Table 4 Summary of recommendations to improve study quality and fill research gaps

Lack of rigorous study designs

Additional studies using rigorous designs are needed, preferably with random assignment

Evaluated interventions not representative of all TVET programmes in existence

Additional studies are needed to evaluate outcomes of apprenticeship programmes, technical education, and vocational education. This is a further need for additional TVET studies from LMICs in Asia, Africa, and Europe.

Inadequate descriptions of interventions

Provide a detailed description of the intervention (including any components) in such as way that the intervention could be replicated (e.g., duration and frequency of sessions, curriculum, stipends, and trainee accreditation).

Lack of studies testing intervention components

Analysis of the effects of individual intervention components is needed in future studies.

Lack of reported information on all key variables

Provide adequate descriptions of the study sample (e.g., age, race, SES, gender, and % with work experience at baseline) and other relevant variables that may influence treatment effects (e.g., whether the intervention was theoretically underpinned, and implementation issues).

Inconsistency in defining, measuring and reporting outcomes

Improve descriptions of employment-related outcome measures, and agree on definitions. Consistently and clearly measure and report study outcomes. Where applicable, measure and report outcomes separately (for example, differentiate part-time and full-time employment).

Lack of studies assessing long-term outcomes

Measure and report outcomes related to employment at multiple time points following the intervention, beyond 18 months and preferably over several years.

Lack of studies assessing intermediate outcomes

Studies need to examine relevant intermediate outcomes (such as work attitudes) via consideration of the theory of change underpinning the intervention.

Lack of studies assessing net employment outcomes

Measure and report net employment outcomes to demonstrate whether job creation is additional or not. Explore whether this is possible to do with sufficiently powered cluster designs.

Inadequate reporting of data needed to compute effect sizes

Report relevant information for all outcomes, such as the sample size, mean and standard deviation (pooled, treatment and/or control) and frequencies/probabilities of an event occurring in each of the groups (as opposed to the difference between the groups), regardless of whether the results of statistical tests for that variable were statistically significant, or the results of other analyses are presented.

Inadequate reporting of information needed for assessments of risk of bias and study replication

Provide a detailed description of the study in such a way that allows (a) confident judgements to made about the trustworthiness of the reported analyses and (b) replication of the study.

Under-development of methods for calculating comparable effect sizes from studies using complex econometrics methods

Further research on these methods is required.

Under-development of quasi-experimental evaluation designs involving complex econometric methods

Develop methods for measuring (and ultimately reducing) the bias incurred by the use of such techniques.