Returns to Educational Attainment in Urban Ghana: The Role of Job-to-Job Transition

Abstract

It is well established that additional educational attainment above lower secondary yields higher returns. We find that these returns are mediated by the job transitions, especially for individuals with vocational education. We then study whether the job-to-job transition explains the differences in the returns to educational attainment and find evidence of positive returns of the movement from employee to self-employed. We show that successful employee voluntary move to self-employment to maximize their earnings, while early spell of self-employment tends to have a lasting negative effect on earnings. Our analysis partially explains why self-employment is the top final destination for many workers in developing countries.

Keywords

Returns to educational attainment job-to-job transition school-to-job transition vocational training earnings

I Introduction

Until recently, little theoretical and empirical effort has been directed towards understanding job transitions in Africa, or how such transitions affect the relationship between education and earnings. Though African governments and policymakers have been very proactive in improving skills for industrial development and in promoting employment for skilled yout,¹ research in Africa has been lagging in informing policy on the potential returns to educational attainment and on facilitating school-to-work transition. Other important aspects of labour market and education, especially school-to-work transition and employment dynamics, are poorly understood. One reason for this deficiency is the lack of longitudinal dataset that traces individuals from education to the current job. Until recently, empirical research on education and employment relationship in Africa has been characterized by the returns of years of schooling (Montenegro and Patrinos, 2014). Despite the number of studies that estimates this relationship, it is unclear as to why we observe heterogeneous returns to schooling, and much less is known about how education affects the occupation status in the first job as well as the entire career.

We extend basic research on the returns to education in Africa to include the concept of job-to-job transition and report new findings on the effect of first job and job-to-job transition patterns in urban Ghana. We were able to do this for the first time, as we had access to a unique and large representative survey, that is, the World Bank Skills Towards Employability and Productivity (STEP), which contains rich set of information on first job after education and current job. We report several findings. We discover that there are systematic differences in the returns between those who transit from school to job and those who transit from one job to another. Returns to education tend to be higher in the sample of school-to-job transition. Building on the findings that the returns to education differ by the type of job transition, we develop four patterns on job-to-job transitions and estimate a mediation model in order to assess whether the transition can be a mechanism that explains differences in the returns to educational attainment. We establish that job-to-job transition mediates the relation between education and earnings at mid-career, especially for individuals with vocational education. We also find that the transition from employee to self-employed on average yields higher returns. Our findings show that the first job as self-employed may have lifetime lasting negative effect on later earnings.

Our article adds to the nascent literature on employment dynamics in developing countries. Our contribution, relative to the existing work, is to construct employment dynamics of school-to-job and job-to-job transition from a cross-sectional data and discipline the quantitative analysis in order to examine job transitions as the mechanism through which education may influence earnings. Though our econometric estimations are limited by the nature of the data, our results are intuitive and replicate several priors that we observe in the labour market in Africa and elsewhere. Our findings, thus, partially explain the increasing self-selection into self-employment (Maloney, 2004; Perry et al., 2007) and corroborate with the finding of Hemmer and Mannel (1989) who argued that self-employment is a sector to enter once skills and connections have been accumulated through several years of working. Our findings that self-employment in the first job can have a lifetime negative effect on earnings is consistent with existing findings in the literature of scaring effect of unemployment (Arulampalam, 2001; Ellwood, 1982; Gregory and Jukes, 2001; Ruhm, 1991) and signalling theories (Scherer, 2004, De Cuyper et al., 2009) and corroborates with previous findings equalizing self-employment to disguised unemployment in Africa (Earle and Sakova, 2000; Ranis and Gollin, 2013).

Our article builds on the extensive literature that examines the returns to education in developing countries; for instance, Psacharopoulos and Patrinos (2004) have shown that the returns to investment in education are higher for every year of schooling, and that tertiary education has the highest rates of return. Evidence also suggests that global trend in the returns to education is decreasing over time but there are mixed patterns across different educational attainment within or across countries; for instance, Montenegro and Patrinos (2014) show that returns to educational attainment has been increasing in Ghana, while in Madagascar, the returns to primary, secondary, and tertiary education have been decreasing, increasing, and fluctuating, respectively. In our article, we look at a wide array of educational attainment, with a clear difference between general and vocational education. Our results support the findings in Altonji (1993) that shows a non-linearity in the returns to educational attainment. We document that the returns to tertiary general education are larger than tertiary vocational, while at the lower level of attainment, upper secondary vocational education yields higher returns than upper secondary general.

This article is organized as follows: in the next section, we describe the context and the data. The third section discusses our strategy for estimating the returns of additional educational attainment above lower secondary. The fourth section presents earning gaps estimates and the fifth section discusses our main mechanism. The sixth section concludes.

II Context and Data

1 Context

Our analysis focuses on Ghana because the country provides a perfect setting to study the returns to educational attainment and how job mobility serves as a mechanism to explain the effect of education on earnings. Similar to many African countries, Ghana is endowed with large reserves of natural resources, and for decades, it has heavily relied on cocoa and gold to function its economy. Unlike many African countries, Ghana has been very successful in developing democratic institutions during the early 1990s and, thus, maintaining political stability. These factors have permitted Ghana to spur economic growth and reach middle-income status. Specifically, Ghana launched structural reforms in 1984 that sustained the economy for more than two decades, as shown in Figure 1. According to Osei and Jedwab (2017), the structural adjustment reforms were characterized by a reduction in public expenditure, privatization of state enterprises, currency devaluation, and the collapse of the urban sector. Nevertheless, since 2001, Ghana economy has grown by an annual average of 6.4% with a pick of 14% in 2011 due to the launch of commercial oil production. As can be seen, this significant level of accomplishment was accelerated by not only a favourable external environment but also the improvement in the domestic business environment.

Figure 1.

Source: Penn World Tables (version 10) and Author’s calculation.

What is important for this article is that during these years, economic growth did not generate manufacturing employment, but led to a dramatic expansion of the service sector and contraction of the government sector. More specifically, data reported by Osei and Jedwab (2017) indicate that employment shares of industry decreased from 62% in 1992 to 40% in 2010, while that of the service sector increased from 25% to 41%. When the government sector contracted, the low productive service sector became the employer of the last resort. According to Baah-Boateng (2016), unemployment of the youth aged 15–24 has increased from 11.4% in 2010 to 13.5% in 2013, with higher increase in urban areas (from 16.2% to 20%), whereas between 2010 and 2013, unemployment declined by 3.1%. The structural change has brought new earnings opportunities for all categories of workers, with the highest increase to self-employment (off-farm), who recorded an average earnings growth of 80% between 2005 and 2012. Public wage increased by about 65% and private wage by 40% (Honorati and Silva, 2016). Nevertheless, the underlying development of the informal sector and self-employment coincided with large income gaps and a high ratio of jobless growth.

Finally, an additional indicator of the strength of the Ghanaian case is that education coverage and access have achieved marked progress during the last years: the country has achieved near universal access to primary education. As part of its strong commitment in favour of skills development for employability and employment, the Ghanaian government has been developing a new national education strategic plan that would span from 2016 to 2030 (a detailed description of the TVET reform is given in Appendix 1). Previous institutional reforms for skills development has been documented in Darvas and Palmer (2014). Recent data indicate that of the 8.9 million of students in the Ghanaian education system, basic school accounts for 86.9%, while senior schools and tertiary education account for 9.1% and 3.5%, respectively. The share of Technical and Vocational Education and Training (TVET) is only 0.5%. The low enrolment in TVET shows that only a small proportion of the school population acquires skills for meaningful employment. There are many reasons for the low shares of TVET. One of them is that informal apprenticeship is historically important in Ghana, as it equips many more workers with skills than do formal public TVET. Darvas and Palmer (2014) claim that the informal apprenticeship system trains more than 440,000 youth at one time, which is about 10 times higher than the formal public TVET. A significant analysis and discussion on education system and labour market in Ghana was recently presented by Darvas et al. (2017) who reported fewer than 10% of workers with technical and vocational skills acquired their skills through public TVET institutions. On a final note, we highlight that the lack of ability to find formal jobs limits many graduates from basic and senior schools to self-employment or jobs in the informal sector.

2 Data

Our analysis is based on the STEP Skills Measurement Survey for individuals living in urban Ghana, aged 17–64 and not currently in formal education. Workers in the armed forces are also excluded in the analysis. The analytical sample includes about 2,480 observations. Table 1 presents summary statistics where the first two columns present mean and standard deviation for the analytic sample. We also present the same statistics for the school-to-job transition sample and the job-to-job transition sample, since we are interested in the subsample of workers who are currently in their first job and those who have changed jobs.² Table 1, panel A, reveals that the percentage of no formal education, primary, and lower secondary in the job-to-job transition is higher than the school-to-job transition. Inversely, the school-to-job sample is somehow highly educated than the job-to-job sample and the analytic sample. Only the percentage of upper secondary vocational closely matches in the three samples.

Table 1.
Descriptive Statistics

Analytical sample
School to job transition
Job to job transition

Mean Std. Dev. Obs Mean Std. Dev. Obs Mean Std. Dev. Obs

Panel A. Education variables

No formal education 0.23 0.42 2480 0.03 0.18 239 0.12 0.33 126

Primary 0.11 0.31 2480 0.09 0.28 239 0.12 0.32 126

Lower secondary 0.36 0.48 2480 0.21 0.41 239 0.44 0.50 126

Upper secondary Voc 0.04 0.21 2480 0.06 0.24 239 0.05 0.23 126

Upper secondary Gen 0.15 0.36 2480 0.23 0.42 239 0.17 0.37 126

Tertiary Voc 0.05 0.22 2480 0.13 0.34 239 0.05 0.22 126

Tertiary Gen 0.05 0.22 2480 0.24 0.43 239 0.05 0.21 126

Panel B. Labor market variables

employed 0.82 0.38 2474 1.00 0.00 239 1.00 0.00 126

Employee 0.34 0.47 2029 0.54 0.50 239 0.38 0.48 126

Self employed 0.62 0.49 2029 0.36 0.48 239 0.62 0.48 126

Formal 0.89 0.31 2480 0.77 0.42 239 0.87 0.33 126

Informal 0.85 0.36 2026 0.65 0.48 239 0.83 0.38 126

Earnings (hourly) 2.11 2.82 1831 3.23 3.39 199 2.09 2.75 118

Experience 16.35 12.02 2187 10.13 10.12 236 18.50 11.41 126

Employee (First job) 0.55 0.50 1824 0.68 0.47 203 0.57 0.50 126

Self employed (first job) 0.38 0.49 1824 0.28 0.45 203 0.43 0.50 126

Employee to employee 0.32 0.47 1265 0.32 0.47 126

Employee to self-employed 0.25 0.43 1265 0.25 0.43 126

Self-employed to employee 0.06 0.24 1265 0.06 0.24 126

Self-employed to self-employed 0.37 0.48 1265 0.37 0.48 126

Panel C. Other variables

Age 35.44 10.80 2480 32.14 10.45 239 36.11 10.75 126

Household size at age of 12 5.48 4.05 2480 5.45 4.51 239 5.62 3.87 126

Has worked at age of 15 0.21 0.41 2468 0.17 0.37 239 0.20 0.40 126

Economic status at age of 15 4.89 2.11 2468 4.95 1.88 239 4.98 2.11 126

	Analytical sample	School to job transition	Job to job transition
Panel A. Education variables
No formal education	0.23	0.42	2480	0.03	0.18	239	0.12	0.33	126
Primary	0.11	0.31	2480	0.09	0.28	239	0.12	0.32	126
Lower secondary	0.36	0.48	2480	0.21	0.41	239	0.44	0.50	126
Upper secondary Voc	0.04	0.21	2480	0.06	0.24	239	0.05	0.23	126
Upper secondary Gen	0.15	0.36	2480	0.23	0.42	239	0.17	0.37	126
Tertiary Voc	0.05	0.22	2480	0.13	0.34	239	0.05	0.22	126
Tertiary Gen	0.05	0.22	2480	0.24	0.43	239	0.05	0.21	126
Panel B. Labor market variables
employed	0.82	0.38	2474	1.00	0.00	239	1.00	0.00	126
Employee	0.34	0.47	2029	0.54	0.50	239	0.38	0.48	126
Self employed	0.62	0.49	2029	0.36	0.48	239	0.62	0.48	126
Formal	0.89	0.31	2480	0.77	0.42	239	0.87	0.33	126
Informal	0.85	0.36	2026	0.65	0.48	239	0.83	0.38	126
Earnings (hourly)	2.11	2.82	1831	3.23	3.39	199	2.09	2.75	118
Experience	16.35	12.02	2187	10.13	10.12	236	18.50	11.41	126
Employee (First job)	0.55	0.50	1824	0.68	0.47	203	0.57	0.50	126
Self employed (first job)	0.38	0.49	1824	0.28	0.45	203	0.43	0.50	126
Employee to employee	0.32	0.47	1265				0.32	0.47	126
Employee to self-employed	0.25	0.43	1265				0.25	0.43	126
Self-employed to employee	0.06	0.24	1265				0.06	0.24	126
Self-employed to self-employed	0.37	0.48	1265				0.37	0.48	126
Panel C. Other variables
Age	35.44	10.80	2480	32.14	10.45	239	36.11	10.75	126
Household size at age of 12	5.48	4.05	2480	5.45	4.51	239	5.62	3.87	126
Has worked at age of 15	0.21	0.41	2468	0.17	0.37	239	0.20	0.40	126
Economic status at age of 15	4.89	2.11	2468	4.95	1.88	239	4.98	2.11	126

Source: Author’s calculations using STEP data.

Panel B, which reports labour market statistics, shows that urban Ghana has a high employment rate, with the majority of workers being self-employed. However, a comparison of the subsamples shows some striking features. The school-to-job sample has more employees than self-employed, that is, 54% as compared to 36%. In the job-to-job sample, however, employee represents only 38% while self-employed accounts for about 62%. The job-to-job sample also has the highest share of informal sector than the school-to-job sample, which corroborates with the findings on self-employed. It is interesting to note that the school-to-job sample has substantially lower working experience and working hours than the job-to-job transition sample, but their hourly earnings is relatively higher.

Panel B also provides means of the job-to-job transition patterns where we identified four transitions: employee to employee, employee to self-employed, self-employed to employee, and self-employed to self-employed. Workers are more likely to remain in their first occupation than to move. Self-employment to self-employment represents 37% of the transitions, followed by employee to employee (32%) and employee to self-employment (37%). The flow from the self-employment to employee represents roughly 6%. Statistics for household characteristics at age of 15 are summarized in panel C of Table 1. We observe no substantial difference among the samples, expect that 20% of workers in the job-to-job transition sample have worked at the age of 15 against 17% in the sample of school-to-job transition.

III Empirical Strategy

We follow Mincer (1974) in estimating the following specification:

Y_{i} = α + β_{1} A_{i} + β_{2} X_{i} + β_{3} X_{i}^{2} + ε_{i},

(1)

where Y_i represents the log of hourly earnings of individual i; A_i represents schooling; X_i is quadratic function for the potential work experience, which is measured as age—years of education—6; and ε_i is the error term. In a standard Mincer style human capital earnings equation, the variable educational attainment is measured by the years of schooling. However, the aim of our article is to estimate the effect on earnings of obtaining an additional degree above lower secondary education. We, therefore, consider an alternative specification which allows us to introduce dummy variables. Our model is

Y_{i} = α + {\sum^{​}}_{6}^{j = 1} β_{1 j} E_{j i} + β_{2} X_{i} + β_{3} X_{i}^{2} + ε_{i},

(2)

where E_j represents the dummy variable indicating the highest educational level j. The STEP survey provides information on the highest level of educational attainment compatible with the International Standard Classification of Education (ISCED) 1997 levels (UNESCO, 2012). We identify the highest educational attainment for no formal education, primary, upper secondary vocational, upper secondary general, tertiary vocational, and tertiary general. Our reference category is lower secondary education since we are interested in measuring the returns of additional educational attainment above lower secondary.

The above equations are generally estimated using OLS. However, correct identification of β_1j depends on two key assumptions. First, ‘all individuals have the same ability and education or training’ is the only factor responsible of their differences. However, in reality, this assumption is difficult to meet due to the variety of background characteristic of individuals. The violation is this assumption, therefore, leads to the endogeneity between schooling and earnings and biased OLS estimates (Card, 1995; Griliches, 1977). To partially tackle this issue, we take advantage of our dataset which provides a rich set of control variables on household observable characteristics and proxy for underlying abilities that might be correlated with the likelihood to complete education. These variables include household size at age 12, economic situation at age 15, and working experience at age 15. The second concern is the sample selection bias. In our data, roughly 82% of individuals were employed, while 7% are unemployed and 12% inactive. Given our interest in simultaneously modelling the returns to education and the determinant of employment, we estimate Equation (2) using the Heckman (1974) two-step procedure to address the sample selection bias.

IV Returns to Education

1 Baseline Regression

Table 2 reports the Heckman second stage regression results of Equation (2), focusing on education variable. Two findings standout from the comparison of the estimated returns across the education achievement groups against lower secondary education. First, we find that additional degree above lower secondary has higher returns, except for the upper secondary general whose gap is positive but not statistically significant; for instance, compared to lower secondary, the returns to upper secondary vocational degree are 26.5%, while the returns to upper secondary general are very low and not statistically significant. Compared to general education, TVET is expected to equip learners with specific skills needed at the workplace. This finding provides support for exposing general secondary education to vocational and practical subjects. Second, the gaps in the returns to educational attainment are much stronger for tertiary education degree holders. This indicates the non-linearities of the returns to education. Relative to lower secondary education, the returns to tertiary general education are estimated at 59.7% while the returns of tertiary general degree are quite high, at 129.3%.³ A comparison of the two set of results reveals that at the lower level of educational attainment, vocational education yields higher returns than general education, whereas at tertiary levels, general education generates stronger returns. More importantly, the large difference between upper secondary vocational and tertiary vocational degree may help us to understand the importance of expanding TVET learning opportunities vertically and creating seamless pathways to higher levels of attainment.

Table 2.
Additional Returns Above Lower Secondary

(1)

Education (Reference is Lower secondary)

No formal education –0.142

(0.0963)

Primary –0.0535

(0.107)

Upper secondary Voc 0.265*

(0.137)

Upper secondary Gen 0.0677

(0.0918)

Tertiary Voc 0.597*

(0.148)

Tertiary Gen 1.293*

(0.146)

Observations 1,676

	(1)
Education (Reference is Lower secondary)
No formal education	–0.142
	(0.0963)
Primary	–0.0535
	(0.107)
Upper secondary Voc	0.265*
	(0.137)
Upper secondary Gen	0.0677
	(0.0918)
Tertiary Voc	0.597***
	(0.148)
Tertiary Gen	1.293***
	(0.146)
Observations	1,676

Source: Author’s calculations using STEP data.

Note: This table reports marginal effects from the wage model with the Heckman selection. Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05 and * p < 0.1.

2 Results across subsamples

Next, we explore heterogeneity in the returns to education by repeating the previous analysis for various subgroups. One natural question in regard to the effect of education on earnings is whether the returns of educational attainment in the school-to-job sample differs from the job-to-job transition sample. In Table 3, we present estimates for the school-to-job sample in column 1 and the job-to-job transition sample in column 2. Our findings indicate that the returns to upper secondary vocational and upper secondary general degrees—relative to lower secondary degree—are positive but not statistically significant among the school-to-job sample.⁴ However, tertiary education generates stronger and positive returns. Consistent with the previous findings, we find that the returns to tertiary general are higher than the tertiary vocational. The results in column 1 show that the returns to tertiary general degree are 178.6%, while we estimate the returns to tertiary vocational at 80.7%. When we look at the sample of job-to-job transition (column 2), we find that the return to upper secondary vocational degree becomes slightly larger and statistically significant while the returns of upper secondary degree remain not significant but get smaller. This implies that upper secondary vocational education is highly rewarded among workers who changed their job or sector. Another important finding is that the marginal effects of tertiary vocational and general education are smaller in the sample of job-to-job transition, indicating that the returns to tertiary education—relative to lower education—are stronger for those who remain in the same occupation. Comparing these two results, we see, for instance, that the differential between tertiary general education and lower secondary education is 178.6% in the sample of school-to-job transition and 106.2% for job-to-job transition. Taken together, these results demonstrate that higher educational attainment is strongly rewarded for those who find a stable first job, and that that upper secondary vocational tends to produce higher returns than upper secondary general. The finding of larger effect of upper secondary vocational education in the sample of job-to-job transition can be explained by job matching, as a result of job-to-job transition, or by skills demand due to technological change. TVET provides industry-specific skills for workers with lower general education, suggesting higher returns for those who find better match for their skills. Since our sample of job-to-job transition is older than in the school-to-job transition, a possible interpretation can be that lower supply of the upper secondary vocational can result in higher returns for those with higher mobility. We will discuss about this mechanism in the next section.

Table 3
Additional Returns Above Lower Secondary—Results by Subsamples

School-to-job Job-to-job

subsample Subsample

(1) (2)

Education (Reference is Lower secondary)

No formal education 0.369 –0.092

(0.488) (0.114)

Primary –0.019 –0.109

(0.409) (0.122)

Upper secondary Voc 0.308 0.317*

(0.303) (0.163)

Upper secondary Gen 0.320 0.131

(0.287) (0.103)

Tertiary Voc 0.807 0.532*

(0.330) (0.179)

Tertiary Gen 1.786* 1.062*

(0.329) (0.200)

Observations 197 987

	School-to-job	Job-to-job
subsample	Subsample
(1)	(2)
Education (Reference is Lower secondary)
No formal education	0.369	–0.092
(0.488)	(0.114)
Primary	–0.019	–0.109
(0.409)	(0.122)
Upper secondary Voc	0.308	0.317*
(0.303)	(0.163)
Upper secondary Gen	0.320	0.131
(0.287)	(0.103)
Tertiary Voc	0.807**	0.532***
(0.330)	(0.179)
Tertiary Gen	1.786***	1.062***
(0.329)	(0.200)
Observations	197	987

Source: Author’s calculations using STEP data.

Note: This table reports marginal effects from the wage model. Robust standard errors in parentheses. *** indicates p < 0.01, ** indicates p < 0.05 and * p < 0.1.

V Contribution of Job-to-Job Transition

The previous section provides a range of evidence that additional educational attainment above lower secondary has positive effect on earnings. We also find that these additional returns are generally stronger for workers in the school-to-job sample than the job-to-job sample, and upper secondary vocational is highly rewarded in the job-to-job sample, as compared to upper secondary general. This section examines the mechanism that could potentially explain these results. We are particularly interested to find out why the additional returns to educational attainment above lower secondary education vary.

There are multiple channels through which education could affect earnings. The most obvious is that education or training can make workers more productive through improved cognitive, behavioural, and other skills, thereby increasing their earnings. We propose to study another plausible mechanism, that is, job-to-job transitions. To do so, we construct four transition patterns using information on the first job and the current job for worker with more than one job in their career. Our data allow us to capture the first job in terms of occupations, namely wage employee and self-employed. For the current job, these occupations can be matched with the information in the first job to study the transition from self-employment to wage employee and vice versa. In order to better capture these heterogeneities, we create the following job transition groups from first job to the current job: employee to employee; employee to self-employed; self-employed to employee; and self-employed to self-employed. Figure 2 presents the distribution of log earnings for four transition groups. The four distributions do not match up exactly, suggesting that job transition may have a differential effect on earnings and, therefore, serve as a channel through which educational attainment affects earnings. We investigate this mechanism into two steps.

Figure 2.

Source: Author’s calculations using STEP data.

1 Establishing our mechanism

To estimate the extent to which job-to-job transitions can mediate between education and earning, we follow two steps: first, we re-estimate the educational returns for the school-to-job and job-to-job transition sample, which serves as the benchmark for comparison. In the second step, we re-estimate the same model as in step 1, including the job-to-job transition as an additional control variable. The estimated educational attainments coefficients from the second step models show the impact of education controlling for the job-to-job transition. Following Osterman (2006), we then test whether the inclusion of job-to-job transition variable changes the magnitude or statistical significance of the educational attainment variable. If it does, then one can conclude that the job-to-job transition serves as a mediating factor or a channel by which education and earnings are linked.

We present the findings in Table 4. The results are very interesting. First, we discover that the job-to-job transitions have differential effect on earnings. Compared to the worker without job transitions, we find that the employee-to-self-employed transition yields higher earnings, whereas the earnings for the other job transitions are negative signs. The findings indicate that the employee-to-self-employed transition leads to 18.4 percentage point higher earnings than those who have not changed job, whereas the employee-employee and self-employed-to-employee transitions lead to 33.2 and 31.8 percentage point lower earnings. The earning gap for self-employed-to-self-employed transition is marginal and not statistically different from zero. The significance of the marginal effects for the job-to-job transition variables indicates the importance of job-to-job transition variable in explaining the differences in earnings in the mid-career. Second, our result shows that the magnitude of the marginal effect of education attainment changes visibly compared to column 1. This suggests that the job-to-job transition serves as a mediator between education and earnings and has a differential effect on the educational attainments. One unanticipated finding is that job-to-job transitions mostly mediate vocational education, while the mediation for general education is quite small; for instance, we find that 14.2% of the upper secondary vocational returns is due to job transitions. On the other hand, job transitions mediate only 5.37% for upper secondary general and 9.50% for tertiary general.

Table 4
Exploring Possible Mechanism

Reference (1) Controls for Job-to-job transition (2)

Education (Reference is Lower secondary)

No formal education –0.096 –0.100

(0.091) (0.077)

Primary –0.104 –0.092

(0.123) (0.114)

Upper secondary Voc 0.331* 0.284*

(0.074) (0.092)

Upper secondary Gen 0.149 0.141

(0.131) (0.141)

Tertiary Voc 0.587* 0.514*

(0.107) (0.123)

Tertiary Gen 1.316* 1.191*

(0.067) (0.090)

Transition (Reference is School-to-job)

Employee to Employee –0.332***

(0.096)

Employee to Self-employed 0.184*

(0.098)

Self-employed to Employee –0.318**

(0.142)

Self-employed to Self-employed –0.031

(0.080)

Observations 1,184 1,184

	Reference (1)	Controls for Job-to-job transition (2)
Education (Reference is Lower secondary)
No formal education	–0.096	–0.100
(0.091)	(0.077)
Primary	–0.104	–0.092
	(0.123)	(0.114)
Upper secondary Voc	0.331***	0.284***
(0.074)	(0.092)
Upper secondary Gen	0.149	0.141
(0.131)	(0.141)
Tertiary Voc	0.587***	0.514***
	(0.107)	(0.123)
Tertiary Gen	1.316***	1.191***
	(0.067)	(0.090)
Transition (Reference is School-to-job)
Employee to Employee		–0.332***
	(0.096)
Employee to Self-employed		0.184*
	(0.098)
Self-employed to Employee		–0.318**
	(0.142)
Self-employed to Self-employed		–0.031
	(0.080)
Observations	1,184	1,184

Source: Author’s calculations using STEP data.

Note: This table reports marginal effects from the wage model with the Heckman selection. Robust standard errors in parentheses. *** p < 0.01, ** indicates p < 0.05 and * p < 0.1.

2 Robustness analysis

The results in the previous analysis suggest that the returns to additional educational attainment above lower secondary are higher and the gaps increases at tertiary level. We also found that job-to-job transition mediates the relation between education and earnings. Our main argument is that education causes higher wages in mid-career via first job and job-to-job transition. While the results of the direct link between education and earnings are consistent with our prior results, there are some other alternative mechanisms that might explain to differential in the returns to education. This section considers other possible explanations of the results.

Structural Reforms

In 1984, Ghana launched structural reforms that sustained the economy for more than two decades, as shown in Figure 1. Between 1984 and 2010, Ghana’s average growth rate was 5.2%, which favoured Ghana to attain a middle-income status in 2007. This implies that the economic growth after structural reform could have had a signalling effect that higher educational attainment can be highly rewarded in the years to come. For instance, government measures of austerity to reduce public deficit led to a severe drop of the income in the 1970s and early 1980s and nourished the informal sector. Second, there is a risk that return to education in some sector may decrease due to technological change and economic reform.

Motivated by Figure 1, we identified two cohorts of workers who entered the job market before and after 1984. We find that 59.40% of our sample started to work after 1984 while 40.60 entered before 1984. Then we investigate whether the difference between the two cohorts can be reconciled with the educational earnings gaps observed in the job-to-job transition sample. For our analysis to be valid, the cohort of those who entered job market after 1984 should have higher educational attainment. This assumption is validated when we look at Table 5. Moreover, we observe that the gaps in educational attainment between the two cohorts are the highest in general secondary education. The variable cohort may, thus, be a valid indicator for signalling future higher returns if it is positively correlated with earnings and affects the magnitude of education.

Table 5
Cohort in Labor Market

Education Cohort of 1984 in labour market

Before 1984 After 1984 Total

No formal education 67.31 32.69 100

Primary 47.79 52.21 100

Lower secondary 37.89 62.11 100

Upper secondary Voc 26.36 73.64 100

Upper secondary Gen 16.23 83.77 100

Tertiary Voc 29.01 70.99 100

Tertiary Gen 22.83 77.17 100

Total 40.6 59.4 100

Education	Cohort of 1984 in labour market
No formal education	67.31	32.69	100
Primary	47.79	52.21	100
Lower secondary	37.89	62.11	100
Upper secondary Voc	26.36	73.64	100
Upper secondary Gen	16.23	83.77	100
Tertiary Voc	29.01	70.99	100
Tertiary Gen	22.83	77.17	100
Total	40.6	59.4	100

Source: Author’s calculations using STEP data.

Column 2 of Table 6 presents the results from a specification which includes the cohort variable as an additional explanatory variable. We find that the cohort variable is not significant and surprisingly negative. The marginal effects of educational attainment are virtually unaffected when controlling for cohort difference. Our results do not appear consistent with this explanation for two reasons: contraction of formal employment due to structural reform and proliferation of the informal sector.

Table 6

Alternative Mechanism

	Reference (1)	Controls for Cohort 1984 (2)	Controls for Undereducated (3)
Education (Reference is Lower secondary)
No formal education	–0.096	–0.099	–0.092
	(0.091)	(0.092)	(0.094)
Primary	–0.104	–0.108	–0.098
	(0.123)	(0.122)	(0.117)
Upper secondary Voc	0.331***	0.332***	0.328***
	(0.074)	(0.072)	(0.073)
Upper secondary Gen	0.149	0.147	0.145
	(0.131)	(0.131)	(0.128)
Tertiary Voc	0.587***	0.577***	0.580***
	(0.107)	(0.109)	(0.106)
Tertiary Gen	1.316***	1.299***	1.308***
	(0.067)	(0.069)	(0.065)
Cohort of 1984		–0.068
		(0.095)
Undereducated			–0.031
			(0.052)
Observations	1,184	1,184	1,184

Source: Author’s calculations using STEP data.

Note: This table reports marginal effects from the wage model with the Heckman selection. Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05 and * p < 0.1.

Mismatch

The quality of matching skills with job could be a reason for higher returns for worker with educational attainment above lower secondary. Tertiary educated workers could have been especially likely to find well suited jobs for their skills, as quality of matching is better in higher education. Fortunately, the STEP data provide information to measure if worker’s skills match with their current job. Skill mismatch is based on self-report on the difference between education and education needed for the job. Workers were asked ‘what minimum level of formal education do you think would be required before someone would be able to carry out this work?’. We use this input to classify workers as ‘undereducated’ if their education level is lower than required for the job.

For a mismatch feature to explain a part of the returns to education, two things must be true. First, there must be an inverse relationship between educational attainment and the likelihood to mismatch. Second, the mismatch variable must be significantly associated with earnings. Evidence on the first criteria is presented in Table 7, which indicates the share of underemployed decreases with higher educational attainment. To get a clearer picture of the second hypothesis, we repeat the estimation of our previous analysis using undereducated as an additional control variable. Column 3 of Table 6 reports the estimated results and shows that skills mismatch has no significant effect on earnings differentials. More importantly, the coefficients on the education variable are less affected with this control variable included compared to job-to-job transition. Therefore, skills mismatch, at least in the sample, does not appear to play a mediating role between educational attainment and earnings.

Table 7

Labour Mismatch

Education	Undereducated
Education	No	Yes	Total
No formal education	87.81	12.19	100
Primary	75.74	24.26	100
Lower secondary	88.9	11.1	100
Upper secondary Voc	93.64	6.36	100
Upper secondary Gen	97.12	2.88	100
Tertiary Voc	100	0	100
Tertiary Gen	100	0	100
Total	89.84	10.16	100

Source: Author’s calculations using STEP data

VI Relation Between Education and Job-to-Job Transition

In this section, we explore the how education affects employment dynamics via two main mechanisms, namely: (a) the time and the type of the first job; (b) the transition from first job to other jobs. We hypothesize that, by reducing the time to get a find job, education attainment could have a positive effect on wage employment. In a similar way, educational attainment could favour the movement or transition to a sector/occupation where expected returns are higher. We begin by empirically exploring the effect of education on the first job.

1 Education and first job

One way in which education could lead to higher earnings is via the first job. To estimate the effect of educational attainment on the first job, we undertake two exercises. First, we look at the probability to obtain the first job as an employee or to work as a self-employed. Second, we estimate how educational attainment affects the time to obtain the first job. For the two exercises, we perform a series of regressions in which the covariates variables are the quadratic approximate age at the time of the first job, gender, father’s and mother’s education, socio-economic status at age 15, household size at age 15, a dummy of working experience before 15, and a dummy of regions.

We present the results in Table 8. Column 1 corresponds to the impact of education on the first job as wage employee, while column 2 focuses on the impact on self-employment. The estimates suggest that additional educational attainment above lower secondary increases the probability to be an employee, while it reduces the probability to be self-employed. As expected, tertiary education has the strongest probability to obtain a wage employment, relative to lower secondary education. It is followed by tertiary vocational, upper secondary, and upper secondary vocational. In column 3, we report the effect of education on the time it took to find the first job. The results confirm that additional educational attainment above lower secondary decreases the time to obtain the first job, and the magnitudes are larger for higher attainment. The estimates suggest that upper secondary vocational education reduces the time to get the first job by 5.9 percentage point, relative to lower secondary. To reconcile the two findings, we re-estimate in column 4 the regression model on the time to obtain the first job in which we add a dummy variable of the being self-employed in the first job. We observe that the variable self-employment in the first job enters the model significantly and positively, suggesting that the transition from school to self-employment requires more time than the transition to wage employment for individuals with the same characteristics. One possible explanation for this finding is that self-employment requires initial capital, and time to obtain credit, to find a workshop, as well as to build its own network. Another explanation is that an individual who could not find a first job as wage worker tends to engage in self-employment because there is no unemployment benefit. This evidence confirms previous findings that have equated self-employed to a disguised unemployment, as a large fraction of self-employed have shown to have similar characteristics as unpaid family worker and unemployed (Earle and Sakova, 2000).

Table 8.
Education and First Job

Employee (1) Sefl-employed (2) Time to obtain the first job (3) Time to obtain the first job (4)

Education (Reference is Lower secondary)

No formal education –0.178 –0.0180 1.131 1.204

(0.119) (0.118) (1.785) (1.787)

Primary –0.134 0.132 –3.203 –3.294

(0.111) (0.112) (1.465) (1.456)

Upper secondary Voc 0.512* –0.401 –5.956* –5.437*

(0.174) (0.171) (1.778) (1.807)

Upper secondary Gen 0.720* –0.724* –5.365* –4.663*

(0.115) (0.119) (1.242) (1.268)

Tertiary Voc 0.896* –0.895* –9.578* –8 794*

(0.182) (0.193) (1.473) (1.499)

Tertiary Gen 1 921* –1.782* –11.68* –10.72*

(0.350) (0.353) (1.267) (1.315)

Sefl-employed 2.896***

(1.105)

Observations 1,485 1,485 1,410 1,407

R-squared 0.123 0.127

	Employee (1)	Sefl-employed (2)	Time to obtain the first job (3)	Time to obtain the first job (4)
Education (Reference is Lower secondary)
No formal education	–0.178	–0.0180	1.131	1.204
	(0.119)	(0.118)	(1.785)	(1.787)
Primary	–0.134	0.132	–3.203**	–3.294**
	(0.111)	(0.112)	(1.465)	(1.456)
Upper secondary Voc	0.512***	–0.401**	–5.956***	–5.437***
	(0.174)	(0.171)	(1.778)	(1.807)
Upper secondary Gen	0.720***	–0.724***	–5.365***	–4.663***
	(0.115)	(0.119)	(1.242)	(1.268)
Tertiary Voc	0.896***	–0.895***	–9.578***	–8 794***
	(0.182)	(0.193)	(1.473)	(1.499)
Tertiary Gen	1 921***	–1.782***	–11.68***	–10.72***
	(0.350)	(0.353)	(1.267)	(1.315)
Sefl-employed				2.896***
				(1.105)
Observations	1,485	1,485	1,410	1,407
R-squared			0.123	0.127

Source: Author’s calculations using STEP data.

Note: In this table, each coefficient represents a marginal effect from separate regressions. All estimates control for age in the first job, gender, marital status, highest educational level for both parents, working experience at age of 15, socio-economic status at age of 15, and pre-school attendance. Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05 and * p < 0.1.

2 Education and job-to-job transition

So far, we have documented strong positive impact of additional educational attainment above lower secondary on the probability to find the first job as a wage earner and the time to get the first job. However, the fact that additional educational attainment beyond lower secondary leads to better job as wage employee in the first job faster than self-employment need not imply that self-employment gives low value to higher educational achievement throughout the entire career. Our assumption has been that self-employment serves as an entry point for those who could not find a job quickly. If, for example, the education returns get higher in the formal sector, then workers who expect higher return to their education can move to formal sector in the mid-career. In the remaining part of this section, we explore how education attainment affects the workers’ flow across these job-to-job transition paths.

We predict the probability of the job-to-job transition using a multinomial probit model with educational attainment as the main predictor. Column 1 of Table 9 reports the marginal effects of the employee-to-employee transition, relative to the school-to-job transition. The most interesting result is that probability of the employee-to-employee transition is high among individuals with upper secondary general education and lower among those with tertiary and education. The results in column 2 suggest that upper secondary vocational is the only educational attainment that is likely to increase the probability of employee-to-self-employed transition significantly. It is interesting to note that all additional educational attainment above lower secondary considerably decreases the probability of self-employed-to-employee transition (column 3) or self-employed-to-self-employed transition (column 4). Another important point evident is that the gaps increase with attainment. This is in agreement with the results presented so far, that is, higher educational attainment increases the probability of school-to-job transition where the returns are higher. One explanation for the persistent effects of higher education is that graduates holding these degrees are more likely to find a first job that matches their skills. Remember that we found that they spend less time in job search. Better matching in the first jobs offers more security throughout the career so that the stayer rate is very high.

Table 9.
Education and Job-to-Job Transition

School to job Base category Employee to Employee (1) Employee to Self–employed (2) Self–employed to Employee (3) Self–employed to Self–employed (4)

Education (Reference is Lower secondary)

No formal education –0.057** –0.019 0.027 0.033 0.016

(0.023) (0.048) (0.034) (0.039) (0.035)

Primary 0.000 –0.020 –0.042 0.024 0.037

(0.031) (0.043) (0.029) (0.036) (0.035)

Upper secondary Voc 0.068 0.009 0.095* –0.070*** –0.102*

(0.049) (0.056) (0.057) (0.026) (0.053)

Upper secondary Gen 0.109* 0.083 0.054 –0.065* –0.182*

(0.031) (0.039) (0.033) (0.024) (0.031)

Tertiary Voc 0.238* –0.038 0.063 –0.081* –0.182*

(0.062) (0.054) (0.070) (0.027) (0.058)

Tertiary Gen 0.435* –0.046 –0.005 –0.089* –0.295*

(0.092) (0.063) (0.104) (0.025) (0.023)

Observations 1,187 1,187 1,187 1,187 1,187

	School to job Base category	Employee to Employee (1)	Employee to Self–employed (2)	Self–employed to Employee (3)	Self–employed to Self–employed (4)
Education (Reference is Lower secondary)
No formal education	–0.057**	–0.019	0.027	0.033	0.016
	(0.023)	(0.048)	(0.034)	(0.039)	(0.035)
Primary	0.000	–0.020	–0.042	0.024	0.037
	(0.031)	(0.043)	(0.029)	(0.036)	(0.035)
Upper secondary Voc	0.068	0.009	0.095*	–0.070***	–0.102*
	(0.049)	(0.056)	(0.057)	(0.026)	(0.053)
Upper secondary Gen	0.109***	0.083**	0.054	–0.065***	–0.182***
	(0.031)	(0.039)	(0.033)	(0.024)	(0.031)
Tertiary Voc	0.238***	–0.038	0.063	–0.081***	–0.182***
	(0.062)	(0.054)	(0.070)	(0.027)	(0.058)
Tertiary Gen	0.435***	–0.046	–0.005	–0.089***	–0.295***
	(0.092)	(0.063)	(0.104)	(0.025)	(0.023)
Observations	1,187	1,187	1,187	1,187	1,187

Source: Author’s calculations using STEP data.

Note: This table reports marginal effects from multinomial probit regression where school-to-first job transition is the excluded category. All estimates control for age in the first job, gender, marital status, highest educational level for both parents, working experience at age of 15, socio-economic status at age of 15, and pre-school attendance. Reported standard errors (in parentheses) are corrected for heteroscedasticity. The symbols ***, ** and * signify significance at the 1%, 5%, and 10% levels, respectively.

VII Conclusion

This article explores the returns of an additional educational attainment above lover secondary in Urban Ghana using data from the World Bank STEP survey. We show that additional attainment leads to significant returns, except for upper secondary general whose gains are not statistically different from lower secondary. In particular, we compare the return to additional educational attainment between the sample of workers moving from school to the job against that of job-to-job transition. We find that the additional returns of tertiary general and tertiary vocational are higher in the school-to-job sample

The literature on the returns to education has proposed several channels through which education affects earnings, including productivity and innovation. We were able to test additional mechanisms such as the pattern of transition between the first job and the current job, the signalling effect of higher future returns, and the matching quality. We find evidence that controlling for educational attainment, job-to-job transition has mediation effect on earnings; hence, this channel explains the returns we observe. The findings indicate that the transition from employee to self-employed has the highest returns. This finding enhances our understanding of the self-selection into self-employment in developing countries. We do not find a discernible effect for the other mechanisms. We, therefore, conclude that structural reforms of 1984 and the matching quality are not driving our findings.

Some caution is needed, however, in interpreting our results. One reason is that our transition variable does not include information on the wage in the first sector. In addition, since we do not have information on the employer, our transition variable may apply to those who remain in the same sector or occupation. Nevertheless, by documenting several channels that explain differences in earnings, we can suggest policies that could improve the school-to-job transition, career counselling, and business incubator.

Footnotes

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

TVET Reforms in Ghana.

The current educational structure of Ghana comprises of 6 years of primary school and three years of junior high/secondary school, which together makes up the basic education level; another 3 years of senior high school, which forms the secondary level; and lastly, 2–years of tertiary level education. Vocational and technical education is organized at all the three levels of Ghana’s educational system: the basic level, the secondary level, and lastly, the tertiary level. Three different types of vocational and technical education are structured across these three levels. These are the pre-vocational, vocational, and technical.

The pre-vocational aspect of TVET take place at the basic school level. Pupils at this level are exposed to a variety of practical activities in the vocational field; hence, they become acquainted with and encourage their interest in vocational matters. This opportunity is offered to pupils at this level to enable them to make informed decisions and choices in their career paths (choose either vocational or general fields). This also offers pupils who will not pursue further education the opportunity to gain paid self-employment in industry, commence, and agriculture with the trade they have acquired through the training. They could as well move into the informal sector for apprenticeship training.

At the senior high (secondary) level, vocational training, which is markedly pursued in Ghana, is organized using a blend of two approaches. The first approach is the parallel system in which vocational and technical institutions coexist with the senior high school system. Graduates from the basic level can pursue/choose the senior high schools or the vocational technical institutes. The vocational technical institutes train and impart practical skills and training leading to the delivery of craftsmen, artisans, technicians, and middle-level people in agriculture, technology, commence, science, and industry. The second approach is the core curriculum. It is used in the orthodox senior high school system after the level of basic education. After graduating the basic level, graduates opting for the senior high school system are exposed to a core curriculum and a collection of elective subjects, which can be in the field of vocational technical. Students desiring to pursue career paths in vocational technical have the option of choosing a minimum of three elective subjects in field of vocational technical. The students are required to study these three electives in addition to the four core (compulsory) subjects. At this level, the goal of vocational technical is to equip the youth with relevant productive skills training making them capable of fulfilling the country’s manpower needs in the domains of industry, technology, agriculture, commence, and business.

At the tertiary level, vocational technical education in Ghana is practically oriented. This is the highest level of vocational technical education and it is organized within the post-secondary or tertiary institutions. The universities, polytechnics (now technical universities), and other pre-service training institutions (post-secondary) underneath sectorial ministries offer this training. Technical and vocational education at the tertiary level offers trainees with practical knowledge and vocational/trade skills essential for industrial, agriculture, scientific, commercial, and technological skills, among others. The goal is to train and equip the human resource base of the country (especially the youth) to match the supply of skilled labour with demand. The other post-secondary or pre-services training institutions include colleges of nursing, agriculture, and teacher training. Institutes and schools of journalism, communication, professional studies, and forestry, among others. Duration of courses are between 2 and 4 years and certificates, diploma, or degrees are awarded.

Notes

References

Altonji

J.G.

1993: The demand for and return to education when education outcomes are uncertain. Journal of Labor Economics, 11, 48–83.

Arulampalam

2001: Is unemployment really scarring? Effects of unemployment experiences on wages. Economic Journal, 111, 585–606.

Baah-Boateng

2016: Youth employment and unemployment challenges in Ghana: Where are we and what do we do? African Center for Economic Transformation.

Card

1995: Earnings, schooling, and ability revisited. Research in Labor Economics, 14, 23–48.

Darvas

and Palmer

2014: Demand and supply of skills in Ghana: How can training programs improve employment and productivity? World Bank.

Darvas

, Favara

and Arnold

2017: Stepping up skills in urban Ghana: Snapshot of the STEP skills measurement survey. World Bank.

De Cuyper

, Notelaers

and De Witte

2009: Transitioning between temporary and permanent employment: A two-wave study on the entrapment, the stepping stone and the selection hypothesis. Journal of Occupational and Organizational Psychology, 82, 67–88.

Earle

J.S.

and Sakova

2000: Business start-ups or disguised unemployment? Evidence on the character of self-employment from transition economies. Labour Economics, 7, 575–601.

Ellwood

D.T.

1982: Teenage unemployment: Permanent scars or temporary blemishes? In Freeman

R.B.

and Wise

D.A.

, editors, The youth labor market problem: Its nature, causes, and consequences. University of Chicago Press.

10.

Gregory

and Jukes

2001: Unemployment and subsequent earnings: Estimating scarring among British Men 1984–94. The Economic Journal, 111, 607–25.

11.

Griliches

1977: Estimating the returns to schooling: Some econometric problems. Econometrica, 45, 1–22.

12.

Heckman

1974: Shadow prices, market wages, and labor supply. Econometrica, 42, 679–94.

13.

Hemmer

and Mannel

1989: On the economic analysis of the urban informal sector. World Development, 17, 1543–52.

14.

Honorati

and Silva

S.J.D.

2016: Expanding job opportunities in Ghana. World Bank.

15.

Maloney

W.F.

2004: Informality revisited. World Development, 32, 1159–78.

16.

Mincer

J.A.

1974: Schooling, experience and earnings. National Bureau of Economic Research.

17.

Montenegro

C.E.

and Patrinos

H.A.

2014: Comparable estimates of returns to schooling around the world (Policy Research Working Paper No. 7020). World Bank.

18.

Osei

R.D.

and Jedwab

2017: Structural change in a poor African country: New historical evidence from Ghana. In Mcmillan

M.S.

, Rodrik

and Sepúlveda

, editors, Structural change, fundamentals, and growth: A framework and case studies. International Food Policy Research Institute and the World Bank.

19.

Osterman

2006: The wage effects of high performance work organization in manufacturing. Industrial and Labor Relations Review, 59, 187–204.

20.

Perry

G.E.

, Maloney

W.F.

, Arias

O.S.

, Fajnzylber

, Mason

and Saavedra-Chanduvi

2007: Informality: Exit and exclusion. World Bank.

21.

Psacharopoulos

and Patrinos

H.A.

2004: Returns to investment in education: A further update. Education Economics, 12, 111–34.

22.

Ranis

and Gollin

2013: Macroeconomic perspectives. In Hino

and Ranis

, editors, Youth and employment in Sub-Saharan Africa: Working but poor. Routledge, Taylor & Francis Group.

23.

Ruhm

C.J.

1991: Are workers permanently scarred by job displacements? American Economic Review, 81, 319–24.

24.

Scherer

2004: Stepping-stones or traps? The consequences of labour market entry positions on future careers in West Germany, Great Britain and Italy. Work, Employment and Society, 18, 369–94.

25.

UNESCO 2012: International standard classification of education: ISCED 2011. UNESCO.

26.

World Bank 2018: The world development report 2018: Learning to realize education’s promise. World Bank.