply between regionswithin a

country. So his research is based on effects of somemacroeconomic

variablesuponregionalhousingpricesbyconstructingaregionspecific

housingsupplyanddemandfunctionofUnitedStates.

Topel and Rosen (1988) examine the extent towhich housing

investmentdecisionsaredeterminedbycomparingcurrentassetprices

withcurrentmarginalcostsofproduction.Theyarguethatcurrentasset

prices are sufficient statistics forhousing investment if shortrun and

longrun investment supplies are the same. If changes in the level of

8

constructionactivity impactthecostofproduction,thensupply is less

elastic in the short run than the long run. This divergence between

shortterm and longterm elasticity indicates that current asset prices

are not sufficient and buildersmust form expectations about future

pricesinordertomakeinvestmentdecisions.

Besidesthesetheoreticalstudiesabouthousingmarket,thereisa

huge literature based on empirical analysis of housingmarket in the

countrylevelinthelightoftheseabovetheories.

SincethehousingmarketofUnitedStates isthemostadvanced

one in theworld, there is somuch empirical analysis about housing

marketaboutthewholecountryaswellasaboutwithinthecountry.

The housing supply and housing demand studies will be

presentedinlatersections.

Other than focusing the supply and demand analysis, the

interaction between the income and price iswidely searched. Joshua

Gallin(2006)searcheswhetherthereisalongrunrelationshipbetween

housepricesand incomebyusing95UnitedStatesmetropolitanareas

for23years.Manyhousingmarketobservershavebecomeconcerned

9

thathousepriceshavegrowntooquicklyandarenowtoohighrelative

topercapitaincomes.Gallinadmitsthatundertheideathatthereisa

longrun relationship between prices and income, prices will likely

stagnate or falluntil they arebetter alignedby income.However,he

finds that with the standard tests, there is little evidence for the

cointegration of housing prices and income in 95 United States

metropolitanareasfor23years.

UnlikeGallin,Malpezzi(1999)findsthathousepricechangesare

not randomwalksandareat leastpartlypredictable. Inhiswork,by

constructingasimplemodelthattestswhetherpricestendtorevertto

some equilibrium ratio of house price to income. Furthermore, he

investigates how supply conditions affect both the equilibrium price

and the time path of adjustment to equilibrium in 133United States

metropolitan areas from 1979 through 1996.According to his results,

the stringency of the regulatory environment was a particularly

powerfuldeterminantof the equilibriumhouseprice to income ratio.

Also, faster rates of population growth and of income growthwere

associatedwithhigherconditionalpricechanges,suggestingalessthan

perfectlyelasticshortrunhousingsupply.

10

CHAPTER3

HOUSINGINVESTMENTTHEORY

Housingstockdependsondepreciatednumberofdwellingsand

numberofhousingcompletionsasinperpetualinventory.

Itisacommonassumptionthathousingsupplyisinelasticinthe

shortrun than in the longrun,sincehousingcompletions isrelatively

smallerthanhousingstock.(Kenny,1998)AlsoTopelandRosen(1988)

explained the reason of this assumption by the high costs of

constructionactivitywhenrapidchangesoccur.So,intheshortrun,the

demandforhousingdrivenbytheexogenousfactorswilldeterminethe

priceofhousingrelativetoothergoodsandservices.

11

Figure1.Shortrun Figure2.Longrunequilibrium equilibrium

InFigure 1, for any levelofhousepricesbelowP1, there is an

excessdemandforhousingandforanylevelofhousepricesaboveP1,

there isanexcesssupply forhousing.From thegraph, it isquiteclear

thatunderconditionsofshortrunequilibrium,anystimulustohousing

demandwillresultinarisemoreinhousepricesrelativetoothergoods

and services than house dwellings as mentioned in Kenny (1998).

Hence, the microeconomic studies of house market predict a very

strongrelationshipbetweentheargumentsofhousingdemandfunction

andtherealpriceofhousingintheshortrun.

However, in the longrun,asudden increase indemandresults

again rise in house prices, this time construction firms will find it

P

P1

S

H H2

DD2

P*

S0

S

D

D

H1 H

12

profitabletosupplymorehousingunitstothemarketwhichmakesthe

supplycurvemoreelastic.

3.1.HousingSupply

Muchof the literaturehas focusedon thedeterminantsofnew

housing supply, particularly the supply of single family detached

homes,andtherenovationandrepairdecisionsofhomeowners.Ithas

focusedonaggregatedatabecause there isso little informationwhere

theunitofobservationisthebuilder,investor,orlandlord.Inaddition,

sincehousingisadurablegood,housingsupplyisdeterminednotonly

by the productiondecisionsofbuildersofnewunitsbutalsoby the

decisionsmade by owners of housing (and their agents) concerning

conversionoftheexistingstockofhousing.(Dipasquale,1999)

While modeling supply side of the market, Poterba (1984)

assumes that thehomebuilding industry is composedof competitive

firms and that the industrys aggregate supply depends on its input

pricesandtherealmarketpriceofhousing.Assumingthereare limits

to supply of any factor of production (such as lumber), increases in

13

demand for construction increase the equilibrium price of structures.

Poterbadefines supply asnet investment in structures, ignoring land

prices;heacknowledges the importanceof landbutomits land inhis

empiricalstudiesbecauseofthedataissuesforhisempiricalwork.

Adisadvantageofacoststructurebasedonrisingsupplyprice

aloneisthatitdoesnotmaketheMarshalliandistinction,inwhichthe

longertheperiod,thefewerthingsthatyouareholdingconstantwhile

you analyze the response of amarket to an external shock, between

shortrunandlongrunsupplyresponses:theindustrysupplycurveis

fixed,andhasno timedimension.Thisassumptiongivesan industry

versionof theadjustmentcost theoryof investment,but isunlikely to

bevalid,becausesupplyislikelytobemoreinelasticintheshortrun.

Therefore,thenatureoftheshortandlongrunsupplyconditions

offactorsofproductiontotheindustryisspecified.Thus,forexample,

labordoesnotmovecostless inandoutof the industry.Neitherdoes

capital.Shortrunfactorsuppliesarelesselasticthanlongrunsupplies.

Togointhisdirection,itrequiresintroducingadditionalstatevariables

into the analysis, which increases the complexity of the model,

especially forempiricalwork. InsteadTopelandRosen (1988)adopta

14

more tractable alternative where supply conditions of factors are

approximately incorporated into an expanded cost function which

includestherateofchangeofindustryoutput.Shortrunoutputsupply

inelasticityisimpliedbycostpenaltiestorapidchangesinthelevelof

constructionactivity.

A complete model of the dynamics of new housing supply

requires detailed specification of supply dynamics for all factors of

productiontotheindustry.Byallowingmarginalcosttovarywithboth

the levelofoutputand its rateofchange,TopelandRosen (1988)cut

throughtheimmensecomplications.

Inhousing literature, there isa large literatureonmodeling the

housingsupplyofnewhomes.WhileTopelandRosen(1988)modelthe

housing investment under the assumption of perfect foresight, they

focus on housing supply. On the supply side of the market, the

representative building firms maximizes discounted profits over an

infinite horizon. Since themarket isperfectly competitive,profits are

definedas

, ,

edt 3.1.1

15

whereP(t) is theprice foroneunitofhousing stockat time t, is

grossinvestmentinhousingattimet,Crepresentsthecostsattimet

andisapositiveconstantrepresentingtheinterestrate.Furthermore,

theindustryscapitalevolutionequationis

3.1.2

Thecostfunctionisspecifiedas

, , , 3.1.3

TotalcostCattimetisafunctionofthelevelofproduction,the

change in production and a number of cost function shifters

representedbya factory.Note that the inclusionof thechangeof the

gross investment level is the difference between the cost function in

Poterba (1984),who includes only the level of investment, andTopel

andRosen (1988),who includeboth the leveland thechange ingross

investment. Third change in the gross investment level denotes the

adjustmentcostthatthefirmfaceswhenchangingitsoutputlevel.

They imposethatC istwicecontinuouslydifferentiableand

thatmarginal costs are positive and increasing in the level of gross

investmentIandthattheadjustmentcostsareincreasing.

16

0,

0

/ 0, / 0

Furthermore, the nonnegative constraints for the derivative

cost function (C2andC22)prevent the infiniteproduction sinceas the

rateofchangeofinvestmentincreases,thecostalsoincreases.

Given these assumptions, we can solve the maximization

problem of the representative building firm by constructing the

Hamiltonianequationandtakingthefirstderivativeswithrespectto,

and .Thenecessarycondition for theoptimalpath isgivenbyEuler

equation.

/

3.1.4

If

0, in other words there is no adjustment cost, firms should

choose such that the price equals to themarginal cost. In such a

situation, therighthandsideofaboveequation (3.1.4)reduces tozero

andcurrentpricesbecomesufficientinordertodetermineproduction.

When the change in appears as an argument in the cost

function, therebecomesadifferencebetweenpriceandmarginal cost

17

that consists of the right hand side of equation (3.1.4). By the

linearizationofeulerequation,wecanderive,

1

3.1.5

where the terms inandarederivativesof thecost function

evaluatedatstationarypoint, / and /.

If the crucialparameter is zero then the above equation

(3.1.5) tells us that the investment is a function of exogenous cost

shiftersandtheprice.

By rewriting the equation (3.1.5) slightly different,we can

havethefollowingexpression,

1 16

where thescanbeobtained from theequation (3.1.5). In themodel

without adjustment costs = 0, that is, changes in exogenous cost

shiftersareimmediatelyreflectedinthelevelofinvestment.Inthecase

wherethereareadjustmentcosts( 0),thereisalagbeforethenew

levelofinvestmentisreached.

18

In the literature, Topel and Rosenmodel is used for different

purposes. Kenny (1999) has considered the potential effects of

asymmetric adjustment costs on the dynamics of housing supply by

utilizing from the Topel and Rosen (1988) supply model with the

flexible adjustment costs function advocated in Pfann (1996). His

empirical results suggest Irish housing supply is unit elastic in

equilibrium in the longrun and also in the Irish housing market,

adjustment costs associatedwith an expansion inhousing output are

greaterthantheadjustmentcostsassociatedwithacontraction.

Furthermore,Kenny (1998) summarizes the housingmarket in

Irelandwhere his estimations about housing supply and demand is

basedonTopelandRosen(1988)housingmodels.Healsoexaminesthe

monetarypolicydevelopmentsabout Irishhousingmarketby looking

deeply the banking channels and also the inflation policy effects on

housingprices.

TopelandRosens(1988)ideassuchasthesupplyrestrictionson

constructionactivityarenotonlyusedinestimationsofsupplymodels

butalsousedinsettingupanequilibriumassetpricingmodelbetween

housepricesandrents(AyusoandRestoy,2006).Theyapplytheirown

19

constructedmodeltoSpain,UKandUS.Andtheyconcludethatsharp

increases inhouseprices leadtopricetorentratiosaboveequilibrium

bymid2003inthosecountries.

Hakfoort and Matsyiak (1997) examine the determinants of

unsubsidizedhousingstarts inNetherlandsbyestimating thesupply

sideofthePoterba(1984)modelandthesupplysideoftheTopeland

Rosen(1988)model.Theformermodelyieldsasupplyelasticityofthe

order1.6whilethelatteryieldsashortrunelasticityof2.3andalong

runelasticityof6.

3.2.HousingDemand

Mostoftheliteratureforthedemandsideofthehousingmarket

isbasedontheestimationofpriceelasticityofdemand.Asmentioned

before,Palmquist (1983)estimates thedemand for thecharacteristicof

housingbyusing thehedonicdemand theory. Heestimates theprice

elasticityofdemandforlivingspacewhichcomesoutunitarywhilethe

other characteristics are more inelastic. The crossprice effects are

20

significantwhiletheexpenditureandincomeelasticitiesarefoundtobe

inelastic.

Theempiricalresearchfordemanddiffereitherinvariablesused

fortheestimationorinthemethodchosenfortheestimation.JamesR.

Follain,Jr.(1979)examinestheeffectofanincreaseindemandonlong

run price of housing by finding the price elasticity of the longrun

supplyofnewhousingconstructioninperiod19471975.Heshowsthat

demand function depends on longrun price of a unit of housing,

permanent income ofhouseholds, interest rate and theprice of other

goods. Follainuses real value ofprivate residential construction as a

quantityinsupplyfunctionbyapplyingOLSandTSLSmethods.

Dipasquale and Wheaton (1994) estimates demand equation

whichiscomposedofstockofsinglefamilyunitsasafunctionofrent

index, age expected homeownership rate, permanent income per

household, price index of single family housing, annual user cost of

homeownership,andtotalhouseholds.Theycomparetwoeconometric

models for actual households as for tenure choice and age expected

households as forboth tenure choice andhousehold formation.They

find that all elasticities arehigherwhen age expectedhouseholds are

21

used thanwhen actualhouseholds areused.The regionaldifferences

withinacountryareseennotonly for thesupplysideof thehousing

but also for the demand side of it.Alan K. Reichert (1990) searches

effectsofsomemacroeconomicvariablesuponregionalhousingprices

byconstructingregionspecifichousingequations.Hederivesdemand

function in theway of assuming utilitymaximization on the part of

homeowners andwealthmaximization on the part of investors. The

demandequationiscomposedofthequantityofnewhousingsoldasa

lefthand sidevariable and realhousingprices indexofnewhousing

quality, resident income, average employment rate, average loan to

value ratio, realmortgage interest rate, themeasureofacceleration in

regionalhousingpricesandseasonaldummyvariablesforeachspecific

region.

In housing economics literature, the demand for housing is

normally derived inmultiperiodmodelwhere consumersmaximize

utility subject to an intertemporal budget constraint. These models

incorporatevariousfeaturesofhousingmarketincludingthelargecost

of housing relative to the current disposable income and hence the

dependenceofhousingdemandsthesavingsinearlierperiodsandalso

theprice.(Kenny,1998)

22

Considera simpledemand functionwhich ignores the frictions

generatedbytheheterogeneityofunitsandthematchingofbuyersand

sellers.(TopelandRosen,1988)Undertheassumptionofperfectcapital

market,theinversedemandequationofTopelandRosen(1988)model

becomes;

3.2.1

where is the rental price of a housing unit, is a vector of

exogenousdemandshifters, bethestockofhousingcapitaland

23

3.2.3

wherekisthelifeofabuilding.

Inthenextperiod, thepriceofahouse isstillsumof therental

pricesbutthereisadepreciationsinceyoudidnotsellthehouseinthe

previousperiod.Also,youhaveadepreciatedincomeandincomethat

areexposingtotheinterestgain.

3.2.4

3.2.5

Equation (3.2.5) is the same with the equation (3.2.2), just

written indiscretetime.Furthermore,thevalueofhousingstockmust

beboundedsothatthediscountedfuturepriceofcapitalconverges:

lim

0 3.2.6

By taking the integral of equation (3.2.2)with respect to t

undertheboundarycondition,wecanwrite;

3.2.7

24

Above equation (3.2.7) tellsus that theprice of a house is the

accumulationofalldiscountedrentalincomethroughitslife.

Hence, the completemarketdynamicsof stocks andprices are

describedbytwolineardifferentialequations:

1 3.2.8

3.2.9

Given the initial conditions 0 and 0with the boundary

condition (3.2.6), by differentiating (3.2.9) with respect to t and

substitutingfrom(3.1.2)yields

1 3.2.10

where .

Thisdemandmodel(TopelandRosen,1988)hasitsoriginsinthe

workofWalras (1954)andmuch laterbyFriedman (1963)andTobin

(1969).Theydealwitha linear structure foranalytical tractabilityand

presentadeterministic (perfect foresight) formulation to illustrate the

key ideas.Toavoid expositorydistractions,whicharewell treated in

25

the literature, they also ignore the special and peculiar income tax

provisionsofhomeownership.

This demand part of the Topel and Rosen (1988) model

completes the housing supply model since the market should be

thoughtsimultaneously.

3.3ImplicationsofTheory

Topel andRosen (1988) housing investment theory provides a

frameworktoanalyzethepossibledeterminantsofthehousingsupply

as well as the allowance of shortrun and longrun analysis in my

empiricalwork. Inaddition,TopelandRosenmodelalsocontains the

expected present value theory of asset pricing which supports my

empiricalanalysisandbecomessuitablefortheTurkishhousingmarket

inthewayofhouses,notbeingonlyconsumptiongoodbutalsoapart

ofahouseholdwealth.Thereforetheirmodel isakindofanextended

versionofPoterbas(1984)model.

26

As in thismodel,bynotomitting the longrun relations, short

runrelationscanbefoundandbeinterpretedinmystudywiththehelp

ofVectorErrorCorrectioneconometricmethodologywhichprovidesus

tostudyonshortrundynamicsbyrestrictingthevariablestoconverge

to their cointegrating longrun relations. (Known asRestrictedVector

AutoRegression).

Inmy empirical framework, the cost indexbehaves likeoneof

theelementofthecostfunctioninTopelandRosen(1988)model,which

isdenotedasy(t).Because,thecostindexhastheconstructionmaterial

prices and in the Topel and Rosen (1988) model the cost shifter is

definedas the factorprices thatare supplied to the industry, thecost

indexcanbeusedasacostshifter.Theotherdynamics,representedas

gross investment level is composed of the quantity of dwellings,

constructed for the defined period, because the investment level

depends on the change of capital stockwith the depreciated capital,

equation(3.1.2).Lastly,therateofchangeoftheinvestmentisaddedto

themodelbecauseoftheslowadjustmentmechanismofthemarketin

the shortrun, so it isused in the shortrun empirical analysis. Inmy

empiricalframework,thelongrunerrorsthatcanbefoundbyJohansen

Cointegration econometric methodology and used in the restricted

27

vector autoregression model, and also the first differences of the

variables are the representatives of the rate of change of gross

investmentlevelintheshortrunanalysis.

According to the equilibrium equation (3.2.10),when demand

side shifter, , increases under the assumption that the other

variables stay the same, the investment level, , increases since

is positive and is negative. Moreover, as

increases, the capital stock increases. Inmy empirical study, the

demand side shifters are population and income. So as population

increases,theneedforhousesincreasessoquantitydemandedincreases

andasincomeincreases,thedemandofhousesincreases.Ontheother

hand, when the supply side shifter, increases, the price of the

investment, , increases then decreases since again

ispositiveand isnegative. Inmyempirical framework,

the supply side shifter is the cost index since it includes factorprices

affectingthesupplyandascostindexincreases,thedesireforbuilding

willdecreaseduetolessprofit.Hence,ascostindexincreases,thelevel

of investmentandso thecapitalstockwilldecrease.Furthermore, the

interest rate, in the equation (3.2.10) affects both the supply side

andthedemandside.Theinterestratehasanegativerelationshipsince

28

isnegative.Inthisstudy,theinterestratehasalsoanegativeeffecton

the quantity of dwellings for both sides of the market. The other

variable,affectingboth thedemandand thesupply, is thepriceof the

investment, . The effect of price to the demand side and to the

supply side is different.According to the supply equation (3.2.8), as

priceincreases,thelevelofinvestmentincreasessinceispositiveand

so the capital stock increases. However, according to the equation

(3.2.9), as the price of investment increases, the capital stock directly

decreasesdue to the fact that ispositiveand isnegative.The

value,which isusedasaproxy for theprice,hasanegativeeffecton

quantity demanded since as prices increases, less people can buy

houses. However, the value has a positive effect on the supply of

housessincebuildingahousemaybecomemoreprofitablethanbefore.

29

CHAPTER4

HOUSINGMARKETINTURKEY

Housing was not ranked among the most important socio

economic issues inTurkeyuntil theearly1960s.Themainreasons for

thislackofinterestmaybesummarizedasfollows.First,themigration

fromruraltourbanareaswasrelativelyslowandtherewasnomarked

deficit in the housing supply at least quantitatively until that era.

Second, the slowpaceof industrializationdidnotmake theworkers

housing question an important source of discontent before the early

1960s.Finallyuntilthebeginningoftheplanneddevelopmentperiod,

housing had not been taken up within the broader context of its

positionrelativetothewholeoftheeconomy.Therefore,itseffecttothe

economywaslargelyneglected.(Keles,1990)

30

After 1960s, transition from the traditional family to nucleus

familyandrapidrisingofpopulationincreasesthedemandforhouses,

especially the housing type called apartments which have smaller

gardens andmore than one floor. Due to Turkeys problems about

economics such as low level of Gross National Product per capita,

chronical high inflation and high interest rates, enough savings for

house building and buying can not be formed. The implemented

policiesabouthousing isnotefficientenoughtosolvetheproblemsof

Turkeyhousingmarket. In thepast, land is allowed tobuildbut the

infrastructure is not constructed for a living place, this reduces the

investmentdesireoftheinvestors.Also,theunavailabilityofmortgage

credits causes more people not to be able to buy houses for long

periods. So, building shanty houses (gecekondu) and unhealthy,

unplanned urbanization spread widely. The promises before each

electionandfrequentlyacceptedconstructionforgivenesscausetoraise

theproblemexponentially.(Gurbuz,2002)

In addition, deficient municipal income is not enough to

construct infrastructure services to the new streets and new counties

wheretherearealreadylotsofshantyhouses.Furthermore,deficiency

in communication between the municipalities who construct

31

infrastructure and theutilityunitswhoprovide electricity andwater

causeswastefulexpenditure.

Increasing investment to the infrastructure services with the

renovationinhousingpolicyin1980smaintainstheconstructionsector

to rally.Collective housing fund, housing aid fund to the employees

and especially the Turkey Emlak Bank had assumed the role of the

leader for the construction sector.With the guidance of these funds,

housesuppliesandcooperatives,whicharesupportedbythemortgage

credits, increase rapidly. Housing Development Administration of

Turkeystartstobuildhousesforthelowincomefamilieswithfacilities

inpayment.Thishelps reducing the inequalitybetweendemand and

supplyinTurkey.

Therewas seen a significantdecline inhousing investments in

themiddleofthe1970sandalsointhebeginningofthe1980swiththe

effect of the crisis seen induring 1970s. Sincehousing investment is

oneofthemostimportantexpenditureofahouseholdandithasahigh

portion in the expenditure of a household, this investment is an

importantsourcefortheotherinvestmentsotherthantheinfrastructure

and utility investments. (Malpezzi, 1990) In the 1980s, the housing

32

investment increases, especially with the help of government

investment, and then starts to decline in the last years of twentieth

centuryand inthebeginningoftwentyfirstcentury.Byobservingthe

figure3,wecaneasilynoticethatafter1998,theratioofhousingfixed

investmenttothegrossfixedinvestmentisrapidlydeclining.

Figure3:Theshareofhousinginvestmentingrossfixedinvestments1998currentprices

With this decrease investment in housing, Turkey housing

investmentsislowerthantheinvestmentratiosofdevelopedcountries,

whereasin1988sthisratioisneartothedevelopedcountries(Eraydin

etal.,1996)

05

1015202530

Housing Investment / Total Investment (% percent)

*Expectation ** Target of the government

Source: DPT, State Planning Organization, www.dpt.gov.tr

33

The literature abouthousing inTurkey iswidelybased on the

inefficiencyofthehousingpolicies;littleempiricalanalysisisdonedue

tothedeficiencyofdata.However,asthehousingsectorimportanceis

understood, various data collection increases and more studies are

done.ForexampleoneofthelateststudiesisdonebySari,Ewingand

Aydin(2007).Theyinvestigatetherelationbetweenhousingstartsand

macroeconomic variables in Turkey from 1961 to 2000. They use

generalized variance decomposition approach for examining the

relations between housingmarket activity and prices, interest rates,

output,money stock and employment.Their results indicate that the

effect of thehousingmarket on output is notnecessarily reflected in

labormarket.Moreover,theshocksto interestrates,outputandprices

havenotableeffectsonhousingactivityinTurkey.

Theprovisionofhousingfinanceindevelopingcountriesisoften

problematic,because of thevolatilemacroeconomic environment and

thelackoflegalandregulatoryframeworkthatsupportscollateralized

lending.ErolandPatel (2004)evaluateTurkishgovernmentshousing

policy for financing the public sector housing and discuss the

appropriatetypeofmortgagesfromthelendersperspective.According

totheirresults,wageindexedpaymentmortgages(WIPM)arefoundto

34

be desirable mortgage instruments in periods of persistent high

inflation from the lendersperspective.Thereasonbehind this finding

is that WIPM eliminate the real interest rate risk, credit risk of

adjustable rate mortgages and the wealth risk of the fixed rate

mortgages.

Another research paper on the Turkish real estate market is

basedon the idea that thehousing isbothan incomedecrease for the

tenantsandanincomeproviderforthelandlords.Sohousinghassome

kind of wealth effect for the households that can affect the whole

economy.BinayandSalman(2008)discusstheextentofwealtheffects,

affordability, financial deepening and creditmarket risks in Turkish

realestatemarket.Theyusepricerentratio to testwhether there isa

realestatepricebubble inTurkeyornot.Asaresult, theydonot find

enoughevidencesupportingthatthereisarealestatebubbleinTurkey,

contradictingwhatmanybelieve.

Therefore,thereisnodirectandcollectivestudywhichisbased

on formulating both the supply and thedemand side of theTurkish

housingmarket.So, thisstudyaims todetermine the factorsaffecting

thehousingsupplyanddemandinTurkey.

35

CHAPTER5

ECONOMETRICMETHODOLOGYANDDATA

In thischapter,econometricmethodology that is foundsuitable

to use in this study is introduced with the data descriptions.

Furthermore,econometricmodelisbrieflyexplained.

5.1Methodology

This chapter presents and discusses a brief review of the

empiricalmethodology employed. In section 5.1.1,we brieflypresent

PhillipsPerronUnitRootTests. In 5.1.2, JohansenCointegrationTest

andVectorErrorCorrectionMethodsarepresented.

36

5.1.1PhillipsPerronUnitRootTest

A stationary time serieshas a constant longrunmean, a finite

variance(timeinvariant)andatheoreticalcorrelogramthatdiminishes

as lag length increases.On theotherhand, foranonstationaryseries,

there exists no longrun mean and its variance is time dependent.

Therefore, under the condition of nonstationarity, to use classical

statisticalmethods such as ordinary least squares (OLS),usual ttests

andFtests,areinappropriate.However,inordertodecidethepresence

ofunitrootswhichcanbedefinedasatendencyforchangesinasystem

topersist, inotherwordsnonstationarity inasystem,only lookingat

the sample correlogram is unreliable. A formal test to detect the

possible presence of unit roots is developed by Phillips and Perron

(1988)

The distribution theory supporting the DickeyFuller tests

assumes that the errors are statistically independent and have a

constant variance. In using thismethodology, caremust be taken to

ensurethattheerrortermsareuncorrelatedandhaveconstantvariance.

Phillips and Peron (1988) developed a generalization of the Dickey

37

Fullerprocedurethatallowsforfairlymildassumptionsconcerningthe

distributionoftheerrors.

ThePhillipsPerontestisexplainedinEnders(1995)asfollows:

Suppose that we observe observations 1,2,...,T of the {yt}

sequenceandestimatetheregressionequation:

2

Fortunately,thechangesareminor;simplyreplace with,

with,and with.Thus,supposewehaveestimatedtheregression:

2

where , , and are the conventional least squares regression

coefficients.

PhillipsPerronderiveteststatisticsfortheregressioncoefficients

underthenullhypothesesthatthedataaregeneratedby

wherethedisturbancetermissuchthat 0.

38

There is no requirement that the disturbance term be serially

uncorrelated or homogenous. Instead, the PhillipsPerron test allows

the disturbances to be weakly dependent and heterogeneously

distributed.

ThePhillipsPerronstatisticsmodifytheDickeyFullertstatistics

byallowingforanadjustmenttoaccountforheterogeneityintheerror

process.

The appropriate critical values are given inMacKinnon (1991)

samewiththeDickeyFullertestcriticalvalues.

5.1.2.JohansenCointegrationTest

Thesequences{yt}and{zt}arecointegrated,iftheyareintegrated

of thesameorder, letussayd,or I(d),and their residualsequence is

stationary. It is a known fact thatOLS estimation procedure can be

appliedifthevariablesinvolvedinthemodelareI(0).Theviolationof

thisassumptioncausesus toobtainspuriouscorrelation (Grangerand

Newbold,1974).Whiledealingthisproblem,Davidsonetal.(1978)state

39

thatfittingtheregressionbyusingthefirstdifferencesofthevariables

would result in a loss of valuable information about the longrun.

Therefore, they propose an error correction mechanism (ECM) by

combining the first differences of the shortrun and undifferenced

valuesof the longrundynamics.However,EngleandGranger (1987)

provethatthismethoddevelopedbyDavidsonetal.(1978)istrueifthe

variablesinthemodelarecointegrated.

A theoretically more satisfying approach is developed by

Johansen (1988) to consider the cointegration relationshipwhen there

aremorethantwovariables.ThisprocedureisexplainedinWatsonand

Teelucksingh (2002) as follows; xt is composed of (n,1) vector of I(1)

variableswhosevectorautoregressive(VAR)representationisgivenas,

(5.1.2.1)

whereare(n,n)matrices.Itcanalsobewrittenas,

(5.1.2.2)

where

and

40

ThepurposeoftheJohansenprocedurecanbestatedasfollows;

1. To determine the maximum number of cointegrating

vectors

2. To obtain the maximum likelihood estimators of the

cointegrating matrix () and adjustment parameters () for a given

valueofr.

The rank of the matrix , r, is equal to the number of

independent cointegrating vectors. There can be at most n1

cointegratingvectorsandifr=0,itisaknownfactthatthevariablesare

not cointegrated and equation (5.1.2.2) is VAR model in first

differences. If r=n, the vector process is stationary. For 0

41

Equation (5.1.2.2) isdenotedasavectorerror correctionmodel

(VECM). When there are r cointegrating vectors, r error correction

terms appear in each of the n equations. For instance, in the first

equation(explainingx1t),xt1consistsofterms,

. .

It isknown that thenumberofcointegratingvectors isequal to

thenumberofsignificantcharacteristicsrootsofthematrix.Suppose

theordered characteristic rootsof thematrix are; .

Toobtainthenumberofcharacteristicrootsthataredifferentfromzero,

Johansen proposes the following tests, that are based on trace and

maximumeigenvaluestatistics,respectively,

ln 1 (5.1.2.4)

, 1 ln1 (5.1.2.5)

whereistheestimatedvaluesofcharacteristicroots(eigenvalues)of

theestmatedmatrixandTisthenumberofusableobservations.

The trace statistic testswhether the number of cointegrating

vectorsislessthanorequaltoragainstageneralalternativewhilethe

42

alternative hypothesis for maximum eigenvalue statistic is r+1. The

criticalvaluesforthesestatisticsarecalculatedbyJohansenandJuselius

(1990)withthehelpofsimulation.

5.1.3VectorErrorCorrectionModel(VECM)

A vector error correction model (VECM) is a restricted VAR

designed for use with nonstationary series that are known to be

cointegrated. The VEC has cointegration relations built into the

specificationsothatitrestrictsthelongrunbehavioroftheendogenous

variablestoconvergetotheircointegratingrelationshipswhileallowing

forshortrunadjustmentdynamics.Thecointegrationtermisknownas

theerrorcorrectiontermsincethedeviationfromlongrunequilibrium

iscorrectedgraduallythroughaseriesofpartialshortrunadjustments.

Formally, the (nx1) vector , , , has no error

correctionrepresentationifitcanbeexpressedintheform:

(5.1.2.6)

where,

43

0=an(nx1)vectorofintercepttermswithelementsi0

i=(nxn)coefficientmatriceswithelementsjk(i)

=isamatrixwithelementsjksuchthatoneormoreofthejk0

=an(nx1)vectorwithelements

Note that the disturbance terms are such that may be

correlatedwith

Thekeyfeaturein(5.1.2.6)isthepresenceofthematrix.There

aretwoimportantpointstonote:

1. If all elements of equal zero, (5.1.2.6) is a traditional

VAR in first differences. In such circumstances, there is no error

correction representation since xt does not respond to the previous

periodsdeviationfromlongrunequilibrium.

2. Ifoneormoreofthejkdiffersfromzero,xtrespondsto

the previous periods deviation from longrun equilibrium. Hence,

estimatingxtasaVAR in firstdifferences is inappropriate ifxthasan

errorcorrection representation. The omission of the expression xt1

entails a misspecification error if xt has an errorcorrection

representationasin(5.1.2.6)

44

5.2.Data

All the data are obtained from the Turkish Statistical Institute

(TurkStat)overtheperiod19702007(annually).

First variable is the quantity of dwellings (q) which is the

number of buildings including apartment houses, houses, other

buildings (commercial, industrial, medical and social, cultural,

religious,administrativeandother).Itistakenascompletedorpartially

completednewbuildingsandadditionsbyuseofbuildingaccordingto

occupancypermitsfromTurkStat.Thisvariablerepresentsthequantity

demanded and quantity supplied in the equilibrium. Also, the

Occupancy permit is preferred in this study since it is a certificate

which must be given to building owners by municipalities to be

constructed in boundaries ofmunicipalities and itmust be given to

building owners by governorships if the construction is out of

boundariesofmunicipalities.

Thesecondvariableistheinterestratewhichdirectlyaffectsboth

thesupplysideandthedemandside.TheCentralBankoftheRepublic

ofTurkey (CBRT)nominaldiscount interest ratesareusedasaproxy

45

for all other interest rates, since the aim of CBRT for this discount

interest rate isbeingabenchmark for theother interest rates, suchas

deposit or loan rates. Under the assumption that the inflation

expectationsareequal to theactual inflation, it is transformed to real

valuesbyusingtheFishersRulewhichis;

1 1 1

where,isthenominalinterestrate

istherealinterestrate

istheexpectedinflation.

The thirdvariable isGrossDomesticProductatconstantprices

(1987),calculatedbyTurkStat.Itisusedasaproxyforrealincome.

Theothervariableisthevalueofdwellings,whichistakenfrom

TurkStat.Thedescriptionofthedataisasfollows:

Unitprice ofm2 are calculated four times at a year byprovince for reinforced concreteandbearingwall construction,forbuildingthattheiruseofbuildingsincludeapartmenthouses,houses, other buildings (commercial, industrial, medical andsocial,cultural,religious,administrativeandother)byprovince.Value is multiplication of floor areas indicated in OccupancyPermits.Thecostoflandisexcluded.

46

Thisvaluem2isturnedintorealtermswiththebaseyear1970by

using the Fishermethod. This real value is used as a proxy for real

housingprices.

Anothervariable isBuildingsConstructionCostIndex,which is

calculated by Turk Stat. Since the aim of forming a building

constructionpriceindexistodeterminethequantitiesofinputsusedin

building construction and to show the yearly cost changes of these

quantitiesofinputs,thisindexcanbeusedascostinthesupplysideof

themarket.This studybegan in1989andand resultswerepublished

first inNovember 1992. 1991wasdetermined as thebaseyear and a

weightedLaspeyresindexformulawasusedinthiscalculation.

Theindexisconstructedasbelow;

Outofa totalof295 items in thebuildingsconstructioncost index,20encompassworkmanship,7aremachinery,146 are construction materials and 122 are installationmaterials. Price of these items are gathered from 24provinces (for every item prices are collected from 3separateestablishments)Thesepricesarecollectedonthe15th of the last month of every quarter from 1292establishments which are producers, wholesalers orretailers who do business with construction firms andcontractors.

47

The Turkey BuildingCost Index starts from 1991 till 2007 but

IstanbulConstructionCostIndexstartsfrom1970to2007.Sobyusing

IstanbulConstructionCostIndexasaleadingvariable,TurkeyBuilding

Cost Index can be generated for the years before 1991. The detailed

calculationisintheappendix.Thisindexisalsoturnedintorealterms

to the base year 1970 by using Fishermethod. In addition, another

studyisestablishedbyusingtheIstanbulConstructionCostIndexsince

this index is highly correlatedwith Turkey BuildingCost Index and

IstanbulConstructionCostIndexstartsfrom1970to2007.Thisindexis

transformed intoreal termsbyusingFishersRuleandIstanbulactual

inflation.

InBuildingCostIndex,presetmaterialsandlaborarecalculated

within the preset weights. The weights and the materials are not

changedduringtheyears.Inaddition,thesepricesaretakenfromthe

producers.On theotherhand, invalue eachbuildings cost are taken

fromthebuilderwithoutinterestedinwhatthematerialsareandhow

much the labor costs to thebuilder.So thematerialsand theweights

probably change over time.Another difference is that value is taken

and calculated for each citywhile building cost index is taken from

presetfourregions.

48

The last variable is the population. Mid year population

(populationonJuly1)estimateistakenfromTurkStat.Thiscoversdata

relatedtotheresultsobtainedinGeneralPopulation.

5.3.EconometricModel

The credit restrictionshave a crucial impacton the signof the

effectofvariousvariablesonhousingdemand.Under theassumption

ofperfectcapitalmarkets,i.e.nocreditrestrictions,boththecurrentand

future income and the expected increase in real house prices have a

positiveeffectonhousingdemandduetothefactthatwhenthecurrent

andfutureincomeofahouseholdincrease;hewantstobuynewhomes

in order to increase his monthly income by taking rent from each

additional home if the benefit from consumption or the return from

alternative investments are less than the housing investment; or he

wantstobuyanewhomeinordertoraisehisstandardofliving.Also,a

householdwhowants tomaximizehisprofit fromahousebuysnew

homeswhenhousepricesareexpectedtoincrease.Sinceaspopulation

rises, theneed for thehouses increase, there isapositive relationship

betweenhousingdemandandpopulation.Conversely,thedemandfor

49

housing is negatively related with the interest rate because higher

interest rates increase the cost of borrowing as well as the cost of

housingservices.

As the credit restrictions increase, theeffectof thevariableson

housingdemandmayvary.Forexample,afallinfutureincomehasan

immediateeffectonfutureconsumption.Sincethehouseholdswantto

smooththeirconsumption,fromnowontheystarttosave.Sincethere

isashortageofalternativesavings(otherthanhousingmarket)because

of the credit restrictions, currentdemand forhousing increaseswhen

futureincomefalls.

Inthisstudy,itisassumedthatthequantitydemanded(qd)isa

function of real value (p), real income (y), real interest rate (c) and

population(n)

Qd=f(P,Y,R,N) (6.2.1)

where Qd=quantityofdwellings

P=realvalue

Y=GDPat1987prices

R=realCBRTdiscountrate

N=population

50

The impacts of these variables are basically based on the

consumerbehaviors.However,thechangesofexogenousfactorsonthe

equilibrium level of value will also depend on how the supply of

housing adjusts both to the changes in demand and to the other

exogenousfactors.

Inthesupplysideofthehousingmarket,thereisabuilderwho

wantstomaximizehisprofit.Sothemoretheconstructioncost,theless

thebuilderwantstocontinuebuilding,whichmeansthereisanegative

relativerelationbetweenhousingsupplyandconstructioncostlikethe

relationbetweensupplyand interestrate.Ifthe interestrate increases,

thecostofbuildingnewhousesincreases,becausetheyhavetoaccept

topaymoreinterestforhavingenoughcapitalforbuildingahouse.On

theotherhand,increaseinboththecurrentandthefuturehouseprices

will increase current supply of housing due to the fact that selling

houses may be more profitable than the other investments. The

populationand the incomeofahouseholdhavenotadirecteffecton

supplyofhousingbutanindirecteffectthroughthehousingdemand.

In this study, it is assumed that thequantity supplied (qs) is a

functionofrealvalue(p),realinterestrate(r)andrealcost(c)

51

Qs=f(P,R,C) (6.2.2)

where Qs=quantityofdwellings

P=realvalue

R=realCBRTdiscountrate

C=realconstructioncostindex

So,inmyanalysistheinstrumentalvariablesarepopulationand

incomeforthesupplysideofthemarketandconstructioncostforthe

demandsideofthemarket.Thesevariableshelpmetoestimatesupply

anddemand.TheexpectedsignsofthevariablesareasintheTable1.

Table1.ExpectedSignsinDemandandSupply

Variables ExpectedSignsinDemand

ExpectedSignsinSupply

RealHousePrices +RealIncome + NoDirectEffectPopulation + NoDirectEffectRealInterestRate ConstructionCost NoDirectEffect

52

CHAPTER6

ESTIMATIONRESULTS

Thedescriptive statisticsof the levelsand logarithmic formsof

allthevariablesinvestigatedinthisstudyaregivenintheappendix2.

The timesseriesplotsof the levelsof thevariablespurport tobenon

stationaryprocesses.However,toobtaintheexactintegrationlevelsof

thevariables,only considering theplots isnot reliable.Therefore, the

PhilipsPerronUnitRoottestsareapplied.Theresultsofthesetestsare

giveninTable2inAppendixC.

According to the Phillips Perron test results (Table 2), all the

variablesareof integratedoforder1,I(1)atthe0.1significance level.

Therefore, the results of Phillips Perron unit root tests can be

53

interpreted not to preclude the validity of employing the Johansen

Cointegrationprocedureforoursample.

The logarithmic forms of the variables are also checked by

PhilipsPerronunitroottestandweobservetheresultsthatcanbeseen

fromTable3inAppendixC

AccordingtoTable3results,allthevariablesareofintegratedof

order1,thatis,thefirstdifferenceofallthevariablesarestationaryat

90%confidence level.Sinceallthevariablesare inthesameorder,we

canuseJohansenCointegrationprocedure.

6.1 Empirical Results of Housing Supply AndDemand

In this section the empirical analysis is employed for two

differenteconometricmodels.Inthefirstmodel,allthevariablesarein

the level formswhile in the secondmodel, in order to observe the

elasticitythevariablesareinlogarithmicforms.Insidethefirstmodel,

thereexistalso twodifferentanalysisbasedondifferent cost indexes,

54

TurkeyBuildingConstructionCostIndexorIstanbulConstructionCost

Index.

6.1.1.LevelDataAnalysis

In this part, we employ the Johansen (1988) cointegration

procedure to investigate the presence of a longrun relationship

between thevariablesofhousingmarketbyusingTurkeyCost Index.

Consideringtheresultsthatallthevariablesareofintegratedoforder1,

I(1),weconsiderallthevariablessimultaneously.Wetestthenullofno

cointegrationbyusingboth the Johansenmaximumeigenvalue (max)

andtrace(trace)statisticsforaVARmodelwithaconstantandwithout

trend.

In table 4 (Appendix C), eigenvalues (i), the maximum

eigenvalue(max)andtraceeigenvalue(trace)statisticsarereported.The

appropriatelaglengthsfortheVARmodelareselectedaccordingtothe

sequential modified likelihood ratio (LR) test, final prediction error

(FPE)andAkaikeInformationCriterion(AIC).

55

FromTable4,itisseenthattherearetwocointegratedvectorsin

ordertoexplainthelongrunrelationbetweenthehousingvariablesat

the0.05significancelevel.Thisresultisconsistentwithourexpectations

sincethehousingmarketconsistsofsupplyanddemandsides.

The economic assumptions tell us that in supply side of the

market,theincomeandpopulationdoesnotaffectthequantitywhilein

the demand side of the housing market, the instrumental variable

should be the cost. In order to test whether these restrictions are

significant in themodel or not, the chi square (2) is estimated and

interpreted.TheTable5inAppendixCshowsthesestatisticsforsome

specifiedrestrictions.

CE1standsforthedemandside,CE2standsforthesupplyside.

Inthefirstrestriction,Ionlyimposetherestrictionsthatareconsistent

with the economic theory during analysiswhich tells cost has not a

directeffectondemandand incomeandpopulationdonotaffect the

supplydirectly. BylookingattheresultsofTable5,atthe30%level,

the first restrictions are appropriate. In the second restriction, in

addition to the first restriction, I restrict the coefficient of quantity

supplied to take thevalueof 1, inordernot tonormalize the supply

56

model.Accordingtotheresults,thesecondrestrictionisappropriateat

the99%conficencelevelInthethirdrestriction,therestrictionsthatare

used in the second restriction is still valid and also the coefficient of

quantitydemanded isrestricted to thevalue1.At the99%confidence

level, third restriction is found significant.Therefore, sincepvalueof

the second restriction is thehighest,whichmeans the restrictions are

themost appropriate among these three restrictions, the second one

shouldbeusedduringtheanalysis.

By using the second type of restrictions and normalize the

demand side,wewill get the resultswhich can be seen in Table 6

(AppendixC)

These results (Table6)provideus toanalyze themarket in the

longrun. So, according to these results, all the variables forming the

supplyanddemandsidesaresignificantin99%confidencelevelinthe

longrun.Inthedemandsideofthehousingmarket,aspricesincrease,

the quantitydemanded of buildingsdecrease. In addition, as income

rises,thequantitydemandedincreasessincethehouseholdshavemore

money tobuyhouses.Populationalsoaffects thequantitydemanded

positively in the long run,due to the fact thataspeople increases the

57

need for houses increases. Also the interest rate effect is negative,

meaning, as interest rate increases the households are lesswilling to

buyahouse.Thesesignsareconsistentwithourexpectationsaswellas

theeconomictheory.

Inthesupplysideofthehousingmarket,allthesignsarealsoas

expected.Asprices increase, thequantity supplied rises;on theother

handcostsandinterestratesarenegativelyrelated.

Thecoefficientsofthevariablesmaynotbeinthesamescalethat

we face in reality, due to the fact that these variables are real, not

nominal.Ontheotherhand,ifweobservethecoefficientmeaningsby

transforming the variables into nominal terms,we can interpret the

coefficientsasfollows;

Firstlythedemandsideofthemarketisanalyzed.Therealvalue

at2007 is0.0006, ifwedecide to increase thevalue0.0025%, inother

words 1.5x108, then the quantity of dwellingwill decrease by 1. By

applying theFisher formulawith theactual inflationof2007,wecan

have the nominal values. The results tell us thatwhen themean of

58

dwellingvalues increaseby9,495YTL inTurkey, thedemand for the

dwellings decrease by 1. The same procedure is applied to the real

interest rate.When thenominal interest rate increasesby1point,100

basepoints, thedemand to thedwellingsdecreases by 64units.The

population and income can directly be interpreted. When the

populationincreasesby1,000peoplethenthedemandincreasesby6.44

numbers of dwellings. Also, when the GDP increases 1.63x108 to

1.64x108,1,000dwellingsaredemanded.

Secondly,whenwe lookat thesupplyside,wecansee that the

responseofbuilderislesssensitivethantheresponseofahouseholdto

thechangesofvalues.Thesameprocedurethatisappliedaboveisalso

doneforthesupplyside.So,ifthemeanofdwellingsvaluesincreaseby

35,140YTL, builderswant to build onemore dwelling. Furthermore,

whenthenominalinterestrateincreasesby1point,100basepoints,44

less dwellings are supplied.Whenwe observe the effect of real cost

which is 141.2 in 2007,by applying the sameprocedurewehave the

followingresult:whenthenominalcostindexincreasesby1,044,which

is the 2percent ofnominal cost index in 2007, the quantity supplied

decreasesby840units.

59

Whenwe lookat thevectorerrorcorrectionmodelresults from

Table7 (AppendixC)whichprovideus toobserve themarket in the

shortrunsimplybydifferencing thedata,wecansee thatmostof the

variables are not significant in the shortrun. In the vector error

correctionmodel,thesequenceisimportantinordertounderstandthe

pathofthespeedofadjustmentwhichcanbedecreasingorincreasing,

inotherwordswhether theshortrundynamicsconverge to the long

rundynamicsbyfollowinganincreasingpathoradecreasingpath.In

myanalysis,thequantityofdwellingsiswrittenfirstandthentheother

variables follow thequantity.Thecoefficientof theerror termsshows

thespeedofadjustment,soaccording to theresulting table, theshort

rundynamics follows an increasingpath in order to converge to the

longrun equilibrium. In addition, the error correction terms are

significant at 99% confidence level. Other than the error correction

terms,therealcostindexandtherealinterestratecomeoutsignificant

at85%confidencelevel.Despitetheinsignificanceinthevalue,wecan

noticeintheshortrunthevalueoftheresponse(coefficientis9.4x105)is

lessthaninthelongrun.ThissupportstheideasofthemodelofTopel

andRosen (1988) since in theirmodel; the shortrunpriceelasticity is

lesselastic, inotherwordssmallercoefficient, than the longrunprice

60

elasticity. This shows that the adjustment costs for a change in the

housingmarketaresignificantinTurkey.

In thedemand sideofTopelandRosen (1988)model,demand

sideshiftersarepopulationand incomewhichcomeoutsignificantat

99%confidencelevelinthelongrun;howeverintheshortruntheyare

insignificant.Furthermore,priceandcommonvariable,interestrate,are

allsignificantinthelongrunandbehaveinthesamemannerasitdoes

inthemodelinthewayofcoefficients.

Now,thesameanalysisisappliedbyonlychangingthevariable,

TurkeyConstructionCost Index to IstanbulConstructionCost Index.1

Firstly,wetestthenullofnocointegrationbyusingboththeJohansen

maximumeigenvalue(max)andtrace(trace)statisticsforaVARmodel

withaconstantandwithouttrend.

From Table 8 inAppendixC, it is seen that there are two co

integratedvectorsinordertoexplainthelongrunrelationbetweenthe

housingvariablesatthe0.05significancelevel.

1Thisanalysiscanbeinterpretedasroboustnesscheck.

61

TherestrictionsarethesamewithTable5whereCE1standsfor

thedemandsideandCE2standsforthesupplyside.Bylookingatthe

results of Table 9 in Appendix C, the first restrictions are not

appropriate for this analysis; at the 99% confidence level, the second

restriction is appropriate and third restriction is significant at 75%

confidencelevel.Therefore,sincepvalueofthesecondrestrictionisthe

highest,thesecondoneisusedduringthisanalysis.

By using the second type of restrictions and normalize it, the

associatedresultsarefound,showninTable10(AppendixC).

Theseresultsprovideustoanalyzethemarketinthelongrunby

using IstanbulConstructionCost Index.So,according to theseresults,

allthevariablesformingthesupplyanddemandsidesaresignificantin

99%confidencelevelinthelongrun.Inthedemandsideofthehousing

market, value and interest ratehave a negative relationshipwith the

quantitydemanded.Ontheotherhandpopulationandincomehavea

positive relationship with the quantity demanded. These signs are

consistentwithourexpectationsaswellastheeconomictheory.

62

Inthesupplysideofthehousingmarket,allthesignsarealsoas

expected.Asprices increase, thequantity supplied rises;on theother

handcostsandinterestratesarenegativelyrelated.

Whenwelookatthevectorerrorcorrectionmodelresults(Table

11AppendixC)whichprovideustoobservethemarket intheshort

run,we can see thatmost of the variables are not significant in the

shortrun.Errorcorrectiontermthatiscomingfromdemandequation,

the lag ofquantitydwellings and lag ofpopulation are significant at

90% confidence level.Despite the insignificance,we cannotice in the

shortrunthepriceoftheresponse(coefficientis5.4x106)islessthanin

the longrunwhenIstanbulConstructionCostIndex isused.Thisalso

supportstheideasofthemodelofTopelandRosen(1988).Thisshows

againthattheadjustmentcostsforachangeinthehousingmarketare

significantinTurkey.

Thedifferencebetweentheanalysis,madebyusingTurkeyCost

Index,andtheanalysis,madebyusingIstanbulCostIndexcanbeseen

in the shortrun results. In the former one, all error termswith real

interestrateandtherealcostaresignificantwhileinthelatteranalysis,

theerror termcoming fromdemandequationand the lagofquantity

63

and lagofpopulationaresignificant.On theotherhand, the longrun

analysis

6.1.2LogarithmicFormAnalysis

Inthissection,allthevariablesaretransformedintologarithmic

termsthenthesameanalysiswithsection7.1.1isapplied.Sinceallthe

variables are in the same order, we can test whether there is a

cointegrationrelationbetweenthevariablesbyusingboththeJohansen

maximumeigenvalue(max)andtrace(trace)statisticsforaVARmodel

withaconstantandatrend.

AccordingtotheTable11inAppendixC,atthe0.05significance

level, twocointegratedrelationshipbetween thesevariablesare found

outinthelongrun.

Therestrictionsareasdescribed in7.1.1.,whereCE1stands for

thedemandside,CE2standsforthesupplyside.InTable12(Appendix

C), the first and second restrictions are significant at 1% confidence

64

level.However,thesecondrestrictionisappropriateat99%confidence

level.

Byusingthesecondtypeofrestrictionsandnormalizeit,wecan

findtheresultspresentedinTable13(AppendixC).

Theseresults(Table13)provideustoanalyzethemarket inthe

longrun elasticities. So, according to these results, all the variables

formingthesupplyanddemandsidesaresignificantin99%confidence

level in the longrun.Thepriceelasticityofdemand is 4.97while the

priceelasticityofsupplyis1.5inthelongrun.Thesecoefficientsmean

that when the prices increase by 1%, the demand to the buildings

decrease by 4.97% on the other hand the supply of the buildings

increaseby1.5%.Furthermore,theincomeelasticityofdemandis10.28,

thatis,whentheincomeofahouseholdincreaseby1%,thedemandof

buildingsincreaseby10.28%.Thishighcoefficientshowsusthatwhen

the income of a household increases, the buying a house iswidely

preferredinTurkey.

65

Whenwe look at the vectorerror correctionmodel results in

Table14 (AppendixC)whichprovideus toobserve themarket in the

shortrun, we can see that the logarithmic forms of real price, real

interestrate,realcostandrealincomeareinsignificant.Converselythe

error correction terms that are coming from demand and supply

equations,thelagofquantitydwellingsandlagofpopulationwiththe

constant term are significant at 99% confidence level. Despite the

insignificance,wecannoticethepriceelasticity is0.13,meaningwhen

thepriceofadwellingincreasesby1%,thequantitysuppliedincreases

by0.13% in the short run.This shows that therearehighadjustment

costsforachangeintheshortruninTurkeysinceforinstanceinUSA

the shortrunprice elasticityof supply is1.0 (TopelandRosen, 1988)

while it is 3 in the longrun during the period 1963 to 1983 with

quarterlydata.TheseresultsareallconsistentwiththemodelofTopel

andRosen(1988).

According toHakfoortandMatsiyak (1997), inNetherlands the

shortrun price elasticity of supply is 2.3 while the longrun price

elasticityofsupplyis6overtheperiod1977to1994withquarterlydata.

On theotherhand,Follain (1979) finds the longrunpriceelasticityof

supply for United States as 1.48 over the period 1947 to 1975. In

66

additionDipasquale andWheaton (1992) finds the price elasticity of

supplyforthelongrunasatleast1.2byusingtheirconstructedmodel.

Inmostof thedeveloping countries,housingmarketdatadoes

not exist completely, however for United States and within United

States; data about housingmarket data is one of themost available.

Hence there isahuge literatureabout finding thepriceelasticities for

UnitedStates.AccordingtoPalmquist(1983),intheshortruntheprice

elasticity of demand is approximately unitary while the income

elasticityisinelasticforUnitedStates.Reichert(1990)findstheincome

elasticity of demand is 3.78 inUnited States over the period 1975 to

1987,withquarterlydata.Heexamines thepriceelasticityofdemand

for the specific regions in United States and finds that the price

elasticityofdemandchangesbetween0.13and0.22withinthecountry.

Green et. al. (1999) estimate the elasticity of housing supply

based upon contemporaneous price change for 44 United States

metropolitan areas over the period 1979 to 1996. According to his

findings,thepriceelasticitiesareintherangeof38.6to0.6

67

6.2.LimitationsofResults

This is the firststudy thatattempts toanalyze thedemandand

supply relationships in the real estate market of Turkey using a

structuralmodel.However there are serious limitations to this study

due to lack of appropriate data. For instance, since the realmarket

housingpricedoesnotexistforTurkey,valuepereachdwellingisused

asaproxyfortheprice.Thevalueisakindofcostthatistakenfromthe

builderwithoutinterestedinwhatthematerialsareandhowmuchthe

labor costs to thebuilder.So, thevalueper eachdwellinghasahigh

correlationwith the cost index;however theyarenot the same.They

haveslightdifferenceswhicharediscussedindetailinSection6.1.

Anotherlimitationofthisstudyisthatinordertousethevalue

data,numberofallthebuildings,suchasresidential,commercial,social

culturalbuildings,aretakenasthequantityofbuildings.Asaresultof

this restriction,we cannot focuson thedynamicsofhousingmarket.

Furthermore, the number of buildings data is constituted annually,

whichmeansforalongperiod,1970to2007,only38dataexists.Infact

thenumberofdwellingsstartsfrom1961buttheinterestratedoesnot

existbeforetheyears1970.

68

Inmy empirical framework, all the variables are transformed

intoreal termsbyusing theFishersrule.Fishersrule isbasedon the

real interest rate, nominal interest rate and inflation expectations.

However, in Turkey the expectation survey data starts in 2001 in

Turkey. So, in this study it is assumed that inflation expectations are

equaltotheactualinflation.

69

CHAPTER7

CONCLUSION:

Thisstudyhasattemptedtomodelthedemandandsupplysides

of the Turkish real estate market using a structural model and an

econometric framework which clearly distinguishes the long and

shortruninformationamongarelevantsetofeconomicvariables.

In this study Topel and Rosens (1988) housing demand and

supplymodelsareusedduetothefactthat inthesemodelsshortand

longrun elasticities are different; shortrun price elasticity is more

inelasticwhich fits the Turkish real estatemarket structure since the

adjustmentcostforshortrunequilibriumishigh.

70

In addition, since all the variables used in this study are of

integrated of order 1 (I(1)), in order not to lose information by

differencing data, cointegration analysis is found appropriate to be

used. Johansen Cointegration test is preferred because there has not

been found a significantweaknesson this test so far. In addition the

VectorErrorCorrectionModel (VECM) isused to find the shortrun

relationsby imposing some restrictionsonVARmodel.Furthermore,

VECMtakesintoaccountthelongrunrelationswhilefindingshortrun

relations,whichisconsistentwiththeTopelandRosen(1988)housing

investmenttheory.

Inthisstudysincethemarketpriceofahousedoesnotexist in

Turkey,thevalueisusedasaproxyforprice.Inaddition,thevalueof

buildings isnotdivided into theuseofbuilding types,sowecannot

observe thedynamicsof theresidentialbuildingsbut thedynamicsof

realestatemarketintheaggregate.Furthermore,becausethenumberof

buildingsdata is formedannually,annualdata isused for theperiod

1970to2007

All the variables, which are taken from Turkish Statistical

Institute,are transformed into real formsbyusingFishers ruleunder

71

theassumptionofactualinflationisequaltotheinflationexpectations.

Theempiricalstudyisdividedintotwogroups,leveldataanalysisand

logarithmicformanalysis.Accordingtothebothoftheanalysis,interest

rate, value, income and population are found to be significant in

explainingthequantitydemandedofdwellingsinthelongrunwiththe

expectedsignsandforthesupplyside,value,costandinterestrateare

foundtobesignificantinexplainingthequantitysuppliedintheshort

runwith the expected signs.On theotherhand, in the shortrun, the

variables,thosearesignificant,aredifferentforthetwoanalyses.

According to the results of the logarithmic form analysis, the

longrun price elasticity of supply is 1.5 while the shortrun price

elasticityof supply is0.13.This shows that therearehighadjustment

costs for a change in the shortrun in Turkey. These results are all

consistentwiththemodelofTopelandRosen(1988).Furthermore,the

longrunpriceelasticity is 4.97which ismoreelasticcomparingwith

the longrun price supply elasticity, that is, consumers are more

sensitivethatthebuilders.

72

SELECTBIBLIOGRAPHY

Ayuso,JuanandFernandoRestoy.2006.HousePricesandRents:AnEquilibriumAssetPricingApproachJournalofEmpiricalFinance,13/3:371388

Binay, kr and Ferhan Salman. 2008. ACritique on TurkishReal

EstateMarket, Turkish Economic Association, Discussion Paper2008/8

Campbell,JohnY.andPierrePerron.1991.PitfallsandOppurtunities:

WhatMacroeconomicstsShouldKnowAboutUnitRootsNBERTechnicalWorkingPaperNo.100

Case,KarlE.andRobertShiller.1989.TheEfficiencyoftheMarketfor

SingleFamily Homes The American Economic Review, Vol.79,No.1:125137

Davidson,J.H.,Hendry,D.H.,SrbaF.andS.Yeo.1978.Econometric

Modelling of theAggregate TimeSeries Relationship BetweenConsumers Expenditure and Income in theUnitedKingdomEconomicJournal,88:661692

73

Dipasquale,Denise.1999.WhyDontWeKnowMoreAboutHousingSupply?JournalofRealEstateandEconomics,18:1:923

Dipasquale,Denise andWilliamC.Wheaton. 1994. HousingMarketDynamics and the Future ofHousing Prices Journal ofUrbanEconomics,35:127

Enders,Walter.1995.AppliedEconometricTimeSeriesJohnWiley&

Sons,Inc.

Engle,Robert F. andC.W.J.Granger. 1987. Cointegration and Error

Correction: Representation, Estimation and TestingEconometrica,Vol.55,No.2:251276

Eraydin,Ayda,AliTrelandAlperGzel.1996.KonutYatrmlarnn

EkonomikEtkileriT.C.BabakanlkTopluKonutdaresiBakanl,KonutAratrmalarDizisi3.

Erol, Isil and Kanak Patel. 2004. Housing Policy and Mortgage

Finance inTurkeyDuring theLate1990s InflationaryPeriod ,InternationalRealEstateReview,Vol.7No.1:98120

Eisner,Robert andRobertH. Strotz. 1963. Determinants ofBusiness

Investment in Impacts of Monetary Policy, by Commission onMoneyandCreditEnglewoodCliffs,H.J.PrenticeHall

Follain,JamesR.Jr.1979.ThePriceElasticityoftheLongRunSupply

ofNewHousingConstructionLandEconomics,Vol.55

Friedman,Milton. 1963. Price Theory: A Provisional Text Chicago:

AldinePublishingCompany

Gallin,Joshua.2006.ThelongrunRelationshipbetweenHousePricesand Income:Evidence fromLocalHousingMarketsRealEstateEconomics,Vol.34/3:417438

74

Gould,JohnP.1968.AdjustmentCostsinthetheoryofinvestmentofthefirmReviewEconomicStudies,35:4755

Granger, C. W. J. and P. Newbold. 1974. Spurious Regressions inEconometricsJournalofEconometrics,2:111120

Green, Richard K., Stephen Malpezzi and Stephen K. Mayo. 1999.Metropoplitian Specific Estimates of the price elasticity ofhousingandtheirresourcesDraft

Gurbuz, Ayhan. 2002. potekli Konut Kredisi ve Trkiyede

UygulamasTCMBUzmanlkTezleri

Hakfoort,JaccoandGeorgeMatysiak.1997.HousingInvestmentinthe

NetherlandsEconomicModelling,Vol.14:5015

Holly, Sean andNatasha Jones. 1997. House prices since the 1940s:

Cointegration, demography and asymmetries EconomicModelling,14/4:549565

Johansen, Soren and Katarina Juselius. 1990. Maximum Likelihood

EstimationandInferenceonCointegrationWithApplicationstothe Demand for Money Oxford Bulletin of Economics andStatistics,Vol.52/2:169210

Johansen,S.1988.StatisticalAnalysisofCointegrationVectorsJournal

ofEconomicDynamicsandControl,12:231254

Keles, Rusen. 1990. Housing Policy in Turkey inHousing Policy in

DevelopingCountries,GilShidlo:140172

Kennedy,Peter.2003AGuidetoEconometricsBlackwellPublishing

75

Kenny,Geoff.1999.AsymmetricAdjustmentCostsandtheDynamicsof Housing Supply Central Bank of Ireland, Technical Paper3/RT/99.

..1999.Modelingthedemandandsupplysidesofthehousing

market:evidence from IrelandEconomicModelling,Vol.16:389409

. 1998.The Housing Market a nd the Macroeconomy:

Evidence from Ireland.CentralBankof Ireland,TechnicalPaper1/RT/98

Leeuw,Frankde.1971.TheDemand forHousing:AreviewofCross

Section Evidence The Review of Economics and Statistics,Vol.53,No.1:110

Lucas,RobertE., Jr.1967.Optimal InvestmentPolicyand theFlexibl

AcceleratorInternationalEconomicReview,8:4755

MacKinnon, J.