The returns to learning later in life: a summary of CEE work Leon Feinstein, Fernando Galindo- Rueda, Andrew Jenkins, Anna Vignoles, Alison Wolf See: CEE DPs 19, 36, 37, 39
Background Labour force has become more educated in UK and most OECD countries Big increase in supply of skilled labour Yet still high rate of return to schooling in the UK and elsewhere Best estimates for UK are within the range of 10%-15% gain in wages per year of schooling
International Comparison of Returns to Schooling (OLS Estimates) Country Males Females GB 0.127 0.006 0.130 0.006 USA 0.074 0.004 0.096 0.005 W. Germany 0.036 0.002 0.043 0.004 Ireland 0.085 0.006 0.090 0.008 Australia 0.051 0.004 0.052 0.006 N. Ireland 0.174 0.011 0.146 0.011 Pooled (28) 0.048 0.001 0.057 0.001 Note: Robust standard errors are in italics. Controls: year dummies, union status, marital status, age, age-squared and, in pooled estimation, country-year dummies. Source: P. Trostel et al (2002)
OECD participation in education at 17 in 2002 110 100 90 80 70 60 50 40 30 20 10 0 Canada (1) Belgium Czech Rep. Sweden Germany Finland Japan Norway France Poland Korea Austria Netherlands Slovak Rep. Hungary Switzerland Australia Denmark Spain United States Ireland Iceland Italy Luxembourg United Kingdom New Zealand Portugal Greece Mexico % in education at age 17 Turkey (1) Source: OECD, 2004
International Comparisons of Level 3 Qualifications Source: Steedman, McIntosh and Green (2004)
The problem with vocational qualifications Part of problem may be with vocational route CEE research has found few (or negative) labour market effects from some lower level vocational qualifications Nil return to Level 2 qualifications (on employment and wages), Dearden et al. (2000), McIntosh (2002) Positive impact from Level 2 for particular groups and industries (Dearden et al., 2004)
Lifelong learning Lifelong learning seen as one policy solution to this problem Skills for life strategy improving basic skills of 2.25m adults by 2010 free literacy and numeracy or IT courses Adult Learning Grant Career Development Loans Encourage workers back into education/ training
CEE work on lifelong learning: data Mostly uses NCDS data National Child Development Study (NCDS) data: a cohort of individuals born in 1958 Data on learning between sweeps 5 and 6, at age 33 (in 1991) and 42 (in 2000)
Defining lifelong learning Narrow definition of lifelong learning If individual acquired a qualification between the ages of 33 and 42 Clearly too narrow for the broad lifelong learning agenda but has the advantage of being precise About one-third of the sample acquired at least one qualification between 1991 and 2000 of the 15% of NCDS respondents doing a course in 2000, 67% were on a course leading to a qualification
Who is more likely to undertake lifelong learning leading to a qualification? More educated and able workers Mathematical ability is particularly important for women Public sector, unionised and large firm workers Family background/type of schooling is not important (for a given level of initial education) Females also more likely to undertake LLL on average Source: Jenkins, Vignoles, Wolf and Galindo-Rueda (2003)
Who is more likely to undertake lifelong learning? More educated and able workers are undertaking LLL not low achievers HOWEVER, the model explains very little
The wage effects of lifelong learning No evidence of across the board wage premiums Weak evidence of a pay off to some academic qualifications Evidence that those with no qualifications in 1991 do benefit from lifelong learning (13% higher wages and 10-14% higher wage growth on average) Academic qualifications pay-off most highly for this no qualification group
The employment effects of lifelong learning For individuals who were not in employment in 1991, lifelong learning positively affected the probability of being in employment in 2000 Result only significant for women Very few men (<300) were not in employment in 1991
How can lifelong learning assist women returners? Methodology - what are the effects of lifelong learning on transitions into employment amongst women? Focus on women out of work in 1991 (sample size 1,443) Proportional Hazards Model of Transitions into Employment Monthly data on Employment Status from 1991 to 2000 Control for number of children, age of youngest child, partnership status and other factors Source: Jenkins 2005
Characteristics of women out of employment in 1991 95% had a child (68% among in work women) Tended to have young children More likely to have a partner Less qualified One fifth no qualifications (10% of in work women) 17% high level qualifications (29% of in work women)
Methodological concerns Endogeneity: Women who are more likely to return to work earlier may also be more likely to undertake lifelong learning Estimation strategy Rich controls including age 11 ability, prior education level, family background, attitudes towards work when young Different models to control for unobserved heterogeneity No evidence of unobserved heterogeneity
Methodological concerns Timing issues Women may have acquired qualifications after returning to work Time varying covariates needed
Results Lifelong learning leading to qualifications had strong positive effects on likelihood that a woman would return to work Initial education level (in 1991) had surprisingly weak effects, after controlling for other factors
Results Model 1 Model 2 Hazard. Ratio z Hazard. Ratio z Education Level in 1991 (base no qualifications) NVQ level 1 1.49 3.31 1.48 3.11 NVQ level 2 1.38 2.90 1.44 3.09 NVQ level 3 1.41 2.62 1.53 2.90 NVQ level 4 1.40 2.37 1.61 2.99 NVQ level 5 1.58 2.54 1.97 3.40 Learning leading to Qualification, 1991 to 2000 1.48 3.05 1.50 3.05 Controls: Marital status, children, health, work experience, attitudes towards work As for model 1 Plus ability, family background, school, region Number of Observations 1276 1271 Log likelihood -5944.21-5896.34
Results Women who undertake lifelong learning make the transition back into the labour market more quickly Not necessarily causal Lifelong learning for these women might signal their intention and motivation to return to work
Evidence on lifelong learning Access to lifelong learning is not equal More educated and able workers are more likely to undertake lifelong learning Lifelong learning does not result in large economic benefits to the individual Acquiring formal qualifications later in life gives no wage gain Some economic gains for workers with no qualifications, unemployed males and women who are out of the labour market
Evidence on training The UK has relatively high levels of work related training By some measures, 40% of UK population receive training in a given year, compared to 35% in US and 30% in Germany UK adults also receive more hours of training compared to other countries
Provision of training in England Off-thejob On-thejob 1999 % 2000 % 2001 % 52 59 55 79 83 78 Base/Coverage: all employers with 5 or more employees: n = 3431 Source: D. Spilsbury, Learning and Training at Work 2001. DfES Research Report 334 (2002).
CEE work on training: data NCDS data cohort of individuals born in 1958 We use data on training between sweeps 5 and 6, age 33 and 42 Number of work related training spells lasting more than 3 days between 1991 and 2000 47% males, 36% of females received training Source: Feinstein, Galindo-Rueda and Vignoles 2004
Problems in the training literature Possible biases (i) correlation between training and unobserved permanent characteristics (e.g. ability) (ii) joint decision to train by firms and individuals based on unobservable characteristics of firm and worker (iii) temporary shocks to wages that may be correlated with training participation
Data and methodology Methods to tackle these biases Adequate controls (family background, ability, early schooling, firm characteristics, work history etc.) Difference equations (wage growth between 1991-2000 allows for fixed unobserved characteristics correlated with training and wages) Instrumental variables Control function model
Data and methodology Instruments Prior training 1981-1991 (from NCDS) Distance to training provider (FE college), based on person s postcode in 1991.
Who undertakes training? More educated workers Younger workers Professional workers More motivated/ ambitious workers Lower tenure workers Permanent workers Full time workers Newly recruited workers Source: Feinstein, Galindo-Rueda and Vignoles 2004; Jenkins 2005
Who undertakes training? Workers in large firms Workers in occupations that experience more rapid change, e.g. IT Public sector and unionised workers Workers in different sectors (especially electricity/ gas, public administration, health, hotels, financial services) Source: Feinstein, Galindo-Rueda and Vignoles 2004; Jenkins 2005
Results Males Average impact of training around 4-5% on wage growth age 33-42 Hides differences between the trained and untrained Gain for trained is high: 12% No potential gain for untrained Females Results not robust No effect from training on wage growth age 33-42
Male Summary Results Treatment: receving employer provided training lasting more than 3 days in duration between 1991 and 2000 Average treatment effect Treatment effect on the treated Treatment effect on the non-treated OLS 1981- IV 1981- Model OLS IV estimate 1991 1991 Estimate 0.048 0.041 0.120-0.030 0.050 0.05-0.02 S.E 0.013 0.048 0.065 0.071 0.047 0.018 0.118 Dependent variable is change in log wages between 1991 and 2000 *Instrument is whether person received training between 1981 and 1991
Female Summary Results Treatment: receving employer provided training lasting more than 3 days in duration between 1991 and 2000 OLS Change Wage 1991-2000 IV Change Wage 1991-2000 OLS Change wage 1981-1991 OLS Wage IV Wage Model 2000 2000 Estimate -0.006 0.090 0.627 1.613 0.145 S.E 0.013 0.021 0.531 0.921 0.026 *Instrument is distance to nearest FE training provider in 1991
Conclusions Training is not given randomly More productive, more able and more educated workers get more training Evidence of selectivity Lifelong learning leading to formal qualifications does not pay off YET short spells of work related training give large wage gains Training pays off for these workers because they are especially selected and because firms know what skills they need to acquire
Policy implication Adults that leave school with poor prospects cannot necessarily improve them with lifelong learning We cannot leave the problem of poor skills to firms Although training does lead to gains in skills and wages, employers tend to train already advantaged workers