Rediscovering the Role of Rediscovering the Role of Noncognitive Elements of Noncognitive Elements of

Human CapitalHuman CapitalJerry Miner Seminar,Jerry Miner Seminar,Syracuse UniversitySyracuse University

5 April 20125 April 2012Henry M. LevinHenry M. Levin

Non-CognitiveNon-Cognitive

Non-cognitive refers to:Non-cognitive refers to: Attitudes and values.Attitudes and values. Social and emotional behavior. Social and emotional behavior.

Cognitive refers to: Cognitive refers to: Knowledge and its application.Knowledge and its application. Cognitive and non-cognitive may interact as in Cognitive and non-cognitive may interact as in

process of problem-solving.process of problem-solving.

Dominance of Cognitive Skills in Dominance of Cognitive Skills in Economic ExplanationsEconomic Explanations

Nothing inherent in G Becker or T W Schultz that Nothing inherent in G Becker or T W Schultz that limits human capital to cognitive skills.limits human capital to cognitive skills.

Becker highlights the influence of families on the Becker highlights the influence of families on the “knowledge, skills, values, and habits” of children. “knowledge, skills, values, and habits” of children. Human Capital (1964):21.Human Capital (1964):21.

Can capitalize the income stream of any human Can capitalize the income stream of any human attribute (even physical characteristics) into a attribute (even physical characteristics) into a measure of human capital as implied by Irving measure of human capital as implied by Irving Fisher’s broader framework in The Theory of Interest Fisher’s broader framework in The Theory of Interest (1930).(1930).

Economists Ignore Noncognitive Economists Ignore Noncognitive ViewView

Human Capital interpretation of link between Human Capital interpretation of link between education and earnings developed in early education and earnings developed in early 1960’s as contained in educational attainment.1960’s as contained in educational attainment.

Rate of return on educational attainment as an Rate of return on educational attainment as an unopened black box.unopened black box.

Opening of black box led to estimation of Opening of black box led to estimation of Educational Production Functions. Educational Production Functions.

50 Years of Educational Production 50 Years of Educational Production FunctionsFunctions

Virtually all production functions assumed that test Virtually all production functions assumed that test scores are the objective function of schools.scores are the objective function of schools.

All used cognitive measures of outcome with notable All used cognitive measures of outcome with notable exception of Card and Krueger 1992 who used exception of Card and Krueger 1992 who used income and found “right signs” for school effects on income and found “right signs” for school effects on income.income.

Cognitive exclusivity is summarized by the many Cognitive exclusivity is summarized by the many articles showing league tables of educational articles showing league tables of educational production function results limited only to student production function results limited only to student achievement (Hanushek summaries).achievement (Hanushek summaries).

James Heckman InfluenceJames Heckman Influence

In the last decade or so, Heckman has In the last decade or so, Heckman has introduced non-cognitive variables into introduced non-cognitive variables into explanation of educational outcomes and explanation of educational outcomes and income.income.

Psychologists and economists are working Psychologists and economists are working jointly to discover specific non-cognitive jointly to discover specific non-cognitive variables that are important for educational variables that are important for educational success and earnings.success and earnings.

Purpose of PresentationPurpose of Presentation

Emphasize importance of non-cognitive influences on Emphasize importance of non-cognitive influences on

school outcomes and life outcomes.school outcomes and life outcomes. Urge greater balance in assessments between Urge greater balance in assessments between

cognitive and non-cognitive outcome measures.cognitive and non-cognitive outcome measures. Rationale.Rationale. Evidence.Evidence. Next steps.Next steps.

Inkeles on Socialization of Inkeles on Socialization of CompetenceCompetence

““Effective participation in a modern industrial and urban society requires Effective participation in a modern industrial and urban society requires certain levels of skill in the manipulation of language and other symbol certain levels of skill in the manipulation of language and other symbol systems, such as arithmetic and time; the ability to comprehend and systems, such as arithmetic and time; the ability to comprehend and complete forms; information as to when and where to go for what; skills in complete forms; information as to when and where to go for what; skills in interpersonal relations which permit negotiation, insure protection of one’s interpersonal relations which permit negotiation, insure protection of one’s interests, and provide maintenance of stable and satisfying relations with interests, and provide maintenance of stable and satisfying relations with intimates, peers, and authorities; motives to achieve, to master, to persevere; intimates, peers, and authorities; motives to achieve, to master, to persevere; defenses to control and channel acceptably the impulses to aggression, to defenses to control and channel acceptably the impulses to aggression, to sexual expression, to extreme dependency, a cognitive style which permits sexual expression, to extreme dependency, a cognitive style which permits thinking in concrete terms while still permitting reasonable handling of thinking in concrete terms while still permitting reasonable handling of abstractions and general concepts; a mind which does not insist on abstractions and general concepts; a mind which does not insist on excessively premature closure, is tolerant of diversity, and has some excessively premature closure, is tolerant of diversity, and has some components of flexibility; a conative style which facilitates reasonably components of flexibility; a conative style which facilitates reasonably regular, steady, and persistent effort, relieved by rest and relaxation but not regular, steady, and persistent effort, relieved by rest and relaxation but not requiring long periods of total withdrawal or depressive psychic slump; and requiring long periods of total withdrawal or depressive psychic slump; and a style of expressing affect which encourages stable and enduring a style of expressing affect which encourages stable and enduring relationships without excessive narcissistic dependence or explosive relationships without excessive narcissistic dependence or explosive aggression in the face of petty frustration. This is already a long list and aggression in the face of petty frustration. This is already a long list and surely much more could be added (Inkeles 1966: 280-1).surely much more could be added (Inkeles 1966: 280-1).

Inkeles and SmithInkeles and Smith

Index of Modern AttitudesIndex of Modern Attitudes Six countries (Argentina, Bangladesh, Chile, India, Israel, and Six countries (Argentina, Bangladesh, Chile, India, Israel, and

Nigeria) n = almost 6.000Nigeria) n = almost 6.000 Explanatory variables: education, work experience, mass Explanatory variables: education, work experience, mass

media, consumer goods, father’s education, urbanization, skill media, consumer goods, father’s education, urbanization, skill level, length of urban residence, modernity of workplace, level, length of urban residence, modernity of workplace, home and school; and schooling.home and school; and schooling.

Explained 32-62 percent of variance in modernism.Explained 32-62 percent of variance in modernism. Education was most powerful influence—2-3 times more Education was most powerful influence—2-3 times more

powerful than any other. (standard regression coefficients).powerful than any other. (standard regression coefficients).

Why favor cognitive?Why favor cognitive?

Long history of cognitive testing with Long history of cognitive testing with relatively clear constructs, measures, statistical relatively clear constructs, measures, statistical properties.properties.

Widespread adoption.Widespread adoption. Skills logically assumed to be cognitive. Skills logically assumed to be cognitive. (At the time of HC development, economists (At the time of HC development, economists

did not associate with psychologists. That has did not associate with psychologists. That has changed today.)changed today.)

Non-cognitive AmbiguityNon-cognitive Ambiguity

Distinction not clear—interacts with cognitive.Distinction not clear—interacts with cognitive. Constructs and measurement are contested.Constructs and measurement are contested. Labels differ:Labels differ:

Social and Emotional (SEL).Social and Emotional (SEL).

Social and Behavioral Social and Behavioral

Attitudes and ValuesAttitudes and Values

Early non-cognitive InterestEarly non-cognitive Interest

Coleman (1966) included efficacy, motivation, Coleman (1966) included efficacy, motivation, and other non-cognitive variables—didn’t use and other non-cognitive variables—didn’t use them.them.

Bowles and Gintis (1976) and Gintis (1971) Bowles and Gintis (1976) and Gintis (1971) argued that schools mainly emphasize non-argued that schools mainly emphasize non-cognitive development.cognitive development.

Rejected by most economists.Rejected by most economists.

Cognitive Test Scores or Behavioral Cognitive Test Scores or Behavioral SocializationSocialization

If cognitive skills explain effect schooling effect on If cognitive skills explain effect schooling effect on earnings in human capital equations, addition of test scores earnings in human capital equations, addition of test scores should deflate severely the earnings coefficient for should deflate severely the earnings coefficient for schooling.schooling.

Schooling variable is highly errorful (no adjustment for Schooling variable is highly errorful (no adjustment for quality, effort, area of study, retrospective errors).quality, effort, area of study, retrospective errors).

25 studies with 58 estimates over four decades.25 studies with 58 estimates over four decades. School coefficient retains about 82 percent of its value after School coefficient retains about 82 percent of its value after

test scores are added to the equation. test scores are added to the equation.

*ln uExpSchoolY

Has effect of cognitive test score Has effect of cognitive test score increased?increased?

Murnane, Willett, Levy (1995) compared test Murnane, Willett, Levy (1995) compared test score impacts on earnings between 1978 & score impacts on earnings between 1978 & 1986 and found rise in hourly wages related to 1986 and found rise in hourly wages related to math scores.math scores.

Bowles, Gintis,Osborne (2001) reviewed 65 Bowles, Gintis,Osborne (2001) reviewed 65 estimates from 24 studies over 30 years and estimates from 24 studies over 30 years and found no rise.found no rise.

UK study for 1995-2004 also finds no increase UK study for 1995-2004 also finds no increase in returns to cognitive skills.in returns to cognitive skills.

Other evidenceOther evidence

Teenage cognitive skills in predicting later Teenage cognitive skills in predicting later earnings (Murnane et al. 2000).earnings (Murnane et al. 2000).

Murnane et al. equations including test scores Murnane et al. equations including test scores explain no more than 17 percent of variance.explain no more than 17 percent of variance.

NRC study of supervisor ratings of worker NRC study of supervisor ratings of worker productivity found that only about 6 percent of productivity found that only about 6 percent of variance was explained.variance was explained.

Employer NeedsEmployer Needs

NRC study of High School Graduates by NRC study of High School Graduates by Employer Panel (1984)Employer Panel (1984)

Examined Cognitive needs, but also Examined Cognitive needs, but also behavioral and social.behavioral and social.

Interacting in socially appropriate manner; Interacting in socially appropriate manner; respect for opinions, customs, and individual respect for opinions, customs, and individual differences of others.differences of others.

Handling conflict maturely; participation Handling conflict maturely; participation reaching group decisions.reaching group decisions.

NRC study continued.NRC study continued.

Positive attitude toward self.Positive attitude toward self. Self-discipline.Self-discipline. Regular attendance, punctuality, Regular attendance, punctuality,

dependability.dependability. Ability to set goals and allocate time to Ability to set goals and allocate time to

achievement of them.achievement of them. Capacity to accept responsibility.Capacity to accept responsibility.

Employer Employment Survey Employer Employment Survey (1995) U.S. Dept. Ed.(1995) U.S. Dept. Ed.

4,000 employers4,000 employers Recruitment characteristics for hiring. (1-5 Recruitment characteristics for hiring. (1-5

high)high) Applicant attitude (4.6); communication skills Applicant attitude (4.6); communication skills

(4.2). (4.2). Test scores, academic grades, reputation of Test scores, academic grades, reputation of

applicants school (2.4 or 2.5).applicants school (2.4 or 2.5).

Independent Effects of non-cognitive Independent Effects of non-cognitive (Perry Preschool)(Perry Preschool)

Perry preschool—initial test score advantage Perry preschool—initial test score advantage faded in early grades.faded in early grades.

But assessment at age 40 showed: earnings But assessment at age 40 showed: earnings advantage of one third; half the crime advantage of one third; half the crime conviction rate. conviction rate.

Higher graduation rates and more positive Higher graduation rates and more positive attitudes towards schools.attitudes towards schools.

Something other than academic achievement Something other than academic achievement must explain.must explain.

GED--HeckmanGED--Heckman

GED recipients had similar test results as regular high GED recipients had similar test results as regular high school graduates who do not enroll in college.school graduates who do not enroll in college.

GED earnings well below high school graduates and GED earnings well below high school graduates and even lower than dropouts with similar test scores.even lower than dropouts with similar test scores.

Ultimate educational attainments lower than for Ultimate educational attainments lower than for dropouts who do not take GED.dropouts who do not take GED.

Something else happening—lower non-cognitive Something else happening—lower non-cognitive skills that affect employment.skills that affect employment.

Tennessee Class Size ExperimentTennessee Class Size Experiment

Dramatic reductions in class size for 1-4 years, Dramatic reductions in class size for 1-4 years, K-3.K-3.

Modest increases in achievement.Modest increases in achievement. Dramatic increases in high school graduation Dramatic increases in high school graduation

rates. (disadvantaged from 70 percent (no rates. (disadvantaged from 70 percent (no reduction of class size to 88 percent for 4 years reduction of class size to 88 percent for 4 years of early class size reductions).of early class size reductions).

Well beyond the impact of prediction from Well beyond the impact of prediction from early achievement advantage.early achievement advantage.

Can non-cognitive be taught?Can non-cognitive be taught?

Experimental study of teaching of executive Experimental study of teaching of executive function (self-regulation)—more important function (self-regulation)—more important than cognitive measures for school readiness.than cognitive measures for school readiness.

Experimental study of Tools of the Mind Experimental study of Tools of the Mind shows dramatic effects on EF. (Barnett 2011)shows dramatic effects on EF. (Barnett 2011)

Bandura cites impressive data-base that shows Bandura cites impressive data-base that shows self-efficacy can be taught.self-efficacy can be taught.

Early Childhood Education Early Childhood Education SummarySummary

Nores and Barnett (2010)—38 studies Nores and Barnett (2010)—38 studies reviewing 30 pre-school interventions in 23 reviewing 30 pre-school interventions in 23 countries RFT or quasi-experiments. Found countries RFT or quasi-experiments. Found both cognitive and non-cognitive benefits.both cognitive and non-cognitive benefits.

Camilli et al (2010) meta-analysi of 123 Camilli et al (2010) meta-analysi of 123 experimental preschool studies showed effects experimental preschool studies showed effects on cognitive skills, student social skills and on cognitive skills, student social skills and school progress.school progress.

Durlak (2011) Social-Emotional Durlak (2011) Social-Emotional LearningLearning

213 school based programs, experimental or quasi-213 school based programs, experimental or quasi-experimental. 270,000 children 5-18.experimental. 270,000 children 5-18.

Non-cognitive (SEL) interventions.Non-cognitive (SEL) interventions. Effect sizes: social & emotional skills, .57; attitudes towards Effect sizes: social & emotional skills, .57; attitudes towards

self and others, .23; positive social behaviors, .24; problems of self and others, .23; positive social behaviors, .24; problems of conduct, .22; emotional distress, .24; and academic conduct, .22; emotional distress, .24; and academic performance, .27, rising to .32 for longer term studies.performance, .27, rising to .32 for longer term studies.

Raise student achievement by 11 percentiles.Raise student achievement by 11 percentiles. Equivalent to closing the gap between U.S. and Canada and Equivalent to closing the gap between U.S. and Canada and

rise in rankings from 17rise in rankings from 17thth to 5 to 5thth place. place.

Heckman and ColleaguesHeckman and Colleagues

Use longitudinal (NLSY) to analyze students educational Use longitudinal (NLSY) to analyze students educational investment on earnings and graduation.investment on earnings and graduation.

Creates battery of non-cognitive scores from anti-social Creates battery of non-cognitive scores from anti-social construct using student anxiety, headstrongness, hyperactivity, construct using student anxiety, headstrongness, hyperactivity, peer conflict along with cognitive test scores.peer conflict along with cognitive test scores.

Models developmental path and impacts of investment in Models developmental path and impacts of investment in cognitive and non-cognitive skill on high school graduation cognitive and non-cognitive skill on high school graduation and earnings. Impact of the two kinds of investments shift and earnings. Impact of the two kinds of investments shift from cognitive at younger ages 6-9 to older ages 9-13.from cognitive at younger ages 6-9 to older ages 9-13.

Swedish Study Swedish Study (Lindqvist, E. & Vestman, R. (2011) “The Labor Market (Lindqvist, E. & Vestman, R. (2011) “The Labor Market

Returns to Cognitive and Noncognitive Ability: Evidence from the Swedish Enlistment,” Returns to Cognitive and Noncognitive Ability: Evidence from the Swedish Enlistment,” American Economic Review: Applied Economics, (3), 101-128American Economic Review: Applied Economics, (3), 101-128

Sample: 14703 men distributed evenly over Sample: 14703 men distributed evenly over 1965-74 birth.1965-74 birth.

Universal enlistment age 18 males.Universal enlistment age 18 males. Evaluation of cognitive ability (test) and Evaluation of cognitive ability (test) and

noncognitive ability 70-80 item questionnaire noncognitive ability 70-80 item questionnaire and interview by psychologist.and interview by psychologist.

Wages age 31-40.Wages age 31-40.

Swedish (cont.)Swedish (cont.)

Non-cognitive effects on wages and earnings are higher than Non-cognitive effects on wages and earnings are higher than cognitive for a range of specifications.cognitive for a range of specifications.

Non-cognitive more important at wages and earnings below Non-cognitive more important at wages and earnings below median and cognitive above median.median and cognitive above median.

Change in 1 sd in predictor:Change in 1 sd in predictor: Non-cognitive ES = .33; Cognitive ES = .15 for wages.Non-cognitive ES = .33; Cognitive ES = .15 for wages. Employment Status: 3.3% to 1.1%.Employment Status: 3.3% to 1.1%.

More measurement error in non-cognitive than cognitive measures.More measurement error in non-cognitive than cognitive measures.

Five Factor ModelFive Factor Model

For last two decades considered as the “big five” on the basis For last two decades considered as the “big five” on the basis of many studies.of many studies.

1.1. Openness—inventive and curious as opposed to consistent and cautious.Openness—inventive and curious as opposed to consistent and cautious.

2.2. Conscientiousness—efficient and organized as opposed to easy-going and Conscientiousness—efficient and organized as opposed to easy-going and careless.careless.

3.3. Extraversion—outgoing and energetic as opposed to solitary and reserved.Extraversion—outgoing and energetic as opposed to solitary and reserved.

4.4. Agreeableness—friendly and compassionate as opposed to cold and Agreeableness—friendly and compassionate as opposed to cold and unkind.unkind.

5.5. Neuroticism—sensitive and nervous as opposed to secure and confident.Neuroticism—sensitive and nervous as opposed to secure and confident.

Examples of big five findings.Examples of big five findings.

Major study finds openness predicts SAT.Major study finds openness predicts SAT. Agreeableness and conscientiousness Agreeableness and conscientiousness

predictedpredicted

Peer ratings of team member performance Peer ratings of team member performance beyond controls for specific skills and general beyond controls for specific skills and general cognitive ability.cognitive ability.

Kyllonen and colleagues go well-beyond big Kyllonen and colleagues go well-beyond big five in their extensive study of noncognitive five in their extensive study of noncognitive constructs and measures.constructs and measures.

Most Comprehensive ReviewMost Comprehensive Review

Almlund, M., Duckworth, A. L., Heckman, J. & Almlund, M., Duckworth, A. L., Heckman, J. & Kautz, T. (2011) Kautz, T. (2011) Personality Psychology and Personality Psychology and

EconomicsEconomics, (January 17, 2001 version) ¨IZA , (January 17, 2001 version) ¨IZA Workshop: Cognitive and Non-CognitiveWorkshop: Cognitive and Non-Cognitive

Skills¨, January 25-27, (Bonn, Germany). Available Skills¨, January 25-27, (Bonn, Germany). Available as:as:

http://iza.org/conference_files/CoNoCoSk2011/heckhttp://iza.org/conference_files/CoNoCoSk2011/heckman_j130.pdf.man_j130.pdf.

ConclusionsConclusions

Non-cognitive skills are important contributors Non-cognitive skills are important contributors to educational outcomes and life outcomes.to educational outcomes and life outcomes.

Can be altered by educational interventions.Can be altered by educational interventions. Contribute to cognitive outcomes.Contribute to cognitive outcomes. Impact of test scores is overstated if Impact of test scores is overstated if

noncognitive measures are omitted from noncognitive measures are omitted from analysis (omitted variable bias).analysis (omitted variable bias).

ImplicationsImplications

Ignoring role of non-cognitive skills as Ignoring role of non-cognitive skills as mediating and outcome variables leads to mediating and outcome variables leads to misleading educational policies.misleading educational policies.

NCLB with its focus on test scores only.NCLB with its focus on test scores only. Basing school and teacher evaluations only on Basing school and teacher evaluations only on

value-added in test scores provides only value-added in test scores provides only limited picture of school and teacher effects.limited picture of school and teacher effects.

Next StepsNext Steps

Choose most promising non-cognitive constructs and Choose most promising non-cognitive constructs and measures.measures.

Test these in structural models of learning outcomes Test these in structural models of learning outcomes and life outcomes.and life outcomes.

Non-cognitive outcomes affect cognitive outcomes Non-cognitive outcomes affect cognitive outcomes and possibly reverse—need to evaluate this.and possibly reverse—need to evaluate this.

Start to incorporate these into national and Start to incorporate these into national and international assessments of learning and impacts of international assessments of learning and impacts of educational systems.educational systems.