Monday, 24 October 2016

Using survey data to estimate the value of vocational qualifications

CVER's Steven McIntosh and Damon Morris, from University of Sheffield, look at the value of vocational qualifications

An important part of the research programme at CVER is to investigate the ‘returns’ to vocational qualifications, that is, the wage premiums earned by individuals who hold such qualifications. The results of such research provide important information to policy-makers about the value the labour market places on qualifications, as well as to individuals making decisions about what courses to pursue. Our work will provide up-to-date evidence on this topic, using a comprehensive range of data sets and methodologies.

The characteristics of those who participate in vocational learning will be different to those of individuals who do not, with these differences often unobserved. Thus, it is very difficult to attribute differences in the wage premia to qualifications alone (when motivation, ability, family background etc. also play a role). The challenge is to use the most appropriate methodology, that produces as fair a comparison as possible. Work within CVER continues in this area, considering a range of data sources, both administrative and survey-based, and methodologies, in order to determine what factors affect the estimated results, and which estimates are likely to be the most accurate. 

This blog summarises the findings from the first stage in this research work, estimating the returns to vocational education using survey data. In particular we used Labour Force Survey (LFS) data from the period 1997-2015. The LFS provides detailed information on respondents’ qualifications, allowing us to disaggregate the results by type, level and subject of qualification, as well as by the gender of the learner.

Returns are estimated by comparing the wages of those with a particular qualification (i.e. ‘the treatment group’) and those in a suitable control group without the qualification, whilst controlling for any differences in observable characteristics between them, such as gender, age, ethnicity, region of residence, full-time/part-time status and public/private sector status. The treatment group can be defined as all those with a qualification, or as those individuals who hold the qualification as their highest. We estimated both, but here only summarise the results from the highest qualification specifications (so-called ‘marginal’ returns).

We first compared learners with each qualification to a control group of individuals with no qualifications. Not surprisingly, the wage premium earned relative to this control group was larger, the higher the level of the qualification. Within levels, a consistent pattern was found, in that the qualifications that earned the largest returns were BTEC qualifications. For example, holding other things constant, an individual with an HNC/HND as their highest qualification (a level 4, sub-degree level qualification) earns on average 58% more than an individual with no qualifications, while an individual with an ONC/OND as their highest qualification (a level 3 qualification) earns on average 39% more than an individual with no qualifications. The equivalent return for an NVQ at level 3 is 26%, while at level 2 is just 5%. When estimated separately for men and women, the estimated returns to BTEC, City and Guilds and apprenticeship qualifications tend to be higher for men than for women, while the returns to RSA qualifications (which are secretarial qualifications) are higher for women than for men.

The returns discussed so far are all relative to individuals with no qualifications. But that is not what is of most interest to policy-makers and potential learners, who want to know by how much their wages will increase if they move one level up the qualifications hierarchy. When estimating the returns to each qualification, we therefore changed the control group to contain individuals whose highest qualification was one level below the qualification of interest (holding either academic, vocational, or any qualifications at that lower level). Figure 1 shows the results when the control groups consist of those individuals whose highest qualification is specifically a vocational qualification at one level lower. Thus, for each type of qualification separately, the figure shows how wages increase by moving up the qualifications hierarchy via the vocational route, since the comparison group in each case is vocational qualifications at the level below. We can see that individuals will reach the highest wages through BTEC qualifications, which was not surprising given the findings discussed in the previous paragraph. The interesting thing to note about Figure 1, however, is that the slopes of the lines are quite similar across qualification types. Thus the main advantage of BTEC qualifications lies in the ‘good start’ that they make, with higher returns than the other qualification types at low education levels. Thereafter, wages increase with levels at a similar rate.

Figure 1: Marginal Returns relative to Vocational Qualifications at Level One

We also considered the subject of study of the vocational qualifications. Looking across different types of qualifications, a common pattern was observed, in that the highest returns are always obtained by Engineering and Construction qualifications, followed by Business and Management qualifications. The lowest are received by holders of Caring, Childcare and Hotels and Catering qualifications. These results are illustrated in Figure 2 for the case of NVQ3 qualifications. For each qualification, the circle shows the point estimate of the return, with the green lines around that circle illustrating the confidence interval around that estimate. The point estimate shows our best estimate of the true return, while the confidence interval shows a range, within which we are 95% sure that the true return lies.

Figure 2: Wage Returns to NVQ3 Qualifications by Subject Area

Of course, part of this difference across subjects is due to the fact that, for example, engineering jobs are typically better paid than childcare jobs. For a young person undecided about what career they are going to follow, this is exactly the information that they would want to know; which qualifications lead to a better paid career. But for someone already working in, say, childcare, the comparison with engineers would be irrelevant, and what they would want to know is how much their earnings would rise within childcare if they obtained a higher qualification, relative to someone working in childcare with lower qualifications. Such returns are illustrated in Figure 3, again for the case of NVQ3, though other qualifications produce very similar patterns. The green circle shows the estimate of the return to each subject, with surrounding confidence interval, while the red circle shows the estimate of the return to each subject when the holder works in a job appropriate for that subject. For example, for Childcare say, the red return shows the estimated wage premium relative to unqualified individuals working in childcare, while the green return shows the estimated wage premium relative to unqualified individuals in all other jobs. As we can see, the former is typically larger than the latter, often significantly so (when the confidence intervals do not overlap). The highest returns within occupations, for the case of NVQ3, are now for Hair and Beauty, where individuals with a Hairdressing qualification earn around 33% more than unqualified workers in the same profession.

Figure 3: Wage Returns to NVQ3 Qualifications by Subject Area and Sector of Work

In conclusion, the results summarised above show that there is a wide range of returns to vocational qualifications, varying by type, level and subject of qualification. So one can’t talk about a return to vocational qualifications in a general sense. The key issue is to show which qualifications work best, and for whom. Work in CVER is continuing, to answer such questions.

* * * 
Further reading: "Labour Market Returns to Vocational Qualifications in the Labour Force Survey", CVER Research Discussion paper, October 2016

No comments:

Post a Comment