The effects of state-subsidised housing on poverty in Cape Town

ABSTRACT

Property rights are widely thought to have considerable direct and indirect effects on urban poverty. However, few studies have been conducted and the evidence supporting these claims is scarce, especially in Southern Africa. This article examines effects of property rights in South Africa through a case study of subsidised housing for poor people in Khayelitsha, Cape Town. A difference-in-differences estimation strategy is employed. Results show that housing subsidies are associated with better physical health and (counter-intuitively) more teenage pregnancies. Improvement in health is attributed to better housing quality and environment while increase in teenage pregnancies is attributed to increased privacy. Effects of titling extend to social effects, which have been understudied in the literature. Since titling showed no effect by most measures, it is likely that poverty is driven so strongly by factors such as unemployment and poor location of housing projects that property rights make little overall difference to poverty.

KEYWORDS:

• Property titling
• housing subsidies
• urban poverty
• difference-in-differences
• property rights

1. Introduction

Since the 1970s, governments across the developing world have implemented large-scale slum upgrading and homeownership schemes aimed at increasing tenure security and reducing poverty (Buckley & Kalarickal, 2006). These schemes favour the delivery of property rights to the poor through land or property titling over administrative recognition of occupancy (De Soto, 2000; Leckie, 2004; Durand-Lasserve & Selod, 2009:110–1). Titling can be defined as the process of integrating informal tenure into a formal property rights system by delivering real property rights; that is, provision of freehold or leasehold rights through title deeds (Durand-Lasserve & Selod, 2009:105). Less favoured is the administrative recognition of occupancy. Administrative recognition of occupancy is a process of delivery of temporary rights – through documentation such as occupancy licenses – in which authorities set conditions for occupancy such as building materials (Durand-Lasserve & Selod, 2009:10).

Since 1994 the post-apartheid government in South Africa has made massive investments into housing subsidies through which it delivered millions of houses under freehold title. The South African state promised and committed itself to ‘the establishment of viable, socially and economically integrated communities’ under the Reconstruction and Development Programme’s (RDP) 1994 White Paper on Housing – which became the Housing Act of 1997 (Republic of South Africa, 1994:clause 4.2). The 2004 South African Housing Policy reiterated this position under the Breaking New Ground policy. Between 1994 and 2013, the post-apartheid state claims to have built over 3.6 million houses nationwide and thus to have benefitted more than 10 million people (Republic of South Africa, 2013). In Cape Town, with a population of 3.5 million and where 16% live in informal settlements, the state built 323 000 housing units and provided 140 200 serviced sites (Provincial Government of the Western Cape, 2013; Republic of South Africa, 2013).

Given this large-scale implementation, an important question to ask is are the massive investments made into housing subsidies by the South African government or other governments worthwhile and have they benefitted target populations? The answer is that we do not yet know enough. Despite these massive investments, empirical knowledge on the effects of titling on individuals, households and communities in developing countries is scarce (Durand-Lasserve & Selod, 2009:115). Some scholars have gone on to argue that the claims about the benefits of titling are exaggerated given the scarcity of evidence (Reerink & Van Gelder, 2010:84). Particularly in Southern Africa, hardly any evaluations have been carried out. Besides, studies about the effects on social aspects are even scarcer. Most studies are on economic effects; namely labour participation, housing investments and access to credit.

To address the above-discussed gap in empirical knowledge, this article systematically examines economic and social effects of the subsidised homeownership scheme in South Africa with a focus on Khayelitsha, a township in Cape Town. The article examines the benefits of both titling and homeownership in comparison with the possession of administrative occupancy rights, which shack-dwellers in Khayelitsha possess. It not only examines economic effects but goes beyond this narrow focus to include an examination of other social aspects. This article helps advance our understanding of effects of titling in the Southern African context.

The article will argue that the benefits have been modest at best compared with the massive investments made. In particular, important economic effects are not evident because of macro-economic factors and the location of housing projects away from opportunities for social mobility.

2. Literature

Although there is a rich literature on housing in South Africa, there are few empirical studies enquiring into the claims made of property titling. These studies mainly focus on economic benefits and less on social effects.

In an econometric analysis of the claims that titling increases labour market outcomes, Franklin (2011) has provided evidence to support the claims. Franklin (2011) examined the impact of the South African government’s housing subsidy programme by examining the effects of the new housing policy on labour market participation, earnings and household income in neighbourhoods that received housing subsidies in Cape Town. He demonstrated that labour market participation, earnings and household income were higher among beneficiary households. They were particularly higher for female young adults. Franklin’s findings apply to an aggregation of neighbourhoods. They therefore require further testing in specific neighbourhoods.

Advocates of titling claim that it can lead to an increase in income through its effect on access to credit (De Soto, 2000). In the South African context, there is no evidence of this effect. Beneficiaries of housing subsidies are wary of the risk involved in using their property as collateral (Kingwill et al., 2006:53–65; Boudreaux, 2008; Lemanski, 2011).

Furthermore, proponents suggest that titling increases wealth (De Soto, 2000). The idea is that titling permits home-based entrepreneurial investments, which in turn increase wealth. Qualitative evidence supports the idea that home-based investments increase, but not meaningful wealth. Boudreaux (2008) demonstrated that some title-holders use their homes as secure places of business, generating income that residents use to support their families. Meintjes (2000) finds that homeowners in Lusaka city and Soweto Township feel an obligation to fill up their house with household durables. Despite it being so at the household level, Rust (2006:44–52) shows that, at a macro level, property titles are not enough because they require a functioning secondary property market, sufficient housing stock, people’s ability to afford repayments on housing finance and mortgage lenders willing to move downmarket. These conditions are inadequate in South Africa. In essence, titling may lead to increased home-based entrepreneurial activities but these are found not to be meaningful enough to create long-term substantive and sustained wealth given the macro-level conditions.

Boudreaux further suggests that titling increases tenure security and results in various household investments. She found that title-holders in Langa, Cape Town invested in improving their homes, which raised property values (Boudreaux, 2008). Also, home improvement projects provided entrepreneurial opportunities for a wide variety of local artisans. However, because Boudreaux’s study relied on qualitative evidence, it is deficient of evidence on the extent of improvements and the percentage increase in property values. In addition, it has been shown elsewhere that untitled households may also make improvements which the study failed to acknowledge.

Notwithstanding the scarcity of quantitative analysis, several qualitative studies – with the exception of Payne et al. (2009) – explore the link between low-income housing and emerging relationships in low-income neighbourhoods in South Africa (Lemanski, 2006, 2011; Seekings, 2008b; Harper & Seekings, 2010; Ross, 2010; Muyeba & Seekings, 2012). Payne et al. (2009), who do not support the focus on titling and false dichotomy of de-jure versus de-facto tenure but suggest that tenure security exists on a continuum (see Payne, 2001), found no evidence of effects of titling on access to credit and other economic aspects in South Africa and Senegal. They suggest, along with Napier et al. (2013), that tenure reforms in developing contexts should consider all tenure types and use those that best fit the context.

The rest of the qualitative literature suggests that as a result of obtaining ownership of new low-income housing, residents espouse values of respectability, decency, dignity and belonging as they forge their relationships (Lee, 2005; Ross, 2010). These values are in contrast to dehumanising experiences during apartheid. Further, community-making efforts among subsidy beneficiaries are undermined by persistent segregation (Oldfield, 2004; Lemanski, 2006, 2011). Nevertheless, there is evidence of tolerance as a result of contact in new mixed neighbourhoods (Muyeba & Seekings, 2011). Other factors that constrain community-making include jealousy and gossip (Ross, 2010:160–3), witchcraft (Ashforth, 2005) and violent crime and criminality (Centre for the Study of Violence and Reconciliation, 2010). The literature further identifies coping mechanisms such as kinship support where it has been found that this is not as supportive as previously thought because residents are recognising fewer and more conditional obligations toward kin (Seekings, 2008a; Harper & Seekings, 2010). Domestic fluidity in the sense of individuals moving between households and being members of multiple households is another coping mechanism (Spiegel et al., 1996; Seekings, 2008a). Others turn to privacy, which constrains community-making (Muyeba & Seekings, 2012).

As the foregoing discussion has shown, evidence for many of the claims about the effects of titling remains unsubstantiated. Gaps remain in our understanding of the effects on both economic and social aspects. The quantitative literature and some qualitative-based literature focuses on economic effects, while the qualitative literature largely focuses on describing emerging relationships in neighbourhoods of newly-titled households. None of the quantitative-based literature addresses social aspects. In light of this state of the literature, this article interrogates the propositions that titling translates into positive effects on labour market participation, household per-capita income, wealth, physical health and neighbourhood stability, and lowers teenage pregnancies and school dropout rates. In particular, these propositions are interrogated in the Southern African context in Khayelitsha Township.

3. The study context: Khayelitsha

Khayelitsha (meaning ‘new home’ in isiXhosa) was a township planned in 1983 by the apartheid state to accommodate Africans that were legally in Cape Town and squatting in formal African townships (Cook, 1992:125). The development of Khayelitsha was the apartheid government’s response to the severe shortage of housing for Africans in the Western Cape (Seekings et al., 1990:8). Building of houses and formal occupation occurred by 1985. Ninety-nine years’ leasehold tenure was allowed for Africans who were legally in Cape Town at the time. However, in November 1984, illegal African migrants were allowed to build shacks on site-and-service plots in Site C, Khayelitsha, near the N2 highway (Cook, 1992:125). This was a temporary administrative recognition of occupancy for illegal migrants while they awaited deportation. Site C was first established as a transit camp for transfer of illegal migrants to rural areas (Cook, 1992; Mdewu, 2004; Zonke, 2006). One tap was provided for four sites and one bucket toilet for every two sites. However, adjacent land in Town One, Town Two and Makhaza that had no services of any kind was also occupied.

Following the establishment of Site C, dense clusters of shacks developed – putting a strain on services. By mid-1989, an estimated 13 000 families were squatting on un-serviced sites while the population increased without a corresponding increase in serviced sites. In 1988, the population of Khayelitsha was approximately 189 000. This increased to 305 323 in 1989 and was estimated to have stabilised around 450 000 in late 1990. At the twilight of apartheid, 162 000 people lived in shacks on un-serviced sites in Site C, Site B, Town One, Town Two and other areas of Khayelitsha (Cook, 1992:125–30).

In 1994, following the development of the targeted once-off housing subsidy – discussed in the Introduction – subsidy allocation began and was carried out in phases. To be allocated a subsidy, applicants had to prove using a payslip that they had a maximum monthly household income of R3500 (Republic of South Africa, 1994). Joint spouses who earned R800 and below obtained a capital subsidy worth R15 000. Those who earned between R801 and R1500 obtained a subsidy worth R12 500. Beneficiaries contributed R2479 intended for them to have an economic stake in their asset (Tomlinson, 2006). Beneficiaries began to move into their houses from 1997. Allocation of houses continued over the course of the 2000s.

Although housing allocation generally met targeted households, there were problems with the process of allocation – which had a bearing on the profile of beneficiaries and non-beneficiaries, and therefore on the empirical analysis comparing homeowners and shack-dwellers. Firstly, the waiting list did not work on a first-come, first-served basis. There were many waiting lists which were drawn up during the apartheid years and under the post-apartheid era. Some people were on multiple lists in different locations. Attempts by municipalities and the province to merge lists failed (Tissington et al., 2013:25). Although the Cape Town Municipality kept a waiting list, they alongside it used a system based on allocation to communities in catchment areas of specific building projects. Housing subsidies were allocated according to quotas for each community under a specific housing project. Communities established project committees that were responsible for allocating houses to their members (Franklin, 2011:10; Tissington et al., 2013:26). Through this system, people of Khayelitsha were called to register during a given time period of up to two weeks at a time if they wished to apply. It did not matter how long an individual had lived in the community. Some applicants came from other communities during the registration period and from as far as the Eastern Cape in order to register and were successful in being awarded subsidies. Meanwhile others had received their subsidies through the housing waiting list, the allocation of which was chaotic. There were occurrences of fraud and corruption. Some beneficiaries received more than one subsidy. Some who had not been allocated subsidies invaded houses but were allowed to stay on by the state (Franklin, 2011:10; Tissington et al., 2013:26).

4. Methodology

Despite these problems in subsidy allocation, the case of Khayelitsha provides an opportunity for the investigation of the effects of real property rights (of both titling and homeownership) relative to administrative recognition of occupancy. The existence of a group that benefitted and one that did not at any point in time provides a treatment group and a comparison group. Those who remained in shacks in Site C, Site B, Town One and Town Two are the comparison group while those who moved to serviced sites in Site B, Makhaza and Kuyasa and eventually moved into state-built houses are the treatment group. Further, the allocation of subsidies was done in phases which ensured that there were beneficiaries and non-beneficiaries existing simultaneously over the course of the 2000s.

4.1. Data

Data from the Cape Area Panel Study (CAPS), which is an ongoing panel study in Cape Town, were analysed to examine the effects. CAPS follows the lives of a large representative sample of young adults living in metropolitan Cape Town as they undergo multiple transitions from adolescence to adulthood. However, CAPS also administers household-level questionnaires including a survey for household heads. The first wave commenced in 2002. There have been four subsequent waves with the latest wave at the writing of this article being wave 5, carried out in 2009. CAPS includes a range of aspects of adolescence including schooling, entry into the labour market and housing, neighbourhood, sexual and reproductive health, and family and kin relations. The panel asks questions about whether the household head or anyone in the household received a government housing subsidy to acquire land or build the house. The panel consists of approximately 4750 randomly selected individuals aged between 14 and 22 years in 2002 and between ages 22 and 35 in 2009. Households were sampled using probability proportional to size and stratified according to population group. First, 440 clusters roughly corresponding to the 1996 Census enumerator areas were selected. Then some households were selected from each cluster. From each household, up to three young adults were selected. Since CAPS covers the entire Cape Town, a sub-sample consisting of only those respondents who were from Site C Khayelitsha in 2002 was selected.

Analysis was restricted to those respondents who reported not having a housing subsidy in wave 1 (2002) of CAPS but reported having one in wave 3 (2005). The total sample size in 2002 was 330. By 2005, 111 had obtained subsidies (treatment) and 219 had not (control). The effects were examined four years later using wave 5 data (i.e. in 2009). Only those who reported that they had a housing subsidy and lived in their houses in both 2005 and 2009 were part of the treatment group. Similarly, only those who reported that they lived in a shack in both 2005 and 2009 and reported not obtaining a housing subsidy were part of the control group. The same individuals were tracked since CAPS follows individuals even when they move. Table 1 presents a description of the variables used in the analysis.

Table 1. Variable descriptions.

Variable	Response categories	Description
Labour participation	Group average	‘How many hours do you work on your current job per day?’
Per-capita income	Group average	Log of household income per capita from imputed per-capita income variable
Household durables	Average, scale	‘Does anyone in your household own … ?’ (e.g. radio, television, microwave, refrigerator, car, etc.)
Physical health	Ordinal scale	‘How good is your health?’ (0 = poor, 1 = fair, 2 = good, 3 = very good, 4 = excellent)
Teenage pregnancy	Dummy	‘Have you ever been pregnant?’ (1 = pregnant, 0 = otherwise), restricted to females below age of 18 years
School dropouts	Dummy	Average number of respondents who had not completed matric and were above 18 years (1 = dropout, 0 = otherwise)
Neighbourhood stability	Dummy	‘Did your household move in the past five years?’ (wave 1)/‘Have you moved since the last interview?’ (waves 4 and 5) (1 = moved in last four/five years, 0 = otherwise)
Membership in associations	Composite scale	‘Are you a member of … ?’ (e.g. youth group, political party, church, etc.)

4.2. Analytical framework

In the estimation strategy, difference-in-differences (DID) was used to estimate the effect of a housing subsidy for the outcome variables. The models are of the form:where is any of the outcome variables for observation i at time t, treatment period is a dummy variable with the value of 1 for the post-treatment period and 0 for the pre-treatment period, RDP homeowner is a dummy taking the value 1 if the individual is in the treatment group and 0 if they are in the control group, and the coefficient of the interaction between treatment period and RDP homeowner, , is the Ordinary Least Squares (OLS) regression DID estimator or the logistic regression odds ratio. The vector specifies the pre-treatment covariates or controls.

DID coefficients for labour participation, per-capita income, assets, physical health and membership in associations are estimated using OLS regressions. The DID coefficients for teenage pregnancies, stability and the proportion of dropouts are odds ratios estimated using logistic regressions.

5. Results

5.1. Characteristics of RDP homeowners and shack-dwellers

Table 2 shows that there are no significant differences in terms of demographics and also shows significant differences in some characteristics of the dwelling in both periods, with beneficiaries scoring higher. The lack of demographic differences supports the use of shack-dwellers as the comparison group in the DID estimation.

Table 2. Characteristics of housing subsidy beneficiaries versus shack-dwellers in 2002 and 2009.

	Wave 1 (2002)			Wave 5 (2009)
	Control	Treatment		Control	Treatment
Variable	(N = 219)	(N = 111)	t	(N = 187)	(N = 102)	t
Age	17.7	18.0	−1.16	24.5	24.6	−0.13
Male	0.43	0.43	−0.10	0.42	0.45	−0.50
Education	8.44	8.72	−1.14	10.03	10.26	−0.92
Marital status	0	0	0.72	0.07	0.10	−0.84
Child	0.22	0.27	−0.69	0.32	0.31	0.21
Household size	5.35	5.66	−1.43	5.52	6.06	−1.62
Employment	0.10	0.12	−0.63	0.48	0.43	0.61
Rooms	3.7	3.7	−0.14	3.6	4.3	−3.63***
Piped water	0.61	0.94	−6.60***	0.67	0.93	−5.10***
Flush toilet	0.75	0.93	−4.05***	0.70	0.93	−4.66***

Note: ***p < 0.01.

At baseline (2002) both groups have an average age of 18 years with t(317) = −1.16. At endline (2009), respondents in both groups are on average aged 24 years with t(225) = −0.13. At baseline the proportion of males to females, 0.43 with t(328) = 0.10, for both groups shows no difference, while at endline the proportion of males in the treatment group, 0.45, is slightly more than that in the comparison group, 0.42, but with no statistically significant difference, t(225) = −0.50. However, in both groups and both periods, we see that there are fewer males than females in the sample. I find that, on average, both groups have the same number of years of education. At baseline, beneficiaries score an average of 8.7 years while shack-dwellers average 8.4 years, t(273) = −1.14. At endline, both groups have an average of 10 years with t(271) = −0.92, showing no significant difference between them. There are almost no married respondents at baseline; the proportion of married respondents being 0 for both groups, with t(314) = 0.72. At endline, there seem to be more married respondents among beneficiaries, 0.10, than among non-beneficiaries, 0.07, although the difference is not statistically significant, t(225) = −0.84. Similarly, the proportion of respondents with children at both baseline and endline did not differ between the two groups. At baseline, the proportion for beneficiaries is 0.27 while that for shack-dwellers is 0.22, t(317) = −0.69. At endline, we see an increase to 0.31 for beneficiaries and to 0.32 for shack-dwellers, t(224) = 0.21. It appears that the rate of increase in the proportion of respondents with children was faster among shack-dwellers than among beneficiaries over the same period. In terms of household size, both groups have an average of 5 members at baseline, t(319) = −1.43, while at endline beneficiaries have almost significantly more members with an average of 6.0 while shack-dwellers have 5.5 members, t(286) = −1.62. However, there is no statistically significant difference. Finally, there are no differences in the proportion of employed respondents. Few respondents are employed at baseline, with beneficiaries scoring 0.10 and shack-dwellers 0.12, t(317) = −0.62. The proportions rise to almost half at endline, with beneficiaries scoring 0.43 and shack-dwellers 0.48 but with no statistically significant difference, t(225) = 0.61.

There are differences in the characteristics of the dwelling in both periods, except for number of rooms in which there are no differences only at baseline, t(328) = −0.14. The significant differences are in the household’s most often used source of drinking water and the most often used kind of toilet for the residence. In this case, RDP homeowners have better sanitation facilities than shack-dwellers at both baseline and endline. Among beneficiaries, a proportion of 0.94 have access to piped water versus 0.61, t(328) = −6.60, at baseline and 0.93 versus 0.67, t(286) = −5.10, at endline. At baseline, the proportion of beneficiaries with access to a flush toilet is 0.93 versus 0.75, t(325) = −4.05, and 0.93 versus 0.70, t(286) = −4.66, at endline. This reveals that more beneficiary households lived on serviced sites than non-beneficiary households prior to treatment. It may also indicate that the roll-out of housing subsidies prioritised households living on serviced sites.

Table 3 reports that there are no significant differences in the number of hours worked on the current job, and that beneficiaries are likely to report lower income, more durables, poorer physical health and a higher proportion of teenage pregnancies. There are no significant differences in the proportion of school dropouts, stability of tenure and membership in voluntary associations. At endline, there are no differences in number of hours but beneficiaries are likely to report less per-capita income and more assets. Physical health, teenage pregnancies, proportion of school dropouts and stability are the same. There is significantly less membership in voluntary associations among beneficiaries.

Table 3. Summary statistics on various outcome variables: beneficiaries versus shack-dwellers in 2002 and 2009.

	Wave 1 (2002)			Wave 5 (2009)
	Control	Treatment		Control	Treatment
Variable	(N = 219)	(N = 111)	t	(N = 187)	(N = 102)	t
Labour	0.77	1.00	−0.74	4.02	3.41	1.03
Income	5.56	5.24	3.47***	6.11	5.88	1.97**
Durables	3.89	4.77	3.24***	6.01	6.63	−2.08**
Health	4.13	3.78	2.88***	3.81	3.94	−0.85
Pregnant	0.06	0.40	2.97***	0.28	0.31	−0.32
Dropout	0.71	0.78	−0.78	0.69	0.73	−0.77
Stability	0.86	0.91	−1.34	0.90	0.93	−0.79
Memberships	1.50	1.50	0.06	0.95	0.64	2.82***

Note: **p < 0.05; ***p < 0.01.

On the basis of these mean scores and mean differences, we cannot ascertain that the government housing subsidy was responsible for the observed differences, particularly during the post-treatment period. This is because, from the post-treatment mean differences, exogenous factors that may have affected the parameters of interest during the course of time are not controlled for. Constant factors in terms of demographics are controlled for, however, since the same individuals are observed at different points in time. To control for exogenous factors, DID estimation is employed.

5.2. Effects of the housing subsidy on labour participation, income, durable assets, physical health, membership in associations, stability, teenage pregnancies and proportion of school dropouts: DID estimation

This section reports the DID estimates for the outcome variables. The data show that the housing subsidy has the effect of improving self-reported health and reducing teenage pregnancies among beneficiaries but has no effect on the other hypothesised outcome measures.

As Table 4 reports, the marginal effect on physical health estimated by the coefficient of the interaction term between RDP homeowner and treatment period is significant at 0.50, controlling for demographic and other explanatory factors. The DID estimator represents the estimated improvement in health attributed to being a beneficiary of a housing subsidy. This coefficient means that a housing subsidy increases physical health by 0.5 on a scale of 1 to 5, with 1 being poor and 5 being excellent. The housing subsidy has had an impact of improving self-reported health by 10% among the beneficiaries in the data.

Table 4. Effects of the housing subsidy on labour participation, household per-capita income, household durables, physical health and membership in voluntary associations: difference-in-differences estimation (multivariate OLS regression).

	(1) Hours worked per day		(2) Per-capita household income (log)		(3) Number of household durables owned		(4) How is your health?		(5) Membership in voluntary associations/organisations
	Pre-treatment	Post-treatment	Pre-treatment	Post-treatment	Pre-treatment	Post-treatment	Pre-treatment	Post-treatment	Pre-treatment	Post-treatment
RDP homeowners	3.24	0.62	6.0	6.6	0.21	0.35	2.99	3.04	0.09	−0.02
	(1.85)	(1.86)	(0.41)	(0.41)	(0.08)	(0.08)	(0.52)	(0.52)	(0.10)	(0.10)
Shack-dwellers	3.16	0.06	6.2	6.64	0.16	0.32	3.35	2.89	0.12	0.04
	(1.84)	(1.83)	(0.40)	(0.40)	(0.08)	(0.08)	(0.51)	(0.51)	(0.09)	(0.09)
Difference	0.08	0.68	0.22*	0.05	0.06**	0.03	0.35**	0.14	−0.03	−0.06**
	(0.49)	(0.50)	(0.11)	(0.12)	(0.02)	(0.02)	(0.14)	(0.14)	(0.02)	(0.03)
Difference-in-differences		0.60 (0.70)		0.16 (0.16)		−0.02 (0.03)		0.50** (0.20)		−0.03 (0.04)
Adjusted R^2		0.209		0.244		0.284		0.057		0.123
Observations		464		417		437		436		440

Notes: Standard errors are in parentheses. Only demographic and significant variables are controlled for in each model. Regression of hours controls for sex, marital status, children, education, household size and physical health. Regression of per-capita income controls for age, sex, marital status, children, education, hours of labour and household size. Regression of household durables controls for age, sex, marital status, children, education, employment status and household size. Regression of physical health controls for age, sex, marital status, education, hours of labour and household size. Regression of membership in associations controls for age, household size, education, civic duties and political awareness. Number of observations is a combined total of the baseline and endline observations. OLS = ordinary least squares; RDP = Reconstruction and Development Programme. *p < 0.10; **p < 0.05.

There is no effect of the housing subsidy on the number of labour hours worked. The marginal effect implied by the estimated coefficient on the interaction term between RDP homeowner and treatment period is 0.60 hours. This coefficient corresponds to an increase in the number of hours of labour. However, the estimate is not statistically significant. This means that housing subsidy beneficiaries work a similar number of hours to shack dwellers. The dramatic fall in number of hours worked per day for both housing beneficiaries and shack-dwellers can be attributed to the increase in unemployment as a result of the 2007 global economic recession. Given such macro-economic factors, it is not surprising to find no effect of the housing subsidy on labour hours. Similarly, there is no effect of a housing subsidy on household per-capita income. The DID estimate is an increase of 0.16. This implies an increase of 16% in household per-capita income. However, this was not statistically significant. Likewise, there is no effect on the number of household durables. The estimated treatment effect is −0.02, implying a reduction in the number of household items. This effect is not statistically significant. I find no effect on the proportion of school dropouts. The marginal effect reported in the DID estimate is 0.77 but this is not statistically significant. With regard to membership in voluntary associations, I find no effect as well. The estimated DID coefficient is −0.03, representing a reduction in memberships that is not statistically significant.

Table 5 reports DID estimates for the logistic regression of socially desirable youth behaviour and stability against homeownership. With regard to teenage pregnancies, I find that teenage girls in households of beneficiaries have higher odds of being pregnant than teenage girls who are shack-dwellers. The DID estimates that the odds of being pregnant for a teenager in beneficiary households are 0.02 times as large as the odds for a teenage girl belonging to a household of shack-dwellers when controlling for other factors (0.11 without controls). Although teenage girls from beneficiary households have a significantly higher proportion at baseline, this reduces after treatment while that of shack-dwellers rises during the same period. Nevertheless, it remains higher. The direction of association is opposite to what was expected. This means that being a beneficiary of a housing subsidy is associated with a marginally larger proportion of teenage girls who become pregnant. I also report no effect on stability, with the odds ratio being 1.42.

Table 5. Effects of the housing subsidy on teen pregnancy and proportion of school dropouts: difference-in-differences estimation (multivariate logistic regression).

	(1) Proportion of teenage pregnancies	(2) Proportion of school dropouts	(3) Stability
Treat_period	25.79***	0.53	4.01***
	(30.53)	(0.23)	(2.05)
RDP_homeowner	49.25***	1.51	2.05
	(65.85)	(0.94)	(1.16)
Period_homeowner	0.02***	0.77	1.42
	(0.03)	(0.55)	(1.75)
Pseudo R^2	0.186	0.079	0.176
Observations	116	273	409

Notes: Standard errors in parentheses. Only demographic and theoretically important variables controlled for. Regression of teenage pregnancy controls includes age, marital status, education and per-capita income. Regression of school dropout controls includes age, sex, marital status, children, household size, per-capita income and mother’s education. Controls for stability include age, sex, marital status, children, household size, education, per-capita income, hours of labour and memberships. Number of observations is a combined total of the baseline and endline observations. RDP = Reconstruction and Development Programme. ***p < 0.01.

6. Discussion

As the evidence shows, we should expect that moving shack-dwellers to houses under the housing subsidy would improve their self-reported health by 10%. The effect estimated is for general self-reported health status, which has not been done before. The mechanism seems to be that the improvement in housing quality and environment as shown by the data on access to piped water and toilets actually translates into better health. Shacks in Khayelitsha and other townships are susceptible to fire, flooding and dust; sources of respiratory and waterborne diseases. Once these are improved, health status improves. There are limitations to this finding. One cannot determine on the basis of the measure used whether the identified effect is long term or short term and whether the same effects are observable on both chronic diseases and short-term illnesses. Neither does the measure apply to child health. Previous studies elsewhere have found positive effects of titling on children’s health. Positive effects of titling on short-term health (weight for height) among children but not long-term health (height for age) have been found in Argentina (Galiani & Schargrodsky, 2004; Vogl, 2007). In Peru, however, the improvement raises the risk of obesity among children (Vogl, 2007). Among adults, titled households have a lower occurrence of some chronic diseases although the actual mechanism through which this occurs is not evident (Gandelman, 2010). Research work remains scarce on the effects of titling on physical health among adults using a more objective measure such as data from physical medical examinations.

Counter-intuitively, having a housing subsidy resulted in an increase in teenage pregnancies. Parents may not allocate more resources to education and monitoring of children because they may not have the financial resources to allocate in the first place. Also, with low levels of education, they may not have the kind of cultural capital to transmit to their children. In addition, a house may provide the privacy that is lacking in shacks for teenagers to engage in sexually risky behaviour. The shacks may provide a deterrent for teenagers since the shacks are very close together, rooms are small and the wall materials thin and unstable. In contradiction, elsewhere Galiani & Schargrodsky (2004) found lower teenage pregnancies among untitled households. Caution must also be exercised in interpreting this result. The sample size at baseline was very small and this may have had an influence as the sample size grew larger. Larger sample sizes at baseline may help improve representativeness and generalisability.

The lack of effect on labour market participation is inconsistent with the theory and can be explained in two ways. Firstly, on careful examination we can realise that in an economy with high unemployment (24% in 2010; Statistics South Africa, 2011) and decreasing market demand for unskilled labour, it is not a realistic expectation that an increase in supply of unskilled labour would lead to an increase in the number of hours worked or let alone to an increase in employment status. In this case, the pool of the unemployed increases but people cannot find work even though they have the time. In an extensive study using detailed nationally representative surveys, Banerjee et al. (2008) investigated why so many South Africans were unemployed. They found that large increases in labour supply (especially among women) have not been matched by an increase in demand for labour (Banerjee et al., 2008:716). This mismatch is especially acute for unskilled labour where there is declining demand in mining and agricultural sectors. Unemployment has persisted in part because as demand for labour has been falling, it has not been accompanied by a fall in wages enough to clear the labour market (Banerjee et al., 2008:716–7). In this kind of labour market, increases in labour market participation within titled households (or resulting from any other intervention) would have little or no effect on employment and income. Secondly, a strong criticism of the housing subsidy programme is that housing projects are poorly located, far from Central Business Districts (CBDs) where employment and other opportunities for social mobility are concentrated (Charlton, 2013; Turok 2013:169). Khayelitsha is located 35 kilometres from Cape Town CBD, which necessitates a long and financially costly commute. A return bus or train ride will cost ZAR20. In the broader literature, this finding is similar to that of Galiani & Schargrodsky (2010) but differs from that of Field (2003) who found that titling in Peru leads to an increase in labour participation. The mechanism and economic context in Field’s study are different. In Field’s case, titled households had significantly low tenure security at baseline that they had to have an adult member present at home at all times. In my study, beneficiaries already had considerable tenure security because they had been allocated plots and were provided with services before baseline. Arguably, in Field’s study the Peruvian economy could absorb an increase in labour supply. Perhaps further research can be done in contexts where unskilled labour is in demand and where there is low unemployment.

The finding that titling is not associated with income increases should not be surprising because previous studies find similar evidence. Galiani & Schargrodsky (2010) found no effect of titling on income in Buenos Aires. There is also a resemblance to the findings of Franklin (2011), who found no association between the housing subsidy and income in Cape Town but, when he broke down the analysis by gender, found significant impacts on income for female beneficiaries. These female beneficiaries were able to reallocate their time from home to the labour market. The findings also differ from Moura et al. (2011), Field & Torero (2006) and Field (2003). In all these cases, an increase in income was preceded by an increase in labour market participation which may not have occurred in this case.

Titling did not result in increases in wealth. Although beneficiaries of housing subsidies have significantly more household items, none of this difference can be attributed to the housing subsidy. Perhaps beneficiaries already had enough space such that they had no obligation to fill up the house, as Meintjes (2000) expected.

Finally, and pertinent specifically to the South Africa housing subsidy programme, is the absence of a relationship with neighbourhood stability. In the first place, the neighbourhoods were highly stable with only 10% of the respondents reporting a move in the previous four to five years before the baseline and endline surveys. It is possible that shack-dwellers did not move in anticipation of obtaining a housing subsidy. It is also possible that the reported movements out of the neighbourhoods obscure moves that occur when a household sells the house and moves back into a backyard shack within the same neighbourhood. However, research has shown that these neighbourhoods are in a high state of flux (Seekings, 2008a). A question that needs to be answered is how long do beneficiaries live in their houses after obtaining a housing subsidy and which beneficiaries are likely to live long in the house? For example, many of the beneficiaries in Cape Town who originate from the Eastern Cape do not consider Cape Town as their home. In fact, they look forward to their return. Are other tenure options such as administrative recognition of occupancy more cost-effective in this regard? Such are questions that need to be re-examined for the subsidy programme to improve cost-effectiveness.

7. Conclusion

There is scarcity of empirical research about the effects of titling in Southern Africa. This article examined the effects of titling and slum upgrading in this part of the developing world by evaluating the economic and social effects of subsidised housing in Khayelitsha, a township in Cape Town, South Africa. This is with a view to understanding the benefits of real property rights and ownership of a capital asset among beneficiaries of low-income households. As other studies have done, this study also revealed the methodological challenges in studying effects of titling. Out of eight measures, the housing subsidy improves self-reported health status by 10% and increases the proportion of teenage pregnancies. The actual effects were smaller and fewer than predicted. Titling was not associated with improvements in most of the measures of poverty reduction, supporting the argument that the benefits of titling may be exaggerated if researchers focus on selected variables rather than on a broader set of measures. It is likely that poverty in Khayelitsha is driven so strongly by factors such as unemployment and poor location of housing projects that real property rights make little overall difference to poverty. The positive effect on physical health is small and can more appropriately be attributed to better housing quality and environment rather than to stronger tenure rights. This implies that cheaper policies with wider coverage, such as administrative recognition of occupancy coupled with upgrading, may produce similar results. The effect of increased teenage pregnancies is counterintuitive and inconsistent with other studies in other contexts. In sum, the benefits of the South African housing subsidy programme in Khayelitsha have been modest and raise questions about whether the massive investments are worthwhile.

Acknowledgement

The author wishes to acknowledge the contribution of Professor Jeremy Seekings for his valuable comments and input.