Job sorting and search frictions in the labour market for young black South Africans

Transcription

1 Job sorting and search frictions in the labour market for young black South Africans Gareth Roberts African Micro-Economic Research Umbrella School of Economic and Business Sciences University of the Witwatersrand Johannesburg, South Africa March 2009 CSAE conference paper Draft Acknowledgements This work was carried out with the aid of a grant from the International Development Research Centre, Ottawa, Canada 1

2 Ithemba kalibulali - Zulu proverb, meaning - Hope does not kill; I shall live and get what I want one day Iqaqa aliziva kunuka - Xhosa proverb meaning - People are unaware of their shortcomings Introduction The unemployment rate among young people in South Africa is significantly higher than in other developing countries. Bhorat and Oosthuizen (2007) show that in 2005 half of those in the labour force aged 15 to 24 were unemployed. This is significantly higher than the corresponding rate in the other countries in sub-saharan Africa - where 21% are unemployed, or the 25% in North Africa and the Middle East, and 16% of this age group in the economies of Central and South America. There is no consensus as to why the general rate of unemployment is so high in South Africa. Rodrik (2006), and Kingdon and Knight (2007) suggest that it is caused by labour market regulation. Kingdon and Knight (2001) argue that there are barriers to entry that inhibit the unemployed from entering into self-employment in this country. Banerjee, Galiani, Levinsohn, McLaren and Woolard (2007) point out this unemployment is caused by structural changes to the economy including an increase in the size of the labour force, changes in production technology, and a shift from agricultural production to less labourintensive sectors. Since employment it is not equally spread across the population (Burger and Woolard, 2005; Oosthuizen, 2006) the microeconomic determinants of this unemployment can, however, be isolated. Van der Berg and Louw (2004), and Burger and Woolard (2005) show that there has been a shift in demand toward more skilled labour in South Africa. Pauw, Oosthuizen and van der Westhuizen (2008) confirm this in their finding that employment growth has been the greatest amongst those people with the highest levels of education. It is less encouraging that unemployment has increased amongst even those relatively better-educated members of the labour force (Oosthuizen, 2006). Branson and Wittenberg (2007) suggest that, while young people are generally better educated than their parents are, they are not being absorbed into the economy at the same rate that they are entering the labour force. 2

3 This paper will examine the microeconomic determinants of unemployment in a sample of young black South Africans by comparing the characteristics of those individuals who were able to find employment to those that were unemployed. It will build on the work of, amongst others, Mlatsheni and Rospabé (2002) who examine the determinants of youth unemployment in South Africa; and Kingdon and Knight (2001), and Nattrass and Walker (2005) - who examine the extent to which unemployment in South Africa is voluntary or involuntary. It will focus on two potential determinants in particular: The hypothesis that work experience improves the probability of either staying in employment or finding employment, and the hypothesis that the reservation wages of the unemployed do not inhibit their employment. These two hypotheses are examined jointly because search may be more efficient from unemployment. This implies that work experience could lead to path dependence - where workers who take low paying jobs remain in these jobs while workers who choose to search from unemployment have a higher probability of finding better jobs. The empirical methodology employed in this paper will differ from the existing literature on the issue of youth employment in one important respect: the paper distinguishes between employment-types by the size of the firm. The equilibrium search model of Eckstein and van den Berg (2007) provides the theoretical framework that motivates this departure. It relies on both firm and worker-heterogeneity to explain the observed dispersion of wages in the labour market and the search frictions that lead to long spells of unemployment. This paper characterizes employment in firms with less than eleven employees as small firm employment. It defines jobs in firms with eleven to fifty employees as medium firm employment and jobs in firms with more than fifty employees as large firm employment. These types do not include selfemployment, which is the fourth type of employment in this paper. The unemployed are those members of the labour force that are not in wage or self-employment. Work experience corresponds to the type of employment and is the sum of the time spent working in a type for more than one year. The reservation wage of the unemployed is the minimum amount that they would accept to work, and the reservation wages of the employed in each type are their wages. The paper will first discuss the issue of youth employment, the methods used to analyze this issue, and the existing literature relating specifically to the problem in South Africa. It will then present the Eckstein and van den Berg (2007) model and provide a summary of the literature relating to the issue of youth unemployment in South Africa. The second section of the paper will present descriptions of the data used 3

4 in this paper, and the third section will discuss the results of the regression model used to isolate the determinants of employment for the sample. Literature Survey The youth unemployment problem Rees (1986), in the seminal An Essay on Youth Joblessness, suggests that there are four reasons why youth unemployment is greater than the corresponding rate for adults: seasonality, turnover, aggregate demand, and a general excess supply of labour. The young people in Rees study were more likely to exhibit seasonal employment patterns because they would enter and exit the labour force between the times that they were otherwise engaged in education. They were also more likely to quit, get retrenched, or be discharged from employment because they were less emotionally mature, and less well-equipped to compete with older workers, less risk-averse, and in some cases less dependent on this income. These characteristics diminished the employment prospects of these young people particularly in situations when low aggregate demand in an economy leads to firm down-sizing and reduction in the number of new jobs. O Higgins (2005) shows that the variations in the youth unemployment rate across countries are directly proportional to movements in the adult unemployment rates in these countries, and this is confirmed by Blanchflower (1999) who shows that the unemployment rate for youth is, as a rule of thumb, twice as high as what it is for adults across countries. Changes in the social structure of the economy have also increased labour force participation both Rees (1986) and Blanchflower (1999) note that the substantial increase in female entry into the labour markets of developed countries has significantly amplified the competition for jobs in these labour markets. The increase in labour force participation has been used to explain why minimum wage regulation could lead to a disproportionate reduction in youth employment. Rees (1986) and O Higgins (2005) acknowledge, however, that there is no accord on the role of minimum wage regulations. O Higgins (2005) even goes as far as to argue that there is no strong evidence that youth unemployment rates are closely related to youth/adult relative wage rates, and Blanchflower (1999: 7) argues that the evidence does not seem to suggest that youths are being priced out of jobs in any major way. Blanchflower (1999) rejects all of the arguments mentioned for the higher incidence of youth employment and instead argues that the problem reduces to the problem of solving adult unemployment. This is based on evidence which 4

5 shows that despite a decrease in the relative size of the youth cohort in many developed countries the two times rule of thumb continues to hold. O Higgins (2005) reminds us, however, that youth unemployment is not evenly spread across this cohort - Gender, ethnic origin, regional differences, and both skills and education levels all appear to have an effect on the probability of observing employment. It follows that any shifts in aggregate demand are likely to have an asymmetrical impact on the outcomes of individuals in this cohort with different characteristics. Those who are unemployed because they are younger, less experienced, and do not have any other redeeming qualities are being excluded from employment in favour of older, more experienced workers because they are younger, less experienced, and do not have any other redeeming qualities - the problem is the two times rule of thumb. Furthermore even if, as Heckman and Borjas (1980: 279) find, there is no evidence that previous occurrences of unemployment or their duration affect future labour market behaviour once the heterogeneity across individuals is taken in account, it is difficult to believe that interventions that aim to improve the characteristics of the youth cohort through employment experience will be unrewarding. Levinsohn (2008: 6), for instance, points to the finding that in South Africa there is evidence which shows that once an individual finds a job in the formal sector the likelihood of retaining employment in the formal sector is high. This consequently makes investigating the differences between the individuals in the cohort important because doing so provides us with a better understanding of the redeeming qualities and because it also shows us how being young magnifies any other characteristics that are relatively less desirable in the labour market. More importantly, focussing on these differences within the youth cohort facilitates the construction of remedies which are, by virtue of the greater length of time these individuals will be part of the labour force, likely to have a significant impact on aggregate supply and consequently on aggregate demand in this economy over time. It may therefore even be possible to argue that, at least in the case of many developing countries, solving that adult unemployment problem ultimately reduces to solving the youth unemployment problem. Causality in the labour market Rees (1986) extends the investigation into the problem of youth unemployment in the United States by examining the disparities in the outcomes of white and black youth. The arguments put forward include the hypothesis that structural changes to the economy have reduced employment in sectors which have traditionally employed young black people, a second hypothesis suggesting that these young black people live in areas where there are fewer employment opportunities, the argument that young black people are 5

6 discriminated against, that household welfare benefits serve as a disincentive to work, and the conjecture that the social structure of the community including a high concentration of poverty, substandard housing, female-headed households, high rates of teenage pregnancy, drug use, and crime serve to inhibit the transition. Rees (1986) was, however, unable to determine which of these arguments were valid. Angrist and Krueger (1999: 1280) point out that there are two methodologies that used when attempting to determine causality in labour market outcomes. The Structuralist approach first recovers the primitives of economic theory ([the] parameters determining preferences and technology) and then estimates decision rules derived from economic models ; and the experimentalist approach which operates in reverse by first trying to identify the causal effects from specific events and then generalizing these perhaps with the aid of economic theory or informal reasoning. They believe that the latter approach is better when investigating labour market outcomes because economic theory almost never specifies which secondary variables (other than the primary ones under investigation) should be held constant in order to isolate the primary effects. This, roughly translated, simply means that the structuralist approach requires assumptions. The experimentalist approach, in contrast, performs better in situations where it is reasonable to presume that the omitted variables are uncorrelated with the variables of interest, which is generally possible when the assignment of outcomes is random. The first pitfall associated with the experimentalist approach, however, is that it is not always possible to identify the counterfactual. For instance, if we want to analyse the effect of secondary school enrolment on the probability of employment we have to decide whether to compare this against the current level of secondary school enrolment, where nobody goes to secondary school, or where everyone goes to secondary school. The counterfactual is important because the difference in employment outcomes attributed to different levels of education could merely be a function of the individual s relative level of education this is what Wittenberg and Branson (2007) suggest may be the case in South Africa. Angrist and Krueger (1999) argue that this issue can only be resolved by assuming that there is no general equilibrium effect, which has significant implications where any findings from such studies are used to inform policy. Secondly, the absence of a formal model relating the outcome to a set of variables means that certain characteristics correlated with the outcome may be omitted. This is problematic when, as Ryan (2001: 45) points out, there are unmeasured personal attributes that are correlated with both the outcome under scrutiny and the variable of interest. In these situations the coefficients associated with 6

7 these variables are likely to lead to misleading conclusions about the magnitudes of their effects. This is a problem particularly in those studies of labour outcomes where the exogenous ability of the individuals being examined is not measured. The third issue Angrist and Krueger (1999) point out is the problem of the inclusion of endogenous explanatory variables - where both the dependent and explanatory variable(s) are determined simultaneously. This means that, for example, those studies that attempt compare employment to unemployment using the duration of unemployment must ensure that this measure does not correspond exactly to the state since all of the unemployed have unemployment experience. A fourth concern relates to the interpretation coefficients that show the linear average effect that the independent variables have on the probability of observing an outcome. In, for example, the case of the effect of age on earnings the coefficient on age provides us with an estimate of the average change in earnings that is associated with an additional year while neglecting to distinguish between these years. It is possible that this relationship is not linear, but there is no precise way of determining what this relationship is when working with samples. The youth unemployment problem in South Africa Mlatsheni and Rospabé (2002) investigate the supply-side determinants of youth unemployment in South Africa using a Probit model to estimate the characteristics associated with labour outcomes on a cross-section sample representing members of the national labour force between the ages of 15 to 64. They estimate a series of regressions with a binary categorical dependent variable that distinguishes between three states: unemployment, employment, or self-employment. The authors use the broad definition of unemployment in their analysis because, they argue, the high incidence of unemployment has lead to a substantial number of discouraged job seekers who, despite wanting to work, have given up looking for employment. The traditional narrow definition of unemployment, in contrast, is insufficient in their view because it includes only those individuals in the labour force (people between the ages of 15 to 64) who are not employed but who were looking for work. The broad definition includes those who were not looking for work but who stated that they would like to work given the opportunity. O Higgins (2005) points out that the distinction between these definitions has important analytical implications because labour force participation will vary endogenously with the unemployment rate - since fewer people will enter or remain in the labour force as the probability of finding work decreases. If people leave the labour force because of this the unemployment rate will not accurately report the extent of the problem. Kingdon and Knight (2000: 17) argue that discouraged workers should be included in the definition of unemployment because they are generally less well off than those actively seeking work, because there is no significant differences in the perception of their well-being to those who have been 7

8 looking for work, and because local wage determination takes non-searching workers into account as genuine labour force participants. Mlatsheni and Rospabé (2002) extend the definition of a young person in their analysis. The international literature (Rees 1986, Blanchflower 1999, O Higgins 2001 and Ryan 2001) defines the youth cohort as individuals aged 15 to 24. They argue that this is not appropriate in the South African context because entry in the labour force in this country generally occurs at older ages. Furthermore, the extent of unemployment in this country means that it takes even longer for young people to find their first job than in more developed countries. Mlatsheni and Rospabé (2002) therefore define the young people as those individuals aged 15 to 34. The problem with this approach is that it may confuse identification when attempting to draw a distinction between labour market discrimination based on the age of the individual, and such discrimination based on the characteristics of the individual which are not correlated with age, or the extent to which these characteristics act as a constraint because of the age of the individuals. This is a concern when there are groups within the cohort that exhibit significantly different outcomes. It would appear that Mlatsheni and Rospabé (2002) are merely acknowledging that the problem in this country is one of heterogeneity and not age. In their paper the dependent variables corresponding to the characteristics of the individuals they use include their gender, race, level of schooling, years of participation in the labour force, marital status, headship status, the number of children in the household, the presence of unemployed and employed individuals in the household, and housing tenure. They include distance from the nearest telephone as a proxy for isolation, and a set of regional dummies to capture these effects. Mlatsheni and Rospabé (2002) refer to several variables that they do not include in their analysis but which could also have an effect on the results - these include the reservation wage of the individuals, parental background variables, and neighbourhood effects. Furthermore, they do not include any employer-related variables because Mlatsheni and Rospabé (2002: 8) believe that it is hardly possible to distinguish the specific effects of the supply-side (worker-related) and demand-side (employer-related) factors on the individual s labour market state. The results of their regressions suggest that almost all of the variables they have included (the exception being secondary education) have an affect on the probability of being employed compared to being unemployed, and most (with the exception of the education variables, labour force participation duration, distance from a phone, and the Gauteng, Eastern Cape and Kwazulu-Natal region dummies) also have an affect on self-employment relative to unemployment. In addition, they show that approximately three-quarters of the age gap in employment is explained by differences in their 8

9 productive characteristics - specifically the level of work experience. They attribute the unexplained part of the gap to discrimination. Lam, Leibbrandt and Mlatsheni (2008) examine the role of education on the outcomes of young people. They start by drawing attention to the key puzzle in the issue which is why such a high proportion of young people between the age of 15 and 24 (by their estimation 42% in 2005) stop studying and enter the labour force despite the fact that approximately 60% of this cohort is unemployed? There may be two reasons: resource constraints which inhibit further education, and household income constraints which force these young people to find work. In their analysis they show that this is a puzzle because they find that completing secondary and/or tertiary education has a significant effect on the chance of finding employment. Furthermore, they find that only one in ten young Africans that leave school find work in the first month and that proportion only improves moderately to about one quarter after 12 months, 2 out of 5 after 24 months, and only a half after three years. Lam et al. (2008) use these statistics to invalidate the argument that they are leaving school to generate income. At the same time they find that there is high turnover in the youth labour market in their survey there is a 20% increase in the number of young African men who start work for the first time between the second and third year of their labour force participation. This increase, they argue, is facilitated by the exit from employment of other workers since the proportion of individuals currently employed only increases by 10%. Lam et al. (2008) then analyse the characteristics associated with the probability of being employed in each month after leaving school using Probit regressions. They include race gender, the number of months since leaving school, monthly age, the highest grade attained, scores from a literacy and numeracy exam (LNE) that were conducted with the survey, and a health indicator. The results suggest that race, gender, completed secondary school, and the LNE score all have a significant impact on the probability of finding employment. Voluntary and involuntary unemployment While both studies identify the characteristics of the employed, they are unable explain the high levels of unemployment amongst this cohort. In other words, they do not explain what the unemployed are doing. The traditional neoclassical models of unemployment are similarly unable to explain long spells of involuntary unemployment (Eckstein and van den Berg, 2007). This has lead to the development of a group of models and theories that attempt to rationalize the existence of people that remain unemployed for extended periods of time. According it Eckstein and van den Berg (2007: 532) Search theory postulates that certain events (like wage offers) occur randomly from the point of view of the individual and that this generates probability distributions for observed labour market outcomes like unemployment, job durations and wages. These equilibrium search models are consistent with the 9

10 microeconomic behaviour of these people because they estimate the determinants of the agents decision problems. Eckstein and van den Berg (2007: 532) develop an equilibrium search model which they propose is consistent with the search model of Mortensen; the search-matching-bargaining models of Diamond and Maskin (1979), Mortensen (1982) and Pissarides (1979) and the wage-posting equilibrium search models of Albrecht and Axell (1984) and Burdett and Mortensen (1998). They are able to do this because these models have a common structure and diverge only in terms of how equilibrium is defined and the strategies that the agents in these models employ in the determination of wages, and despite having common identification and estimation problems these model facilitate informative decompositions of individual labour market outcomes including the extent to which wage variation can be explained by individual and firm heterogeneity, and by frictions associated with search from both sides of the market; and the effect that individual decisions and the macroeconomic environment have on the duration and exit out of unemployment. In their basic framework workers of different types sample job offers from firms sequentially at different Poisson rates depending on whether they are employed or unemployed. These rates vary over the type of worker and are determined by the distribution of firms, on the search efforts made by the workers and firms, and on the magnitude and composition of aggregate unemployment and the aggregate amount of vacancies in the market (Eckstein and Van den Berg: 535). Workers maximize their expected lifetime utility given the costs of searching for employment. Firms maximize profits given the productivity of their workers. The parameters of this framework then assume different values to accommodate the three different approaches to the determination of wages in equilibrium. The first approach assumes that the wage offer distribution is equivalent to the worker s marginal (average) productivity. It corresponds to the model of Lucas and Prescott (1974) who propose that there are multiple productivity distributions each associated with a particular group of firms. In each group profits are zero and productivity is exogenous. A worker will only accept the wage offered if it is higher than her reservation wage. Eckstein and Van den Berg (2007) show that this reservation wage is unique. The model can be extended to include search on the job following Burdett (1978) so that the optimal strategy for the worker is to accept a job only if the wage is greater than the current wage - which means that the reservation wage of the currently employed worker is the wage the worker is earning. The second and third approaches examine the implications for the wage in equilibrium when firms do not offer worker s their marginal product i.e. when there are search frictions because firms have monopsony 10

11 power. In this case worker s have to incur costs to find an alternate match and it follows that any match may carry a positive rent or surplus (Eckstein and Van den Berg: 536). This, in turn, implies that with one reservation wage firms can capture the entire rent by merely offering the worker her reservation wage. Empirical studies, however, show that there is a large variance in the wages paid to workers with the same observed characteristics. The search-matching-bargaining equilibrium approach therefore proposes that the wage is the outcome of a decentralized bargaining game between the two parties involved in a match (Eckstein and Van den Berg: 536). The posting-wage-equilibrium approach in contrast suggests that firms have all the bargaining power and post wages before the match. In both the wage paid to the worker is generally less than her marginal product, and in the latter approach the wage (distribution) depends on the determinants of the behaviour of all agents in the market. (Eckstein and Van den Berg: 536). In the posting-wage-equilibrium models there are two sources of reservation wage heterogeneity. The approach by Albrecht and Axell (1984) allows the utility of being unemployed to vary across workers while the approach by Burdett and Mortensen (1998) extends the basic model with search on the job presented above to cases where the wage is less than the worker s marginal productivity. The dispersion of reservation wages in these incarnations of the model lead to dispersion in the wages offered that, in turn, is derived from the presence of search frictions as a consequence of limited information. This model is useful in that it provides this additional rationale for wage dispersion. According to Eckstein and van den Berg (2007) most of the empirical studies in the literature have previously only used productivity differences to explain the difference in observed wages. Since this model allows for heterogeneity across firms and workers, it is similar to Kremer s (1993: 521) O-Ring theory of economic development which proposes that workers of different types are matched together in equilibrium because of complementarities in production so that wages and output rise steeply in skill. This theory can be used to explain why there is a predominance of small firms in poor countries. In this model, however, employers cannot perfectly observe the skills of their workers. This leads to imperfect matching and multiple equilibria in education. The search equilibrium model is also useful in that it draws attention to the effect that the macroeconomic job offer arrival rates for unemployed and employed workers, and the effect that the microeconomic optimal reservation wage of the worker, all have on this dispersion, and consequently employment in equilibrium. The conclusions of this model correspond to those models that attempt to explain the existence of unemployment in developing countries. According to Kingdon and Knight (2001: 3) the dominant view in the literature is that much open unemployment is due to search and is voluntary. The models of Harris and Todaro (1970) and Harris and Sabot (1982) provide the support for 11

12 this view: when the wage in one sector of the economy is above the competitive level for the economy a proportion of workers in the labour force will choose to remain unemployed as long as the expected wage associated with searching for employment in the high wage sector equals what they could be earning in the next highest paying sector which may include self-employment. The expected wage in these models is equal to the wage in the high wage sector multiplied by the probability of finding employment in this sector. Kingdon and Knight (2001) model workers as either being in wage employment, self employment, or unemployment. In South Africa, while the income from wage employment generally exceeds the income from self-employment, the income from self-employment generally exceeds the income of the unemployed. They attempt to explain why workers choose be unemployed in this county when they can be self-employed and provide two answers: that job-search may be more efficient from unemployment, or that the unemployed are prevented from entering self-employment by barriers to entry into this sector. In the first case this unemployment is voluntary, and the second case, it is involuntary. This argument does not explain why these people do not enter into employment where they would earn less than they could in self-employment. Heintz and Posel (2008: 41), in a more recent paper, show that there are significant earnings differentials between formal and informal (unregulated 1 ) employment in South Africa. They, like Kingdon and Knight (2001), suggest that barriers to entry (and mobility) into this employment are the reason why informal employment does not mop up the labour surplus. Kingdon and Knight (2001: 3) acknowledge that some people may choose not to enter employment due to the the possibility of redistribution within the household. They also suggest that the unemployed could misjudge their expected wage by overestimating the probability attached with receiving a particular wage offer. The third reason they provide follows the classical search model-type of Stigler (1962) which serves as the foundation to the equilibrium search model outlined - who proposes that in an imperfectly competitive labour market there is a fixed distribution of expected wages. In this model the voluntarily unemployed will search for a wage in this distribution until the marginal cost of doing so exceeds the expected marginal return. This form of unemployment is correlated negatively with risk aversion. In their data, however, they find that in South Africa there is little support for the idea that unemployed people choose to be unemployed since, they argue, the unemployed are, on average, substantially worse off than the informally employed - in terms of income, expenditure and well-being (Kingdon and Knight, 2001: 13-14). They suggest that this implies that they are unemployed involuntarily despite being unable to determine if this unemployment is caused by unrealistically high expected-wages. They construct a 1 Heinz and Posel (2008) examine the definitions of informal employment. The definition they use includes both informal wage and self-employment 12

13 measure of the extent to which reservation wages are unreasonably high by first predicting what the unemployed would earn in employment given their characteristics and then subtracting this log value from their stated reservation wages (logged). While their data shows that the reservation wages of the unemployed are generally higher that what these people could reasonably expect to earn in employment, they are unwilling to draw any conclusions based on this evidence because they believe that answers to this question in the survey were unreliable. They do however concede that young people have little experience of the labour market and may therefore have inaccurate expectations, biased upwards either because of natural optimism or because they adopt a bargaining stance (Kingdon and Knight, 2001: 12). Nattrass and Walker (2005) examine the effects that reservation wages have by using a survey designed to avoid this problem. They construct a comparable measure of the reservation wages using a Heckman maximum likelihood model to predict what all the respondents would earn. The equation relating selection into employment includes the age, gender, years of education and race of the respondents. They also include the head of the household and the urban/rural background of the respondent because, they suggest, these are good proxies for social capital. The ratio of the stated reservation wage and the predicted wage is used in a Probit of unemployment on employment using a dummy variable that they give a value of 1 if this ratio is greater than 1 and 0 if it is less than 1. The coefficient from the estimate, surprisingly, is negative and they conclude that this shows that there is no evidence that relatively high reservation wage to the predicted wage ratios are the cause of unemployment in their data. 13

14 The Data This paper uses data from the South African Young Persons Survey (SAYPS). The SAYPS is a retrospective survey conducted between July and November It was commissioned by the Centre for Development and Enterprise and conducted in conjunction with the African Micro-Economic Research Umbrella (AMERU) of the University of the Witwatersrand in Johannesburg, South Africa. Representative samples were selected randomly within clusters which were distributed proportionate to the residential spread of young Africans in townships, informal settlements, suburbs, inner-city areas, and tribal areas in three locations: the Johannesburg metro in Gauteng, the ethekwini metro in KwaZulu- Natal, and the town of Polokwane and adjacent rural area of Dikgale in Limpopo (CDE, 2007: 8). These locations were chosen because they permit a comparison between urban and rural settings and between the regional economies of the locations. Table A1 in the appendix shows the number of respondents by their age and gender: Approximately half of individuals were interviewed in Johannesburg and the remainder were distributed equally between the other two areas. The SAYPS first asked the respondents questions about the environment (including the characteristics of the household) when they were born, at age six, age twelve, age fifteen, and at the time of the interview; and their primary and secondary schooling. It then asked the respondents to relate the details of any of the current and past further education activities and spells of employment or unemployment by asking each respondent the following question: We are going to ask you about all the significant activities (jobs, unemployment, education etc) since you were aged 15 and working forwards. Please tell us about all jobs, periods of unemployment and education since then. When you were aged {} what was your main activity? The questions about education distinguish between school and further or higher education where the latter includes any course that the respondent took, and may correspond to job or vocational training. The questions about jobs distinguish between working for someone else and working for oneself: working for someone else includes working for formal, informal or labour broker firms; and working for oneself includes both formal and informal businesses. The survey also asked those respondents who indicated that they were unemployed or had experienced a spell of unemployment to choose the type as one of Unemployed: looking for work and unable to find any ; Unemployed: not at paid work and not looking for work ; Unpaid family worker, helping with domestic and other family work ; or Disabled and unable to work. 14

15 Descriptive Statistics The first part of this section will provide an overview of the respondents in the sample to provide a context for the second part which examines their employment. The purpose of the first part is to convey some sense of the backgrounds of the individuals in this particular sample which extends only to young black people in these three locations. The young people in the sample Over one third (36%) of the respondents were born in the Limpopo Province, almost 30% in Gauteng, and just over a quarter were born in Kwazulu-Natal. Most of the respondents (70%) lived less than 3km from the nearest primary school when they were six years old, and 84% attended the nearest school when they started their formal education. The data shows that only two individuals in the sample did not go to primary school, and approximately 80% of the remaining individuals started their schooling aged 6 or 7. Just 19 (2%) did not complete primary school. However, 40% of the respondents had to repeat at least one year of primary school 35% of the females and 45% of the males and just over 20% only completed primary school when they were 15 or older. A third of the respondents were taught in English even though only 2% of the respondents spoke English at home just over 40% of all the respondents who went to primary school were not taught in their home language. Nine out of ten respondents started secondary school, and 68% of these respondents were taught in English more than two thirds (70%) of the respondents who started secondary school were not taught in their home language. The data shows that a significant proportion (22%) of the respondents started secondary school after they had turned sixteen. It also reveals that more than half (53%) of the respondents who entered secondary school had to repeat at least one year and less than a third of the respondents did not have to repeat any years at school. Three quarters of the respondents were looked after mainly by their mothers or by both parents when they were fifteen and one in ten were looked after by their grandmothers. Most (45%) said that their father was the head of the household and a third said their mother was the head. Almost 60% of these heads were employed as regular wage workers and 6% were unemployed. More than 85% of the heads had not completed matric. Two thirds of the households had fewer than seven members, and just over a third had more than one earner. Approximately half (52%) of the respondents believed that their household was better off than the 15

16 other households in the area they living when they were 15 years old. Table 1 presents the proportions of the respondents by completed years of school. Forty-five respondents completed courses at private further education institutions, sixty-five at technical training colleges, and seventy-five completed a degree or diploma at a university, Technikon, or Teachers Training College. Table 1: Completed years of school by gender Female Male Total Primary or less (Grade 1 to 7) Grade 8 Grade 9 Grade 10 Grade 11 Grade 12 (Completed Secondary School) No % No % No % No % No % No % The SAYPS data shows that 10% of the respondents had become the head of the household at the time of the survey. Substantially fewer (only 44%) of the heads were now employed in regular wage employment. Approximately two fifths of these households also had more than one earner. The data also shows that more than half of the respondents were living in a dwelling or brick structure on a separate stand/yard/farm and most (55%) lived in a dwelling that had tap water. Approximately a third of the respondents said that someone in the household received a welfare grant. A significant number (54%) of the respondents had at least one child more than two thirds of the females and just under a third of the males respondents. Two thirds of the respondents who had a child were not living with the other parent. The majority (60%) of the respondents who had children said that they received child support grants from the government. Just over a third of all the respondents had to help look after their siblings by paying for food, clothes, schooling. Half had to spend more than seven hours a week looking after their siblings, over a quarter had to help look after their parents by sending them money, and almost a half (45%) had to spend more than seven hours a week looking after their parents. 16

17 The respondents were asked about what they wanted to achieve in the next two years and were allowed to give more than one answer. These answers show that, amongst the other goals, more than half of the respondents said that they wanted more education, 60% said they wanted to acquire a job, 18% wanted to keep the job they had and 42% said they wanted a better job or a promotion. Furthermore 54% of the respondents who said that they wanted a job or a better job also said that they wanted more education. A quarter of the respondents wanted to become healthier, and approximately one fifth wanted to have children. Only 12% of the respondents indicated that they thought they were unlikely to achieve their goals in the next two years. This question referred to all of the goals they listed and does not only apply to any one particular goal. However the data shows us that these include, amongst the other goals, a tenth of those that wanted more education, 15% of the respondents who wanted a job, 13% of those who wanted to keep their job and 9% of the individuals who said they wanted a better job. The breakdown of the figures for those that are less optimistic suggest that at least some of the respondents were aware of the lower probability of finding employment than of achieving the other goals listed. This is confirmed when the data reveals that a quarter of those who were unemployed at the time of the survey believed that they were unable to find employment because they did not have the skills, a quarter believed that they did not have the necessary experience for the jobs that were available, and 40% did not think that there were any jobs. Only 5 of these unemployed respondents, however, indicated that they believed that the jobs that are available pay too little. Employment and unemployment Approximately half of the respondents were unemployed. The following table shows the distribution of the activities of the respondents was not evenly spread across provinces: in Kwazulu-Natal the unemployment rate was 58% while in Gauteng proportionally more of the respondents had wage employment. The second column of each province shows the proportion of respondents engaged in the activity for that province. 17

18 Table 2: Current activity by Province Gauteng KwaZulu-Natal Limpopo Total No. % No. % No. % No. % Further or higher education School education Unemployed Working for oneself Working for someone else Total The majority - 85% (913) - of the respondents were in the labour force at the time of the survey. It is a common problem across surveys of this nature that there are missing or implausible values for some of the questions. Table A2 shows that 28 (3%) of these observations there are missing answers to one or more of the questions relating to the firm size or individual characteristics that will be used in this paper. Most of these observations correspond to respondents who were working for someone else at the time of the interview because they did not provide answers to the questions about the size of the firm. The observations that were excluded appear to have been randomly selected by gender, current age and the number of years of education (where this latter is defined as those years of education greater than 7) since the proportion of males, the average age, and the average number of years of education do not change significantly when these observations are omitted. Table 3 shows the proportions by state for the remaining observations Table 3: Proportion of the labour force engaged in activities by Province Gauteng KwaZulu-Natal Limpopo Total No. % No. % No. % No. % Unemployed Working for oneself Working for someone else Total Those respondents who were working for someone else at the time of the survey or had such a job in the past were asked What size is the size of the company? and could choose from one of: < 5 employees; 6-10 employees; employees; employees; employees; and >100 employees. This paper will, as mentioned, distinguish between small firms: 1 10 employees; medium firms:

19 employees; and large firms: > 50 employees. These three categories should provide a more accurate measure of the firm size because they are broader. It follows from these answers that proportionally more of male respondents who were in the labour force at the time of the survey were employed in large and medium size firms, and in self-employment. The unemployment rate for females was significantly higher than the corresponding figure for males: Table 4: Proportion of labour force by employment state 2 Female Male Total No. % No. % No. % Large Medium Self-employed Small Unemployed Total The data also shows that there more individuals (as a proportion of the labour force) were employed in large firms in Gauteng than in the other province, and just over 70% of the respondents in Limpopo who were part of the labour force were unemployed at the time of the survey. The respondents from Limpopo only made up 20% of the labour force observations. Approximately 34% of respondents that were in this group were born in this province. The data shows that 13 of the 106 individuals that had moved from unemployment or education in another province to Guateng had found jobs at small firms, and 26 found work at firms with more than ten employees. Only 1 of the 22 individuals who moved to the Limpopo province from another province (18 moved from Gauteng) was able to find employment in this province - and this was as a cleaner in a hospital. When the wages or the reservation wage of the respondents are used as explanatory variables there are significantly more observations where no answer was given or where the answer that was given was implausible (e.g. negative or very high). The proportions in each province do not change dramatically, but it would appear that more of the unemployed in Kwazulu-Natal did not provide an answer to the question about the minimum wage they would accept to work. The next two tables show that while the exclusion of the non-responders to the wage (including the minimum accepted wage) question has an effect on the mean values, the only concern is the experience (the sum of self-employment and small firm 2 The paper distinguishes between labour force participation by firm size types, self employment and unemployment 19

20 work experience) of the self-employed. The first table shows the average age at the time of the interview, education, and large firm, medium firm and other work experience all in years. The second table provides the same figures for those respondents that provided answers their wages or reservation wages. They show that the mean age and education of those respondents who were employed in large firms is higher than for the other firm sizes, but the mean wage is less than that of the self-employed. It is interesting that the mean number of years of education (which is defined here as the completed number of years of school and further education greater than seven 3 ) is marginally higher than the same figure for workers in small firms. Table 6: Sample with firm size answers 885 observations. Mean age, education and experience by state Male Large Medium Other Age Education (Proportion) experience experience experience Large firm Medium firm Self-employed Small firm Unemployed Total Table 7: Sample with firm size and wage or reservation wage answers 784 observations. Mean age, education and experience by state Male Large Medium Other Age Education (Proportion) experience experience experience Large firm Medium firm Self-employed Small firm Unemployed Total Both tables show that the mean number of years of experience in a firm of a particular size corresponds to the respondent s current employment type 4. This figure for those respondents who were employed in a large firm is significantly higher than for any of the other sizes. The unemployed have the least 3 Only a small number of the respondents said that they had not completed primary school 4 Experience is defined has any experience in the firm type of over one year 20