Double-edged sword: Segmentation within the South African informal sector. Nwabisa Makaluza

Transcription

1 Double-edged sword: Segmentation within the South African informal sector Nwabisa Makaluza

2 Introduction The term informal sector originates from the work of Hart (1973) in his description of the economic activities of low income earners in Ghana. The informal sector, as he saw it, had heterogeneous economic activities of primary, secondary, and tertiary enterprises. He emphasised the importance of the sector because it provides essential services on which life in the city is dependent (Hart, 1973:68). Recognising its importance, Weeks (1975) proposed three major advantages to a developing country having a growing informal sector. Firstly the sector produces goods that are consumed by the lower income earners in the population. Second, workshops may provide the source for a local capital goods industry and provide training to local entrepreneurs. This is supported by Brand s (1986) study of informal workers where she learnt that the skills used in informal activities were self-taught by observing other workers. Thirdly, a growth in the informal sector (relative to the formal sector) would result in a shift towards more labour intensive economic activities. The term informal sector was then adopted by the ILO during its employment missions in Kenya. The ILO had come to realise that it took a long time for the policy recommendations they had made, in its previous missions, to filter down to the poorest sections of the population (Sethuraman, 1976). It became important to identify specific sections in the urban areas in need of more efficient methods to allocate resources. Furthermore the report on Kenya emphasised the importance of distinguishing between the informal sector (consisting of business units) and informal employment (consisting of workers). This distinction between the informal sector and informal employment allows for suitable policy recommendations in developing countries. Fields (1990:50) describes the sector as follows, In between formal sector employment and open unemployment is employment (or underemployment) in the urban informal sector. Further investigations led to the realisation that people who live in countries with no unemployment benefits and other forms of social safety nets need to earn an income of some kind, even if it is at very low returns (Moser, 1978). The informal sector in South Africa remains small despite the high unemployment rate. The two explanations that have been given for its size are that there are possible barriers to entry and/or individuals have a high reservation wage. The presence of social grants, in South Africa compared to other African countries, has been linked to the high reservation wages. This paper aims to analyse the decision making process underlying the choice individuals make to enter into the informal sector by using a theoretical model and empirical analysis will provide a 1

3 description of the characteristics of informal sector workers. The wage structure of the South African economy is discussed followed by a description of the data. The theoretical model demonstrates that at any given time, the informal sector is populated both by those with insufficient non-labour market income sources (who are force into the informal economy because to cope with poverty) and by job-seekers who expect an easier transition into formal employment than via active job search (who are drawn into the informal sector as a transition to the formal sector). A multinomial logistic regression (MNL) is used to identify the characteristics of informal workers. The effect of informality on earnings is estimated on a descriptive basis, using propensity score (PSM) reweighting techniques with complimentary panel regression. The results from the panel regression illustrate that time invariant unobserved characteristics upwardly bias the effect of informality on wages ( which serves as an important factor in the labour market decision making process). Wage structure of South African economy The wage structure of any sector is an important component of the incentives that individuals face in labour market decision making process. Fields (1990) identifies five institutional factors that influence the wage setting in formal enterprises of developing countries. i. Trade unions are powerful often because of their strong ties with the ruling political party. The labour unions keep wage employment high for insiders at the detriment of the unemployed. ii. Minimum wage laws keep wages artificially high. iii. Labour codes may result in inflated labour costs due to requirements such as severance pay and fringe benefits. iv. The pay policy in the public sector, which is not necessarily linked to the level of productivity, may set the pattern for wages in the rest of the economy. v. Multinational companies often pay higher wages partly to maintain uniform remuneration between local labour and expatriates. The informal sector is not subject to any of the five factors therefore this sector can set lower prices for its labour. This explanation has been used to describe the reasons for wage differences between the formal and informal sectors. Because formal sector wages are possibly set artificially high, the 2

4 formal sector may choose individuals who are more productive. This sorting mechanism may result in informal sector workers being less productive therefore it is important to compare similar individuals when trying to understand the preferences of job seekers between these two sectors. In order to have a better understanding of how the informal sector functions it is important to acknowledge the heterogeneity with it. Hart (1973) observed that the informal sector included individuals who used the informal sector as a means of survival as well as workers who viewed it as an opportunity to grow their wealth. These labourers differ from those found in survival enterprises because some financial and human capital is required before one begins their type of work. Fields (1990) labelled the enterprises that were started by these entrepreneurs as the upper-tier informal sector. This form of employment is expected to award higher earnings than from survivalist lower-tier enterprises. Empirical studies suggest that the earnings from this segment are comparable to formal sector wages (Blunch, Canagarajah and Raju, 2001). The structure of the informal sector in post transition South Africa was determined by apartheid laws, which limited the movements of black people in urban areas, where the most economic activities take place. Complementary labour policies restricted black individuals from obtaining semi-skilled occupations these effects are enduring. Many black labour participants have to travel a long distance from townships, at a high cost, to the economic hubs and are characterised by low human capital and small artisan or entrepreneurial base (Altman 2007). The job search model The empirical findings shed light on the duality within the informal sector and the incentives faced by the individuals who choose to enter the sector as a means to transition into formal employment an The basic model describes the actions of a utility maximising job seeker. In their search the job offers arrive at a rate λ I, λ F that follows the Poisson distribution with a wage w k attached to that offer. The wages themselves are random variables taken from the H(.) distribution. The individual chooses between three mutually exclusive states: unemployment, and employment in the informal or formal sector. The decision is based on the value V u, V I F e, V e attached to the different states respectively (which is an increasing function of the wages received from that sector). The individual exits employment if they are fired which occurs at a rate q I, q F. 3

5 An employed person receives a wage, s/he gains from the future utility of remaining in the job (if they remain employed), and the utility of unemployment if they lose their job. Adding these factors and discounting them to the present period yields the utility of employment in the k th sector. V e k w k = 1 1+r dt w k dt + 1 q k dt V e k + q k dtv u k => rv e w k = w k q k k V e w k V u This equation implies that a one period utility in formal (informal) employment is equal to the wage received in the formal (informal) sector less the utility that is lost when a person loses their job. The equation may be rewritten to reflect the change in utility if an individual changes their employment status from unemployed to employed. The change is positively affected by a rise in wages and negatively affected by an increased probability of unemployment, a rise in the one period discounted utility of unemployment, and increasing impatience (denoted by r). The reservation wage of a job-seeker An individual bases their decision to accept or decline a job offer on their reservation wage and the utility of employment will always be larger than the utility of unemployment for higher wages. The reservation wage serves as the lower bound for wages that will result in a person s decision to change their employment status, and may be measured as the one period utility of unemployment. The utility of unemployment, for a job seeker, also depends on the values attached to the job offers. A job offer with a wage that is lower than the reservation wage will be declined thus the individual will have the utility of an unemployed person. A wage offer that is higher than the reservation wage will be accepted resulting in a utility of being employed in the formal sector. In the case of the informal sector, an individual observes income opportunities (that arrive at a rate λ F ). If the opportunity yields a higher return than the reservation wage the job seeker will gain the utility associated with self-employment in the informal sector. x V λ F = V u dh F w F 0 x V λ I = V u dh I w I 0 F + V e w F dh F w F x I + V e w I dh I w I x An unemployed job seeker receives non-wage income (z), utility associated with a (in)formal sector job offer as well as the utility of remaining unemployed if the individual receives no job offers. If job offers from the two sectors do not arrive simultaneously then the expected utility of being unemployed is: 4

6 V u = r dt zdt + λ F dtv λ F + λ I dtv λ I + 1 λ F dt λ I dt V u rv u = z + λ F F V e w F V u dh F w F x + λ I I V e w I V u dh I w I x The one period expected utility of unemployment is equal to the expected returns their investment in the labour market. These expected returns consist of the non-wage income of a job seeker and the average gain that they get if/when they change their employment status. Using the one period utility of employment and the assumption of the reservation wage being equal to the one period discounted utility of unemployment i.e. rv u = x, the above equation simplifies to: x = z + λ F r + q F x wf x dh F w F + λ I r + q I x wi x dh I w I The reservation wage is increased by an increase in the job offer arrival rate and is decreased by an increase in the probability of exiting the sector therefore a person searching for work where there is a shortage (high λ and low q) of their skill will have a higher reservation wage than what the case would be if there was a surplus (low λ and high q). In the South African case, skilled labourers would have a higher reservation wage than unskilled labourers. The wage distributions of the formal and informal sectors are instrumental in determining the individual s reservation wage. The higher wages associated with formal sector employment not only make it the sector of choice but they also have an effect of increasing the reservation wage. If individuals observe that there is a wage premium in the formal sector they may adjust their expectations through the reservation wage. On-the-job search in the informal sector The model is extended to include on-the-job search, at an arrival rate of λ F I, for individuals in the informal sector who aim to find employment in the formal sector. The expansion of the basic model allows us to measure the effect of the different motives for entering the informal sector. The nonwage income (z) demonstrates the ability of the individual to survive without ever changing their unemployment status. A decrease in z corresponds with a need to supplement their income from F employment. The on-the-job search arrival rate λ I serves as a signal of the willingness to enter the informal sector in the upper-tier. An individual makes the decision to change sectors based on their current wage. The formal sector job offer will only be accepted if the worker thinks that stand to gain from higher wages. The expected utility of receiving a formal sector job offer for an on-the-job seeker depends on the value 5

7 of continuing with informal employment (for wages below the current wage) and the value of formal employment (for wages above the current wage). V λi F = w I 0 V e I w I dh F w F F + V e w F dh F w F w I This expected utility enters the individual s evaluation of the value of informal sector employment and the resultant one period utility of being in the informal sector increases by the expected gain from changing sectors. I rv e w I = w I + q I I V u V e w I F F + λ I V e w F I V e w I dh F w F w I When an individual has to choose between the sectors, they consider the utility gained from each of the labour market outcomes. The reservation wage that s/he has reflects the income for which s/he will be indifferent to the available choices i.e. V e I x I = V e F x F = V u. An individual who is able to conduct on-the-job search will then be willing to accept a lower wage offer from the informal sector if s/he will be compensated by the gain from changing sectors. The reservation wage for formal sector employment is the one from the basic model since the utility of formal sector employment remains unchanged. The non-wage income affects both the formal and informal sector reservation wage. A decrease in z means that an individual will accept a job offer at a lower rate regardless of the sector. The nonwage income is a signal of the need for earnings for survival. Any increase in informal employment due to z indicates an increase in the lower-tier and a possible decrease in the wellbeing of the individual. An increase in the formal sector arrival rate increases the proportion of individuals who enter the upper-tier portion of the informal sector. x I = x F λ I F w I V e F w F V e I w I dh F w F 6

8 0 kernel density Data description The Labour Force Survey (LFS) is a rotating panel dataset gathered by Statistics South Africa (Stats SA). The data used in the paper is for the period between 2001 and The LFS is a rotating panel that is collected biannually in March and September. The method which is used to collect the data is a two-stage complex sampling procedure. Workers are informal labourers if they answer an explicit question about whether their main activities are in the sector. Heintz and Posel (2008) argue, that the respondents answered in a way that revealed self-perception rather than the sector that they were employed in. Several complimentary questions are asked in order to verify that the respondent is not employed in the informal sector. The LFS contains information about the number of hours that the respondents worked in the previous week as well as the annual, monthly, and weekly earnings. This information is used to calculate the log of wages. The sample is restricted to the black population group with a resultant sample size of observations (from the pooled dataset) in the multinomial discrete choice analysis. The sample is further restricted to females in the propensity score matching analysis, the sample size is reduced to 9291 (from the panel dataset). Figure 1: Kernel densities of the log of wages for black people log of wages informal - female formal - female informal - male formal - male Formal sector wages are generally higher than income from informal activities; however there is a considerable overlap between the earnings. Women who earn an income in the informal sector that is comparable to the earnings in the formal sector are mainly involved in manufacturing crafts. A notable difference between the informal sector in South Africa and other African countries is that manufacturing activities are underdeveloped Valodia et. al. (2005). The activities are usually part of the lowest value and poorest in quality segments of the manufacturing sector. 7

9 0 0 kernel density kernel density A large part of the informal sector is concentrated in retail services, such as street vendors. According to the data, informal sector workers who earn wages that are much lower than formal sector wages hold elementary occupations in the retail industry. Valodia et. al. (2005) found that these informal traders tend to setup shop close to formal retailers, transport routes and taxi ranks, and are normally found in clusters of traders with similar products. Females in the formal sector have a bimodal wage distribution. The females in the higher mode are more educated and have entered into higher paying occupations in the formal sector. The kernel density graph shows that the distribution of female workers at these higher wages is larger than for males. The sample is partitioned into two groups, those who have and have not completed their secondary education. For individuals who have not completed their secondary education, males earn higher wages than females, regardless of sector. The male informal sector workers earn an income that is comparable to females in the formal sector. The gender wage differentials for workers with higher qualifications close. Figure 2: Kernel densities by education (grade 12 and less left, and more than grade 12 right) log of wages informal - female formal - female informal - male formal - male log of wages informal - female formal - female informal - male formal - male Multinomial discrete choice model To gain a general idea of how labour market outcomes are chosen the multinomial logistic model (MNL) is used. McFadden (1974) proposed a model with a multiple choice latent index and in this way linked the multinomial logistic model to utility maximisation (Imbens and Woolridge, 2007). Suppose individual i chooses option j and the utility (V ij ) associated with this choice is V ij = X ij β + ε ij 8

10 The vector X ij is a combination of state (such as age and gender) and choice variables (such as level of education and marital status). Assume that ε ij is independent over all of the individuals characteristics and has an extreme value distribution. Let individual i choose option j if s/he receives the highest utility from it, such that Y i = j if V ij V il for all l = 0,, J Then Y i follows a conditional logistic model, the MNL is a special case of this model. McFadden proved that if the independence of errors condition is met, the MNL can be derived from utility maximisation. The categories for the dependant variable are non-participants, unemployed (broad definition), informal employment (excluding domestic and agricultural work), and formal employment. Informal sector employment forms the base category. The explanatory variables are the level of education (each level represented by a binary variable), gender, province, age (and age squared), factors that reflect the household structure marital status, household head, household size, number of workers in the formal or informal sectors, children, and the elderly. The participation equations also include the year that the survey was conducted as well as interactions of gender with the household head, the marital status, the number of children, and the elderly. The sources of income accounted for before the individual makes a choice between the labour market outcome categories are the child support and the old age pension grants. Interaction terms of the grants with the elderly and children are also included. The MNL is used to compare the probabilities of the different labour market outcomes for individuals who are 42 years of age and are heads of a household that is structured as follows: five non-participants/unemployed people with an elderly person who receive an old age grant and two children who do not receive the child support grant. The old age grant indicates that the household in question is of a low socio-economic status because it is awarded after a means test based on the elderly person s income and asset value. For this group education is a significant determinant of formal sector work and non-participation. Education The education level attained by an individual is highly correlated with the sector they find employment in. The probability of finding employment in the formal sector is approximately 10% 9

11 no school grade 1 grade 2 grade 3 grade 4 grade 5 grade 6 grade 7 grade 8 grade 9 grade 10/ NTC I grade 11/ NTC II / grade 12/ NTC III diploma degree honours/ masters phd Probability for women who have less than grade 5 worth of schooling. Schooling increases the value of formal sector employment for the labourer; this is evident from the steep increase in the probability of formal sector employment after an individual completes their secondary education. The utility of informal sector employment declines, relative to other labour market outcomes, once an individual completes their secondary education. In general, the returns to labour are lower in the informal sector and due to either signalling or due to lower human capital there are increasing returns to education therefore the informal sector is not attractive to prospective workers with tertiary education because it does not create enough high paying income opportunities (upper tier). Figure 3: Probabilities of the labour market outcomes given education Informal (female) Informal (male) formal (female) formal (male) 42 years of age, KZN, household head, household structure (5 unemployed/nonparticipants, 1 employed in formal sector, 2 children, 1 elderly), other source of income (old age pension). Age The group of individuals depicted in the graph are female heads of their households. These females are also unmarried and therefore represent single parent households. The utility gained in from obtaining secure employment in the formal sector is higher than the less secure informal work. The value of informal sector employment begins to decline closer to the retirement age than the value of formal sector employment. The retirement age is when a poor person is now eligible for the state s old age pension therefore the later decline in informal employment may indicate lower tier activities. There may still be individuals who enter the informal sector in the upper tier as a means of entering the formal sector because utility of formal sector employment is still higher. 10

12 Probability Probability Figure 4: Probabilities of labour market outcomes given age by household head status (females) Informal (head) formal (head) grade 7, female, household head, KZN, household structure (5 unemployed/nonparticipants, 1 employed in formal sector, 2 children, 1 elderly), other source of income (old age pension) Household structure: Number of unemployed/non-participants A large number of unemployed people (and non-participants) form clusters around whichever source of income is available. An increase in the number of people who do not work raises the burden on the household s income. Active job search in the economic hubs from the townships have high transportation costs. Self-employment also requires start-up capital. The lower per capita income that is due to an increase in the number of non-workers in the household is an important constraint which may result in the rise of discouraged work-seekers within the household. Figure 5: Probabilities of labour market outcomes given the number of non-workers Informal formal 42 years of age, female, household head, grade 7, head of the household, KZN, household structure (1 employed in formal sector, 2 children, 1 elderly), sources of income (old age pension) 11

13 Household structure: Number of the elderly Unlike the case for the number of non-workers in the household, an increase in the number of elderly increases the probability of employment. The utility of employment is higher for people who are members of a household with at least one pensioner. The state s old age pension relieves the constraints associated with the costs of job-search and/or physical capital. The relaxation of the budget constraint allows a higher value of employment for women belonging to poor households. This is consistent with the result found by Posel and Casale (2003) about the effect of social grants on labour market outcomes. Figure 6: Probabilities of labour market outcomes given the number of the elderly Informal Informal (grant) formal formal (grant) 42 years of age, female, household head, grade 7, head of the household, KZN, household structure ( 1 employed in formal sector, 1 unemployed/nonparticipants, 3 children) Propensity score matching The MNL described the behavioural labour market outcomes associated with the level of education, age, the number of non-workers in the household as well as the effect of the state funded old age pension. The kernel densities illustrated the distribution of wages of informal and formal sector activities. The kernel densities are useful in obtaining the general idea of the differences in the returns of the two sectors, but the individual characteristics are not controlled for. Propensity score matching (PSM) provides means of controlling for individual characteristics, and making comparisons of the earnings of groups with similar qualities. 12

14 The micro-econometric scenario that the PSM techniques are based on is one of an individual who may choose between the two labour market outcomes. The individual chooses between informal sector employment (D = 1) and formal sector employment (D = 0). In order for a causal relationship to be established, the observational data should replicate the results of experimental data. This may be done by finding an exact replica of a person in the informal sector that works in the formal sector and contrast their wages. It is impossible to evaluate the true causal effect of the treatment on the outcome at the individual level because we do not observe the counterfactual this is The Fundamental Problem of Causal Inference (Holland, 1986). A descriptive analysis on the difference in returns is possible at the population level by using the mean treatment effect of treated individuals, ATT= E Y 1 Y 0 D = 1, X, with the same characteristics. The results from the MNL consistently showed low values, indicated by low probabilities, attached to informal sector employment. The difference in wages from the sectors may form an important component of these values. The initial step is to the impact of the informal sector on the wage rate by adding an informal sector binary variable (counterfactual is formals sector) in a regression. This binary variable is statistically significant therefore we may proceed in finding how important the informal sector is in the determination of wages. The PSM technique relies on two assumptions; the conditional independence assumption (CIA) and overlap. Conditional Independence Assumption The CIA states that unobservable characteristics do not influence the decision to receive treatment. That is, the outcomes Y 0, Y 1 are independent of treatment (D) conditional on the individual characteristics (X) and the decision to receive treatment is independent of the potential outcomes. It follows that if the CIA is valid then E Y 0 D = 1, X = E Y 0 D = 0, X. An example of unobservable differences between individuals employed in the formal and informal sectors is that workers with lower ability or drive are employed in the informal sector due to sorting in the formal sector. The unobservable differences bias the results of the effect of the informal sector on wages. A drawback of the CIA is that not testable therefore one relies on labour market theory to infer its plausibility. Overlap In order for both sides of the CIA equation to be defined over all X simultaneously one assumes 0 < Pr D = 1 X < 1. If any of the individual characteristics are perfect predictors of the 13

15 probability of treatment, then the matching condition cannot be used to estimate the wage effect of informal employment. This is because the people identified by these characteristics will always choose informal sector employment or will never choose it. This means that the missing values generated by the decision to work in a particular sector cannot be estimated, Heckman (1998). One of the properties of PSM is that it reveals the common dimensions of the participation decision. The results from matching are meaningful only if the characteristics that are conditioned on are shared by individuals in two sectors. These shared characteristics form the common support region. If the characteristics that are not included in the common support region are correlated to the effect of informal sector employment; the results from the matching estimation will not be consistent with the unsupported group (Bryson et. al. 2002). Matching techniques require large datasets, however Rosenbaum and Rubin proved that the ATT can be estimated by reweighting the regression with a balancing score. If the choice of treatment is strongly ignorable (the CIA and overlap assumptions hold) then the outcomes are independent of the sector choice given the balancing score. The propensity score, probability of employment in the informal sector, is the chosen balancing score. Once the propensity score is estimated for the common support region, one must choose the most suitable matching algorithm. The method used in this paper is the propensity score regression reweighting technique. This algorithm is appropriate when there is sufficient overlap in terms of reducing the bias of the treatment effect (Dehejia & Wahba, 2002). Partial equilibrium models assume that an individual s decision to be treated is independent of the decisions of other people (Bryson, Dorsett, and Purdon, 2002). A violation of this assumption may occur if, for example, people choose to stay away from the informal sector because they have witness the poor performance of their members of the household who are employed in it. If the peer correlation is to the labour market outcomes then the estimates will be biased. Another assumption that is made in partial equilibrium models is the stable unit treatment value assumption (SUTVA). The term, first used by Rubin (1980), means that the impact of the informal sector on a worker is not dependant on the number of people who are employed in the informal sector. 14

16 kernel density kernel density Results The main analysis aims to estimate the existence of, and measure the wage penalty of informal relative to formal employment. In order to comply with SUTVA the variables that indicate the number of persons employed in the formal and informal sectors were removed from the estimation function of the propensity score. These variables were also perfect predictors of the probability of formal and informal employment respectively. Two common support regions could be found for females if they were partitioned according to education level attainment (grade 9). The group who had completed their education up to grade 9 had high propensity scores and the other group (more educated group) lower scores. The informal workers in the second group are in the upper-tier as they are more likely to enter formal sector employment from the informal sector; this group s total sample size is The effect of the informal sector on this group s wages is 89.26%. PSM ignores the impact of unobservable characteristics on the treatment effect. Assuming that characteristics such as motivation and ability do not change over the period of interest, fixed effects may be included in the regression to account for time invariant unobservable characteristics. The results from the fixed effects regression show that the time invariant unobservable characteristics account for approximately 59.19% of the effect of upper-tier informal employment on wages which is about 36.42%. The lower-tier has a sample size of 3720 and the effect of informal activities in the lower tier is 81.54% before controlling for fixed effects and becomes 28.29% afterwards (the time invariant unobservable characteristic account for 65.31% of the original difference). Figure 7: Kernel density functions of the propensity score (upper- and lower-tiers) propensity score informal formal propensity score informal formal Education, province, age, age 2, birth year cohort, household head, married, household size, children, elderly, old age pension recipient in household, and value of the old age pension 15

17 The results illustrate that individuals who enter the informal sector in the upper tier are willing to accept lower informal sector wages relative to the wages they would have earned in the formal sector. These results confirm the findings of the theoretical model which proposes that upper-tier informal workers hold lower informal sector reservation wages in comparison to the minimum income they are willing to accept from the formal sector. Conclusion The informal sector does not offer the same level of income opportunities as the formal sector. It is plausible to assume that individuals enter into the sector for two reasons: they may enter as a means to transition into the formal sector (upper-tier), or they are forced into the sector as a means of survival (lower-tier). The theoretical job search model suggests that individuals are willing to accept a job in the upper-tier for a lower wage than for the formal sector. This lower income is accepted only if the person will be compensated by the expected value associated with a change of sectors. Empirical analysis identifies the important correlates of the labour market outcomes. Education is a very important component of the sector choice, and if the individual is in the informal sector, which tier they enter. The number of non-working dependants decreases the probability of overall employment because of the credit constraints associated with clustering around a source of income. The old age pension grant relaxes the constraints faced by poor females allowing them to enter the informal sector through self-employment or to find employment in the formal sector. The PSM analysis found that two common support regions may be found for individuals who chose between formal and informal employment. These regions aid in the identification of the tiers and are determined by the level of education attained by the person. After controlling for both observable and time invariant unobservable characteristics it has been shown that the wage penalty associated with upper-tier employment is larger than the wage penalty associated with lower-tier employment. These results are consistent with the finding of the theoretical model which maintains that the reservation wages required for upper tier informal sector employment relatively lower than for lower-tier informal sector work. 16

18 Bibliography Altman, M. (2007) What are the policy implications of the informal sector becoming the informal economy? In IZA/WB conference on Employment and Development. Germany, June 8 & 9. [Online] Available from: [Accessed: 06/01/13] Blunch, N. Canagarajah, S. & Raju, D. (2001) The Informal Sector Revisited: A Synthesis Across Space and Time. Social Protection Discussion Paper Series. [Online] 0119 Available from: [Accessed: 04/01/13] Brand, V. (1986) One Dollar Workplaces: A Study of Informal Sector Activities in Magaba, Harare. Journal of Social Development in Africa. [Online] 1. P Available from: pdf [Accessed: 04/01/13] Bryson, A. Dorsett, R. & Purdon, S. (2002) The use of propensity score matching in the evaluation of active labour market policies. Department of Work and Pensions. [Online] wp4 Available from: e_labour_market_policies.pdf [Accessed: 08/01/13] Dehejia, R. H. & Wahba, S. (2002) Propensity score matching methods for nonexperimental causal studies. The Review of Economics and Statistics.[Online] 84(1) p Available from: pdf [Accessed: 09/01/13] Fields, G. S. (1990) Labour Market Modelling and the Urban Informal Sector: Theory and Evidence. The Informal Sector revisited. Organisation for Economic Co-operation and Development. [Online] p49-69 Available from: [Accessed: 05/01/13] Gunther, I. & Launov, A. Informal employment in developing countries: Opportunity or Last resort? Journal of Development Economics. [Online] Available from: [Accessed: 03/01/13] Hart, K. (1973) Informal Income Opportunities and Urban Employment in Ghana. The Journal of Modern African Studies. [Online] 11(1) p61-89 Available from: [Accessed: 04/01/13] 17

19 Hausman, J. & McFadden, D. (1984) Specification Tests for the Multinomial Logit Model. Econometrica. [Online] 52(5) p Available from: [Accessed: 10/01/13] Heckman, J. J. Ichimura, H. & Todd, P. (1998) Matching As An Econometric Evaluation Estimator. Review of Economic Studies. [Online] 65 p Available from: [Accessed: 09/01/13] Imbens & Woolridge (2007) Discrete Choice Models [Lecture notes] What s New in Econometrics? NBER Summer 07 [Online] Available from: [Accessed: 10/01/13] Kingdon, G. G. & Knight, J. (2001) Why high unemployment and small informal sector in South Africa? Centre for the Study of African Economies. [Online] Available from: [Accessed: 04/01/13] Lechner, M. (2002) Program heterogeneity and Propensity score matching: An application to the evaluation of active labour market policies. The Review of Economics and Statistics. [Online] 84(2) p Available from: [Accessed: 04/01/13] Moser, C. (1978) Informal Sector of Petty Commodity Production: Dualism or Dependance in Urban Development. World Development [Online] 84(2) p Available from: DR/2/92b96ae21bf32ddd65ae600cfa3a3d60 [Accessed: 04/01/13] Rosenbaum, P. R. & Rubin, D. B. (1983) The Central Role of the Propensity Score in Observational Studies for Causal Effects. Biometrika [Online] 70(1) p41-55 Available from: [Accessed: 09/01/13] Sethuraman, S. V. (1976) The urban informal sector: concept, measurement and policy. International Labour Review. [Online] 114(1) p69-82 Available from: nals [Accessed: 05/01/13] Valodia, I. Lebani, L. & Skinner, C. (2005) A Review of labour markets in south Africa: low-waged and Informal employment in South Africa. Human Sciences Research Council [Online] Available from: [Accessed: 07/01/13] 18

20 Weeks, J. (1975) Policies for Expanding Employment in the Informal Urban Sector of Developing Economies. International Labour Review [Online] Available from: nals [Accessed: 05/01/13] 19

21 Appendix Table 1: Multinomial logistic regression of labour market outcomes nonparticipant unemployed formal employment Coef. Std. Err P>z Coef. Std. Err P>z Coef. Std. Err P>z Education grade grade *** * *** grade *** grade *** * grade *** ** grade *** <0.001 grade *** <0.001 grade *** <0.001 grade *** ** <0.001 grade 10/ NTC I *** * <0.001 grade 11/ NTC II / diploma *** <0.001 grade 12/ NTC III *** * <0.001 diploma *** *** <0.001 degree * * <0.001 honours/ masters <0.001 phd female *** *** Province Eastern Cape ** *** Northern Cape ** Free State * KwaZulu Natal * North West * Gauteng Mpumalanga Limpopo age *** *** <0.001 age^ *** *** <0.001 head *** *** *** female*head *** *** married *** ** <0.001 female*married *** *** <0.001 household size *** *** <0.001 number employed (informal) *** *** <0.001 number employed (formal) *** *** <0.001 children *** *** <0.001 child support grant ** <0.001 children*child support grant * <0.001 female*child support grant *** *** elderly *** *** old age pension *** *** elderly*old age pension ** female*elderly*old age pension *** year year ** year ** constant *** *** <0.001 * p<0.1 ** p<0.05 *** p<0.01 Dependant variable base category : informal sector 20