Does strength of tenure rights among the urban poor improve household economies? Contrasting Matero and George in Lusaka city

ABSTRACT

In a perspective shifting approach, Payne showed that tenure security exists on a continuum from weak to strong tenure rights, rather than as de facto versus de jure categories. This article examines claims that stronger tenure rights improve property values, credit access, employment, household income and participation in income-generating activities among the poor. The article contrasts these outcomes between households with leasehold and occupancy rights in two low-income urban neighbourhoods in Lusaka. The paper draws on household survey data (n = 623) and in-depth qualitative interviews. Multiple regression analysis is utilized and propensity score matching methods are employed to overcome selection bias. The results show that leasehold rights are associated with significantly higher property values and less participation in income-generating activities. Poorly performing macroeconomic factors – such as high unemployment – may drive household economies so strongly that stronger rights make little impact on employment status, access to credit and household income.

KEYWORDS:

• Property rights
• propensity score analysis
• urban poverty
• tenure security
• low-income housing
• Lusaka

1. Introduction

The evidence for the efficacy of property rights is at best equivocal despite massive investments by developing countries over the course of about 50 years. Protagonists have claimed that property rights increase property values, access to credit, employment, household income and housing investments (De Soto 1989, 2000). However, some scholars have argued that the claims about the benefits of titling are exaggerated (Reerink and Van Gelder 2010, p.84) and evidence suggesting improvement in household economies on account of tenure security is scarce (Durand-Lasserve and Selod 2009, p. 115; Payne et al. 2009; Reerink and Van Gelder 2010, p. 84). Moreover, few empirical studies have been conducted to validate the claims in the developing world (Durand-Lasserve and Selod 2009, p. 115) and knowledge on Southern Africa is even scarcer despite the existence of policies to strengthen tenure rights.

Until recently, hypotheses were tested using a legal lens that viewed property rights in a dichotomy of de jure versus de facto tenure rights (Payne 2001; Van Gelder 2013). In a perspective-shifting approach, Payne showed that tenure security exists on a continuum from weak to strong tenure rights, rather than as a dichotomy as previously defined. According to Payne (2001), the weak end of the continuum has street homelessness or pavement dwelling, squatter tenancy, squatter ownership in un-regularized settlement and tenancy in unauthorized subdivisions. On the stronger end are documented tenure rights namely squatter ownership in regularized settlement, ownership in an authorized subdivision, legal ownership with unauthorized construction, tenancy with legal contract, leasehold tenure and freehold tenure rights.

This article makes use of Payne’s typology to explore the benefits of stronger tenure rights. It aims to contribute to the existing literature by providing much needed empirical evidence of the impact of stronger tenure rights in a Southern African city. The article investigates whether stronger tenure rights improve household economies compared to weaker tenure rights. It examines the effect of leasehold tenure on property values, credit access, employment, household income and participation in income-generating activities relative to ‘occupancy rights’ afforded to residents in a regularized settlement. To make this examination, it contrasts the case of the transfer of low-cost public rental housing to sitting tenants in Matero Township, which took place in 1996 as part of privatization policies under Zambia’s Structural Adjustment Program (SAP); and the provision of occupancy rights granted to George Compound residents in 1979–1981 which took place under a World Bank-funded slum upgrading project in urban Lusaka. The following hypotheses are tested: Households with stronger property rights (in this case, leasehold rights) have significantly higher

1. property values,

2. access to credit,

3. employment,

4. participation in income-generating activities and

5. household per capita income

than households with weaker formal property rights (occupancy rights). To be considered statistically significant, an alpha level of 0.05 (95 per cent level of statistical significance) will be used.

2. Literature

The body of knowledge on the efficacy of property rights is unclear about many of the effects. Some empirical studies have shown that property rights improve property values (Lanjouw and Levy 2002) and labour supply in Peru (Field 2005, 2007; Field and Torero 2006) and household income in Brazil (Moura and Bueno 2009) but not in Argentina (Galiani and Schargrodsky 2010, p. 710). Studies show that titling does not improve access to credit except in Indonesia, and this is attributed to creditworthiness on the part of the individual (Field and Torero 2006; Durand-Lasserve and Selod 2009, p. 109; Galiani and Schargrodsky 2010, p. 710). Also, titling may allow for increased participation in income-generating activities which leads to increased household income (Prasanna 2007; Boudreaux 2008). The mechanism is two-pronged: through access to credit and through use of premises. Income-generating activities include renting out part of the house, commercial activities such as selling of grocery items and foods on a stand, services such as hair cutting or salon, tailoring or small/medium scale home-base enterprises (Prasanna 2007, p. 4). However, households in informal settlements have been found to participate in income-generating activities even without property title (Riaño 2001, p. 3). Nevertheless, the World Bank, member governments and influential scholars have backed the use of private property rights as a solution to urban slums for more than four decades (Abrams 1966; Turner and Fichter 1972; De Soto 1989, 2000; World Bank 1993; Buckley and Kalarickal 2006).

Studies on urban housing tenure in Zambia have been concerned with the evaluation of the World Bank-funded squatter upgrading and site and service project that took place from 1974 to 1983 (see Rakodi and Schlyter 1981; Hansen 1982; Chisanga 1986; Sanyal 1987; Rakodi 1988; Moser et al. 1997) and the privatization of publicly owned housing that began in 1996 (see Palmer 2000; Schlyter 2002, 2004; Basila 2005; Butcher and Oldfield 2009; Mususa 2010). A comprehensive evaluation of the 1996 Zambia Housing Policy is offered by Makasa (2010) but does not discuss the effects of stronger tenure rights and does not go into detail in evaluating upgrading projects. The study mostly focussed on effects of property rights is a qualitative study offered by Basila (2005) on Mufulira (Copperbelt) in which she found a slight improvement in economic status. She concluded however that housing did not lead to meaningful economic empowerment and did not provide a sustainable solution to economic insecurity. Mususa (2010) offers an ethnographic account of post-privatization experiences of housing in Luanshya, Copperbelt, and shows that the houses and yards that people gained were used for a wide array of informal income-generating activities which allowed families to just ‘get by’.

In respect of Matero, Schlyter (1998, 2002, 2004) has made several qualitative evaluations of life before, during and after privatization of low-cost housing in Matero and provides residents’ understanding and experiences of the mechanisms of privatization, with a focus on women. Butcher and Oldfield (2009) conducted a comparative investigation of privatization in Matero and Valhalla Park in Cape Town, South Africa, by interrogating the homeownership model. They show that ownership in the two case studies is gendered and is contested through everyday experiences.

This study builds on Schlyter’s (2004) and Butcher and Oldfield’s (2009) work on Matero. Schlyter’s (2004) evaluation came too early to measure in quantitative terms the effects of housing privatization. Schlyter’s fieldwork was done 2 years after privatization. Both Schlyter (2004) and Butcher and Oldfield (2009) are selectively focused on women’s experiences of privatization. Their methods are also qualitative with small samples. They also do not provide a control as part of their research design. This study builds on that knowledge by drawing on a survey of households in Matero and George and several in-depth qualitative interviews to evaluate the impact of granting residents leasehold rights.

3. Slum upgrading, SAP and tenure rights in Matero and George

The government of Zambia has always pursued policies to strengthen tenure security as part of ensuring better economic conditions and adequate shelter for its poor urban slum population. Shortly after independence, in the early 1970s, the government implemented slum-upgrading and site-and-service projects with a loan from the World Bank. George compound was an illegal settlement that was upgraded under this slum-upgrading program from 1979 to 1981. Fast-forward to the early 1990s, the Zambian government privatized its stock of publicly held property including existing low-income public housing stock under the auspices of the World Bank, and International Monetary Fund (IMF)-backed SAPs. These policies were a conduit for the transfer of property rights to private citizens.

3.1. Matero

Completed in 1959, Matero developed as a state public rental project for public service workers. It was developed by the British Colonial Administration as a result of the rise in demand for housing among African workers (Mulenga 2003, p. 6; Schlyter 2004). From 1948 to 1959, 5097 (Government of the Republic of Zambia 1996a) houses were constructed. They were occupied by municipal workers (Mulenga 2003, p. 7). At independence in 1964, the government of Zambia took over and placed the houses under the Lusaka Urban District Council (LUDC). It was employment-tied housing until the early 1990s when the policy was changed. Thirty years later in 1996, sitting tenants living in the now derelict low-income public housing in Matero were offered the right to buy the houses and land under 99-year leasehold tenure. According to the Ministry of Local Government and Housing’s Circular number 2 ‘Revised Procedures for Sale of Council Houses’, houses were sold at a 100 per cent discount because they were old and in very poor condition (Government of the Republic of Zambia 1996a). Sitting tenants only had to pay transaction fees which included a transfer fee of K10,500 (US $8) and a surveying fee of K60,000 (US $50) – A US dollar sold for 1310 Zambian Kwacha at the time. Those that paid the fees were to receive their title deeds within 30 days (Schlyter 2004, p. 6).

At the time of privatization, Matero was a poor old working class area where workers paid rent to the council. Many of the houses were dilapidated, had no electricity and had only an external water source and pit latrine.

3.2. George

In contrast, George began as a formal farm occupied by workers but gradually turned into an illegal settlement after independence. The government earmarked George for demolition. However, the state decided to upgrade it for fear of political unrest and loss of political support in the area (Rakodi and Schlyter 1981). The land on which the houses were built was placed under the LUDC. From 1978 to 1981, the state upgraded the settlement through the provision of services such as communal water supply, grading of gravel roads and a clinic. The council also allocated plot numbers. In 1981, the Council issued George residents with occupancy licences valid for 30 years signifying administrative recognition of occupancy. It also began to charge a land occupancy tax which it now calls ground rent.

At the time Matero residents were offered houses, George residents were not offered the public land on which their houses were built. Occupancy rights gave residents of George tenure security and access to services with little control over the use of property. Following the transfer of leasehold ownership, residents in Matero are allowed to sell their land while, under the terms of the occupancy licence, those in George are not allowed to do so because the land belongs to the council (Lusaka City Council 2010). Moreover, George residents are not allowed to rent out their houses or conduct any business on their premises. Transactions however do take place in George despite the restrictions because of weak enforcement of laws.

In sum, the variation in tenure rights in Matero and George presents a research opportunity to understand the effects of stronger tenure rights compared to weaker tenure rights using Payne’s continuum of tenure security typology.

4. Methods and data

To answer the research question, the article uses a mixed methods approach. Methods include use of quantitative data from a household survey – regression analysis and propensity score matching techniques – as well as qualitative data from in-depth semi-structured interviews.

4.1. Household survey

A survey and qualitative interviews were carried out as part of the study because no existing data sets were readily available in both neighbourhoods. The survey was conducted from July to August of 2011 –15 years after the privatization of houses in Matero. Data collection was done using a structured questionnaire that was administered by seven enumerators. Enumerators were trained over the course of a week. A pilot study was first implemented. During the pilot, each enumerator conducted one interview as a final part of training. The questionnaire was reviewed using results of the pilot study. In the study, we interviewed heads of households or their competent proxies. In Matero, we interviewed households in which the head was a direct beneficiary, and who held a title deed or deed of sale.

4.2. Sampling

A sample of 350 households was targeted, out of 5097 households in Matero (Government of the Republic of Zambia 1996a) and 350 holders of occupancy licences, out of approximately 20,000 households in George. Csaja and Blair’s (1996, pp. 126–133) formula was used to arrive at the sample size. A Google Earth map of the neighbourhoods was used to create a sampling frame. Households were selected using systematic interval sampling with a random start. In George compound, the initial plan was to draw a matched sample using a database from the Zambian census of population as a sampling frame – the database is available at the Central Statistical Office and is not publicly available. The study was not permitted access to this database. Notwithstanding, the study proceeded with selecting an unmatched sample of 350 households from George, with the intention of including analysis that would mathematically match respondents using propensity score matching techniques.

The total realized sample size came to 623 households (89 per cent response rate). A total of 312 completed interviews were obtained in Matero and 311 in George. However, 50 survey participants were offered to purchase houses by the council after original owners had failed to complete payment. These were excluded from the analysis. In George, 75 participants had obtained their occupancy titles and entered into housing after 1997. These were also excluded from the analysis because they came after treatment had occurred. The final data set had 498 individuals with 262 participants from Matero and 236 from George. The sample remains representative, but with less precision (wider confidence intervals) than the original calculated sample size. With this sample size, significance for computations assuming normality is at less than 0.07.

Variables used in this paper are summarized in Table 1.

Table 1. Variable descriptions.

Variable	Description
Independent and control variables
Leasehold title	Tenure type, 1 = leasehold title, and 0 = occupancy licence
Age	Age of household head
Gender	Gender of household head. 1 = male and 0 = female
Education	Number of years of education completed by head of household
Employed members of household	Are there other members of this household who have full-time paid jobs? 1 = yes and 0 = no
Occupational skill level	International Standard Classification of Occupations (ILO 2012) skill level 1 (simple, routine, physical, and manual jobs) = 0, skill level 2 (operating machinery and electronic equipment) = 1, skill level 3 (complex technical and practical tasks) = 2, and skill level 4 (complex problem solving, decision making, and creativity based on established theoretical and empirical knowledge) = 3
Father’s education	Number of years of education of father of head of household
Household size	Number of household members residing in the dwelling
Rooms	Number of rooms in the dwelling
Duration	Number of years respondent has lived in the house
Dependent variables
Square root of property value	Two questions. 1. What is the market value of the dwelling (in Zambian Kwacha) according to the last valuation? 2. Please provide an estimate if your house has not been valuated.
Credit access	Have you used this house in any way (e.g. as collateral) to obtain loans or credit? 1 = yes and 0 = no
Employment status	Employment status of head of household, 1 = employed and 0 = not employed
Income-generating activities	Has the house been used for any income-generating activities? 1 = yes and 0 = no
Logged per capita income	What is the total monthly income for this household?

Table 1 is largely self-explanatory but a bit more detail is warranted on the property value variable. All respondents were asked to provide their property value based on the most recent government valuation. A separate question asked those who had not yet valued their houses to make an estimate based on the amount their neighbours had sold their houses for. This question was assumed to arrive at an estimate reflecting the government value because houses were uniform with very similar plot sizes. Valuations had taken place between 1997 and 2011 as part of the process of housing purchase. Property values based on government valuation were provided by 149 (57 per cent) of the respondents from Matero. Government valuations were used rather than the going market rate because they are the most recent official and objective estimate of property value in Matero. Thus, limiting self-reported values was a way of limiting the influence of social desirability bias – over-reporting value – and recall bias – inaccuracy of memory. Nevertheless, property prices have risen considerably since 1997 in Lusaka. This suggests that the difference between Matero and George for the 43 per cent where the going market rate is used may be more significant. Government valuations are not available for George because of the regulation that houses cannot be sold. Participants from George were asked to provide a value for their house using the going rate of houses on the active informal housing market in the neighbourhood (several handwritten posts of ‘House for Sale’ were posted in George). Participants were aware of the prices as many had neighbours who sold their houses.

4.3. Analytical strategy

Ordinary least squares and logistic regression analyses are employed to estimate the effects of leasehold tenure rights on property value, credit access, employment, income-generating activities and household income. The equation to make the regression estimates is as follows:

$Y_i=\alpha +\gamma \mathrm{L}\mathrm{e}\mathrm{a}\mathrm{s}\mathrm{e}\mathrm{h}\mathrm{o}\mathrm{l}\mathrm{d}\mathrm{r}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t}{\mathrm{s}}_i+\beta X_i+{\epsilon }_i.$

Y_i is any of the above-mentioned outcome measures for observation $i$. $\gamma$ is the coefficient which provides an indication of the size of effect of the ‘leasehold rights’ measure. In the equation, ‘leasehold rights’ is a dummy variable that equals 1 for individuals with leasehold rights and 0 for those with occupancy rights. X is any of the covariates controlled for mainly but not limited to background characteristics. The background characteristics include age, gender, number of years of education, father’s number of years of education, mother’s number of years of education and marital status. Finally, ${\epsilon }_i$ is the error term which represents the amount of variation in the outcome variables that is unexplained by the regression model.

4.4. Propensity score analysis

Observational studies such as this, unlike randomized control studies, do not eliminate selection bias; hence, I employ propensity score analysis to eliminate selection bias and achieve comparability of the groups (Rosenbaum and Rubin 1984). In other words, participants from Matero differ from participants in George for reasons other than their tenure status which may determine differences in outcomes. The propensity score balances scores between the two groups on observed characteristics, thereby eliminating selection bias and making the two groups comparable on baseline characteristics. This means that conditional on their propensity scores, participants from both neighbourhoods will have the same distribution of measured or observed baseline covariates.

4.4.1. Estimating the propensity scores

An initial propensity score model was estimated using 107 variables. To estimate the propensity score, a probit regression model was used in which treatment status (possession of leasehold rights vs. possession of occupancy rights) was regressed on baseline characteristics (Rosenbaum and Rubin 1984). Rosenbaum and Rubin (1984) and Austin, Grootendorst and Anderson (2007) recommend that models estimating propensity scores include those variables that affect the outcome or that affect both treatment selection and the outcome. The variables used are plausible predictors of property value, credit access and employment, participation in income-generating activities and household income in households with stronger tenure rights. Because of the need to ensure balance on variables that are predictive of these outcomes, the variables were included in the propensity score model. Additionally, demographic and many other attitudinal variables were included in the model to ensure that many observed characteristics were included as necessary.

4.4.2. Matching on propensity scores

Treated and untreated participants were matched based on similar propensity scores. There were more participants (262) who received leasehold title than there were participants on occupancy title, i.e. more treated than untreated participants. Inferences are therefore made only in the region of common support – the space where the distributions of the propensity scores of the two groups overlap. The objective was to match a treated participant to each participant who did not receive leasehold rights. Participants were matched on the propensity score using nearest neighbour, radius with callipers of width equal to 0.1, kernel and stratification matching. No specific matching technique is superior to another in estimation of treatment outcomes.

4.4.3. Stratification and balance on propensity scores

Quintiles of the estimated propensity scores were computed. The region of common support ranged from 0.08 to 1. Participants in the overall study sample were stratified into five approximately equal-size groups using the quintiles of the estimated propensity score. The distribution of scores across groups is as shown in Figure 1 and the results of the tests of equality of means in propensity scores within each quintile block – shown in Table 2 – are not significant showing that the means are equal.

Table 2. T-tests of equality of the means of propensity scores between treated and untreated groups in each block.

Block of propensity score and minimum	Untreated (n = 129)	Treated (n = 227)	t
Block 1 – 0.08	0.14	0.14	−0.57
Block 2 – 0.2	0.29	0.31	−1.37
Block 3 – 0.4	0.50	0.50	−0.24
Block 4 – 0.6	0.70	0.71	−0.94
Block 5 – 0.8	0.90	0.94	−1.27

Figure 1. Balance of propensity scores across treated (leasehold tenure rights) and untreated (occupancy tenure rights) groups.

Figure 1 and Table 2 show that the extent of overlap is satisfactory and balance is achieved across the two samples within each quintile.

4.4.4. Estimating treatment effects

As noted earlier, the article examines the effect of stronger tenure rights on property value, credit access, employment, participation in income-generating activities and household per capita income by comparing leasehold rights in Matero and occupancy licences in George. Propensity score matching allows one to estimate the average treatment effect for the treated (ATT) that is the average response to treatment for individuals that were assigned treatment – leasehold rights. The ATT are estimated for matched samples derived using nearest neighbour, radius, kernel and stratification matching techniques. The models do not control for unobservable or unknown factors that may be driving observed variation in the outcomes. The assumption of the propensity score models in using observed characteristics is that there are no unobserved factors correlated with the outcome.

4.5. Qualitative data: in-depth semi-structured interviews

In-depth semi-structured interviews complement findings of the quantitative analysis. In-depth semi-structured interviews were conducted with altogether, 35 households – 25 in Matero and 10 in George. Households were selected out of the pool of the survey sample. A purposive sampling technique based on geographical clustering was the selection method. The most recent (2009) Google Earth images of the terrain were used to identify and select clusters and households. First, Matero and George were considered two large clusters. Then, Matero was divided into sub-clusters made up of households. The aim was for the Matero sample to be as geographically representative as possible. The boundaries of the clusters were defined by road networks. Then, households were selected in such a way as to represent concentrations of households within the cluster. One to three households were selected from each cluster depending on its size and density. George interviews were conducted for control purposes. Geographical spread and long-term residency were the main factors in purposively selecting George households.

The method of data collection used was an interview schedule. It was loosely structured in order to allow interviewees to be as expressive as possible. The topics on the guide included questions about the individuals and the dwelling, economic benefits, social benefits and benefits under human capital. The idea behind structuring the questions this way was to avoid biasing the interviewee. This ensured that interviewees provided what they thought was the benefit rather than being asked specifically whether a pre-known theme such as credit access for example was something the interviewee thought was a benefit. However, interviewees were probed to give more details. For example, ‘What businesses do you conduct from this house?’ and ‘Have you ever got a loan for business or for other purposes?’

Each interview was recorded, with the permission of the interviewee, to facilitate analysis. Audio recordings of the interviews were stored on a computer with password encryption. The author transcribed and analysed the interviews. Eleven interviews were conducted in English, seven in Bemba and seventeen in Chinyanja. The author translated the interviews conducted in Bemba and Chinyanja into English during the process of transcribing. The author is a native speaker of Bemba and Chinyanja and received instruction in English from first grade to graduate school.

Thematic analysis was the method of analysis used following transcription of all the interviews.

5. Results

Table 3 presents the characteristics of respondents in Matero and George and compares their mean differences.

Table 3. Summary statistics of Matero versus George household heads in the unmatched sample.

Variable	George, n = 236	Matero, n = 262	t
Property value	35,872,074	70,623,686	−8.30***
Credit access	0.01	0.03	−1.11
Employment status	0.67	0.71	−0.84
Income-generating activities	0.65	0.58	−1.79**
Log of income per capita	11.29	11.67	−4.21***
Age	54	55	−0.89
Male	0.64	0.52	2.80***
Education	7	9	−4.80***
Occupational skill level	1.53	1.81	−2.15***
Marital status	0.56	0.52	0.92
Household size	6.19	6.34	−0.65
Father’s education	6.76	6.95	−0.31
Duration	30.58	27.48	2.75***

*p < 0.10; **p < 0.05; ***p < 0.01.

There are no demographic differences in terms of age, marital status, father’s education, and household size. Respondents in the Matero sample are likely to be more educated with an average of 9 years than George residents who have an average of 7 years (t(484) = −4.80, p = 0.00). It also however means that the typical respondent in both neighbourhoods dropped out of school before completion of High School. The proportion of male heads of household in the Matero sample is lower (0.52) compared to that of George (0.64) (t(495) = 2.80, p = 0.00). This difference may be attributed to the original demographics in George where at the time of settling, only men settled there as unmarried women were prohibited to live in cities at the time. Matero had a section called the married quarters where married workers were allowed to live with their families. Many of the families have remained in their original plots. In fact, respondents in Matero have a slightly lower length of housing tenure compared to George. On average, respondents in Matero have lived there for 28 years compared with 31 years for George respondents (t(479) = 2.74, p = 0.00). Matero residents have slightly less duration because of turnover from their employment-tied housing. Until 1991, sitting tenants had to leave housing when their employment ended. The new tenants who moved in had shorter duration by the time of privatization.

As regards the outcome variables, respondents in Matero report higher property values compared to those from George. Matero participants report that the average value of their houses is K70.6 million (US$13,600) while those in George report an average of K35 million (US$6700) – the dollar sold for K5200 in 2011. This difference is statistically significant at the 1 per cent level with t(275) = −8.30, p = 0.00. Despite their geographical proximity, houses in Matero report a higher property value.

In terms of credit access, there is an insignificant minority of respondents in both neighbourhoods who have used their house as collateral to obtain a loan. Of the Matero sample, 3 per cent are able to obtain loans compared with 1 per cent for respondents from George. This difference is not statistically significant with t(496) = −1.11, p = 0.27.

Matero participants are not different from George residents in terms of employment. Matero respondents have a higher average proportion of employed household heads at 71 per cent while George respondents had 67 per cent. The difference is however not statistically significant (t(376) = 0.84, p = 0.40).

With respect to participation in income-generating activities, Matero respondents have a significantly lower proportion compared to those in George. Matero respondents have a score of 0.58 compared with 0.65 for George residents. The difference is statistically significant at the 10 per cent level t(491) = 1.79, p = 0.07. This means that Matero respondents are less likely to run businesses compared to George respondents.

Matero respondents report significantly higher household per capita income than those in George. The logged per capita income score of Matero respondents is 11.67 compared with 11.30 for George respondents with t(331) = −4.21, p = 0.00. As expected, the variable for household income has many missing observations which are a challenge in many surveys.

Overall, the results show preliminary support for the hypotheses that stronger tenure rights are associated with higher property values for dwellings and household income per capita. Counter-intuitively, there is no support for the hypothesis that leasehold rights are associated with higher participation in income-generating activities, but that they are associated with less. Nevertheless, based on t-tests, we cannot ascertain the strength of these relationships. We cannot also ascertain the magnitude of association in these relationships. Correlation and regression analysis are employed to explore these hypotheses further and to generate a sense of the likely magnitude of the differences and strengths of relationships of association.

Table 4 reports pairwise correlations between tenure rights and each of the dependent variables. The correlation coefficient is a measure of the strength and direction of the relationship. Results are consistent with the t-tests. Leasehold rights are positively and significantly correlated with property value (r = 0.44 [p = 0.00]) and household per capita income (r = 0.23 [p = 0.00]). Leasehold rights are negatively correlated with participation in income-generating activities in general (r = 0.08 [p = 0.08]). This is a very weak relationship which means that tenure rights explain a very small part of the variation in participation in income-generating activities. Nevertheless, the finding means that stronger rights are associated with less participation in income-generating activities. There is no correlation between leasehold rights and access to credit and employment status.

Table 4. Pearson’s pairwise correlations between each outcome variable and leasehold rights.

	Leasehold title	Property value	Credit access	Employed	Income-generating activities	Income per capita
Leasehold title	1
Property value	0.44***	1
Credit access	0.05	0.04	1
Employed	0.04	−0.04	0.06	1
Income-generating activities	−0.08*	0.11**	0.01	0.04	1
Income per capita	0.23***	0.20***	−0.08	0.05	0.02	1

*p < 0.1; **p < 0.05; ***p < 0.01.

Considering the size of the correlation coefficients, the strengths of the relationships are generally weak. They are nevertheless significant. Correlation analysis however does not provide us with an indication of the magnitude of the change in the outcome variables that can be associated with leasehold rights. I employ regression analyses to provide such estimations. I proceed to explore only the hypotheses that I have found to be correlated with leasehold rights.

5.1. Property value

The models in Table 5 show that leasehold rights are associated with higher property value. The bivariate regression in Model A reports property values 2414.45 (about K5.83 million) higher than those for occupancy rights. When demographic characteristics are included in model B, the regression coefficient increases to 2554.51 (K6.5 million). The explained variation increases to 32 per cent (R² = 0.32). When I include employment status, secondary employment and per capita income variables, the coefficient shows property values higher by 2834.31 (K8 million). The explained variation rises to 48 per cent (R² = 0.48) in the final model. In all the models, the regression coefficient is statistically significant at the 1 per cent level indicating that this is not a chance occurrence. The finding remains robust with various controls; the leasehold rights variable does not lose significance when demographic and other variables are included in the models.

Table 5. OLS regression of property value, income-generating activities and household income.

	Property value			Income-generating activities			Household income
	Model A	Model B	Model C	Model A	Model B	Model C	Model A	Model B	Model C
Leasehold	2414.45***	2554.51***	2834.31***	−0.33*	−0.18	0.21	0.38***	0.29***	0.29***
	(289.78)	(333.24)	(406.73)	(0.19)	(0.25)	(0.34)	(0.09)	(0.09)	(0.10)
Age of Household Head		14.20	43.08**		−0.00	−0.00		−0.01*	−0.00
		(13.27)	(16.94)		(0.01)	(0.01)		(0.00)	(0.00)
Male		−34.30	675.92		−0.16	−0.15		−0.16	−0.15
		(458.50)	(435.91)		(0.11)	(0.11)		(0.11)	(0.11)
Education		96.61**	−16.43		0.03	−0.00		0.06***	0.05***
		(48.56)	(55.53)		(0.03)	(0.05)		(0.01)	(0.01)
Married		234.57			0.08	−0.03		−0.15	−0.06
		(413.24)			(0.26)	(0.33)		(0.11)	(0.12)
Household size		5.65	71.68		−0.01	−0.08		−0.12***	−0.13***
		(67.04)	(87.80)		(0.05)	(0.05)		(0.02)	(0.02)
Rooms		1231.71***	757.67*			0.49***			0.06***
		(336.84)	(439.17)			(0.09)			(0.02)
Household durables						−0.23
						(0.26)
Employed			−945.71**			−0.44			−0.18
			(436.44)			(0.42)			(0.13)
Employed members of household			913.33**						0.06
			(454.34)						(0.12)
Log household Income per capita			941.32***
			(315.67)
Occupational skill level 2		−270.53	−691.33		−0.19	−0.51		−0.00	0.19
		(544.20)	(617.56)		(0.34)	(0.46)		(0.12)	(0.12)
Occupational skill level 3		643.09	2097.03*		−0.43	−1.22		−0.07	0.13
		(1043.25)	(1236.81)		(0.64)	(0.84)		(0.27)	(0.33)
Occupational skill level 4		33.88	892.06*		0.50	0.02		−0.15	−0.06
		(463.04)	(513.78)		(0.34)	(0.45)		(0.11)	(0.12)
Constant	5989.33***	1420.39	−9607.34**	0.64***	0.62	−0.18	11.30***	12.04***	11.75***
	(223.24)	(1030.19)	(3910.31)	(0.14)	(0.71)	(1.04)	(0.07)	(0.25)	(0.33)
Observations	277	193	123	493	327	254	333	239	197
R-squared	0.20	0.32	0.48	0.005	0.02	0.17	0.05	0.35	0.42

Robust standard errors in parentheses.

***p < 0.01; **p < 0.05; *p < 0.1.

By interpretation, there is a significant difference in property values between Matero and George households. Where we find stronger rights, we find higher property values. Within the confines of an observational study, we can only say that this may indicate an effect of stronger tenure rights. The evidence is at least exploratory and at best indicative of stronger rights being responsible for higher property values. Section 5.4 shows models controlling for selection bias using propensity score analysis to reduce the impact of these limitations.

5.2. Income-generating activities

Table 5 reports that households with stronger tenure rights are associated with a lower probability of running income-generating activities but this effect disappears when the variable number of rooms is introduced into the equation. This suggests that the observed negative effect of leasehold rights could be explained by the number of rooms, with those with leasehold rights having fewer rooms than those with occupancy rights. In Model A, respondents with leasehold rights are found to have significantly lower odds relative to George respondents by 0.33 but the explained variation is 0. In Model B, the coefficient is not significant and the odds are lower at 0.18 when demographic variables are included. The explained variation rises to 2 per cent. In the final model, leasehold rights are also not statistically significant once the number of rooms variable is included in the equation. The explained variation rises to 17 per cent. The evidence suggests that most of the income-generating activities are rental based. The evidence also suggests that households in neighbourhoods with weak tenure rights are involved in informal economic activities. However, the models do not control for baseline heterogeneity and selection bias. Section 5.4 will also use propensity score analysis to reduce the impact of this limitation.

5.3. Household income

Does the evidence indicate that leasehold rights are positively associated with household per capita income? Yes. Table 5 reports that household incomes are between 29 and 38 per cent higher among Matero respondents compared to George respondents. The bivariate regression in Model A reveals that leasehold rights are associated with a coefficient of β = 0.38. This means that per capita income among Matero respondents is 38 per cent higher relative to those in George. The explained variation is 5 per cent (R² = 0.05). When demographic variables and particularly education are added to the regression equation, the coefficient reduces to β = 0.29 (29 per cent higher). The explained variation increases to 35 per cent which is a good model fit. In the final model, respondents in Matero have per capita incomes 29 per cent higher (β = 0.29). The model fit is good with R² = 0.42.

Although the regression models show that leasehold rights are associated with higher household per capita income, we have already seen that leasehold rights have no effect on the mechanisms through which income should rise. According to the theory, property rights increase access to credit, and employment which should lead to higher income. The regression models may indicate an effect of selection bias. One may argue that Schlyter (2004, p. 7) documented the practice of raising finances for houses in Matero through the extended family in order to make extensions. Such contributions are too small to make an impact. The propensity score model in the next section also addresses selection bias in the estimate of the effect of leasehold rights on household income.

5.4. Results of propensity score analysis

As shown in Table A1, participants from Matero and George are identical in their observed characteristics within each block in almost all of the cases. Only 1 (age in block 4) out of 35 comparisons are significant at the 5 per cent level of statistical significance. Four comparisons are significant at the 10 per cent level which is not an acceptable significant level for the sample size (Israel 1992, p. 5). The quality of the balances is good.

5.5. Treatment effects

The estimated treatment effects of leasehold rights on property value, credit access, employment status, participation in income-generating activities and household income using propensity score matching reveal that leasehold rights are associated with higher property values and less participation in income-generating activities. My computations using propensity score matching methods revealed that property values in Matero are K64 million (US$12,300) and K36 million (US$6900) in George, a 44 per cent difference. When household heads are matched, property values are higher in Matero by about K3 million ($650) to K8 million ($905) on average and that there are about 22 per cent fewer households engaged in income-generating activities than in George.

As Table 6 shows, matching using nearest neighbour techniques yields an ATT of 1962.31 in property values. This means that participants from Matero report an average of K3.85 million (US $740) more in property values than those in George. The radius matching technique yields an ATT of 2169.67, which translates to K4.7 million (US $905). The Kernel matching method yields a similar effect as the nearest neighbour matching technique. Using stratification matching, we see an ATT of 1836.59, which translates to K3.37 million (US $650). Each of these is significant at the 1 per cent level. Although the estimated treatment effects on property values from propensity score matching are lower than those from regression analysis, they both nevertheless support the hypothesis that households with stronger rights have houses with higher property value. The percentage difference is 44 per cent. Privatization of public rental housing and titling opened houses of study participants in Matero to the private market and raised property values.

Table 6. Average treatment effect on the treated (ATT) using nearest neighbour, radius, kernel and stratification matching.

	Matching method
Outcome variable	Nearest neighbour	Radius	Kernel	Stratification
Property value	1962.31*** (944.63)	2169.67*** (575.27)	1961.51*** (861.01)	1836.59*** (901.99)
Credit access	0.02 (0.01)	0.01 (0.02)	0.02 (0.01)	0.02 (0.01)
Employment status	0.03 (0.15)	0.08 (0.10)	0.08 (0.15)	0.08 (0.15)
Income-generating activities	−0.23*** (0.06)	−0.15** (0.09)	−0.24*** (0.04)	−0.27*** (0.04)
Log of household income	0.20 (0.15)	0.15 (0.20)	0.15 (0.16)	0.17 (0.17)

Standard errors in parentheses, ***p < 0.01; **p < 0.05; *p < 0.1, models control for all demographic and behavioural characteristics listed in Appendix.

Table 6 also shows that stronger tenure rights are significantly associated with less participation in income-generating activities. Stronger property rights have an ATT of −0.23 using nearest neighbour, −0.15 using radius, −0.24 using kernel and −0.27 using stratification matching techniques. These are significant at the 1 per cent level indicating that households with stronger tenure rights engage in significantly less participation in income-generating activities of about 15–27 per cent.

The effect on household per capita income is no longer significant as it was in the regression model suggesting that the regression estimate was affected by selection bias.

5.6. Results of in-depth qualitative interviews

Additional to the value-raising property of property title on the market, semi-structured interviews with study participants from Matero revealed that the right to make housing improvements after obtaining title was one of the most important benefits. It likely in part explains the higher values. Original houses came with two rooms – a bedroom and a sitting room – and a kitchen, bathroom and a dug out/PIT toilet outside. Study participants underscored that leasehold rights came with the right to make modifications such as extensions and to build more houses within their large yards.

Extensions were important primarily because original houses were too small to accommodate the typically large families. The average household size is eight people. Here is a conversation with Chongo who had 14 people living in his house.

Chongo:

… Like this here, I have extended [the dwelling]. We were in a two-roomed house with the number of children over ten, how can they sleep? So I extended for children to sleep well.

Author:

So how many are you in the house

Chongo:

roughly about 14

Thus, making extensions to the original house to create room for their families resulted in an unintended consequence of generating higher property values.

George residents engage in small businesses and extend houses wherever they have space to do so despite the fact that they do not possess the legal right to do so. Extension is done without the consent of the council as it restricts residents to use only the land upon which the house is built. I found Bwalya working on extending his house and sat down for an interview with him. His plan was to add a house to rent out in addition to the tenants he already has in another room he had extended. I asked him whether he and his neighbours get approval from the council;

Bwalya:

Normally we don’t, normally people here don’t get approval from the council because those building we just, they’ll [residents] do the planning on their own, they will do everything on their own so it means there is no need for someone to approach the council and seek their approval before anything can be done, yes!

Bwalya reveals that the regulation from the council is openly disregarded but enforcement from the council is also very weak. Almost all my interviewees had not had direct communication from the council for 30 years since they first obtained their occupancy title. The time I was conducting interviews, the 30-year occupancy licences were expiring and up for renewal. Many of the residents were surprised to learn that they had unpaid bills in ground rent. Their ground rent had accumulated to unsustainable levels leading to growing fears about eviction.

6. Discussion

As the wider body of knowledge has shown, knowledge on the effects of property rights on urban poverty is unclear. The evidence from Matero and George tends to corroborate much of what we know in terms of a positive association with property value but highlights the uncertainty on access to credit, employment and household income. Moreover, the apparent negative effect on income-generating activities highlights the relevance of other factors apart from tenure rights in explaining the propensity of residents in low-income settlements to engage in home-based income-generating activities.

It is clear that tenure rights have been found to substantially increase property values. The findings from Matero and George suggest a 44 percentage increase. The percentage difference could be significantly higher given that the study has used property valuations for the period 2009–2011. Moreover, in recent years, land and housing prices have rapidly appreciated in Zambia and in Lusaka in particular. Elsewhere, Dowall and Leaf in their work with land brokers in Jakarta, Indonesia, found an increase of 73 per cent (Dowall and Leaf 1991), and Alston, Libecap and Schneider reported a 100 per cent increase in Brazil (Alston et al. 1996). Jimenez in Davao, Philippines, found a 58 per cent increase in value (Jimenez 1984). Several studies estimate an increase of around 25 per cent, for example Lanjouw and Levy (2002) in Ecuador, Cantuarias and Delgado (2004) in Peru, Dowall (1998) in Indonesia and Friedman et al. (1988) in Manila, Philippines. It needs to be acknowledged that in addition to stronger tenure rights, the differences in property values between George and Matero can be partly explained by their different histories and status within the city. When Matero was built by the colonial government in the 1950s, it was purposely built to house government employees and private sector employees (Schlyter 2004, p. 1). Matero maintained its status as a formal area, while George began as an informal area earmarked for demolition and was then formalized.

Nevertheless, increase in property values also comes with unintended consequences. Durand-Lasserve (2006) has shown that titling and the ensuing increase in property value can lead to reduced tenure security since market-driven displacements are common following titling. Indeed following the transfer of housing, Schlyter (2002, p. 12, 2004) showed that Matero experienced rapid gentrification with relatively higher income families buying up properties and extending them for their families. Petty landlords bought up properties to turn them into extended rental units. ‘Sitting tenants moved out and sold or let their houses to prosperous families’ (Schlyter 2004, p. 12).

The findings also underscore the fact that tenure security matters more for a household’s participation in income-generating activities than that a household has stronger tenure rights. This study supports the findings documented in the literature that weak tenure rights can be sufficient to provide tenure security and support an increase in informal income-generating activities. Research from other parts of Africa supports the view that households in informal settlements engage in income-generating activities in the informal sector despite the fact that regulations do not permit such activities (Prasanna 2007). Prasanna (2007), who studied the effects of tenure security on income-generating activities in informal settlements in Dar es Salaam, concluded that upgrading of land tenure rights alone cannot address the needs of the urban poor; it must be accompanied by the provision of basic infrastructure facilities that support livelihood activities. Also, in the Barrios of Quito, Ecuador, Riaño (2001, p. 3) found that the main use of land in informal settlements is not only supply of shelter but also as centres of economic production such as brick-making, agriculture and housing production.

Nevertheless, for the specific cases of Matero and George, an additional explanation for the increase in income-generating activities can be found in the different effects the decline in formal sector jobs in urban areas had on the neighbourhoods following SAP. In general, there was a rise in the number of urban dwellers relying on the informal sector for livelihoods but Matero and George were affected differently. Matero residents were less affected because a greater number of people were clerks and cleaners in central or municipal government employment – since access to housing was by government or private sector employment. Those in government employment had sufficient security and income so as to reduce the need to engage in informal income-generating activities. Although earlier studies of George Schlyter and Schlyter (1979) and Schlyter (1998, p. 261) showed that George was a working class area whose composition of the inhabitants was all employed working full time, her later study (Schlyter 1999, preface) documents that the 1990s was laden with economic hardship for George inhabitants due to the negative effects of SAP. Many inhabitants of George depended on jobs in the nearby industrial area, which collapsed during SAP. There were likely more individuals in Matero with stable employment than in George following macroeconomic reforms.

From a macro perspective, the presence of a poorly performing macroeconomic environment such as Zambia’s following SAP – high unemployment, low incomes and high prevalence of poverty – stronger rights makes little impact on household economies. Further research is required as to what the optimal economic conditions are for stronger rights to have a positive effect on employment and income either through increases in home businesses and/or tenure security.

This paper does not find evidence supporting the hypotheses that leasehold rights are associated with access to credit, employment and higher household income thereby highlighting the uncertainty in the body of knowledge. Elsewhere, the findings are ambiguous with a leaning towards no effect of stronger rights on credit access (Field and Torero 2006; Boudreaux 2008; Galiani and Schargrodsky 2010; Lemanski 2011). Some studies find positive effects of titling on household employment and hours allocated to activities outside the home mainly through the mechanism of tenure security (Field 2005, 2007; Field and Torero 2006) while others find no significant effects (Rose 2006; Galiani and Schargrodsky 2010). In Senegal, the impact of titling on the economy of families is ‘limited and barely measurable’ which means that leasehold rights barely have an effect on labour market outcomes (Payne et al. 2009).

Notwithstanding, the article demonstrates that a view of property rights as existing on a continuum has important implications for our understanding of how property rights impact households and neighbourhoods of the urban poor in developing countries. The provision of weak tenure rights such as occupancy rights may support informal livelihood activities for poor households, in addition to providing much needed tenure security. Residents engage in these activities as formal rules are not enforced. With Payne’s (2001) perspective, we see that stronger tenure rights attract higher property values, which may mean higher returns from rents even with less participation in income-generating activities. Providing stronger rights would not only raise property values but help stimulate the land and housing market and ultimately benefit the city’s economy. Therefore, a view of tenure security as existing on a continuum from weak to strong rights has much value in how we understand effects of property rights on households of the urban poor.

7. Conclusion

The massive investments by developing countries and international financial institutions over the course of about 50 years have not been in vain, at least in Matero where low-income public rental houses – most of which were in poor condition – were transferred to sitting tenants. Stronger tenure rights are associated with higher property values. Although stronger tenure rights do not necessarily result in greater income-generating activities, tenure security is what matters in the first place for households to participate in income-generating activities. Poorly performing macroeconomic factors – such as high unemployment – may drive household economies so strongly that stronger rights make little impact on employment status, access to credit and household income. Nevertheless, using Payne’s (2001) perspective, the article has demonstrated that a view of property rights as existing on a continuum has important implications for our understanding of how property rights impact households and neighbourhoods of the urban poor in developing countries. We see that the provision of weak tenure rights is sufficient to support the development of informal income-generating activities. Additionally, the article has shown what others have shown before that understanding on the effects of stronger property rights on urban poverty is equivocal. The findings from this study indicate that stronger property rights have a positive effect on property value, but no positive effect on the level of income-generating activities, access to credit, employment and household income.

Acknowledgments

The author wishes to acknowledge the contribution of Professor Jeremy Seekings, Professor of Politics and Sociology at University of Cape Town, Professor Cheryl Doss at Yale University and Professor Cynthia Horan also at Yale University who all provided valuable comments to earlier version of this article. Thanks also go to all the participants interviewed during data collection.

Disclosure statement

No potential conflict of interest was reported by the author.

Additional information

Funding

This work was supported by the Center for Social Science Research, University of Cape Town.

Notes on contributors

Singumbe Muyeba

Singumbe Muyeba is a Visiting Lecturer at Bridgewater State University and Director of Research at the Massachusetts Housing and Shelter Alliance in Boston. His doctoral research at University of Cape Town was on the effects of property rights among the poor in Cape Town, South Africa and Lusaka, Zambia. He is a 2011/2012 Fox International Fellow at Yale University and was NRF Innovation Postdoctoral Fellow from 2013 to 2014. His research interests are in the international political economy of development with a focus on property rights, housing and poverty.

Appendix

Table A1. t-Statistic for outcome variables and demographic characteristics of Matero versus George residents within estimated propensity score blocks.

Variable	Block 1	Block 2	Block 3	Block 4	Block 5
Outcome variables
Property value	1.09	2.31**	1.51	−3.39***	−1.23
Credit access	0.57	0	0.90	0.73	0.47
Employment status	1.10	1.76**	1.54*	0.15	0.95
Income-generating activities	0.05	0.47	0.11	0.02	−2.17**
Logged per capita income	0.91	0.74	1.50	0.39	0.47
Demographic variables
Age	0.27	−1.06	−0.98	2.55**	−1.66
Male	−0.11	0.49	0.28	−0.45	0.68
Education	0.62	1.65	−0.40	−1.93*	1.32
Occupational skill level	−1.11	−0.16	0.50	0.36	−0.69
Married	0.86	1.64	0.30	−0.59	1.99*
Household size	1.33	−0.95

*p < 0.10; **p < 0.05; ***p < 0.01.