THE BENEFITS AND CORRELATES OF LEARNING AT WORK

THE BENEFITS AND CORRELATES OF LEARNING AT WORK: Empirical Evidence Using PIAAC and STEP Data Victoria Levin Marcela Gutierrez Bernal Alexandria Valerio Carla Solis Uehara November 29, 2018

Outline 1. Motivation 2. Conceptual framework: Drivers of learning at work 3. Data 4. Correlates of learning at work 5. Benefits of learning at work 6. Policy implications 1

Outline 1. Motivation 2. Conceptual framework: Drivers of learning at work 3. Data 4. Correlates of learning at work 5. Benefits of learning at work 6. Policy implications 2

What does it take to be a “well-educated” person in the 21 st Century? Cognitive skills Cognitive Involving the use of logical, intuitive and creative thinking Raw problem solving ability vs. knowledge to solve problems Verbal ability, numeracy, problem solving, memory (working and longterm) and mental speed 3 Socioemotional skills Technical skills Socio-emotional Beliefs, personality traits, Behavioral skills Big 5 traits: Openness to experience, conscientiousness, extraversion, agreeableness, emotional stability Behaviors/attitudes: Self-regulation, Grit/perseverance, decision making, self & interpersonal skills Skill Set Technical Involving manual dexterity and use of methods, materials, tools & instruments Developed through VET/university or acquired on the job Related to specific occupations/trades (e. g. engineer, economist, IT specialist, plumber)

Education ≠ basic skills, even for youth Percentage of youth (aged 19 -20) with at least upper-secondary education scoring below minimum literacy proficiency (PIAAC and STEP) Notes: i) Data on Armenia, Bolivia, Colombia, Georgia, Ghana, Kenya, Kosovo, Serbia, Ukraine, and Vietnam: World Bank’s STEP Skills Measurement Surveys; data on rest of economies: Programme for the International Assessment of Adult Competencies (PIAAC) data set; ii) Tertiary education is merged with upper secondary education; iii) STEP surveys are representative of urban areas; iv) The PIAAC sample for the Russian Federation does not include the population of the Moscow municipal area; v) Minimum literacy proficiency threshold is defined as level 2 and above; individuals that score below level 2 are able to identify letters or words but are not able to deploy the full set of cognitive strategies required to integrate, evaluate, or interpret information from a variety of text materials and text complexity. 4

Can learning at work be an equalizer? 1) Does learning at work correlate with other skills, cognitive and socioemotional? 2) Does learning at work pay off? 5

Outline 1. Motivation 2. Conceptual framework: Drivers of learning at work 3. Data 4. Correlates of learning at work 5. Benefits of learning at work 6. Policy implications 6

Potential factors for learning at work Opportunity to Learn Willingness to Learn Ability to Learning at Work 7

Outline 1. Motivation 2. Conceptual framework: Drivers of learning at work 3. Data 4. Correlates of learning at work 5. Benefits of learning at work 6. Policy implications 8

PIAAC and STEP – similar but different § Learning at work Ø Equivalent measure across two data sources: Indicator for learning at work from any source at least once a month § Ability to learn Ø Direct assessment of literacy on the same proficiency scale for PIAAC and STEP: Attainment of at least Level 2 proficiency to indicate the ability to identify information using simple texts § Willingness to learn Ø Socioemotional skills only in STEP: Big Five personality traits (Openness to Experience, Conscientiousness, Extraversion, Agreeableness, and Emotional stability), grit, decision-making § Opportunity to learn Ø Three occupational groups: High-skilled white-collar workers, low-skilled whitecollar workers, blue-collar workers 9

Outline 1. Motivation 2. Conceptual framework: Drivers of learning at work 3. Data 4. Correlates of learning at work 5. Benefits of learning at work 6. Policy implications 10

Learning at work is correlated with development… 11

…for all occupation groups High-skilled white-collar workers 12 Low-skilled white-collar workers Blue-collar workers

Ability to learn: Basic literacy is important for learning at work… Percentage of adults who learn at work at least once a month 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Below Level 2 Above Level 2 Notes: i) Data from the STEP Skills Measurement Surveys, 2011 -2016 and the Programme for the International Assessment of Adult Competencies (PIAAC); ii) Individuals that score below level 2 could be able to identify letters or words but are not able to deploy the full set of cognitive strategies required to integrate, evaluate, or interpret information from a variety of text materials and text complexity. It should be noted that quality jobs require individuals with at least a level 2 of literacy skills proficiency 13

Percentage of adults who learn at work at least once a month . . . for all occupation groups 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% All Below Above Level 2 High Skilled White Collar Low Skilled White Collar All Below Above Level 2 Blue Collar Notes: i) Data from 41 countries using the STEP and PIAAC surveys. Namely: Armenia, Austria, Belgium, Bolivia, Canada, Chile, Colombia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Georgia, Germany, Ghana, Greece, Ireland, Israel, Italy, Japan, Kenya, Korea, Kosovo, Lithuania, Netherlands, New Zealand, Norway, Poland, Russian Federation, Serbia, Singapore, Slovak Republic, Slovenia, Spain, Sweden, Turkey, Ukraine, United Kingdom, United States, and Vietnam; ii) Individuals that score below level 2 could be able to identify letters or words but are not able to deploy the full set of cognitive strategies required to integrate, evaluate, or interpret information from a variety of text materials and text complexity. It should be noted that quality jobs require individuals with at least a level 2 of literacy skills proficiency 14

Ability to learn: Basic literacy correlates with learning at work even when many observables are held constant Net effect of literacy (not controlling for education) Literacy controlling for education 0% 2% 4% 6% 8% 10% % increase in probability of learning at work at least once a month Notes: (a) We include a sample of adults aged 16 -64 in Armenia, Austria, Belgium, Bolivia, Canada, Chile, Colombia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Georgia, Germany, Ghana, Greece, Ireland, Israel, Italy, Japan, Kenya, Korea, Kosovo, Lithuania, Netherlands, New Zealand, Norway, Poland, Russian Federation, Serbia, Singapore, Slovak Republic, Slovenia, Spain, Sweden, Turkey, Ukraine, United Kingdom, United States, and Vietnam; (b) The model includes a set of country dummies, OECD status, age squared, gender, levels of education (with lower secondary as a base group); (c) Minimum literacy proficiency is understood as obtaining level 2 or above in the reading assessment administered by STEP and PIAAC. The surveys characterize literacy as representing a range of skills, from the decoding of written words and sentences to the comprehension, interpretation, and evaluation of texts that are of varying complexity and fit workplace, personal, society, and community contexts; (d) * p < 0. 1, ** p < 0. 05, *** p < 0. 01 15

Willingness to learn: Openness to experience and grit correlate strongly with learning at work Full sample High-skilled white-collar workers Low-skilled white-collar workers Blue-collar workers One-by- All at one once Extraversion +** Conscientiousness +*** Openness +*** +* -* +** +*** +* +*** +** Emotional Stability Agreeableness +* Grit +*** Decision-making +*** +* +* +** Notes: (a) We include a sample of adults aged 16 -64 in Armenia, Bolivia, Colombia, Georgia, Ghana, Kenya, Kosovo, Serbia, Ukraine and Vietnam (b) * p < 0. 1, ** p < 0. 05, *** p < 0. 01; (c) Each regression includes, a set of country dummies, age squared, gender, minimum literacy proficiency (Level 2) indicator. 16

Willingness to learn: Openness to experience and grit correlate strongly with learning at work Full sample High-skilled white-collar workers Low-skilled white-collar workers Blue-collar workers One-by- All at one once Extraversion +** Conscientiousness +*** Openness +*** +* -* +** +*** +* +*** +** Emotional Stability Agreeableness +* Grit +*** Decision-making +*** +* +* +** Notes: (a) We include a sample of adults aged 16 -64 in Armenia, Bolivia, Colombia, Georgia, Ghana, Kenya, Kosovo, Serbia, Ukraine and Vietnam (b) * p < 0. 1, ** p < 0. 05, *** p < 0. 01; (c) Each regression includes, a set of country dummies, age squared, gender, minimum literacy proficiency (Level 2) indicator. 17

Outline 1. Motivation 2. Conceptual framework: Drivers of learning at work 3. Data 4. Correlates of learning at work 5. Benefits of learning at work 6. 18 Policy implications

Hypothesis and methodology • Hypothesis: There is an earnings premium for learning at work, and this premium rises with accumulated experience • Methodology: Augmented Mincer regressions, with a focus on wage -experience profile with and without learning at work • Dependent variable: hourly wage and self-employment earnings • Covariates of interest: Learning at work, Experience (potential), interaction of experience and learning at work • Other covariates: gender, years of education, and fixed country effects • Two-step Heckman selection correction using available exclusion restrictions in each dataset Analysis conducted for PIAAC and STEP separately 19

Learning at work delivers a significant earnings premium, that rises with experience… Notes: The figure is based on a two-step Heckman selection model regression of log hourly earnings that also includes gender, years of education, and fixed country effects as covariates. The first stage of the model included gender, years of education, quadratic polynomial of potential experience, fixed country effects, and a set of exclusion restrictions (for PIAAC: employed partner and number of children under 6; for STEP: number of shocks at age 15, socio-economic status at age 15, and household asset index). Kosovo was excluded from the STEP model as a positive outlier on learning premium.

… and this does not depend on functional form of experience Notes: The figure is based on a two-step Heckman selection model regression of log hourly earnings that also includes gender, years of education, and fixed country effects as covariates. The first stage of the model included gender, years of education, nonparametric model of potential experience, fixed country effects, and a set of exclusion restrictions (for PIAAC: employed partner and number of children under 6; for STEP: number of shocks at age 15, socioeconomic status at age 15, and household asset index). Kosovo was excluded from the STEP model as a positive outlier on learning premium.

For PIAAC countries, learning premium for all occupation groups, but is highest for low-skilled white-collars Notes: The figure is based on a two-step Heckman selection model regression of log hourly earnings that also includes gender, years of education, and fixed country effects as covariates. The first stage of the model included gender, years of education, quadratic polynomial of potential experience, fixed country effects, and a set of exclusion restrictions (employed partner and number of children under 6).

For STEP countries, negative experience premium for lowskilled white-collars, but learning premium for all Notes: The figure is based on a two-step Heckman selection model regression of log hourly earnings that also includes gender, years of education, and fixed country effects as covariates. The first stage of the model included gender, years of education, quadratic polynomial of potential experience, fixed country effects, and a set of exclusion restrictions (number of shocks at age 15, socio-economic status at age 15, and household asset index). Kosovo was excluded from the STEP model as a positive outlier on learning premium.

Learning at work could mediate the relationship between skills and earnings Cognitive skills Socioemotional skills Technical skills 24 Learning at work

For STEP countries, learning premium remains even with controls for literacy and socioemotional skills… Notes: The figure is based on a two-step Heckman selection model regression of log hourly earnings that also gender, minimum literacy proficiency indicator, indicators for above-average values for socioemotional skills (Big 5 personality traits, grit, and decision-making), and fixed country effects as covariates. The first stage of the model included gender, minimum literacy proficiency indicator, indicators for above-average values for socioemotional skills (Big 5 personality traits, grit, and decision-making), quadratic polynomial of potential experience, fixed country effects, and a set of exclusion restrictions (number of shocks at age 15, socio-economic status at age 15, and household asset index). Kosovo was excluded from the model as a positive outlier on learning premium.

… with highest impacts for low-skilled white-collars Notes: The figure is based on a two-step Heckman selection model regression of log hourly earnings that also gender, minimum literacy proficiency indicator, indicators for above-average values for socioemotional skills (Big 5 personality traits, grit, and decision-making), and fixed country effects as covariates. The first stage of the model included gender, minimum literacy proficiency indicator, indicators for above-average values for socioemotional skills (Big 5 personality traits, grit, and decision-making), quadratic polynomial of potential experience, fixed country effects, and a set of exclusion restrictions (number of shocks at age 15, socio-economic status at age 15, and household asset index). Kosovo was excluded from the model as a positive outlier on learning premium.

Outline 1. Motivation 2. Conceptual framework: Drivers of learning at work 3. Data 4. Correlates of learning at work 5. Benefits of learning at work 6. Policy implications 27

Can learning at work be an equalizer? 1) Does learning at work correlate with other skills, cognitive and socioemotional? YES 2) Does learning at work pay off? YES 28

Main implications • Act earlier and do better • Early childhood development is vital to provide children with adequate cognitive and socioemotional skills and to put them on the right path • Quality of basic schooling and positive learning environment can support development of foundational skills to produce learning-ready labor market entrants But don’t stop here 29

Formal workplace training incidence is low % of adults who reported having participated in training Participation in workplace training in the last 12 months, by literacy level (STEP) 25% 20% 15% 10% 5% 0% Armenia Bolivia Colombia Georgia Level 1 and Below Ghana Kenya Ukraine Vietnam Level 2 and Above Notes: i) Information from the STEP Skills Measurement Surveys, 2011 -2016; ii) Individuals that score below level 2 are able to identify letters or words but are not able to deploy the full set of cognitive strategies required to integrate, evaluate, or interpret information from a variety of text materials and text complexity. It should be noted that quality jobs require individuals with at least a level 2 of literacy skills proficiency. 30

Main implications • Act earlier and do better • Early childhood development is vital to provide children with adequate cognitive and socioemotional skills and to put them on the right path • Quality of basic schooling and positive learning environment can support development of foundational skills to produce learning-ready labor market entrants But don’t stop here • Increase learning at work for ready learners • Stimulate formal workplace training, which appears to be underprovided in most countries • Informal learning at work could also be encouraged through mentorship programs, incentivizing knowledge sharing and keeping up to date, and team rotations 31

Thank you! Victoria Levin Senior Economist Education Global Practice vlevin@worldbank. org 32

Annex slides 33

% increase in probability of learning at work at least once a month Ability to learn: The net effect of basic literacy is important for all occupation groups; direct effect for low-skilled white-collars 5% 5% 4% 4% 3% 3% 2% 2% 1% 1% 0% With education Net effect High-skilled white-collar workers Low-skilled white-collar workers Blue-collar workers 34

Primary or Less Lower Secondary Upper Secondary +L 2 (N=44937) -L 2 (N=2901) +L 2 (N=45277) -L 2 (N=9342) +L 2 (N=7529) -L 2 (N=5171) +L 2 (N=2287) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% -L 2 (N=5210) Predicted probability of learning at work at least once a month Ability to learn: Basic literacy mediates the relationship between education and learning at work Tertiary Notes: i) Data from 44 countries using the STEP and PIAAC surveys. Namely: Armenia, Austria, Belgium, Bolivia, Canada, Chile, Colombia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Georgia, Germany, Ghana, Greece, Ireland, Israel, Italy, Japan, Kenya, Korea, Kosovo, Lao PDR, Lithuania, FYR Macedonia, Netherlands, New Zealand, Norway, Poland, Russian Federation, Serbia, Singapore, Slovak Republic, Slovenia, Spain, Sri Lanka, Sweden, Turkey, Ukraine, United Kingdom, United States, Vietnam and Yunnan province in China; ii) Individuals that score below level 2 could be able to identify letters or words but are not able to deploy the full set of cognitive strategies required to integrate, evaluate, or interpret information from a variety of text materials and text complexity. It should be noted that quality jobs require individuals with at least a level 2 of literacy skills proficiency; iii) Number of observations in parenthesis; iv) coefficients obtained after controlling for age, age squared, gender, country and OECD status. 35