Public Disclosure Auhorized ASSESSING BENEFITS OF SLUM UPGRADING PROGRAMS IN SECOND-BEST SETTINGS * WPS3993 Public Disclosure Auhorized Public Disclosure Auhorized Basab Dasgupa and Somik V. Lall Developmen Research Group, The World Bank, Washingon DC 20433, USA Absrac Slum upgrading programs are being used by naional and ciy governmens in many counries o improve he welfare of households living in slum and squaer selemens. These programs ypically include a combinaion of improvemens in neighborhood infrasrucure, land enure, and building qualiy. In his paper, we develop a dynamic general equilibrium model o compare he effeciveness of alernaive slum upgrading insrumens in a second-bes seing wih disorions in he land and credi markes. We numerically es he model using daa from hree Brazilian ciies and find ha he performance of in siu slum upgrading depends on he severiy of land and credi marke disorions, and how complemenary policy iniiaives are being implemened o correc for hese problems. Pre-exising land supply and credi marke disorions reduce he benefi-cos raios across inervenions, and change he rank ordering of preferred inervenions. In he ligh of hese findings, i appears ha parial equilibrium analysis used in ypical cosbenefi work oversaes he sream of ne benefis from upgrading inervenions, and may in fac propose a misleading sequence of inervenions. JEL Classificaion: R21, R31, R38, I31. Key Words: Slum upgrading, preexising disorions, second bes policies Public Disclosure Auhorized World Bank Policy Research Working Paper 3993, Augus 2006 The Policy Research Working Paper Series disseminaes he findings of work in progress o encourage he exchange of ideas abou developmen issues. An objecive of he series is o ge he findings ou quickly, even if he presenaions are less han fully polished. The papers carry he names of he auhors and should be cied accordingly. The findings, inerpreaions, and conclusions expressed in his paper are enirely hose of he auhors. They do no necessarily represen he view of he World Bank, is Execuive Direcors, or he counries hey represen. Policy Research Working Papers are available online a hp://econ.worldbank.org. * This research has been funded by a World Bank Knowledge for Change (KCP) program gran Dynamics of Slum Formaion and Sraegies o Improve Lives of Slum Dwellers. We hank Paulo Avila, Bob Buckley, Dean Cira, Mila Freire, Sonia Hammam, Hyoung Gun Wang and Chrisian Zimmerman for helpful commens. Auhor for correspondence: Somik Lall; Senior Economis, Developmen Research Group, World Bank. Email: slall1@worldbank.org.
1. Inroducion Slum formaion is occurring a unprecedened raes. A repor by he Unied Naions shows ha here are more han 1 billion slum dwellers worldwide, which is abou 32 percen of he global urban populaion (Unied Naions 2003). While he represenaion of slum dwellers varies across regions, here is no doub ha slum formaion is a dauning problem. Slum dwellers accoun for 71.9 % of he urban populaion in sub-saharan Africa, 58% in Souh-cenral Asia, 36.4% in Eas Asia and 32% in Lain America and he Caribbean. The UN Secreary General warns if no serious acion is aken, he number of slum dwellers worldwide is projeced o rise over he nex 30 years o abou 2 billion (Unied Naions 2003). To address he growing problem of slum formaion, many naional and ciy governmens, as well as inernaional financial insiuions have ongoing programs aimed a reducing he rae of fuure slum formaion and improving he lives of exising slum dwellers. The World Bank has disbursed $14.3 billion in sheler lending over he pas 30 years, spread over 278 projecs wih an average size of almos $50 million in 2001 dollars (Buckley and Kalarickal 2005). 1 The UN Millennium Developmen Goals include Ciies wihou Slums as Targe 11, which specifically calls for significan improvemen in he lives of a leas 100 million slum dwellers by he year 2020. While here is an urgen need o scale up inervenions ha improve he qualiy of life for slum dwellers, here is lile clariy on he ypes of inervenions ha are mos effecive or he relaive cos effeciveness of alernae sraegies. 1 The World Bank s urban sheler lending porfolio has moved from financing projec based sies and services and slum upgrading o now include broader housing policy and housing finance loans. 2
Over several decades, sraegies of naional governmens and developmen agencies o achieve beer living condiions of slum dwellers have included sies and services programs, reselemen o new housing developmens, and land iling. Iniially, policies favored sies and services programs where infrasrucure could be provided relaively cheaply on newly developed land. A major reason for he limied success of sies and services programs was he lack of access o housing finance for consrucion of he dwelling uni. Wih he persisence of large slum areas, limied success of slum relocaion programs in erms of low reenion raes and he realizaion ha many slums could no be simply removed, slum upgrading projecs have become more widespread. Slum upgrading ypically involves he provision of a package of basic services, which include clean waer supply, sewage disposal, wase collecion, housing, access roads, sidewalks, lighing, schools, healh poss and communiy ceners. An imporan componen of hese programs is regularizing properies in siuaions of insecure or unclear enure. The underlying logic behind hese inervenions is ha he poor canno afford o make sheler improvemens on heir own due o a variey of income and credi consrains. The focus on explici inervenions raher han on cash ransfers is ha increases in disposable incomes for he poor may no ranslae ino heir access o basic services. This can be for several reasons, which include limied empowermen of he poor (include limied communiy cohesion and social neworking among he poor) and ime delays in expanding service coverage (informaion and coordinaion problems, supply consrains in nework expansion, and weak incenives for providers o improve performance). 3
In order o idenify sraegies ha are useful for improving he lives of slum dwellers, here has been recen ineres in evaluaing he effeciveness of slum upgrading programs. As par of his, he World Bank recenly commissioned a paper o provide guidance on how o esimae he impac of slum upgrading inervenions (Field and Kremer 2005). Relaed o his effor, here have been several iniiaives where baseline daa are being colleced so ha hey can be used o se up rigorous evaluaions of projec oucomes. The fly in he oinmen in his evaluaion sraegy is he assumpion ha inervenions happen in firs-bes seings. Similarly, much of he ex-ane cos-benefi esimaion of hese inervenions also assumes marke clearing. Clearly, his is no he case in mos developing counries where here are pre exising disorions in he land (for example, excessive zoning, developmen conrols) and credi (higher loan raes) markes. These pre exising disorions no only end o reduce he cos effeciveness of inervenions, bu also may in fac change he welfare rank ordering across inervenions. Thus, assessing he benefis of inervenions wihou accouning for pre exising disorions may in fac be misleading. The performance of in siu slum upgrading depends on he severiy of land and credi marke disorions, and how complemenary policy iniiaives are being implemened o correc for hese problems. In his paper, we develop a dynamic general equilibrium model o compare he effeciveness of alernaive insrumens for improving he welfare of slum dwellers in a second-bes seing wih disorions in he land and credi markes. Wih four decisionmaking agens in our economy- households, developers, financial insiuions and he governmen, we analyze how land or credi marke disorions aler he rankings of differen policy insrumens. We also esed he effecs of building caps and infrasrucure 4
bolenecks, and ge similar resuls. However, o mainain breviy, we do no include hese findings in his paper. For he analysis, we lay ou he residenial locaion problem for poor urban households and analyze hree ypes of inervenions. These inervenions include improvemens in land, infrasrucure, and building qualiy. 2 To illusrae he analyical problem, we use examples from hree Brazilian ciies for which daa have been colleced in recen sudies. The analyic sraegy, however, is general and can be applied o a wide range of ciies. We find ha he presence of preexising land supply and credi marke disorions reduce he benefi cos raios across inervenions, and change he rank ordering of preferences across ypes of upgrading packages. In he ligh of hese findings, i appears ha parial equilibrium analysis used in a ypical cos-benefi se up (which does no address pre exising disorions) may be oversaing he sream of ne benefis from inervenions, and may, in fac, propose a misleading sequence of inervenions. These findings are consisen wih research in environmenal economics (Bovenburg and Goulder 1996; Parry and Oaes 1998) and public finance (Ballard and Fulleron 1992; Wildasin 1984), which sugges ha he presence of pre-exising disorions changes he welfare impacs of new policy insrumens. The analysis is of paricular relevance for Brazil as here are more han 1.3 million subsandard housing unis wih 80 percen of hem locaed in meropolian areas (World Bank 2002). In February 2000, he Brazilian Governmen amended he consiuion (Consiuional Amendmen No. 26) and approved housing as a social righ. The hree-ier governmenal suppor srucure wih he federal governmen a he helm of 2 Beruad and Brueckner (2005) also examine he welfare implicaions of one paricular se of land marke disorion he Floor Area Raio (FAR), arguing ha his regulaion encourages sprawl and increases commuing coss for edge residens. 5
affairs, made he Minisry of Ciies (MOC) as he responsible agency for esablishing a naional housing policy. The assurance of housing righs encompasses access o land enure, basic public services and financial services. Our sudy is oulined as follows. Following his inroducion, Secion 2 describes he model and equilibrium condiions. Secion 3 discusses he impacs of slum upgrading programs. Secion 4 concludes. 2. Baseline Model In his secion, we develop a general equilibrium model o examine he effeciveness of alernae slum upgrading policy insrumens ha can be used o improve household welfare. We firs evaluae how hese inervenions perform in firs-bes seings, and hen examine he effecs of pre exising insiuional and regulaory consrains on he effeciveness and relaive rankings of hese inervenions (Secion 3). These consrains include land supply consrains, 3 infrasrucure bolenecks, and credi raioning. We sar wih he assumpion of a monocenric-closed ciy wih no populaion growh. The model has four economic agens: households, developers, financial inermediaries and he ciy governmen. 3 The increase in informal housing unis beween 1991 and 1998 (Morais 2000 based on PNAD/IBGE in World Bank Repor No. 22032 BR, (2002), pp 16) shows he exen of housing defici in Brazil. The repor indicaes ha among 10 meropolian regions, seven is repored o have over 50 percen increase in housing defici wih Recife and Curiiba having 52.8 and 143.4 percen respecively. This sylized fac suppors our consideraion of supply side boleneck in housing marke. 6
2.1. Households We consider an infiniely lived represenaive household who maximizes lifeime uiliy by consuming a composie good subjec o a budge consrain. In period, he household earns fixed wage income and disribues i beween consumpion and saving (S ) in period. The composie consumpion good in period is comprised of non-housing consumpion, (C ) and housing, (H ). 4 We assume ha each household consumes 1 uni of housing (H = 1) wih specific aribues. These aribues are based on he household s hedonic preferences for building srucure (B ), land area (L ) per house wih paricular locaion aribues such as infrasrucure ameniies (A ) and disance (D ) from he ciy cener. 5 We assume ha, wih a given income (w ) and non-housing consumpion (C ), households ry o improve presen qualiy of housing aribues (q ) over las period (q -1 ) 6. By definiion, qualiy of housing is deermined from he combinaion of differen aribues presen in a house. Therefore, ceeris paribus, household can improve qualiy housing by improving any of hese aribues from previous period. Given he budge consrain, his implies ha he household can decide is opimal demand for housing on he basis of her preferred combinaion of hese aribues. The governmen usually seps in o assis when poor households are unable o make improvemens due o limied affordabiliy. The objecive of various governmen programs for slum upgrading is eiher 4 H is a bundle of housing aribues, which include he dwelling uni, infrasrucure aribues, and neighborhood qualiy. 5 See Clap (1980), Clapham e. Al (2004), Mayo (1986), Mills and Simenauer (1996), Reiff e al (2005), and Wolveron (2000) for deails on hedonic esimaion. 6 As similar o he hedonic pricing lieraure (see Clapham e al (2004) for deails) we incorporae he represenaive or sandard dwelling qualiy (q -1 ) for comparison. This sandard can be se eiher by he households hemselves under no governmen inervenion or by he governmen as he social planner when necessary. 7
o improve housing services or infrasrucure so ha i, a leas, mainains previous qualiy level. We consider ha he governmen makes a ransfer paymen, T, o he poor households o cover heir housing qualiy enhancemen program. Household savings (S ) are assumed o be deposied in he bank, and he gross reurn from his saving is (1 r ) S, when deposi rae is r d, 1. + d, 1 We also consider ha beer faciliies are concenraed in he cener of he ciy. Given income and non-housing consumpion in period, households willingness o pay for each of he above-menioned aribues in period, hus, moves in andem wih he disance of he house from he ciy cener, available infrasrucure and housing faciliies. Households incorporae hese individual resource coss in a linear fashion o esimae heir willingness o pay for a house. Households maximize heir lifeime uiliy from he composie good as follows: max A, L, S, B s.. V w + T = β ln ( C = 0 + (1 + r d, 1 ) S q + e 1 e ) = C + p b, B + p l, L + p A, A + S (1) (2) q = q 1 A L B D ξ1 ξ2 ξ3 1 γ (3) Where, β is he discoun facor bounded by 0 β 1, q -1 is he represenaive or he sandard qualiy from he previous period, A represens infrasrucure aribues, L is land area, B is building srucure, and disance of housing from he cener of he ciy is D. The value of he elasiciy of subsiuion, e, deermines households resource allocaion beween consumpion and housing qualiy and is bounded wihin [0, 1]. The parameersξ 1, ξ 2 and ξ 3 represen he share of neighborhood aribues, land area per house and building srucure respecively. The parameer γ represens he shape of he 8
ciy and commuing echnology. For a concenric ciy, we adop he value of 0 γ 2 from Henderson and Venables (2004, pp. 4-5). Assuming no growh and elasiciy of subsiuion, e, being 1, 7 he Euler s equaions wih respec o A : p S B : p a, L : p l, b, ξ1( q = ξ 2 ( q + βq = L ξ3( q + βq = B 1 : β = 1+ r d, + βq A + 1 + 1 ) + 1 ) ) (4) (5) (6) (7) The above prices represen household s willingness o pay or demand for respecive componens in order o mainain heir housing qualiy. 2.2. Developers 8 The developer is assumed o supply building srucure and developed land in he form of housing (H ). We consider ha developers use he available echnology o ransform land and building maerials ino residenial unis by incurring cerain cos. We assume he marke for housing o be perfecly compeiive and he developer s marginal cos is exacly equal o price of each housing uni. The producion funcion of housing is assumed o have consan reurn o scale of he following form: H α 1 α = ψ L B (8) 7 Elasiciy of subsiuion is considered o be 1 for compuaional simpliciy. I does no aler he basic resuls. However, he model can be calibraed for any value of e beween [0,1]. 8 Developers can be eiher in he public or privae secor 9
where, H is he sock of housing in he economy, L is land area in sq. m., B represens building srucure. The parameer ψ represens he echnological parameer and is assumed o be 1. Wih price per house in period, p h, he developer s profi maximizaion problem can be wrien as max Π s.. R = [ p = L p h L l, α B + B 1 α p (1 + r ) R b, ] (9) (10) where, R is he oal demand for finance by he developer o pay for he plo as well as maerials. Based on he developer s maximizaion condiion in he long run, supply of land and building srucure per uni of housing are L B αph = (1 + r ) p l, (1 α) p = (1 + r ) p h b, (11) (12) Noice ha he developer s supply decision is inversely relaed o he loan rae, r. Therefore, any disorion in he credi marke ha affecs he loan rae is expeced o influence he land and building supply decision of he developer. 2.3. Financial Inermediary We assume banks o be financial inermediaries ha maximize profi in a perfecly compeiive environmen. Banks conver heir enire deposis from households ino loans owards he developer wihou any fricion. In reurn, hey charge a loan rae r. To make our model more general and compaible wih our objecive, we consider imperfecions in he credi marke. Such imperfecion originaes from he defaul vulnerabiliy of he borrowers and expeced recovery coss. This assumpion is relevan for Brazil as well as mos developing counries. According o a repor by he World Bank (World Bank 2002), 10
he housing finance sysem (HFS) could caer only 27 percen of oal demand for loans of 23.7 million dollars during 1964-96. I also repors defauls in 30 percen of cases during his period. To jusify raioning in he credi marke, we assume ha he bank has a posiive expeced recovery cos, r c, when a borrower defauls. Larger coverage exposes he bank o larger expeced recovery coss and evenually higher loan raes. Incorporaion of posiive expeced recovery cos in loan raes makes he banks' profi maximizing loan rae differen from he marke clearing loan rae, and leads o credi raioning in he marke. Banks also use credi raioning o hedge agains defaul risk. Thus, banks' profi maximizaion problem can be saed as B max Π = ωr R rd, S rcωr R (13) we ge rd, r = + r ω c (14) where, ω is he fracion of oal demand for loan being supplied by he bank. Banks decide his opimum fracion from he amoun ha saisfies bank's zero profi condiion in he long run. The proporion ω lies beween [0, 1]. In our unconsrained world, ω is 1. A value of ω < 1 indicaes raioning in he credi marke. As menioned above, ω is 0.27 in he case of Brazil for HFS loans. 11
2.4. The Governmen The source of revenue of he ciy governmen is a hree-ier sysem. 9 The ciy governmen collecs a par of is revenue from infrasrucure faciliies, (A ), provided o he households in he previous period. Also, a he beginning of each period he ciy governmen ges some exogenous funding (G ) from he sae and cenral governmen. Wih balanced budge assumpion, he governmen budge consrain hus, akes he following form T p A + G = a 1 1 (15) Where, T is he ransfer o each household in period by he ciy governmen. The governmen makes his ransfer o cover he cos of differen inervenion programs (e.g., supply of A, or in siu upgrading, ec) for housing qualiy enhancemen. For simpliciy we assume no exra sources of revenues for he ciy governmen. The duy of he ciy governmen is o allocae hese funds ino differen developmen projecs for he nex period such as infrasrucure and housing developmen. For welfare enhancemen purposes he governmen can eiher ake up any policy exclusively or is allocaion sraegy can be a combinaion of housing and infrasrucure developmen. 2.5. Equilibrium Equilibrium in his model economy is a sequence of prices{ r }, rd,, ph, pb,, pl,, pa,, =0 allocaions, { C,,, } H B L, sock of financial asses, { S, } =0 R =0 { T, } A G =0,, such ha:, and policy variables 9 We ake his hree ier sysem o represen he hree ier governmen suppor srucure in Brazil. 12
1. The allocaions and income solve he household's dae maximizaion problem [Equaion (1) - (3)], given prices and policy variables. 2. The allocaions solve he firm's dae profi maximizaion problem [Equaions (9) and (10)], given prices and policy variables. 3. The sock of financial asses solves he bank's dae profi maximizaion problem [Equaion (13)], under credi raioning given prices and policy variables. 4. The loanable funds marke equilibrium condiion under credi raioning: R L pl, + B pb, =. dd 5. The housing marke equilibrium condiion saisfies H = f L, B ) for all. 6. The governmen budge balances when G + p A 1 = T 1. ( Afer solving each agen's opimizaion problem we find from a sysem of equaions wih he same number of unknown variables ha The discoun facor β can be esimaed from Equaion (7). The share of land per residence, α, is esimaed from Equaions (5), (6), (11) and (12) as ξ 2 α = ξ + ξ 2 3 (16) From Equaions (5) and (11) we ge equilibrium value of housing qualiy as α q = (1 ξ 2 + r)(1 + β ) p h (17) This reduced form equaion indicaes he direc relaionship beween housing qualiy (q ) and house price (p h ). 13
Given he inpu prices p l, p b and housing price, p h, we ge equilibrium supply of land per residence by insering Equaion (17) ino (5) as L = αph ( 1+ r) pl (18) The amoun of land developed and supplied by he developer decides he availabiliy of land in he ciy. However, land availabiliy is a funcion of availabiliy of loan, ω, hrough he loan rae, r (Equaion 14). Based on available land and is fixed share in producion, we ge equilibrium building srucure wih given prices as (1 α) p B = (1 + r) p h b (19) To mainain he same seady sae housing qualiy (such ha q = q 1 q ), he = equilibrium values of land (Equaion 18) and building srucure (Equaion 19) esimaed above deermine he opimum amoun of infrasrucure in he following way: D A = ξ2 L B γ 1 ξ 3 1 ξ1 (20) From Equaion (10), (18) and (19) we ge equilibrium loan requiremen per house for he consrucion of one uni residence as R = p L l + p b B (21) 2.6. Esimaing Equilibrium Parameer Values We quanify our model using daa for hree municipaliies in Brazil - Brasilia, Curiiba and Recife, based on parameer values developed in recen research. From Serra e al (2004), we use oal housing sock (H ), oal urban developed land, land price per square 14
meer (p l ) boh wih and wihou infrasrucure, o esimae he cos of land and cos of infrasrucure per house. Using land prices, boh wih and wihou infrasrucure faciliies, we esimae he price of infrasrucure ameniies per uni of land (in square meers) and hen conver his ino corresponding average coss per house. These esimaes are provided in Table 1. The average consrucion cos per house (B p ) and oal cos per house (p h ), have been aken from World Bank (2002). 10 Tha repor also shows sae level per capia expendiures on urban developmen in Brazil. The amoun of land, infrasrucure and building srucure per house is considered as he unis of respecive asses and uni prices have been adjused accordingly (see Table 1 for deails). From Equaion (17), housing qualiy is direcly relaed o price, given he parameer values α, β and loan rae, r. This corroboraes our consideraion of he share of each resource in oal cos as heir respecive shares in qualiy. From his, we can assume ha he share of land (ξ 2 ), building (ξ 3 ) and infrasrucure (ξ 1 ) in housing qualiy will be similar o heir relaive conribuions o housing price. Table 2 presens he esimaed parameer values from our model. Based on he parameer values given in Table 2, we esimae he equilibrium values of he following variables (presened in Table 3). The above seady sae values indicae he requiremens of each resource o mainain he equilibrium housing qualiy a he ciy cener (D=1 mile around he cener). In he following secion we calibrae our model and discuss he demand for each inervenion in a firs bes seing, and hen evaluae he change in demand for hese inervenions in he presence of various marke disorions. For our numerical analysis, we consider hree differen siuaions wih sub opimal housing qualiy. The seady sae 10 Table 36: Cos Break Down of Urban Upgrading, Recife, 1998 (pp. 71) 15
value of qualiy derived from our model may have qualiy equivalence of 1.25, 1.5 and 1.75 respecively wih respec o each of hese subsandard siuaions. These hree hypoheical siuaions help us undersand how demand for alernae inpus changes in order o reach he seady sae equilibrium qualiy level. Based on hese changes, we also esimae he social welfare o cos raio for each governmen inervenion and rank hem accordingly o idenify he bes possible inervenion in a consrained seing. The social welfare o cos raio has been presened as he welfare gain per uni of money (real for Brazil) spen on he respecive resource. 3. Impacs of Slum Upgrading In his secion we sar our analysis wih he assumpion ha he governmen uses slum upgrading o improve welfare of he urban poor. These welfare programs mainly focus on improving housing qualiy, which will improve qualiy of life for households living in sub-sandard residenial unis. As menioned earlier, we consider hree such subsandard siuaions. According o our consideraion, he seady sae has a qualiy equivalence of 1.25, 1.5 and 1.75 respecively as compared o hese hree subsandard siuaions. The governmen inervenes o improve social welfare from hese inferior siuaions o he seady sae equilibrium. Given he fixed consumpion of non-housing iems, he governmen can improve welfare o he seady sae sandard by improving land availabiliy, up grading building srucure, or improving infrasrucure faciliies. These improvemens can be done in siu wihou relocaing he households from heir curren dwelling unis, or may involve relocaion of households o areas where more land is available. In principle, relocaion of households away from he ciy cener or CBD will 16
increases heir demand for alernaive resources o mainain he same welfare level. In wha follows is a comparison of social welfare and cos beween in-siu upgrading and relocaion of households. We calibrae our model for differen disances (D ) and qualiy equivalence (k). 11 3.1. In-Siu Upgrading The following analysis represens governmen s opion for In Siu upgrading. 12 Under his policy he governmen provides land o households, who, in urn, upgrade heir own house wih available loans from he bank. Public inervenions may also focus on infrasrucure and building qualiy. We compare he changes in social welfare o cos raios and demand for resources under In Siu upgrading in order o find ou ordering of each inervenion. Nex, we compare In Siu upgrading o relocaion sraegies. We also examine how he social welfare of various improvemens fares in he presence of pre exising disorions. When he governmen allocaes resources o households o achieve he equilibrium arge, i produces differen oucomes under marke disorions. For example, when he land marke is disored, he governmen can inervene on he supply side and address his problem. However, if he disorion in he land marke is ransmied from he credi marke, hen only addressing disorions in land marke will no help unless he imperfecions in credi marke are also correced. 11 The mehodology for esimaing social welfare and requiremens of differen inervenions o reach he equilibrium welfare from our subsandard siuaion is provided in he appendix. 12 We consider 1 insead of 0 miles from he cener o define he cener of he ciy in order o avoid compuaional complexiies. 17
Table 4 provides he respecive invesmens required for each inervenion a various qualiy equivalence, k. We calibrae hese resource requiremens for differen qualiy equivalence arges as menioned earlier. Table 5 shows he ordering of social welfare o cos raio when he governmen ops for upgrading In Siu. Based on his, he social planner ranks each policy accordingly. According o his Table, he mos effecive policy for Brasilia should be building upgrading as compared o land iling or infrasrucure for small or medium sized inervenions (k =1.25, 1.5). However, when large-scale improvemens are needed (say k=1.75), building upgrading will no longer be he bes policy o improve qualiy. Raher, increasing land supply becomes mos effecive sraegy. Given our se up infrasrucure developmen is never he bes opion for Brasilia. For he oher wo ciies, however, infrasrucure appears o be he bes sraegy relaive o he oher wo inervenions for any degree of qualiy enhancemen. A closer look suggess ha he second and he hird ranked policies inerchange places for higher level of qualiy equivalence. Apparenly, in a perfec world, one can survive wihou caring much abou his reshuffling beween second and hird ranks. However, in a world wih infrasrucure bolenecks, i urns ou o be more crucial for he social planner in picking up he second bes. Along wih Table 5, a closer look a Figure 1 (he curve wih disance = 1 mile) suggess ha he relaionship beween he demand for land and infrasrucure in Brasilia is perfecly inelasic. This indicaes ha even a very large change in infrasrucure may no be able o mainain he overall qualiy a he same level. We find similar relaionship 18
beween building upgrading and infrasrucure in Brasilia. Alernaively, we can suppor why infrasrucure developmen is mos preferred in Curiiba or Recife. From Figure-1, we see ha he demand for infrasrucure in hese wo ciies is more inelasic (see C 1 and R 1 ) before he poin of inflexion is reached. The household is ready o subsiue any amoun of land in exchange for one uni of infrasrucure o cross his hreshold value of infrasrucure requiremen. 3.2. Relocaion Sraegies We now look a a governmen s policy, which includes relocaion from he cener and compare i wih he In Siu program described above. For his analysis, we assume ha he governmen fixes is qualiy equivalence arge a 1.25. One way o relocae he poor successfully is by providing hem wih more land or improved faciliies ha will leave hem no worse off han in heir presen locaion. Table 6 shows he corresponding change in demand for each inervenion beween in siu upgrading and relocaion a differen disances. Table 7 compares he social benefi cos raios beween In Siu upgrading and relocaion of households a various disances in Brasilia, Curiiba and Recife. As saed earlier, he benefi cos raio has been presened in his able as welfare gain per real spen on a paricular resource under differen invenions. The Table shows ha he social benefi cos raio drops drasically due o a policy shif from In Siu upgrading o relocaion. Furher he social welfare-cos raio of each inervenion reduces wih an increase in disance from he ciy cener. 19
Ineresingly in Brasilia, improving building srucure urns ou o be bes soluion over land supply or infrasrucure improvemens under in siu program wih seady sae qualiy equivalence 1.25. However, under relocaion programs, land supply becomes he bes policy for same level of qualiy equivalence and remains he bes opion wih increase in disance. For he oher wo ciies, he bes soluion is infrasrucure developmen in boh, in siu as well as relocaion. However, similar o higher qualiy equivalence, he second and hird ranked policies iner change places from in siu o relocaion. The significance of his oucome becomes prominen under marke disorion for infrasrucure. In ha case, due o unavailabiliy of he firs bes soluion, he second bes policies will be differen beween he in siu and relocaion programs. Figure 1 suggess ha given he infrasrucure faciliies, he demand for land shifs verically upward quie significanly in each ciy due o a shif in policy of in siu upgrading o one ha involves relocaion of households. The gap beween he wo curves in each figure indicaes he effec of increasing disance beween he residence and he cener of he ciy on subsiuion beween infrasrucure and land. 3.3. Pre Exising Disorions: Land Supply Consrains The analysis so far assumes ha here are no pre exising disorions ha could influence he performance of slum upgrading. However, in pracice here are many binding consrains such as unresponsive land supply and credi raioning, which make i imporan o assess slum-upgrading insrumens in a second bes seing. In his secion, we examine he implicaions of a small se of pre exising disorions. We sar by examining he effec of consrained land supply, which effecively means ha he 20
availabiliy of developable land is fixed. 13 Therefore i is no possible for he governmen o supply addiional land o slum dwellers as ciywide land supply is seriously consrained. When land supply is consrained, marke price are expeced o bid up significanly -- depending upon he severiy of he supply problem, and leads o a decrease in he benefi cos raios of governmen inervenions (see Proposiion 1). Proposiion 1: The benefi cos raio decreases due o land supply consrains. Proof: From he relaionship beween qualiy and he residence aribues, ξ1 ξ2 ξ3 1 γ q q A L B D and o mainain he same qualiy, + 1 = = q = q+1 q we ge from Equaion (6) ha ξ 2 (1 + β ) q pl, = L Now, if L is reduced such ha L 1 <L, hen ξ 2 ( 1+ β ) q ξ 2 (1 + β ) q p 1l = > = pl. L1 L MU Land Given he marginal uiliy from land, Pl QED P l In such siuaions wih a binding land supply consrain, he seady sae qualiy can be mainained by increasing supply of building or infrasrucure. We prove ha under such binding condiions, he household demands more building srucure per uni increase in disance han he siuaion wihou any land supply consrain (see Proposiion 2). 13 Land supply consrains could be boh due o naural facors (elevaion, locaion, ec) as well as policies (resricive land use and zoning) 21
Proposiion 2: The rae of change in demand for building qualiy wih increase in disance is posiive and he rae of change is higher under land supply consrains. Proof: From he relaionship beween building wih land and infrasrucure, given qualiy, we ge γ 1 1 D ξ3 B = [ ] ξ1 ξ2 A L 1 ξ γ 2 3 δb ( γ 1) D ξ3 = [ ] > 0 {since γ > 1 and A, L, D > 0 ξ1 ξ2 δd A L Now, if he available land per residence is L 1 such ha L 1 < L, hen γ 2 1 ξ3 γ 2 1 ξ3 ( γ 1) D ξ ( γ 1) D 3 ξ3 [ ] > [ ] QED ξ1 ξ2 ξ1 ξ2 A L A L 1 We find same resuls for he change in demand for infrasrucure under such disorion. Figure 2 represens he effecs of land supply consrains on he demand for building srucure o mainain he seady sae sandard. The curves show ha he household s demand for building qualiy improvemen increases in an exponenial fashion wih increase in disance beween he residence and he cener of he ciy. The problem ges exacerbaed in he presence of land marke disorions. The gap beween he wo curves in each figure indicaes he effec of land marke disorion on demand for building srucure due o each uni increase in disance beween residence and he cener of he ciy. We find ha he effec is severe in Brasilia (from B-1 o B-2) and modes in Recife (R-1 o R-2) and in Curiiba (from C-1 o C-2). The increase in demand for infrasrucure wih an increase in disance beween he residence and he cener shows a difference in preferences across hree ciies. While he demand for infrasrucure in Brasilia is almos perfecly elasic wih respec o disance, i shows ha households adjus heir housing qualiy by demanding more infrasrucures in Curiiba and Recife. However, he reacions of households in he presence of land marke 22
disorion are no same in hese wo ciies (Figure 3). The gap beween he wo curves in each figure indicaes he effec of land marke disorion on demand for infrasrucure for each uni increase in disance beween residence and he cener of he ciy. The higher marginal uiliy of land in Brasilia makes i easier under no land supply consrain o reach he required qualiy even wih sligh improvemen in land availabiliy per uni of housing (see Table 8). However, under land scarciy, when provision of exra land is no possible, he governmen is lef wih he wo oher opions. In such a siuaion he second bes opion for Brasilia should be improving building qualiy. The siuaion does no arise in he case of Curiiba and Recife since infrasrucure developmen remains he mos effecive opions under land supply consrain. Table 9 shows he changes in demand for building or infrasrucure under binding land consrain. To keep he household a he same welfare level, resource requiremens increase dramaically wih land supply consrains. 3.4. Pre Exising Disorions: Credi Raioning From our model, we find ha land supply or housing qualiy improvemen decisions are a funcion of he loan rae. Thus, when credi marke imperfecions disor he ineres rae we can expec a ransmission of such a disorion in resource markes ha are dependen on he loan rae. In his par of he analysis, we examine how credi marke imperfecions ranslae ino supply side bolenecks in he land and housing markes. 23
Earlier in his paper, we menioned ha he housing finance sysem (HFS) exended loans o abou 27 percen of 23.7 million requiremens beween 1964 and 1996 (World Bank 2002). In he same repor i has also been repored how he housing defici increased from 1991 o 1998. We are ineresed in puing hese pieces ogeher and use our analyic framework o find ou wheher credi marke imperfecions influence he funcioning of he land marke. In our model we assume ha developers use bank loans o fund new land developmen and purchase building maerial. As a resul, credi raioning hinders land developmen and consrucion by developers. Proposiion 3: Sringen credi raioning disors oher resource markes. Proof: From he developers opimizaion problem we ge ha L and B are inversely relaed o he bank loan rae, r. L B αph = (1 + r ) p l, (1 α) p = (1 + r ) p h b, Also, from bank s opimizaion we ge he loan rae is inversely relaed o fracion of oal demand for credi supplied, ω as r d, r = + ω r c. Now, given r d and r c, as lim ϖ 0 r L, B QED Proposiion 3 shows ha sringen credi raioning leads o disorions in he land marke. In oher words, land or housing supply decisions are consrained due o scarciy 24
of developer finance. Table 10 shows how sringen credi raioning reduces he equilibrium land supply. Figure 4 shows he effec of credi raioning on he change in supply of land. The curves show ha credi raioning affecs land supply in he same fashion as a land marke disorion does. This implies ha he imperfecion in he credi marke ransmis o he land marke and creaes supply side bolenecks. Figure 5 shows how sringen credi raioning influence demand for resources such as land under a relocaion program. B-1, R-1 and C-1 represen land supply when he loan rae is in equilibrium for Brasilia, Recife and Curiiba respecively. B-2, R-2 and C-2 represen land supply in he same ciies under a higher loan rae. We see ha he demand for land increases as a households moves from he ciy cener o he periphery. This reflecs he compensaing variaion in erms of land provision required o make he household no worse off as i moves ou of is presen locaion. The gap beween he wo curves in each figure (wihou and wihou credi raioning) indicaes a credi marke disorion effecively lowers he availabiliy of developed land and exacerbaes he unme demand for land. From his analysis, i becomes eviden ha i is difficul o address land marke problems wihou evaluaing consrains in linked markes. Raioning in he credi marke ransmis similar disorions in he housing supply decision of developers. 4. Conclusions There is increasing emphasis on he imporance of slum upgrading insrumens as a susainable approach o improve he lives of slum dwellers. However, here is no consensus of wha paricular se of insrumens works bes, and how he effeciveness of 25
alernae insrumens changes when here are pre exising disorions in he land and credi markes. One of he objecives of our paper is o provide a more realisic assessmen of wha upgrading projecs are likely o achieve if hey are no par of a larger se of reforms ha relax various disorions ha hinder he funcioning of he land and housing markes. In his paper, we develop a dynamic general equilibrium model which includes households, developers, financial insiuions and he governmen o evaluae he effeciveness of alernaive insrumens. Our findings are based on daa from hree Brazilian ciies, bu he approach developed here can be generalized and is relevan for mos developing counries where land and housing markes are subjec o disorions from excessive zoning and developmen conrols. We believe ha here are hree main reasons ha he general equilibrium approach we propose here will provide beer insighs relaive o a parial equilibrium assessmen of slum improvemen programs. Firs, ypical parial equilibrium analysis is based on a households marginal benefi and marginal cos wihou aking supply side consrains ino accoun. Under in siu upgrading, i is usually assumed ha households can keep increasing heir consumpion of one resource so long he marginal benefi exceeds he marginal cos (see for example, Heikkila 2004 who proposes a concepual framework for applicaion o Brazilian ciies). The argumen is difficul o defend if here are supply side bolenecks. Second, marginal benefis of program inervenions are calculaed from a households uiliy funcion, while he marginal cos is based on he governmen s expendiure / cos funcion. Fulfilling individual household demand (from heir assessmen of marginal benefis), solely based on heir preferences may no yield a 26
socially opimal soluion. This discrepancy requires ha he problem be recas in social benefi- social cos framework. Third, resuls from a general equilibrium framework improve upon parial esimaes as i becomes possible o assess he reacion of various decision makers households, governmen, and financial inermediary o every policy shock For example, we show how a developer s or household s decision is moivaed by a decision aken in he banking secor. A comparison on upgrading in siu vs. involving relocaion based on our model shows ha he social benefi cos raios across inervenions drop dramaically if households are relocaed from heir original locaions. The siuaion is made worse if here are pre exising land marke disorions. The welfare analysis presened here suggess ha hese pre exising disorions no only end o reduce he cos effeciveness of individual insrumens, bu also in fac change he welfare rank ordering across inervenions. Thus, assessing he benefis of inervenions wihou accouning for pre exising disorions is likely o be misleading. Furher, he choice of preferred insrumen (infrasrucure, housing qualiy, land provision) depends on ciy specific characerisics and he severiy of he underlying supply side bolenecks. We also find ha disorions are ransmied across markes for insance, land and building supply decisions of developers are a funcion of ineres raes and any disorions in he credi marke ha increase effecive ineres raes also reduce land and housing supply. Thus disorions in he credi marke exacerbae consrains in he land marke. In his conex, he effeciveness of projec level upgrading inervenions is likely o be enhanced if hese are accompanied by insiuional and regulaory reforms. 27
References Amann, E and W. Baer (2005): Economic Orhodoxy Versus Social Developmen? The Dilemmas Facing Brazil s Labor Governmen, Working paper 05-0108, Deparmen of Economics, Universiy of Illinois a Urbana-Champaign. Ballard, C and D. Fulleron 1992. Disorionary Taxes and he Provision of Public Goods. The Journal of Economic Perspecives, 6, 3. 117-131. Beraud, A. and J. Brueckner (2005), Analyzing Building Heigh Resricions: Prediced Impacs and Welfare Coss, Regional Science and Urban Economics, 35, 109-125. Bovenburg, A. L. and L. H. Goulder (1996) Opimal Environmenal Taxaion in he Presence of Oher Taxes: General- Equilibrium Analyses. The American Economic Review, 86, 4 985-1000. Buckley R. and J. Kalarickal eds (2005). Thiry years of World Bank Sheler Lending: Wha Have We Learned? World Bank. Clapham E, P. Englund, J. Quigley and C. L. Redfearn (2004): Revisiing he pas: Revision in Repea sales and Hedonic Indexes of House Prices, Working Paper No. W04-005. Fisher Cener for Real Esae and Urban Economics, Universiy of California, Berkeley. Field, E and M. Kremer (2005) Impac Evaluaion for Slum Upgrading Inervenions. World Bank Haurin, D., P. Hendersho and D. Kim (1991): Local House Price Indexes: 1982-1991, Journal of he American Real Esae and Urban Economics Associaion, 19, 451-72. Heikkila, E. (2004) Concepual Framework: Evaluaion of Policy Alernaives For Promoing Equiable Access o Urban Land in Brazil, World Bank Ciies Alliance-IPEA, Enabling Sraegy for Moving Urban Upgrading o Scale in Brazil. Hendersho, P and T. Thaibodeau (1990): The Relaionship beween Median and Consan Qualiy House Prices: Implicaions for Seing FHA loan limis, Journal of he AREUEA, 18, 323-34. Henderson, J. V and A. J. Venables (2004) The Dynamics of Ciy Formaion: Finance and Governance, UK ESRC Cenre for Economic Performance a LSE. Mayo, S. K. (1986) Sources of Inefficiency in Subsidized Housing Programs: A Comparison of U.S. and German Experience, Journal of Urban Economics, 20, 229-249. Mills, E. and R. Simenauer (1996): New Hedonic Esimaes of Regional Consan Qualiy House Prices, Journal of Urban Economics, 39, 209-15. Reiff, L. O, and A. L. Barbosa (2005). Housing Sock in Brazil: Esimaion Based on a Hedonic Price Model, BIS Papers No. 21, Real Esae Indicaors and Financial Sabiliy, pp.257. Parry, I W H and W. Oaes (1998): Policy Analysis in a Second Bes World. RFF Discussion Paper 98-48. Resources for he Fuure. 28
Serra, M. V, D. E. Dowell, D. Moa and M. Donovan (2004): Urban land markes and Urban Land Developmen: An Examinaion of hree Brazilian Ciies: Brasilia, Curiiba and Recife, working paper 2004-03, Insiue of Urban and Regional Developmen, Universiy of California a Berkeley. Unied Naions (2003) The Challenge of Slums: Global Repor on Human Selemens. UN HABITAT, Nairobi. Wildasin, D. (1984): On public goods provision wih disorionary axaion. Economic Inquiry 22,2 227-43 Wolveron, M. L. (2000): Hedonic Esimaes of Regional Consan Qualiy House Prices, American Real Esae Sociey. World Bank (2002). Brazil Progressive Low-Income Housing: Alernaives for he Poor, Repor no. 22032 BR, World Bank LAC FPSI. 29
Table 1: Resource Availabiliy, Corresponding Uni Resource Price and Toal Coss/house Brasilia± Curiiba Recife Toal cos/ house (in Reais) 7185 7185 7185 Availabiliy of Building srucure/house (in sq. m) 32 32 32 Cos of building srucure/house (in Reais) 2800 2800 2800 Uni Land price wih infrasrucure ( in Reais/ sq. m) 164 109 102 Uni Land price wihou infrasrucure (in Reais/ sq. m) 157 38 44 Uni Price of infrasrucure (in Reais/ sq. m) = [land price wih infra land price wihou infra] 7 71 58 Availabiliy of Land area/house ( in sq. m)* 26.74 40.23 42.99 Cos of land/house (price of land wihou infra * land area/house) 4197.84 1528.72 2395 Cos of infrasrucure/ house 187.17 2856.28 1990 Raio of infrasrucure cos/house cos (p a /p h) ) 0.026 0.397 0.277 Raio of land cos/ house cos (p l /p h) 0.584 0.213 0.333 Raio of building cos/house cos (p b /p h ) 0.389 0.389 0.389 Credi availabiliy ω 0.27 0.27 0.27 Source: Land price/ square m, wih and wihou infrasrucure, has been aken from Serra e al (2004), Table 24. Toal cos of a residence and building srucure cos (cos of bricks plus maerials for a 32 sq.m srucure) is aken from Table 36: Cos Break Down of Urban Upgrading, Recife, 1998 (pp. 71), in Repor No. 22032 BR, Brazil Progressive Low-Income Housing: Alernaives for he Poor ha repors he COHAB esimaes based on Habia-Brazil projecs in he Recife Meropolian Region. Noe 1: Availabiliy of Land area/house = oal cos of land plus infrasrucure/ price of land plus infrasrucure ± The lower infrasrucure price in Brasilia, as compared o he oher wo ciies can be aribued o exorbianly high prices of unserviced land. Since infrasrucure is no a problem in Brasilia, a higher price for serviced land reflecs he disorion in marke for unserviced land. Table 2: Esimaed Equilibrium Parameer Values Values Parameers Descripion Brasilia Curiiba Recife α Share of land /house 0.6 0.35 0.46 β Discoun facor 0.91 0.91 0.91 r Real loan rae 0.099 0.099 0.099 r d Real deposi rae 0.027 0.027 0.027 ξ 1 Share of Infrasrucure/qualiy 0.03 0.4 0.28 ξ 2 Share of land/qualiy 0.58 0.21 0.33 ξ 3 Share of building/qualiy 0.39 0.39 0.39 30
Table 3: Seady sae Values of he variables per uni of housing Variables Descripion Values Brasilia Curiiba Recife Q Qualiy of housing 3540.94 5704.85 4771.33 L Average land area per house 0.935 1.497 1.256 B Average building srucure per house 0.934 1.518 1.261 A Infrasrucure per house 8.989 0.539 0.554 R Toal loan requiremen per house 6538.35 6538.35 6538.35 Noe: For easy comparison of differen siuaions wih respec o he equilibrium as base, we sandardize he value of land, building and infrasrucure wih respec o heir availabiliy per uni of house. For example, he uni of building srucure available per uni house is 32 square m (based on he source menioned a he boom of Table 1). We consider i as uni building srucure. Similarly for land and infrasrucure across hree ciies. Table 4: In Siu Programs wih Differen Qualiy Equivalence Targes and Resource Requiremens per uni of house (Disance=1 mile from he cener) Resource requiremen per uni of house Brasilia Curiiba Recife For a qualiy equivalence, k=1.25 Land 1.373 4.332 2.469 Building srucure 1.655 2.689 2.235 Infrasrucure 4 1.5 E 0.941 1.228 For a qualiy equivalence, k=1.5 Land 1.880 10.321 4.291 Building srucure 2.642 4.293 3.566 Infrasrucure 6 6.6 E 1.484 2.355 For a qualiy equivalence, k=1.75 Land 2.453 21.505 6.845 Building srucure 3.922 6.373 5.295 9 Infrasrucure 1.1 E 2.182 4.085 Noe: See noe wih Table-3 for deails abou uni of each resource 31
Table 5: Social Welfare gain per Real spen on each of he following inervenions for Differen values of Qualiy Equivalence (Disance=1 mile from he cener) Social Welfare gain per Real spen on Brasilia Curiiba Recife For a qualiy equivalence, k=1.25 Land 61.448 (2) 56.603 (2) 62.071 (2) Building srucure 76.419 (1) 49.771 (3) 58.672 (3) Infrasrucure 0.124 (3) 139.460 (1) 150.201 (1) For a qualiy equivalence, k=1.5 Land 53.848 (2) 26.936 (3) 41.358 (3) Building srucure 57.458 (1) 35.359 (2) 42.562 (2) Infrasrucure 0.000 (3) 100.241 (1) 90.677 (1) For a qualiy equivalence, k=1.75 Land 48.161 (1) 15.963 (3) 31.348 (3) Building srucure 45.148 (2) 29.405 (2) 34.664 (2) Infrasrucure 0.000 (3) 84.189 (1) 63.229 (1) Noe: Figures in he parenheses indicae he respecive rankings of each policy based on social welfare o cos raio. Highes value indicaes mos preferred as an inervenion. Table 6: Resource Requiremens for Relocaing Households a Various Disances from he Cener (Qualiy Equivalence, k=1.25) Resource requiremen per uni of house Brasilia Curiiba Recife In siu Upgrading, D=1 mile around cener Land 1.373 4.332 2.469 Building srucure 1.655 2.689 2.235 Infrasrucure 4 1.5 E 0.941 1.228 For Relocaion o a Disance, D=2.5 miles away from cener Land 3.025 38.385 9.898 Building srucure 5.358 8.707 7.234 Infrasrucure 10 6.5 E 2.958 6.308 For Relocaion o a Disance, D=5 miles away from cener Land 5.499 199.941 28.290 Building srucure 13.031 21.175 17.592 Infrasrucure. 15 6.8 E 7.036 21.748 For Relocaion o a Disance, D=7.5 miles away from cener Land 7.799 525.007 52.292 Building srucure 21.915 35.611 29.585 18 Infrasrucure 5.8 E Noe: See noe wih Table-3 for deails abou uni of each resource 11.679 44.861 32