THE INCOME AND PRICE ELASTICITY OF DEMAND FOR HOUSING IN GHANA: EMPIRICAL EVIDENCE FROM HOUSEHOLD LEVEL DATA

Transcription

1 160 SAJEMS NS 19 (2016 No 2: THE INCOME AND PRICE ELASTICITY OF DEMAND FOR HOUSING IN GHANA: EMPIRICAL EVIDENCE FROM HOUSEHOLD LEVEL DATA Francs Tandoh Department of Economcs, Unversty of Zululand Dev Datt Tewar Faculty of Commerce, Admnstraton and Law, Unversty of Zululand Accepted: January 2016 Abstract Housng s a challengng ssue n Ghana, due to the rsng demand and sluggsh supply whch has led to a defct of more than two mllon housng unts. Ths study amed to estmate and analyse the determnants of the demand for housng n Ghana. The estmated elastctes show that owner and rental demand for housng s prce and ncome nelastc. Permanent ncome elastctes were greater n each case than current ncome elastcty. The quantle regresson showed that permanent ncome, current ncome and prce were sgnfcant for all quantles of housng unts consumed. It s recommended that all these factors be taken nto account when addressng the housng supply challenges facng Ghana to help clear the exstng defct and to provde for the antcpated ncrease n demand due to ncreasng ncome, snce demand for housng n the country s ncome nelastc. Key words: demand, elastcty, Ghana, hedonc, permanent ncome, quantle JEL: D11, R21 1 Introducton Demand for housng 1 n Ghana s ncreasng progressvely as a result of demographc, economc and socal factors. The ncreasng populaton s one of the factors whch accounts for rsng demand n the country. Ghana s populaton grew from a lttle more than 2.3 mllon n 1921 to 26.3 mllon n Populaton estmates for 2014 show that close to 56 per cent of the populaton s between the ages of 15 and 60 years old (Ghana Statstcal Servces [GSS], Smth and Searle (2010 argue that youthful populatons ncrease demand for housng. Fgures from GSS (2012, show that the growng populaton, coupled wth a reducton n household sze, has resulted n an ncrease n the number of households. The number of households nearly doubled wthn 14 years, from 3.7 mllon n 2000 to 6.6 mllon n Increasng ncome per capta and mproved access to educaton, partcularly tertary educaton, also account for the rsng demand for housng n Ghana. Gross Domestc Product (GDP per capta ncreased from US$218 n 1970 to $1841 n 2013 (GSS, As ncome rses, demand for housng also ncreases for two man reasons: the nvestment motve and the ncrease n the demand for housng servces. Fnally, changng cultural norms and the farly stable poltcal and economc envronment n Ghana also contrbute to the ncreasng demand for housng. Hgh nflaton rates coupled wth the contnued deprecaton of the currency motvate Ghanaans to nvest n real estate as a hedge aganst rsng nflaton, thus ncreasng the nvestment demand for housng (Bank of Ghana [BoG], For nstance, the natonal currency of Ghana, the Ghana Ced, deprecated by 31 per cent aganst the US dollar n the year Inflaton n the same perod was 17 per cent and house prces ncreased by 35.2 per cent n the same perod. Investng n property s used as a hedge aganst the hgh nflaton rate, so Ghanaans would demand a lot of housng unts, not only for consumpton but for nvestment purposes as well. How to cte DOI: ISSN:

2 SAJEMS NS 19 (2016 No 2: Despte the growng demand for housng, supply s not able to respond approprately. For nstance, n the year 2000 the demand for housng outstrpped supply by about 1.5 mllon housng unts. 2 Ths ncreased to over two mllon n a perod of 14 years n 2014 as shown n Table 1-1. Ths s due partly to the fact that most housng s suppled by ndvdual households who take between fve and 15 years to buld a sngle unt (BoG, It s estmated that 4.9 mllon Ghanaans lve n slums scattered around the major ctes and towns and that such settlements are growng at a rate of 1.83 per cent per annum (Government of Ghana [GoG], Table 1 Demand and supply of housng n Ghana Year Housng demand Housng supply Housng defct ,678, , , ,410,096 1,226,360 1,184, ,708,250 2,181,975 1,526, ,467,136 3,392,745 2,074, ,601,500 4,207,003* 2,394,497 Source: GSS (2012 & 2014 In spte of the extensve lterature on the demand for housng n developed countres and some developng countres, there s a paucty of emprcal studes on Ghana. Malpezz and Mayo (1987 and Asedu and Kagaya (1991 conducted emprcal studes on demand for housng n Ghana, but each focused on a sngle cty, so ther fndngs cannot be generalsed to the whole country. Besdes, these studes are too old to represent current realtes. Although much has been wrtten on the housng market n Ghana (Obeng-Odoom, 2011, 2010, 2009; Karley, 2008; Asedu & Arku, 2009, none of these studes attempted to emprcally estmate the elastcty (the degree of responsveness of dependent varable subject to a unt change n the ndependent varable of the varous determnants of the demand for housng. Ths study therefore flls ths gap n the lterature. Second, the desgn and mplementaton of approprate urban plannng polcy depend heavly on prce and ncome elastcty for estmatng future ncreases n the demand for housng. Malpezz and Mayo (1987 argue that varatons n the assumptons of ncome and prce elastcty ncrease housng subsdy costs by 73 per cent and lead to a deadweght loss to socety of 400 per cent. Therefore demand elastctes are a vtal nput n effcent and approprate housng polcy outcomes. Ths study ams to provde these elastctes for Ghana. Thrd, almost all demand estmatons n lterature utlse the OLS or other mean-based regresson methods that may clad the effects of the varous quantles of housng demand, thereby ether underestmatng or overestmatng the ncome and prce elastctes of the varous consumer unts; thus wrong estmates can lead to polcy falure. Ths study provdes the ncome and prce elastcty of the varous quantles of housng consumpton to ad effectve housng demand polcy and plannng n Ghana. In developed countres, almost every house s tradable but ths s not the case n Ghana. Ownershp of some houses s not clearly defned: the house belongs to the past, current and future kn of the famly. No ndvdual has the power to trade such houses, for any reason. These houses sometmes serve as the famly s source of unty. They exst for consumpton, socal and cultural purposes and not as an nvestment. The socal relatons of housng cause some households to reman n the same house, no matter what the sale of that unt would have fetched for the household and how much t would have fetched f redesgned. Although the socal relatons of housng have dwndled n the bgger ctes (UN Habtat, 2011, households n the majorty of towns and vllages stll clng to ths belef. Ths clearly llustrates that n developng and some emergng economes the functonng of the housng market s not the same as n developed countres. Hence the demand condtons are dfferent. The rest of the paper s organsed as follows: Secton Two outlnes the theoretcal and emprcal framework; Secton Three presents the owner demand estmates; Secton Four presents a

3 162 SAJEMS NS 19 (2016 No 2: dscusson for renter demand estmates; Secton Fve comprses a comparatve analyss of elastctes; whle Secton Sx concludes the study. 2 Theoretcal and emprcal framework The theoretcal framework used housng as a consumer good as elaborated n Zabel (2004. Ths theory s premsed on three fundamental assumptons: that the consumer optmses utlty gven prce and ncome constrants; the consumer chooses an unobservable homogenous commodty called housng, and that a perfectly compettve housng market exsts. The ndvdual household n market j wth the utlty functon (U depends on the quantty of housng consumed (q, consumpton of other commodtes (C and household demographc characterstcs (A, such that the utlty for each household s assumed to be constant, although households have dfferent demographc characterstcs. The consumpton decson of an ndvdual household s therefore charactersed by the maxmsaton of the ndvdual s utlty functon subject to the budget constrant stated as: Maxmse utlty Max U( q, C, A Subject to: qc j j j j (1 Budget constrants Ij = Cj + pj q (2 where I s the ncome of the household, the prce C (other commodtes s normalsed to one and p j s the prce ndex of housng whch s allowed to dffer across markets. Solvng the budget constrants n terms of C j gves Cj = Ij pj qj substtutng equaton (3 nto the utlty functon (1 generates wj = Max : U( Ij p j. qj, qj, A (4 where w j represents the ndrect utlty functon. A soluton of equaton (4 gves the nverse demand functon (3 p = w q w (5 I Ths therefore gves the demand functon of housng of the form: qj = ( I, pj, A It s beleved that the demand for housng depends on ncome, prce and the demographc characterstcs of the household. Ths has been the fundamental approach n estmatng demand (Megbolugbe, Marks & Schwartz, Equaton (6 s the demand functon to be estmated f all the varables are known. In realty q s unobservable rather than the value of a house (v, whch s a functon of the characterstcs (z of the housng unt (number of rooms, lot sze, etc. and a parameter whch allows for varaton n the values across locaton (ϕ. Therefore the values of housng unt k consumed by household n a market j can be stated as: v kj = v( z k, φ j (7 Equaton (7 s the hedonc functon expressng the value of a house based on ts characterstcs. From equaton (7, one can obtan the parameter ϕ j usng hedonc prce modellng gven the observed value of v, and z. Addtonally f z s defned as a standard unt, then the prce kj k * k (6

4 SAJEMS NS 19 (2016 No 2: ndex can be constructed as: * v( z j ; φ j p j = 100 (8 * v( z j ; φ1 The prce ndex s constructed based on the predcted value of the hedonc estmatons for the locaton or market, where the prce ndex for a market where j = 1 s taken as the base. Ths therefore means that the value of a housng unt (v k n a housng market (j for an ndvdual ( can be expressed as: v kj = q j. p j (9 Equaton (9 postulates that the value of a housng unt n a gven market s the product of the prce and the quantty of housng unts consumed. Therefore the quantty of housng unts consumed s shown as: v kj q j =. (10 p j Now that q j, P j are known and the ndvdual characterstcs are observable, t s expected that equaton (6 can confdently be estmated. Prevous studes (Zabel, 2004 advocate usng permanent rather than current ncome. There s a rch lterature on the demand for housng as a consumer good. Ths lterature focuses on dfferent thematc areas: demand for housng servces; demand for housng characterstcs; the smultaneous demand for housng quantty and tenure; and demand for housng space and dwellng type. The demand for housng servces s modelled as the servces per unt of housng produced by a housng unt as homogenous goods, called housng servces (Rouwendal, Studes n ths category nclude Polnky and Ellwood (1979 who reconcled the dfferences between mcro and aggregate data n explanng the demand for housng servces, whle Zabel (2004 argued that estmated prce and ncome elastctes of demand for housng were ambguous because there was no clear defnton of housng servces. To resolve ths ambguty and to demonstrate that a coherent measure of housng servces requres a constant prce n all markets whether or not neghbourhood characterstcs are ncluded, an emprcal study that classfed housng demand determnants nto neghbourhood characterstcs and structural characterstcs was conducted. Ths approach, based on demand for housng characterstcs, examnes the specfc attrbutes and neghbourhood characterstcs of housng besdes ncome and prce as determnants of housng. Ioanndes (2002 argued that a neghbourhood has a substantal effect on the demand for housng and that the socal nteracton effect s much stronger than a household s prevous consumpton. Ioanndes and Zabel (2008 jontly modelled the demand for housng and neghbourhood choce. Ths enrched Ioanndes and Zabel s (2003 study that modelled the demand for housng on the assumpton that neghbourhood choce was exogenously determned (neghbourhood choce was determned outsde the housng demand framework. They argued that jont modellng of housng demand and neghbourhood choce leads to a consstent estmate despte ts complextes, and concluded that gnorng the smultaneous nature of neghbourhood effects and demand for housng based the estmates. The thrd category of housng demand s the smultaneous demand for quantty of housng and tenure choce. Rapaport (1997 stated that a household's housng demand decson was a smultaneous choce between prce and quantty gven the communty of choce, and that the communty factor was usually omtted from housng demand models. Therefore the demand for housng was modelled as a 14, dependent on the communty and/or tenure status decson and that the prce of housng was communty- and tenure-specfc. House prce reflected the avalable supply of housng gven the communty and the tax rate, whle the prce of housng was endogenously determned. Ras, Gameren and Eggnk (2005 used a three-step procedure to

5 164 SAJEMS NS 19 (2016 No 2: analyse the demand for housng n the Netherlands. The quantty of housng servces demand was smultaneously modelled for owner occupers and renters. Garca and Hernandez (2008 jontly modelled housng tenure and housng type usng mcro data from Span. Demand for housng space and dwellng type assumes that the consumer measure the attrbutes of each of the avalable alternatves. Bajar and Kahn (2005 estmated households wllngness to pay for housng attrbutes and tenure as well as the dwellng type. Ths study adopts the frst category and borrows from Zabel (2004 and Fontenla and Gonzalez (2009 n the emprcal model to be estmated. Ths s due to the fact that ths model s the most approprate n the lght of the avalable data. Furthermore, the emprcal estmates for the second category requre extensve data on neghbourhood characterstcs. Ths presents many complextes n the estmaton procedure (Ioanndes & Zabel, The thrd category requres the smultaneous estmaton of the housng demand and communty choce, but due to the extensve nature of the data requred to do so, t s practcally mpossble to adopt ths methodology. The fnal category requres usng panel data, so that the choce of households can be observed over tme. Ths requrement makes t mpossble for ths study to adopt ths category. Recent studes ndcate that the determnants of housng demand n some developng countres were arrved at utlsng the frst category. Therefore, ths category was adopted for ths study due to ts flexblty wth cross sectonal data (Twar & Parkh, 1999; Fontenla & Gonzalez, The emprcal estmaton of the demand for housng follows three steps: permanent ncome s estmated n lne wth the lterature (avalable upon request; the hedonc prce model for housng s estmated to help construct a general prce ndex of housng n Ghana (also avalable upon request; and fnally, the demand for housng model s estmated gven the quantty, prce and other varables. The quantty of housng consumed (q j from equaton (6 s stated as a functon of permanent ncome (I p, transtory ncome (I t, and housng-prce (P j, and other varables are specfed n logarthmc terms. That s, the estmated demand equaton s of the form: ln( H QTY j = β0 + β1 ln( I P + β2 ( I T + β3 ( H ndex + β4 ( HH age + β5 ( HH WED + β6 ( HH SIZE + β7 ( HH SEX + β8 ( HH ELE + β9 ( HH GRA + β ( HH + β ( HH + β ( L + ε 10 UNI 11 SEC 12 URB Heteroskedastcty-consstent standard error estmators for OLS regresson (hc3 were employed (Hayes & Ca, 2007; Long & Ervn, Table 2 provdes a descrpton of the varables. The subscrpt represents the th household. The varables prefxed wth HH denote varables that descrbe the head of the household. The varables prefxed wth H denote those that descrbe the housng unt and those prefxed wth L descrbe locatonal varables. Varable H QTY I C I P I T H INDEX HH WED HH SIZE HH SEX HH ELE HH GRA HH SEC HH UNI L UR Source: Author Table 2 Descrpton of housng demand varables Quantty of housng unts consumed by a household Current ncome of the household Permanent ncome of the household Transtory ncome of the household Estmated prce ndex of housng Descrpton A dummy 1 f the household head s marred and 0 otherwse The total number of people n the household A dummy 1 f the gender of the household head s male and 0 otherwse A dummy 1 f the hghest educaton of the household head s elementary school and 0 otherwse A dummy 1 f the hghest educaton of the household head s a post graduate (master s and doctorate and 0 otherwse A dummy 1 f the hghest educaton of the household head s a secondary educaton and 0 otherwse A dummy 1 f the hghest educaton of the household head s unversty degree and zero otherwse (Honours Household stays n the urban localtes (11

6 SAJEMS NS 19 (2016 No 2: Permanent ncome equaton estmaton Varous authors have used permanent ncome as the deal ncome for housng demand estmaton wth ts justfcaton long establshed (see Ioanndes & Zabel (2003; Fontenla & Gonzalez (2009. Current ncome I s made up of transtory (I T and permanent (I P ncome: I = I P + I T Therefore, to obtan permanent ncome, current ncome of the household s regressed on the household demographc characterstcs. The functonal equaton s therefore stated as: ' ln I = a + X β + Dγ + ε (13 Where I t s the observed ncome of the household, x " # s the personal characterstcs that determne permanent ncome, and D s a vector of regonal dummes. The predcted values become the permanent ncome whle the resduals become the transtory ncome. 4 Owner-occuper households 4.1 Descrptve statstcs The descrptve statstcs show that, on average, Ghanaans consume 3,434 unts of housng per year wth a standard devaton of The skewness value shows that a smaller number of households consume more housng unts than the average unt. Ths means that the quantty of housng consumed s not evenly dstrbuted. The kurtoss value confrms the non-normalty of the quantty of housng unts demanded. The average household sze of fve shows that owner-occupers have bgger households compared wth other tenure status consdered n ths study, snce the overall average household sze n Ghana s four. The statstcs for graduate educaton are hghly skewed, showng only a few graduate household heads. The coeffcent of skewness shows that the ncome of most of these households falls below the mean ncome. Owner occupaton was hgher n the rural areas (58 per cent than the urban areas (42 per cent. Fnally, Table 3 shows that 67 per cent of owner-occuper household heads are marred. The mean prce ndex for owner-occuped households was and the average housng unts consumed by a household was 7.5. Owner-occuper households have an average current monthly? annual? ncome of about Gh 600eds (equvalent of US$600 wth a standard devaton of 2.6 and a permanent average ncome of Gh 650 wth a standard devaton of 1.5. Ths shows that there s more nequalty n current ncome than there s n permanent ncome. Varable Observaton Mean Standard devaton Table 3 A descrpton of owner occuper n Ghana Mnmum Maxmum Skewness Kurtoss I C I P L URB H INDEX H SIZE H QTY HH AGE HH GRA HH ELE HH SEX HH WED Source: Estmaton (12

7 166 SAJEMS NS 19 (2016 No 2: Owner occuper elastctes The estmated elastctes for owner-occuper households show that sx varables are sgnfcant determnants of demand for owner-occuper housng demand. Except for prce, all the sgnfcant varables were postve, as shown n Table 4. The estmated permanent ncome elastcty shows that demand for housng n Ghana s ncome nelastc (housng demand s less responsve to changes n permanent ncome. Ths could be due to three possble reasons. The property market n Ghana s currently n ts nfancy. There s also no socal housng, snce the houses that dd exst were dvested or reserved for publc sector workers. The only alternatve to ownng a home s prvate rental housng. Ths alternatve presents many problems, ncludng substandard housng, hgh rent advances and sgnfcant shortages n some areas, leadng to hgher rents. Ths suggests that the best opton for a household s to move to owner-occupancy whch they can do f ther ncome ncreases. The next possble explanaton for the lack of ncome elastcty s borrowed from Brounen et al. (2012, who argued that homeowners beneft from hgh nflaton snce house prces rse n an nflatonary perod and the value of owner-occupers property also ncreases. Snce Ghana s a hghly nflatonary country, households demand owner- occupaton as a hedge aganst nflaton as well as for the captal gans assocated wth t. The fnal reason s the need to hedge aganst the rsk of rent ncreases. Sna and Souleles (2003 argued that the hedge aganst future hgh rent payment forces households to move to owner-occupaton. The acute shortage of housng unts n Ghana especally standard, qualty housng means that households who rent, pay hgh rents. In order to hedge aganst hgher rents n the future, households would move to owner-occupancy wth an ncrease n ncome. Ths fndng corresponds wth the fndngs of other studes. For example, Malpezz and Mayo (1987 found owner ncome-elastcty for Egypt to be 0.17, whle Mehta and Mehta (1989 estmated owner-ncome elastcty for Inda to be 0.2. The quantle estmaton shows that the ncome elastcty of the demand for the varous quantles of housng unts s less than the OLS (mean estmates. The ncome elastcty of the lowest 25 per cent of quantles s greater than the ncome elastcty for the top 25 per cent of housng consumers, although the elastcty for both s less than the medan elastcty (0.5Q shown n Table 4. The quantle elastctes show the need to account for the dfferent unts n the elastcty estmaton so that relable polcy nferences can be drawn from t wth lttle varaton. The prce elastcty of demand for owner-occupaton shows that the prce of housng does not necessarly affect the demand for owner-occuper housng n Ghana sgnfcantly. Ths could be ascrbed to the mode of housng fnance systems n Ghana. Most households fnance ther houses wth savngs or use the ncremental housng method; 4 therefore an ncrease n the prce of housng does not affect the consumpton of housng snce the nterest rate effect does not come nto play. Increases n house prces n Ghana are the result of ncreases n the value of land and the cost of buldng materals (Bank of Ghana, As such, an ncrease n house prces leads to ncreases n the cost of alternatves or substtutes for ownng, that s, n prvate rental prces whch are already very hgh n Ghana due to the acute shortage of housng and the hefty requrement of twoto fve-year advance payments. An ncrease n the prce of housng leads to a slowdown n the rate at whch households add to ther buldng, thus restrctng the quantty of housng unts owneroccupers can consume, although ths wll be very margnal. The dynamcs of the housng market do not make the usual shft from owner-occupaton to rental housng easy: the costs nvolved also deter most owner-occupers n Ghana from mantanng ther status despte house prce ncreases. Therefore the elastcty truly reflects the fundamental characterstcs of the housng market n Ghana. The estmated prce elastcty s less than the lterature for developng countres proposes. Mayo (1981 proposed 0.38 to 0.37 and the recent fndngs of studes by Fontenla and Gonzalez (2009 confrmed ths. Follan, Lm and Renaud (1980 acheved a near zero prce elastcty of owneroccupaton n South Korea, whle Malpezz and Mayo s (1987 estmate for owner- occuped elastcty n Caro and Manla was close to 1 or unty n absolute terms. The prce elastcty

8 SAJEMS NS 19 (2016 No 2: estmates were also n lne wth current estmates for Spanwhch range from to (Garca & Hernandez, 2008 and the USA where Zabel (2004 found elastcty of usng 1993 data and for 2001 data. The varaton n the estmates of the prce elastcty s due to varatons n the defntons of the prce and the models used n the estmaton, as well as the procedure and data used n obtanng the prce ndex (Zabel, Mayo (1981 argued that the dfferent demographc characterstcs across countres make t dffcult to make comparsons, therefore t s preferable to restrct dscusson to the country under study. The OLS estmates show that the gender of the household head, household sze and the age and martal status of the head of the household were not sgnfcant determnants of the demand for housng n Ghana. However, the quantle estmates show that the gender of the household head was nsgnfcant only n the lowest 25 per cent of housng consumers. The sze of the household was sgnfcant n all the quantles despte not beng sgnfcant n the OLS estmaton. The elastcty from the level of educaton shows that elementary educaton was an nsgnfcant determnant of the demand for housng n the top 25 per cent of consumpton unts but unversty educaton was sgnfcant at all levels of consumpton. Table 4 Owner occuper demand elastcty (OLS and quantle Varable OLS 0.25Q 0.5Q 0.75Q HH SEX HH SIZE ***.0076***.0058* HH AGE *** HH URB.0439***.0306***.0445***.0576*** HH GRA.0003*.0004***.0005***.0002** HH ELE.0027**.0022*.0026**.0021 H INDEX *** *** *** *** # I P.1719***.1548***.1620***.1329** # I C.1682***.1113***.1431***.1311*** HH WED ** R 2 = R 2 = 30 R 2 = 20 R 2 = 22 Ramsey RESET F(3, 5436 = Prob > F = #Permanent ncome (I pand current ncome (I C were estmated n the model separately Source: Estmaton 4.3 Rural and urban owner demand elastctes A breakdown of the demand estmates (both OLS and quantle for the rural and urban locatons n Ghana shows that permanent ncome was the major determnant of the demand for housng n urban localtes, whle current ncome was the major determnant of demand for housng n rural ones. Ths fndng s consstent wth the condtons prevalng n these two localtes. Most of Ghana s rural dwellers are subsstence farmers. Ths means that they have no permanent and consstent source of ncome or expectaton of one; they therefore consume housng based on ther current ncome (usually related to harvests from ther farmng actvtes. In contrast, most urban dwellers work n the formal sector and earn a regular ncome so they have expectatons of a permanent ncome for consumpton purposes. Therefore the permanent ncome elastcty for the urban areas was greater than the current ncome. The permanent ncome elastcty for the medan urban housng unt consumer was wthn the range advocated by Malpezz (1999. The wde dfference n permanent ncome elastcty between urban and rural localtes mght account for the smaller lower level of the overall elastcty reported above. The elastcty of the educatonal varables shows that elementary educaton was only sgnfcant n the urban areas and not the rural ones. Even n the urban areas the quantle estmates show that t determned consumpton only n the lowest 25 per cent of unts. Whle permanent ncome and level of educaton were the major determnants of the demand for housng n urban localtes,

9 168 SAJEMS NS 19 (2016 No 2: current ncome, age of the household head, gender of the household head and sze of the household were the major determnants of the demand for housng n the rural localtes n Ghana (Table 5. Table 5 OLS and quantle elastctes for owner demand (urban and rural locatons Urban Ghana Rural Ghana Varable OLS 0.25Q 0.5Q 0.75Q OLS 0.25Q 0.5Q 0.75Q HH SEX *** HH SIZE * ***.0119***.0090** HH AGE **.0153***.0104***.0157** HH GRA *** HH ELE.0063**.0082** HH INDEX -.087*** -.079*** -.075*** -.086*** -.104*** -.118*** -.107*** -.095*** # I P.3335** **.5986*** # I C.2749***.2464***.2947***.2029***.0952***.0599***.1107***.1070*** HH WED *** R 2 = 0.70 R 2 = 0.43 R 2 = 0.20 R 2 = 0.25 R 2 = 0.56 R 2 =0.17 R 2 = 0.27 R 2 =0.45 #Permanent ncome (I pand current ncome (I C were estmated n the model separately. Source: Estmaton 5 Rental households 5.1 Descrptve statstcs The descrptve statstcs show that rental Ghanaan households on average consumed 5,684 unts of housng per year wth a standard devaton of The value of the skewness s 3.7 whch shows that a smaller number of households consumed more unts than the average. Ths means that the quantty of the housng consumed was not evenly dstrbuted. The kurtoss value confrms the non-normalty of the quantty of rental housng demand. There were extremes n the rental housng market. Statstcs for household sze agan show that sze was not evenly dstrbuted across households. The average household sze of three shows that rental households had fewer members than owner-occupers. Statstcs relatng to educaton show hgh skewness of graduate educaton n Ghana and rentng. Only a few rental household heads had a graduate qualfcaton. In terms of basc educaton, the statstcs show that t was normally dstrbuted across heads of rental households n whch almost half had elementary educaton. Martal status was normally dstrbuted among households (Table 6. Varable Observaton Mean Table 6 Descrptve statstcs of the rental households n Ghana Standard devaton Mnmum Maxmum Skewness Kurtoss H QTY HH SIZE HH SEX HH AGE HH GRA HH ELE L URB H INDEX I P I C HH WED Source: Estmaton

10 SAJEMS NS 19 (2016 No 2: Rental elastctes The rental ncome elastcty shown n Table 7 ndcates that rentng s ncome nelastc n Ghana. Permanent ncome elastcty s greater than current ncome elastcty and both are statstcally sgnfcant. The rental ncome elastcty reveals three mportant aspects of the Ghanaan housng market n general and the rental market n partcular. The hgh postve ncome elastcty shows the rental arrangement whch prevals n the country. To rent n Ghana, one needs to make a payment at least two years n advance. Ths amount often exceeds current ncome so people have to borrow money n order to afford the consumpton of the housng unts they desre. Most Ghanaans use ther permanent ncome to rent, therefore when t ncreases, they ncrease ther level of consumpton, snce they can now pay hgher nstalments on the loans they took out to rent the unt they occupy. Ths elastcty s smlar to the fndngs of Malpezz and Mayo (1987 for Kumas, the second largest cty n Ghana. It also compares favourably wth the fndngs of Garca and Hernandez (2008. The quantle breakdown shows that ncome elastcty for the medan housng consumer s less than the average elastcty (OLS estmate (Table 7. Ths means that the ncome elastcty provded by the OLS model overestmates the elastcty for the majorty of the populaton. Therefore, any polcy based on the OLS estmaton would margnalse the majorty of the populaton of housng consumers n the country. The lower 25 per cent of consumers rent out of current rather than permanent ncome, as ther permanent ncome s not statstcally sgnfcant. Rental prce elastcty estmates show that renters respond massvely to changes n prce. Rental elastcty s clearly greater than the prce elastcty estmate for owner occupers. Ths confrms Goodman and Kawa s (1984 asserton that the hgh transacton cost of swtchng tenure from ownershp to rentng makes the prce elastcty of rentng greater than that of ownng. The estmated renter prce elastcty for ths study corresponds wth the fndngs of Zabel (2004. Table 7 Demand elastcty estmates for renters Varable OLS 0.25Q 0.5Q 0.75Q HH SEX ** HH SIZE * *** * HH AGE HH URB *** *** *** *** HH GRA *** *** *** *** HH ELE * H INDEX *** *** *** *** I P ** *** * I C *** *** *** *** HH WED ** R 2 = R 2 = 0.34 R 2 = 0.44 R 2 =0.36 Ramsey RESET: F(3, 5436 = Prob > F = Source: Estmaton Age was expected to have a negatve effect on rentng snce younger households save to buy a home or to make the payment needed by the mortgage market (Cho, 2010, but t was found not to be statstcally sgnfcant. The explanaton may be that the effect of age on rentng s not mportant n the rental market n Ghana. Inhertance and the long perod of transton from rentng to ownershp (f ndeed t eventually occurs mght be the reason. The age elastcty contradcts the fndngs of Asedu and Kagaya (1991, who undertook a smlar study n Kumas. Whle the current study found age to be statstcally nsgnfcant but negatve, ther study found t postve whch contradcts the lterature but sgnfcant. The reason for ths mght be the more lmted scope of Asedu and Kagaya s (1991 study area compared wth the scope of the current study whch consders the rental sector n the country as a whole. The estmated results show that educaton correlates postvely wth rental housng demand. Ths can be explaned by the fact that most hghly educated people work outsde ther places of orgn. They manly work n ctes and

11 170 SAJEMS NS 19 (2016 No 2: towns where t s expensve to own a home, whch explans why unversty graduates have the hghest postve relatonshp wth rentng. Martal status shows an nverse relatonshp wth rentng, but t was statstcally nsgnfcant except at the medan consumpton level. 5.3 Rural and urban rental demand elastctes The breakdown of rural and urban statstcs confrms the popularty of rentng n urban localtes relatve to rural localtes. Permanent ncome elastcty shows that urban households rental decsons were nfluenced by permanent and current ncome, except at the lowest 25 per cent of housng consumers. These households based ther rental decson only on current ncome. In the rural localtes, however, rental decsons were based on current ncome, except the medan consumer unts (Table 8. The prce elastctes show that both urban and rural rental decsons were manly nfluenced by the prce of rentng. Ths s the only varable that showed sgnfcance n both localtes and at all quantles of housng consumpton. The hghest prce elastcty was recorded at the top 25 per cent of housng consumers. The estmated elastctes show that urban consumers were more senstve to prce changes than those n the rural localtes. The quantle estmates for the two localtes show that whle the top 25 per cent had the hghest elastcty value for prce n the rural localtes, n the urban localtes, the lowest 25 per cent dsplays the hghest elastcty. In the rural areas the elastcty was reduced wth ncreasng quantle, whle n the urban areas t decreased from the lowest quantle, ncreased from the medan quantle, and peaked at the hghest quantle. The results show that unversty educaton, current ncome and prce were the man determnants of rentng n urban localtes. These were sgnfcant n all quantle and OLS estmates for the urban localtes. In the rural localtes, however, prce was the man determnant and, to some extent, current ncome as well. Age was consstently shown to be nsgnfcant n both rural and urban localtes and at all quantles. The sze of the household was only sgnfcant at the medan and top 25 percentle of consumers n rural localtes. The gender of the household head showed postve and sgnfcant effects n all quantles n the urban areas, although t was nsgnfcant n the OLS estmates. In the rural localtes however, only the OLS and the medan estmates for household sze were sgnfcant. The household sze showed a postve effect n the nstances where t was sgnfcant, whch means that male-headed households were more lkely to rent than female-headed households. Rental by graduates was not applcable n the rural localtes and was thus omtted n the estmatons. Elementary educaton, however, was only sgnfcant at the lowest 25 percentle of urban consumers. Martal status was consstently shown to be an nsgnfcant factor n rental housng demand n both rural and urban localtes and at all quantles, except the medan consumer n the rural localtes. Table 8 Rental demand elastcty estmates for rural and urban localtes Urban localtes Rural localtes Varable OLS 0.25Q 0.5Q 0.75Q OLS 0.25Q 0.5Q 0.75Q HH SIZE ***.0282* HH SEX *.01418**.0203**.0247* ***.0004 HH AGE HH GRA.0011***.0014***.0010***.0010*** (omtted omtted omtted omtted HH ELE ** HH INDEX -.707*** -.698*** -.652*** -.717*** -.633*** -.662*** -.603*** -.54*** I P.3591** **.3874** ** I C.2601***.2149***.2270***.2500***.3378*** ***.2520** HH WED *** Source: Estmaton R 2 = 0.92 R 2 = 0.53 R 2 = 0.50 R 2 = 0.50 R 2 = 0.40 R 2 =0.26 R 2 = 0.20 R 2 =0.15

12 SAJEMS NS 19 (2016 No 2: A comparson of elastctes 6.1 Owner demand elastctes The Chow test (a test of whether the coeffcents estmated over one group of data was equal to coeffcents estmated over another for the owner-occuper households shows that there were sgnfcant dfferences between the ncome elastctes for the overall estmates for the country and the varous localtes. Ths s evdent n Tables 3 and 5. For nstance, there s a sgnfcant varaton between the permanent ncome estmates for the overall country and the varous localtes. The estmated coeffcent for the urban localtes s greater than the overall estmates (0.33 for urban and 0.17 for urban. The estmated coeffcent for the rural localtes s not sgnfcant. Ths and others mght have contrbuted to the sgnfcant dfferences obtaned from the Chow test. Generally, the dfferences are due to the rural/urban settng n the Ghanaan economy. Thereare vast dfferences between rural and urban housng unts and housng condtons. Ths means that a polcy or an nference based on the estmates for the whole populaton wll not acheve the desred results. Hence there s the need to separate the estmates for the demand for owner-occuped housng elastctes from rental demand elastctes. Other varables that showed sgnfcant dfferences ncluded the sze of the household and the level of educaton (Table 9. However, the prce elastctes are the same whether one consders the whole country or a partcular localty. Ths means that the effect of a prce change on owner demand s smlar natonwde. Age, martal status and the gender of the household head also have smlar elastctes, rrespectve of the localty or the segment of the populaton consdered. Table 9 Comparatve analyss of owner demand elastctes Varable ᵪ2 Test Decson I C 46.2( Sgnfcant dfference I P 5.67( Sgnfcant dfference HH SIZE 17.3( Sgnfcant dfference HH SEX 1.45( No sgnfcant dfference HH AGE 0.51( No sgnfcant dfference HH GRA 6.78( Sgnfcant dfference HH ELE 5.77( Sgnfcant dfference HH WED 0.15( No sgnfcant dfference H INDEX 0.18( No sgnfcant dfference Source: Estmaton Table 10 Comparson of rental demand elastctes Varable ᵪ2 Test Decson I C 0.18( No sgnfcant dfference I P 2.02( No sgnfcant dfference HH SIZE 0.23( No sgnfcant dfference HH SEX 2.40( No sgnfcant dfference HH AGE 0.03( No sgnfcant dfference HH GRA 41.6( Sgnfcant dfference HH ELE 0.48( No sgnfcant dfference HH WED 2.88( No sgnfcant dfference H INDEX 2.84( No sgnfcant dfference Source: Estmaton

13 172 SAJEMS NS 19 (2016 No 2: Rental demand elastctes The test for smlarty among the elastcty for the varous rental demands for housng shows that wth the excepton of graduate educaton, all other varables had the same effect whether or not the whole populaton or specfc localtes are taken nto account. Ths means that a change n ncome had the same effect on rental housng n the country, whether one consders a rural or an urban localty. The same apples to the prce and other elastctes (Table 10. The smlartes can be explaned by the fact that the whole rental market has a smlar pattern n terms of provders and the type of housng, as well as housng condtons. 7 Concluson Ths study amed at fllng a gap n the lterature n relaton to the housng stuaton n Ghana, provdng both the mean and quantle housng demand estmates for the country. It was undertaken because the housng market n Ghana, lke most Afrcan countres, s practcally dfferent from that of the developed and other emergng economes. The study estmated and analysed the demand elastctes for owner-occuper housng unts and rental housng unts. Estmates for rural and urban locatons were conducted employng both the OLS and quantle regresson technques. The estmated elastctes show that owner demand for housng s prce- and ncome-nelastc. Permanent ncome elastcty was greater than current ncome elastcty and the prce elastcty was close to zero. The quantle breakdowns show the sgnfcance of permanent and current ncome and prce. The quantle estmates show that the OLS (mean sometmes overestmates the ncome elastctes (See Table 4. The rural and urban breakdown of varables shows that besdes prce, permanent ncome s the man determnant of ownershp n the urban localtes, whle current ncome s the man determnant of ownershp n the rural localtes. Prce effect s greater for rural localtes than for urban localtes. Permanent and current ncome, graduate educaton, urban locaton, household sze and prce are the factors that determne the demand for rentng. These varables have more of an mpact on hgher quantles of housng demand than on lower quantles. The rural and urban estmates exhbt the same trend as the overall estmates. Ths means that for polcy effcency the varous quantles of the housng estmates must be used nstead of the OLS (mean estmates. The results based on the quantle estmates show that stakeholders n the housng ndustry need to take nto account not only localty dfferences but quantle dfferences as well, f they serously wsh to mprove polces governng the housng stuaton n Ghana for the beneft of her populaton. Endnotes 1 The Unted Natons (UN defnes a house as a structurally separate and ndependent place of abode such that a person or group of persons can solate themselves from the hazards of clmate such as storms and the sun. 2 A housng unt s a separate and ndependent place of abode ntended for habtaton by a sngle household or one not ntended for habtaton but occuped as lvng quarters by a household at the tme of the census (UN Statstcs Dvson. 3 It must be noted that some houses are completed earler than the stated perod but ths s rare. 4 That s, they buld as and when money becomes avalable. References ASIEDU, A.B. & KAGAYA, S A cross-secton analyss of household rental housng demand Kumas cty, Ghana. Hokkado Unversty, Japan. ASIEDU, A.B. & ARKU, G The rse of gated housng estates n Ghana: Emprcal nsghts from three communtes n metropoltan Accra. Journal of Housng and the Bult envronment, 24(3: BAJARI, P. & KAHN, M.E Estmatng housng demand wth an applcaton to explanng racal segregaton n ctes. Journal of Busness and Economc Statstcs, (231: BANK OF GHANA The housng market n Ghana. Accra: Ghana. BROUNEN, D., EICHHOLTZ, P.A.M., STRAETMANS, S. & THEEBE, M.A.J Inflaton protecton from homeownershp: Long-run evdence. SSRN: :

14 SAJEMS NS 19 (2016 No 2: FOLLAIN, J.R., LIM G.-C. & RENAUD, B The demand for housng n developng countres: The case of Korea. Journal of Urban Economcs, 7. FONTENLA, M. & GONZALEZ, F Housng demand n Mexco. Journal of Housng Economcs, 18:1-12. GARCIA, J.A.B. & HERNANDEZ, J.E.R Housng demand n Span accordng to dwellng type: Mcro econometrcs evdence. Regonal Scence and Urban Economcs, 38: GHANA STATISTICAL SERVICES The 2010 populaton and housng census: Summary of fnal report. Accra: Ghana. GHANA STATISTICAL SERVICES Ghana lvng standards survey report of the ffth round (GLSS 6. Accra: Ghana. GHANA STATISTICAL SERVICES Ghana lvng standards survey report of the ffth round (GLSS 6. Accra: Ghana. GOODMAN, A.C. & KAWAI, M Functonal form and rental housng market analyss. Urban Studes, 21: GOVERNMENT OF GHANA Growth and poverty reducton strategy: Natonal Development Polcy Commsson. Accra Ghana. HAYES, A.F. & CAI, L Usng heteroscedastcty-consstent standard error estmators n OLS regresson: An ntroducton and software mplementaton. Behavor Research Methods, 39: IOANNIDES, Y.M Resdental neghbourhood effects. Regonal Scence and Urban Economcs, 32: IOANNIDES, Y.M. & ZABEL, J.E Neghbourhood effects and housng demand. Journal of Appled Econometrcs, 18: IOANNIDES, Y.M. & ZABEL J.E Interactons, neghbourhood selecton and housng demand. Journal of Urban Economcs, 63: KARLEY, N.K Ghana resdental property delvery constrants and affordablty analyss. Housng Fnance Internatonal, 22 (4: LONG, J.S. & ERVIN, L.H Usng heteroscedastcty consstent standard errors n the lnear regresson model. Amercan Statstcan, 54: MALPEZZI, S Economc analyss of housng markets n developng and transton economes. In: Mlls, E.S., Cheshre, P. (eds. Handbook of regonal and urban economcs. Pp MALPEZZI, S. & MAYO, S.K The demand for housng n developng countres: Emprcal estmates from household data. Economc Development and Cultural Change, 35(4: MAYO, S.K Theory and estmaton n the economcs of housng demand. Journal of Urban Economcs, 10: MEGBOLUGBE, I.F. MARKS, A.P. & SCHWARTZ, M.S The economc theory of housng demand. Journal of Real Estate Research, 6: MEHTA, M. & MEHTA, D Metropoltan housng market: A study of Ahmadabad. New Delh: Sage. OBENG-ODOOM, F Real estate agents n Ghana: A sutable case for regulaton? Regonal Studes: OBENG-ODOOM, F Urban real estate n Ghana: A study of housng-related remttances from Australa. Housng Studes, 25(3: OBENG-ODOOM, F Prvate rental housng n Ghana: Reform or renounce? Journal of Internatonal Real Estate and Constructon Studes, 1(1. OBENG-ODOOM, F Prvate rental housng n Ghana: Reform or renounce? Journal of Internatonal Real Estate and Constructon Studes, 1(1: POLINKY, A.M. & ELLWOOD, D.T An emprcal reconclaton of mcro and grouped estmates of the demand for housng. Revew of Economcs and Statstcs, 6: RAPAPORT, C Housng demand and communty choce: An emprcal analyss. Journal of Urban Economcs, 42: th RAS, M., GAMEREN E. V. & EGGINK, E The demand for housng servces n the Netherlands. 45 congress of the European Regonal Scence Assocaton, Amsterdam:1-22.

15 174 SAJEMS NS 19 (2016 No 2: ROUWENDAL, J On housng servces. Journal of Housng Economcs, 7: SINAI, T. & SOULELES, N Owner-occuped housng as a hedge aganst rent rsk. Workng paper 9462, NBER. SMITH, S.J. & SEARLE. B.A The blackwell companon to the economcs of housng: The housng wealth of natons. Unted Kngdom: Wley-Blackwell: TIWARI, P. & PARIKH, J Effectve housng demand n Mumba (Bombay metropoltan regon. Urban Studes, 36(10: UN-HABITAT Housng profle n Ghana, Unted Natons Human Settlements Programme, Narob, Kenya. ZABEL, J.E The demand for housng servces. Journal of Housng Economcs, 13(1:16-35.