The employment environment for youth in rural South Africa: A mixed-methods study

ABSTRACT

South Africa has high youth unemployment. This article examines the predictors of youth employment in rural Agincourt, Mpumalanga Province. A survey of 187 out-of-school 18 to 24 year olds found that only 12% of women and 38% of men were currently employed. Men with skills/training were significantly more likely to report employment, mostly physical labour (adjusted odds ratio: 4.5; confidence interval: 1.3, 15.3). In-depth interviews with 14 of the youth revealed that women are perceived more suitable for formal employment, which is scarce, informing why women were more likely to pursue further education and yet less likely to be employed. Ten key informants from local organisations highlighted numerous local youth employment resources while, in contrast, all youth in the sample said no resources were available, highlighting a need for the organisations to extend their services into rural areas. Because these services are focused on entrepreneurship, programmes to increase financial literacy and formal employment opportunities are also needed.

KEYWORDS:

• Youth
• employment
• rural

1. Introduction

Persistently high youth unemployment is one of the biggest social and economic challenges in South Africa. Several analyses of the 1995 October Household Survey and the 2005 Labour Force Survey have concluded that unemployment is highest and rising the fastest in the younger cohorts; the broad unemployment rate among 20 to 24 year olds is estimated to be 52.3%, compared with 36.0% for 25 to 29 year olds (Yu, 2013). Unemployment in South Africa also disproportionately affects youth who are black, female, and/or living in a rural area (Mlatsheni & Rospabe, 2002; Lam et al., 2010). In the rural Mpumalanga Province of South Africa, over 70% of the population is under the age of 35 and the unemployment rate is higher than the national average (Mpumalanga Provincial Government, 2011).

Research from South Africa demonstrates high youth unemployment is associated with negative health and economic outcomes in both the short and long term. In the short term, it is associated with higher crime, substance use, HIV risk, and ceasing to search for jobs (Booysen & Summerton, 2002; Rankin et al., 2012). In the long term, a gap between youth finishing school and finding work impacts future wages and thus long-term well-being (Freeman & Wise, 1982; Posel & Casale, 2011; Rankin et al., 2012).

HIV rates in South Africa are also among the highest in the world and the burden of HIV falls disproportionately on young women (UNAIDS, 2014). In 2012, 17.4% of 20 to 24-year-old women compared with 5% of men were infected with HIV (Shisana et al., 2014). Young women are three or four times more likely to be HIV-infected than men their age (Pettifor et. al, 2005). Unemployment among young women in South Africa can increase their vulnerability to HIV by increasing their economic dependence on male partners and potentially increasing unsafe sexual behaviours (Booysen & Summerton, 2002; Shisana et al., 2014). Understanding the factors associated with youth employment in rural South Africa may help inform interventions to improve the socio-economic status (SES) and HIV risk of youth.

The aim of this study was to identify factors associated with employment for out-of-school youth in rural Mpumalanga, South Africa, and to help identify promising strategies to increase employment and reduce HIV vulnerability. It also aligns with an existing cash transfer programme in the Agincourt sub-district incentivising young women to stay in school (MacPhail et al., 2013). Being in school is associated with decreased HIV vulnerability, but once youth graduate or drop out and are unable to find work, the HIV risk may increase (Pettifor et al., 2008). In South Africa, it is particularly difficult for many youth to find work because the economic growth in rural areas may be slow and sectors that do grow rapidly may require workers with higher skills than youth would graduate with (Burns, Godlonton, et al., 2010; Mpumalanga Provincial Government, 2011; International Labour Office, 2012). Despite these difficult employment circumstances, some youth do find work.

This mixed-method study used both quantitative data – to find out how the youth who find work differ from their peers – and qualitative data – to describe the pathways through which these differences could lead to employment. The characteristics we hypothesised could predict employment were informed by prior studies. For demographics, being married or being a parent could increase the odds of youth employment (Mlatsheni & Rospabe, 2002). Completing high school and passing matric are also likely to increase youth’s ability to obtain work (Mlatsheni & Rospabe, 2002; Lam et al., 2010; Ardington et al., 2013). However, training certificates or diplomas may be a more important predictor of employment than education because they can signal to the employer the presence of more concrete skills (Mlatsheni & Rospabe, 2002; Burns, Edwards, et al., 2010; Rankin et al., 2012). Beyond qualifications, it seems that youth with relatives or friends who are working may have more information on job vacancies, thus increasing their chances of employment. Finally, with regards to psychology, youth who are more confident in their entrepreneurial skills are probably more likely to be employed, because employed youth in South Africa are concentrated in self-employment (International Labour Office, 2012; Rankin et al., 2012). Also, employed youth often have greater life satisfaction compared with unemployed youth (Mlatsheni & Rospabe, 2002; Witte et al., 2012; Lloyd & Leibbrandt, 2013).

2. Methods

We conducted semi-structured quantitative surveys (n = 187) and in-depth interviews (IDIs) (n = 14) with 18 to 24-year-old youth, and key informant interviews (n = 10) with government and non-profit organisation leaders from the rural Mpumalanga Province. Youth participants were sampled from the Agincourt Health and Socio-Demographic Surveillance System (HDSS), now covering some 110 000 people in 31 villages. The HDSS conducts an annual census update to collect health, socio-economic and population information on all residents in the study area; descriptions of the cohort have been published previously (Kahn et al., 2007, 2012). We randomly sampled out-of-school youth from one of the largest and smallest villages covered by the HDSS to maximise the breadth of employment experiences in the sample. Although the Agincourt HDSS study area is rural, agriculture is not a predominant form of local employment due to the region’s arid climate.

2.1. Procedure

The study and consent procedures were approved by the Institutional Review Board of the University of North Carolina at Chapel Hill, USA and the Human Research Ethics Committee (Medical) of the University of the Witwatersrand in Johannesburg, South Africa. Data were collected from June to July 2013. Using the HDSS sampling frame, we randomly sampled 447 households that met the following eligibility criteria: at least one household member was 18 to 24 years old, not currently enrolled in school, and living in the area consistently for at least the past six months. We sampled youth living in the area in order to capture youth employed locally rather than youth engaged in migrant labour. Of the nearly 600 individuals contacted by the fieldworkers, over 300 were ineligible because they had moved out of the area or re-enrolled in school since the last census (Figure 1). It is common in this area for students who failed Grade 12 to repeat it, which could explain the high proportion of our sample that were unexpectedly still in school (Lam et al., 2011). Despite evening and weekend data collection, more women were enrolled than men, probably due to the higher prevalence of labour migration among men (Collinson et al., 2006). Surveys were conducted in Shangaan and lasted one to two hours.

Figure 1. Flow chart of sample recruitment. Note: Starting with a random sample of 447 households and 686 individuals, contact was made with 596 individuals, 187 of whom were eligible and all of whom consented to participate.

Participants for the IDIs were recruited from the quantitative sample. Individuals were purposively sampled on employment status, gender, and age to maximise the diversity of our IDI sample along key socio-demographic lines. Interviews were with one respondent and one trained interviewer, audio-taped, conducted in Shangaan, lasted one to two hours, and then were transcribed and translated into English. To sample key informants, we inquired with local staff and researchers and used government websites as well as recommendations from other informants. These interviews were conducted in English and transcribed by the author, and lasted one to three hours. Information from the IDIs and key informant interviews were analysed carefully (Section 2.4).

2.2. Measures

Measures included in the quantitative semi-structured survey were largely adapted from existing youth employment and household socio-economic questionnaires, many from sub-Saharan Africa (Bandiera et al., 2009; The World Bank, 2013; Medical Research Council, 2014). The completed survey had 10 sections – a household roster, household assets, demographics, education history, employment history, job environment perceptions and expectations, professional network, loans and savings, job skills and training, financial literacy – and three scales to assess self-confidence in entrepreneurship, empowerment attitude, and overall life satisfaction.

2.2.1. Outcome: Employment

Employment status was assessed by asking participants whether the respondent had performed a range of employment activities in the past four weeks (e.g. construction, office work). If a respondent reported receiving compensation for any activity, they were considered employed. The vast majority of employed respondents reported informal work so this was collapsed with the formal work category.

2.2.2. Independent variables: Demographics

Because a very small proportion of the sample reported being widowed, divorced or cohabiting, we collapsed marital status into a binary variable of those currently married and those not. Being a parent was also a binary variable generated from responses to the question of ‘How many living children do you have?’

2.2.3. Independent variables: Education, training and skills

Educational attainment was assessed on a range from no education at all to completing university. The majority of the sample clustered around less than high school or completing high school, and therefore education was coded as a binary variable indicating completing high school or not. Passing the school-leaving matriculation examination (matric) was assessed separately and only the respondents who reported not taking matric at all were coded as missing. Respondents were also asked whether they had completed a post-secondary diploma or a diploma in progress, probably through a training college that often offers both formal and vocational education and some connections to employment (McGrath, 2004). For skills, respondents were asked whether they possessed any of a range of working skills (e.g. computer skills) and were also able to give additional skills not listed. The skill list was adapted from the World Bank Urban Youth Employment Project Eligibility Screen (The World Bank, 2013). Finally, respondents had the opportunity to list general or business trainings or apprenticeships they had completed. Respondents who completed any training in these categories were identified with a binary variable. Respondents were also asked whether they were seeking work and their desired work sector at age 30.

2.2.4. Independent variables: Social networks

If a respondent reported having an employed household member or employed friend (within or outside the study area), they were coded as having a professional network. Counting employed household members, regardless of relationship, as a respondent’s professional network is consistent with previous analyses (Burns, Godlonton, et al., 2010).

2.2.5. Independent variables: Psychosocial

Psychological scales used in the survey were adapted from the Brac Uganda Adolescent Development Program (Table 1; Bandiera et al., 2009). The first scale assessed self-confidence in entrepreneurial tasks (Cronbach’s alpha = 0.83) by asking participants to rank, on a scale of one to 10 (10 = high), their ability to do 10 tasks (e.g. ‘manage financial accounts’). Using the same scaling system, the empowerment attitude scale (Cronbach’s alpha = 0.76) asked respondents to rank how true 10 statements were for themselves (e.g. ‘I often make plans for the future’). The overall life dissatisfaction scale (Cronbach’s alpha = 0.68) asked participants to rank on a scale of one to seven (7 = high) their dissatisfaction with aspects of their lives (e.g. job, house). An item from this scale asked about satisfaction with friends but the responses had so much heterogeneity that the item dropped the scale’s alpha to 0.58 and so was excluded. For each of the three scales, answers were summed into a continuous score and reverse coded where necessary.

Table 1. Items in psychology scales.

Self-confidence in Entrepreneurial Tasks (10–100)

Rate your ability to do task (1: cannot do activity, 10: definitely can)

Run your own business

Identify business opportunities to start up new business

Obtain credit to start up new business or expand existing business

Save in order to invest in future business opportunities

Make sure that your employees get the work done properly

Manage financial accounts

Bargain to obtain cheap prices when you are buying anything for business (inputs)

Bargain to obtain high prices when you are selling anything for business (outputs)

Protect your business assets from harm by others

Collecting the money someone owes you

Empowerment Attitude (10–100)

Rate how true each statement is to yourself (1: not at all, 10: a lot)

If I start working on a task, I definitely see the end of it no matter how difficult it is

While doing any task, it is important for me to do it better than others

If I have the chance, I would make a good leader

I want to be a respected person in my village

I do not care what others think about my success or failure

I am in control of what happens in my life

I save regularly

A person can get rich by taking risks

I often make plans for the future

I believe that my future is determined by luck no matter how hard I work

Overall Life Satisfaction (7–42)

Rate how happy you are with each aspect of your life (1: completely happy, 7: not at all happy)

Your education level

Your family

Your friends

Your job or employment prospects

Your earnings/income

The house you live in

Life as a whole

Notes: Relevant items were reverse coded in the final scale. The friends item was dropped from the Overall Life Satisfaction scale because it showed very low variability and improved the Cronbach’s alpha.

2.2.6. Socio-demographic characteristics

Age, SES, and village of residence were used as control variables in these analyses. SES was measured using asset quintiles, which range from one to five (one being the lowest quintile). This measure was derived from the HDSS SES index based on household assets, on which we collected detailed data (Collinson et al., 2009). All respondents identified as black/African and all but two identified as South African. An indicator variable denoted village of residence.

2.2.7. Descriptive variables

To inform future intervention development, information was also collected on respondents’ financial situation and overall financial literacy. Specifically, respondents were asked whether they had savings; the amount and source of savings; their motivation for saving; and whether they had a bank or post office account. Respondents also completed a six-question financial literacy test (Table 2) for which scores ranged from zero to six (6 = high).

Table 2. Items in the financial literacy test.

What are the things you need to know in order to make a budget?
NOTE: Can list more than one
Income^a
Expenditure^a
Savings
Don’t know
Other
Is there any difference in the interest rate of a current account and savings account in a bank? If so, which one gives a higher interest rate?
Savings account^a
Current account
Same interest rate
Suppose you have deposited 100 Rand in the bank for an interest of 10 Rand per year. If you withdraw all the money after 2 years, how much will you get?
120 Rand
Out of five of the things being mentioned to you, mention any three that you think is important while taking a loan?
Name of the loan officer
Rate of interest^a
Time of meetings
Fine or delay in repayment^a
Number of instalments^a
Suppose you need to take a loan of Rand 1000 and you have two choices. In one you pay an interest of Rand 10 every month and in the other you pay an interest of Rand 120 at the end of the year. Which one has a higher interest rate?
First option (monthly)
Second option (yearly)
Same interest rate for both^a
What will happen to the price of firewood if the price of kerosene increases?
Increase
Decrease^a
Unchanged

Notes: ^aCorrect answers.

2.3. Statistical analysis

All models were stratified by gender. Adjusted and unadjusted logistic regression models were used to assess the relationship between each independent variable and employment, the key outcome of interest. Hypothesised confounders were age, asset quintile, and village of residence and were left in adjusted models if they were statistically significant. Referent groups were chosen based on a priori hypotheses to produce estimates above the null. For example, we regressed employment on education with not completing high school as the referent and we ran the models, separately for males and females, with and without measures of the hypothesised confounders.

2.4. Qualitative interview guides and analysis

The IDIs were semi-structured with questions about the economic environment, the interviewee’s employment experience, barriers to employment, and resources to aid in the school-to-work transition. One set of a priori codes were established based on the interview questions (e.g. one code was used when youth mentioned what they or other youth did for employment and another code was used when youth mentioned specific employers by name). The a priori codes were updated iteratively as coding progressed, allowing for new themes to emerge (e.g. different employment opportunities for males and females). The Dedoose software program was used for all qualitative coding (SocioCultural Research Consultants, 2013).

Key informant interviews were semi-structured with questions on the economic environment for youth, what their organisation did to help youth make the school-to-work transition, and what more they thought was needed to foster youth employment. The transcripts were read multiple times, and summarised to capture key points relating to each interview topic. Findings from the IDIs and key informant interviews informed our interpretation of the quantitative findings.

3. Results

3.1. Descriptive analyses

Descriptive information on the youth sample is presented in Table 3. The average age was 22.3 years and just over half of the sample was female (54%). Marriage was relatively rare, although nearly three times as many women (17.9%) reported being married as men (6.2%). Over 70% of women reported having at least one child compared with less than 20% of men. Of the 10 key informants, four were women and six were men and they held a range of positions in their respective organisations (four were youth leaders, five were regional managers for national programmes, and one was a national leader).

Table 3. Characteristics of the sample.

	Total (n = 187)	Men (n = 81)	Women (n = 106)
Controls
Age	22.3 ± 1.8	22.1 ± 2	22.4 ± 1.7
Asset quintile
1	20 (10.7)	5 (6.17)	15 (14.15)
2	47 (25.13)	24 (29.63)	23 (21.7)
3	37 (19.79)	17 (20.99)	20 (18.87)
4	23 (12.3)	13 (16.05)	10 (9.43)
5	40 (21.39)	13 (16.05)	27 (25.47)
Missing	20 (10.7)	9 (11.11)	11 (10.38)
Independent variables
Marital status
Never married	149 (79.68)	69 (85.19)	80 (75.47)
Married	24 (12.83)	5 (6.17)	19 (17.92)
Cohabiting	11 (5.88)	5 (6.17)	6 (5.66)
Divorced/separated	3 (1.6)	2 (2.47)	1 (0.94)
Number of living children
0	95 (50.8)	67 (82.72)	28 (26.42)
1	61 (32.62)	8 (9.88)	53 (50.00)
2	27 (14.44)	6 (7.41)	21 (19.81)
3	4 (2.14)	–	4 (3.77)
Educational attainment
Some primary	5 (2.67)	2 (2.47)	3 (2.83)
Completed primary	6 (3.21)	5 (6.17)	1 (0.94)
Some high school	69 (36.9)	25 (30.86)	44 (41.51)
Completed high school	102 (54.55)	46 (56.79)	56 (52.83)
University or technikon	5 (2.67)	3 (3.70)	2 (1.89)
Passed matric	105 (56.15)	45 (55.56)	60 (56.6)
Completed diplomas outside school	40 (21.39)	15 (18.52)	25 (23.58)
Completing diplomas outside school	19 (10.16)	5 (6.17)	14 (13.21)
Employment
Currently employed	44 (23.53)	31 (38.27)	13 (12.26)
Seeking work in past 3 months	129 (68.98)	64 (79.01)	65 (61.32)
Have working skills	86 (45.99)	50 (61.73)	36 (33.96)
Received skills training	47 (25.13)	30 (37.04)	17 (16.04)
Received business training	42 (22.46)	23 (28.4)	19 (17.92)
Network
Relatives working	168 (89.84)	74 (91.36)	94 (88.68)
Friends working	75 (40.11)	40 (49.38)	35 (33.02)
Psychosocial
Self-confidence in entrepreneurship (maximum = 100)	81.5 ± 13.6	82.7 ± 13.9	80.6 ± 13.4
Empowerment attitude (maximum = 100)	84.2 ± 11.4	84.0 ± 13.4	84.3 ± 9.7
Life dissatisfaction (maximum = 42)	27.0 ± 9.1	27.2 ± 8.9	26.9 ± 9.3

Notes: Data presented as number (%) or mean ± standard deviation.

3.2. Employment

Less than one quarter of respondents reported employment in the past four weeks. The proportion of men reporting employment (38.3%) was more than twice as high as the proportion of women (12.3%). Nearly 80% of men reported currently seeking work compared with just over 60% of women.

3.3. Education

More than half of respondents reported completing high school (57.2%) or passing matric (56.2%). More men (60.5%) reported completing high school than women (54.7%), but the proportions who reported passing matric were similar (55.5% vs. 55.7%).

3.4. Skills and training

Women were more likely than men to have completed (23.6% vs. 18.5%, respectively) or currently be completing (13.2% vs. 6.2%) a diploma outside of school, whereas men were more likely to report having received skills or training (66.6% vs. 47.2%). The most common diplomas were in computing, security, and medical services, although there was a wide range. The most cited training opportunities were related to physical labour (e.g. forklift, plumbing). When respondents were asked what sector they desired to be working in when they are 30 years old, 48.54% of women said clerical work. By comparison, the two most common responses for the men were mining (16.2%) and clerical work (16.2%).

3.5. Psychosocial

Women and men were similar in their high levels of self-confidence and empowerment. Out of 100 possible points, where a higher score meant higher self-confidence in entrepreneurship, men scored 82.7 on average (standard deviation [SD]: 13.9) and women had an average score of 80.6 (SD: 13.4). Similarly, out of 100 points on the empowerment attitude scale, men scored an average of 84.0 (SD: 13.4) and women 84.3 (SD: 9.7). Interestingly, men and women also shared similarly moderately high levels of overall life dissatisfaction. Out of 42 possible points, where a higher score signalled higher dissatisfaction, men scored 27.2 (SD: 8.9) on average and women scored 26.9 (SD: 9.3) on average.

3.6. Finances and financial literacy

The financial characteristics and literacy of the sample are presented in Table 4. Fewer than half of the sample reported having savings, although more than half of these reported their savings amount as zero, and all but one of these respondents reported keeping them at a bank. Among those who did report some savings, the mean amount saved was 488 Rand or about $46. For both men and women, the most common source of savings was parents (63.6%) but the second most common source was earnings from work for men (27.50%) and ‘other’ for women (24.4%). Most respondents cited future expenditures (55.2%) as their main motivation to save, but, beyond this, men were more likely than women to say future study (27.5% vs. 6.7%) and women were much more likely than men to say future child expenditures (27.7% vs. 2.6%). None of the respondents picked wedding expenses as a reason to save. Approximately half of the respondents reported having a bank account and less than 10% reported having a post office account. Only one respondent reported ever taking out a loan. Overall, the scores on the financial literacy test were low; of a maximum score of six points, the average score was less than three for men and women.

Table 4. Financial characteristics of the sample.

	Total (n = 187)	Men (n = 81)	Women (n = 106)
Finances
Have savings	87 (46.52)	40 (49.38)	47 (44.34)
Savings amount (Rand)	488.1 ± 600.8	508.3 ± 734.0	474.9 ± 513.5
Source of savings
Parents	54 (63.53)	23 (57.50)	31 (68.89)
Earnings from work	14 (16.47)	11 (27.50)	3 (6.67)
Other	17 (20.00)	6 (15.00)	11 (24.44)
Motivations to save
Future expenditures	44 (55.16)	23 (58.97)	21 (44.68)
Bad times	9 (10.47)	4 (10.26)	5 (10.64)
Further study	19 (22.09)	11 (28.21)	8 (17.02)
Wedding expenditures	0	0	0
Child expenses	14 (16.28)	1 (2.56)	13 (27.66)
Have a bank account	95 (50.80)	43 (53.09)	52 (49.06)
Have a post office account	16 (8.56)	6 (7.41)	10 (9.43)
Financial literacy
Test score (0–6)	2.77 ± 0.85	2.83 ± 0.75	2.73 ± 0.92

Notes: Data presented as number (%) or mean ± standard deviation.

3.7. Logistic regression models

Table 5 presents results for the unadjusted and adjusted logistic regression analyses stratified by gender. Men who reported having working skills or training were much more likely to be employed, compared with those with no working skills (adjusted odds ratio [aOR]: 4.5; 95% confidence interval [CI]: 1.3, 15.3). For women, the only variable significantly associated with employment was greater life dissatisfaction (aOR: 1.1; 95% CI: 1.0, 1.2).

Table 5. Odds ratios for odds of employment for demographic, training, education, skills, social network, and psychosocial predictors.

	Men (n = 81)		Women (n = 106)
	Unadjusted OR (95% CI)	Adjusted OR (95% CI)	Unadjusted OR (95% CI)	Adjusted OR (95% CI)
Demographic
Marital status
Not married	1.00	1.00	1.00	1.00
Married	1.08 (0.17, 6.86)	1.81 (0.22, 15.0)	1.44 (0.36, 5.84)	1.02 (0.18, 5.67)
Number of living children
None	1.00	1.00	1.00	1.00
At least one	1.26 (0.39, 4.05)	1.40 (0.33, 5.84)	4.91 (0.61, 39.6)	3.57 (0.42, 30.3)
Education, training, and skills
Educational attainment
Less than high school	1.00	1.00	1.00	1.00
High school or more	0.55 (0.22, 1.37)	0.62 (0.22, 1.77)	0.96 (0.30, 3.08)	1.63 (0.42, 6.34)
Did not pass matric	1.00	1.00	1.00	1.00
Passed matric	0.90 (0.20, 4.14)	1.16 (0.24, 5.73)	0.85 (0.16, 4.65)	1.07 (0.17, 6.63)
No completed diplomas	1.00	1.00	1.00	1.00
Completed diplomas	0.77 (0.24, 2.51)	0.66 (0.16, 2.74)	0.97 (0.24, 3.83)	0.73 (0.13, 4.07)
No diplomas in progress	1.00	1.00	1.00	1.00
Diplomas in progress	0.38 (0.04, 3.60)	0.40 (0.04, 4.21)	0.51 (0.06, 4.28)	1.08 (0.11, 11.1)
No working skills/training	1.00	1.00	1.00	1.00
Have working skills/training	4.09 (1.35, 12.4)*	4.49 (1.32, 15.3)*	1.94 (0.59, 6.38)	1.89 (0.52, 6.92)
Social network
No relatives/friends working	1.00	1.00	1.00	1.00
Relatives/friends working	0.91 (0.32, 2.55)	1.31 (0.40, 4.28)	1.13 (0.32, 3.94)	1.09 (0.28, 4.18)
Psychosocial
Self-confidence in entrepreneurship	0.99 (0.96, 1.02)	0.99 (0.95, 1.02)	1.03 (0.98, 1.08)	1.04 (0.98, 1.10)
Empowerment attitude	0.97 (0.92, 1.03)	0.97 (0.92, 1.03)	1.00 (0.93, 1.08)	1.00 (0.93, 1.08)
Life dissatisfaction	1.04 (0.98, 1.09)	1.03 (0.97, 1.10)	1.13 (1.04, 1.23)**	1.13 (1.03,1.24)*
Control variables
Age		1.02 (0.79, 1.32)		1.26 (0.83, 1.91)
Asset quintile		1.23 (0.81, 1.85)		0.65 (0.40, 1.04)
Village		0.31 (0.09, 1.08)		0.62 (0.16, 2.38)

Notes: The control variable estimates are illustrative because they were similar across all tests. CI = confidence interval, OR = odds ratio.

*p-value < 0.05; **p-value < 0.01.

Although most of the independent variables did not have a significant association with employment, the magnitude and direction of some point estimates do warrant attention. The aOR for completing high school was above one for women (aOR: 1.6; 95% CI: 0.4, 6.3) but not for men (aOR: 0.6; CI: 0.2, 1.8), and women with skills and training also seemed to have higher odds of employment (aOR: 1.9; CI: 0.5, 6.9). Finally, the aOR was quite high for women who reported having children (aOR: 3.6; 95% CI: 0.4, 30.3), indicating that mothers in our sample may be more likely to work than childless women. However, it is important to note for all of these point estimates that the CIs were wide due to the small sample size.

3.8. Qualitative findings

Results from the IDIs provide context for the regression results. Overall, respondents felt there were many barriers to youth employment in Mpumalanga. The three most commonly mentioned barriers were lack of skills, lack of information on job openings, and an overall lack of jobs. In the words of a 20-year-old male respondent:

Lack of qualifications and experience, that’s why we are not getting jobs. They [employers] will tell you that they want 10 years’ experience or 5 years’ experience and a driver’s license, or a certain diploma or a degree, only to find that you don’t have any of these.

When asked about what employment opportunities existed for youth in the rural area, respondents consistently and spontaneously outlined different employment opportunities for men versus women. The most commonly listed forms of employment were construction (e.g. brickmaking) for men and hair braiding or paid domestic work for women. Other types of employment were occasionally listed without gender specifications, and these included selling things in the market, working in the tourism/game reserves, or doing odd jobs (e.g. fetching firewood). Because unemployment is very common among youth in this setting, we also asked the IDI respondents what youth generally do during the day. Most responses fell into one of the three categories, as outlined by this 23-year-old female respondent: ‘Some when they wake up they do women’s duties, some … go to sell their stuff and some when they wake up they bathe and go to the tavern [bar].’ Many respondents also said youth do nothing during the day or spend time with friends.

Given the apparent gender differences in employment opportunities, respondents were also asked whether they thought women had a harder time finding jobs compared with men. Most respondents felt that women had a harder time finding work and one respondent, a 21-year-old female, explained it in the following way:

Young women have a harder time than men because there are lots of jobs outside that need power and those jobs are for men. For example, a woman can’t pick up a brick to build a house. Women need easy jobs like working with a computer.

Other respondents said young women can have a harder time finding work because they have young children, lack skills, or there is stigma against women in many professions.

The key informant interviews provided helpful context for the youth comments in the IDIs. Notably, all of the key informants listed resources available for youth to help them in the school-to-work transition, whereas no youth we interviewed mentioned any such resources, even when asked. The resources listed by the key informants ranged from government agencies to non-profit organisations as well as private companies. The support these groups provide included: training, both formal (e.g. vocational diplomas, entrepreneurship development programmes) and informal (e.g. talks on CV writing); resources (e.g. computer labs, business development grants, scholarships); and activities (e.g. community events, after-school activities, youth-led education campaigns).

A few key informants mentioned that, although there may be resources in the urban areas in the province, rural youth may lack information about or transportation to these resources. In the words of one key informant: ‘Those opportunities are not situated where those people are … is there a newspaper in [rural areas], I don’t know?’ Only one organisation was identified as using local volunteer leaders to collect information on opportunities for youth and distribute them in rural areas. Further, there seems to be a lack of communication between the organisations, as expressed by one key informant: ‘I can’t even tell you where the [organisation] offices in Bushbuckridge are.’

4. Discussion

In this study we used both quantitative and qualitative methods to examine factors associated with youth employment in rural South Africa. Overall youth unemployment was very high and the data indicate that factors associated with youth employment differ for men and women. The national average of youth unemployment in South Africa (just over 50%) is significantly lower than the 75% unemployment in this sample but this higher estimate is in line with recent region-specific data from Mpumalanga (Yu, 2013). There were more than twice as many men employed in our sample as women, and the odds of employment were highest for men who reported having skills or training. Men were more likely to report having skills than women and women were more likely to pursue further education than men. Further education did not seem to increase the odds of employment for women.

Gender differences in the type of skills and training that men and women have in the study reflect the opportunities that are common in the area. If men are concentrated in construction jobs, it makes sense they would benefit from skills and training, most of which were in physical labour. If women are only considered eligible for more formal employment, this could explain why they are more likely to seek diplomas and why they appear to be more likely to find work if they have completed high school. However, if, as many of the IDI respondents asserted, construction jobs are more prevalent than formal jobs in the region and it is only appropriate for women to do the latter, this could explain why twice as many men in the sample were employed compared with women.

Although these data are cross-sectional, which prevents us from making causal inferences, the associations reported here lend themselves to informative interpretations. That training and skills were significantly associated with employment for men could indicate that learnerships, or paid training provided by employers, are effectively helping youth get jobs (Mayer, 2011). Alternatively, the significant relationship between working skills and employment could indicate that these employed youths received training on the job and now have working skills, changing the directionality. However, the most commonly cited employment barrier for youth was a lack of skills and experience, so it makes sense that youth who reported having skills were more likely to be employed.

For women, only greater life dissatisfaction was significantly associated with higher odds of employment. Although it is possible that people who are more dissatisfied may be more motivated to seek employment, it is not clear why this would apply only to women. Further, past research supports this finding. A 2012 report from the International Labour Office emphasises that employed youth in developing countries often fare worse than youth who are not in education, employment, or training because employed youth may have to work in order to survive and so in desperation take jobs that keep them stuck in a cycle of poverty (International Labour Office, 2012). This strain may be especially salient for the women in the sample because over 70% of them are mothers, and the majority are unmarried, and thus are burdened with providing for themselves and their children, while fewer than 30% of men reported having children. Indeed, mothers in our sample had nearly 3.5 times higher odds of employment compared with childless women.

The South African labour market is a difficult one for youth to enter. South Africa’s economy is shifting to value more skilled labour (Banerjee et al., 2008; Burns, Edwards, et al., 2010). Also, South African employers face high costs for layoffs due to labour regulations and therefore want to verify a worker’s productivity before hiring (Rankin et al., 2012; Yu, 2013). The youth interviewed in the IDIs repeatedly reiterated this point, saying that employers require training certificates and work experience. Given the shifting economic context in South Africa, it is fitting that the three most common resources provided by the organisations the key informants represented were: entrepreneurship development, computer access and training, and more formal skills training. This is in line with current best practices in youth employment promotion, although youth entrepreneurship programmes are probably not a silver bullet because youth ventures are higher risk, require capital investment that youth do not have, provide lower wages, and do not decrease the risk of future unemployment (Betcherman et al., 2007; Burns, Edwards, et al., 2010; International Labour Office, 2012). Also, given that financial literacy was very low in this sample, it appears that funding for entrepreneurial projects must be preceded by financial education.

Fundamentally, although the key informants identified many employment resources for youth, every single one of the youth in our sample said there were no employment resources for youth in their area. This demonstrates a clear need to connect more rural youth to local resources already available, most of which were within 45 minutes of where the youth lived. It is possible, however, that youth are aware of the resources but do not find them valuable, suggesting that an evaluation of local employment programmes could be useful.

The main limitations of this study stem from the relatively small, although random, sample and the cross-sectional design. This study was a pilot study intended for hypothesis and intervention generation. Also, the survey questions were adapted from many sources and some measures were not previously validated in the Southern African context. Finally, because our analyses did not include youth who had out-migrated for employment-related reasons, we were unable to study factors associated with finding employment outside the study area. However, our results are generalisable to the large youth population residing in the study area.

5. Conclusion

In this study we identified factors associated with youth employment in rural Mpumalanga, specifically training for men and possibly education for women. Our qualitative findings showed that improving youth awareness of and access to employment resources already available in the local area is a feasible next step. However, because many of these resources are focused on developing youth entrepreneurs, it is important these programmes are paired with financial education. Further, in an economy that is shifting to value more skilled labour, formal-sector job creation also remains a critical long-term priority, especially for women who have less access to informal jobs and for whom economic insecurity can increase vulnerability to HIV (Booysen & Summerton, 2002; Mpumalanga Provincial Government, 2011; Shisana et al., 2014). Increasing youth employment in rural South Africa would improve the local economy and community, and would also probably decrease women’s vulnerability to HIV (Booysen & Summerton, 2002).

Acknowledgements

The authors wish to thank all participants for contributing to this study. They also wish to acknowledge Rhian Twine, Floidy Wafanawaka, Amanda Selin, and the staff and fieldworkers of the Medical Research Council/Wits University Rural Public Health and Health Transitions Research Unit (Agincourt) at the University of Witwatersrand in South Africa for facilitating data collection.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This study was supported by a pilot grant from the Carolina Population Center (http://www.cpc.unc.edu). This research also received support from the Population Research Training grant [T32 HD007168] and the Population Research Infrastructure Program [R24 HD050924] awarded to the Carolina Population Center at The University of North Carolina at Chapel Hill by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (to AW, MR). Additionally, this research received support from a Predoctoral Ruth L Kirschstein National Research Service Award from the National Institute on Drug Abuse [F31 DA036961] (to AW). The Agincourt HDSS is supported by the UK Wellcome Trust [Grants 058893/Z/99/A, 069683/Z/02/Z, 085477/Z/08/Z] and the University of the Witwatersrand and Medical Research Council, South Africa. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.