1 Cognitive Skills: Determinants and Labour Market Outcomes W. Craig Riddell University of British Columbia WISE conference Xiamen, China December 12-13,

Slides:



Advertisements
Similar presentations
The Well-being of Nations
Advertisements

Chapter 5 Human Capital Copyright © 2010 by The McGraw-Hill Companies, Inc. All rights reserved. McGraw-Hill/Irwin.
Written by: Brahim Boudarbat, Thomas Lemieux, and Craig Riddell Analyzed by: Mico Radulovic, Ben May, Dajana Sormaz, Bryan Taylor, and Arjun Arulambala.
Random Assignment Experiments
A “Best Fit” Approach to Improving Teacher Resources Jennifer King Rice University of Maryland.
Correlation and regression Dr. Ghada Abo-Zaid
3.2 OLS Fitted Values and Residuals -after obtaining OLS estimates, we can then obtain fitted or predicted values for y: -given our actual and predicted.
DEMOGRAPHIC TRANSITION AND ECONOMIC DEVELOPMENT AT THE LOCAL LEVEL IN BRAZIL Ernesto F. L. Amaral Advisor: Dr. Joseph E. Potter Population Research Center.
Income distribution, labour market returns and school quality REDI3x3 Income Distribution Workshop 4 November 2014 Rulof Burger.
Pooled Cross Sections and Panel Data II
BACKGROUND RESEARCH QUESTIONS  Does the time parents spend with children differ according to parents’ occupation?  Do occupational differences remain.
Simple Linear Regression
Fall 2006 – Fundamentals of Business Statistics 1 Chapter 13 Introduction to Linear Regression and Correlation Analysis.
So far, we have considered regression models with dummy variables of independent variables. In this lecture, we will study regression models whose dependent.
Linear Regression and Correlation Analysis
1 Simple Linear Regression Chapter Introduction In this chapter we examine the relationship among interval variables via a mathematical equation.
Chapter 13 Introduction to Linear Regression and Correlation Analysis
Autocorrelation Lecture 18 Lecture 18.
Stat 112: Lecture 9 Notes Homework 3: Due next Thursday
PIAAC data in Canada and the United States Satya Brink, Ph.D
Simple Linear Regression. Introduction In Chapters 17 to 19, we examine the relationship between interval variables via a mathematical equation. The motivation.
Learning Policy Directorate, HRSDC 1 IALSS 2003 Literacy and Labour Force and HRSDC Implications of Findings Part II Presented by Satya Brink, Ph.D. Director,
Regression and Correlation Methods Judy Zhong Ph.D.
Hypothesis Testing in Linear Regression Analysis
Haripriya Gundimeda Associate Professor Department of Humanities and Social Sciences Indian Institute of Technology Bombay Human capital estimates for.
The Retirement Prospects of Immigrants: Getting Worse? Presentation to PMC Winnipeg Node Meeting September 29, 2009 Derek Hum Wayne Simpson.
PIAAC Programme for the International Assessment of Adult Competencies A European Survey on Skills at Work (ESSW) Workshop on “Exploring possibilities.
The Changing Economic Advantage from Private School* Francis Green *Talk based on: Green, F., S. Machin, R. Murphy and Y. Zhu (2010). The Changing Economic.
Research methods in adult development
1 Least squares procedure Inference for least squares lines Simple Linear Regression.
COURSE: JUST 3900 INTRODUCTORY STATISTICS FOR CRIMINAL JUSTICE Instructor: Dr. John J. Kerbs, Associate Professor Joint Ph.D. in Social Work and Sociology.
Welfare Reform and Lone Parents Employment in the UK Paul Gregg and Susan Harkness.
Parents’ basic skills and children’s test scores Augustin De Coulon, Elena Meschi and Anna Vignoles.
Growing Up and Moving On: Family Involvement in Transition Lauren Lindstrom, Ph.D. University of Oregon Youth Transition Program Conference February 16,
Comments on: Higher Education and Social Mobility in the United States: A Glimpse Inside the Black Box? Lars Osberg Economics Department Dalhousie University.
Correlation Analysis. A measure of association between two or more numerical variables. For examples height & weight relationship price and demand relationship.
HAOMING LIU JINLI ZENG KENAN ERTUNC GENETIC ABILITY AND INTERGENERATIONAL EARNINGS MOBILITY 1.
International Adult Literacy and Skills Survey (IALSS)
MBP1010H – Lecture 4: March 26, Multiple regression 2.Survival analysis Reading: Introduction to the Practice of Statistics: Chapters 2, 10 and 11.
Chapter 4 Linear Regression 1. Introduction Managerial decisions are often based on the relationship between two or more variables. For example, after.
1 REGRESSION ANALYSIS WITH PANEL DATA: INTRODUCTION A panel data set, or longitudinal data set, is one where there are repeated observations on the same.
Chapter 11 Correlation and Simple Linear Regression Statistics for Business (Econ) 1.
Competition to get into a Post- Secondary institution has greatly increased and the change in the school curriculum has, and will continue to, affect.
Accounting for the Effect of Health on Economic Growth David N. Weil Proponent/Presenter Section.
(1)Combine the correlated variables. 1 In this sequence, we look at four possible indirect methods for alleviating a problem of multicollinearity. POSSIBLE.
Chapter 13 Repeated-Measures and Two-Factor Analysis of Variance
1 Literacy & Numeracy Do They Really Matter? W. Craig Riddell University of British Columbia Ministry of Training, Colleges and Universities Toronto, Ontario.
Chapter 6: Analyzing and Interpreting Quantitative Data
Chapter 8: Simple Linear Regression Yang Zhenlin.
Unit 2: Research & Statistics n Psychology deals with many experiments and studies n WHO? Every experimenter must decide on a SAMPLE, which is a group.
1 Cognitive Skills: Determinants and Labour Market Outcomes W. Craig Riddell University of British Columbia RCEA Labour Workshop Rimini, Italy August,
Education Quality and Economic Growth G Balasubramanian National conference of Sahodaya Complexes 24 th December 2012.
1 The Training Benefits Program – A Methodological Exposition To: The Research Coordination Committee By: Jonathan Adam Lind Date: 04/01/16.
Does Aid Matter? Measuring the Effect of Student Aid on College Attendance and Completion By SUSAN M. DYNARSKI Source: The American Economic Review, Vol.
1 Researchers : 10 altogether; 3 from Denmark, 2 from Finland, 2 from Norway, and 3 from Sweden. Scientific board: professor Antero Malin, Finland (project.
Slide Slide 1 Chapter 10 Correlation and Regression 10-1 Overview 10-2 Correlation 10-3 Regression 10-4 Variation and Prediction Intervals 10-5 Multiple.
Necessary but not sufficient? Youth responses to localised returns to education Nicholas Biddle Centre for Aboriginal Economic Policy Research, ANU Conference.
Choosing and using your statistic. Steps of hypothesis testing 1. Establish the null hypothesis, H 0. 2.Establish the alternate hypothesis: H 1. 3.Decide.
Targeting Fertility and Female Participation Through the Income Tax Ghazala Azmat (Universitat Pompeu Fabra) Libertad González (Universitat Pompeu Fabra)
Methods of multivariate analysis Ing. Jozef Palkovič, PhD.
Productivity Commission Anthony Shomos Productivity Commission Australian Conference of Economists 29 September 2009 Links between literacy and numeracy.
Chapter 13 Simple Linear Regression
Linear Regression.
More on Specification and Data Issues
More on Specification and Data Issues
Regression Analysis Week 4.
Motivation THIS TALK: 1. Documents a stagnation in the schooling attainment at age 25 of Spanish cohorts born after Can we explain the poorer.
Figures adapted from the TIEDI Analytical Report #11:
More on Specification and Data Issues
Presentation transcript:

1 Cognitive Skills: Determinants and Labour Market Outcomes W. Craig Riddell University of British Columbia WISE conference Xiamen, China December 12-13, 2008

2 Introduction Most research on human capital uses indirect measures like education and work experience These are inputs into the production of human capital, not outcomes -- skills, competencies and knowledge Relatively little is known about the relationship between direct measures of skills and labour market outcomes Advances in data collection allow us to explore these linkages

3 Presentation draws on research using Canadian surveys IALS 1994 and IALSS 2003 Data combine methods of educational testing with household survey techniques In IALSS 2003, skills were assessed in four domains: –Prose literacy, Document literacy, Numeracy, Problem-solving Key feature: these are skills used in daily activities, not ability measures Provide measures of the skills of the adult population

4 eg: Document literacy: skills required to find and use information contained in text formats such as job applications, payroll forms, bus schedules, maps, graphs Numeracy questions changed considerably between 1994 and 2003 Problem solving only assessed in 2003 survey Prose and document literacy: about 45% of questions were identical in two surveys Scores in remaining questions were rescaled to match differences observed in identical questions We treat prose and document scores as being comparable in the two years

5 Sample sizes: IALS IALSS Sample restrictions: Drop immigrants, aboriginals, students, those with missing information on education Resulting sample sizes: IALS IALSS Earnings sample drops self-employed, unemployed, non- participants, wage outliers Correlation among test scores: –Prose and Document.96 –Document and Numeracy.92 –Document and Problem solving.92 Factor analysis: one principal component placing essentially equal weight on all four scores

6 Determinants of skills I Simple heuristic model Individuals start life with ability and parental resources These plus formal schooling influence cognitive skills before legal school leaving age After that age students decide whether to remain in school; more skilled individuals face lower costs After HS, continuation depends on individual choice and admission decision; both depend on cognitive skills Thus cognitive skills and schooling are jointly determined

7 Determinants of skills II Table 3 reports log skills regressions Small gender difference Essentially no relationship between skills and age in cross-section (e.g. impact of 1 year at age 30 = - 0.1%) G&R (2003) also found no relationship between literacy skills and age or experience with IALS 94 data Strong relationship with formal schooling, but concave Suggests skills acquired via parenting and schooling, ‘locked in’ thereafter

8 Possible omitted variables bias Literacy skills and education may be influenced by ability Ideal control would be IQ type measure at a young age Columns 2 and 3 add proxies for unobserved ability Parental education only matters if parents dropouts (OLS2) Other parental influences (occupation, immigrant, working) minor Ease of learning mathematics in high school (OLS 3) Modest decline in coefficient on schooling (< 10%) OLS and IV estimates (columns 4 & 5): IV > OLS IV is province in which respondent attended HS interacted with age Regressions include controls for current province of residence

9 Table 3: Log of Document Literacy Regressions VariableOLS 1OLS 2OLS 3OLS 4IV Female-0.014**-0.012** *-0.012**-0.016** (0.0046)(0.0045) (0.0044)(0.0051) Years of Schooling0.058***0.054***0.053***0.024***0.052*** (0.0053)(0.0049)(0.0048)(0.0008)(0.013) Schooling Squared *** -- (0.0002) -- Age0.0035***0.0064***0.0066***(0.0068)*** (0.0008) (0.0036) Age Squared *** ***-0.011*** (0.0009) (0.0039) Mother’s Education Less than High School ***-0.038***-0.028*** (0.0059)(0.0058) (0.016) Some Post Secondary * -(0.0067)(0.0066)(0.0065)(0.011) BA or More (0.0102)(0.0098)(0.010)(0.016) None Reported ***-0.069***-0.048*** (0.0123)(0.0122)(0.011)(0.024)

10 Table 3: Log of Document Literacy Regressions continued VariableOLS 1OLS 2OLS 3OLS 4IV Father’s Education Less than High School ***-0.030***-0.029*** (0.0067) (0.012) Some Post Secondary (0.0073)(0.0071)(0.0073)(0.012) BA or More * (0.0083)(0.0082) (0.020) None Reported ***-0.055***-0.065*** (0.012) (0.011)(0.023) Immigrant Mother (0.0077)(0.0076)(0.0077)(0.011) Immigrant Father-0.015**0.016** (0.0071)(0.007)(0.0071)(0.011) Good Math Grades-0.027*** - -(0.0055) - Teachers Too Fast *** - -(0.0057) - Constant5.08***5.09*** 5.29***5.04*** (0.04)(0.038) (0.020)(0.12) Observations R-squared

11 Age and cohort effects I Flat skills - age profile in cross-section may reflect age and cohort effects e.g. 35 year old in 2003 may differ from 25 year old in 2003 both because she is older and comes from an earlier cohort Use IALS94 and IALSS03 to create synthetic cohorts e.g. observe skills of year olds in 1994 and year olds in 2003 Each is a random sample of birth cohort Also examine cohorts by education To ensure education doesn’t change, focus on individuals over age 25

12 Age and cohort effects II Cohorts defined for age ranges: –26 to 35 in 1994(35 to 44 in 2003) –36 to 45 in 1994(45 to 54 in 2003) –46 to 55 in 1994(55 to 64 in 2003) –56 to 65 in 1994 (65 to 74 in 2003) –65+ in 1994(74+ in 2003)

13 Age and cohort effects III First compare the same age groups in 1994 and 2003, shows cohort effects holding age constant e.g year olds in both years: 10th percentile increases from 223 to 248; 90th percentile declines from 363 to 354; median increases from 299 to 306 Breakdowns by education group: –dropouts: 10th percentile increases from 135 to 172 –college & university grad: 90th percentile drops by 10% Within cohort movements: –26 to 35 year olds in 1994: 10th percentile rises form 223 to 232, 90th declines from 363 to 344

14 Figure 2: Document Literacy, Age 26-35

15 Figure 3: Document Literacy, Age 36-45

16 Figure 4: Document Literacy, Age 46-55

17 Figure 5: Document Literacy, Age 56-65

18 Figure 6: Document Literacy, less than high school

19 Figure 7: Document Literacy, high school

20 Figure 8: Document Literacy, post-secondary non-university

21 Figure 9: Document Literacy, university

22 Figure 10: Document Literacy, age 26 to 35 in 1994

23 Figure 11: Document Literacy, age 36 to 45 in 1994

24 Figure 12: Document Literacy, age 46 to 55 in 1994

25 Figure 13: Document Literacy, age 56 to 65 in 1994

26 Figure 14: Document Literacy, less than high school, age 26 to 45 in 1994

27 Figure 15: Document Literacy, high school, age 26 to 45 in 1994

28 Figure 16: Document Literacy, university, age 26 to 45 in 1994

29 Table 4: Pooled Regressions Including Cohort Effects VariableOLS10th QuantileMedian90th Quantile Years School0.083***0.13**0.086***0.043*** (0.0067)(0.0052)(0.0031)(0.0028) School Squared-0.002*** *** *** *** (0.0002) (0.0001) Female * (0.0058)(0.0083)(0.0049)(0.0036) Age Group * *** (0.013)(0.018)(0.011)(0.0083) ** ***-0.12*** (0.016)(0.025)(0.015)(0.011) ***-0.084**-0.099***-0.17*** (0.021)(0.031)(0.019)(0.014) ***-0.18*** -0.30*** (0.033)(0.039)(0.026)(0.016) Cohort * *0.055*** (0.010)(0.018)(0.0097)(0.0078) Cohort ** **0.082*** (0.016)(0.025)(0.016)(0.012) Cohort **0.052*0.055**0.13*** (0.029)(0.032)(0.023)(0.014) Cohort *** (0.033)(0.039)(0.025)(0.015) Constant4.97***4.39***4.98***5.48*** (0.054)(0.047)90.025)(0.021) Observations14734

30 Table 5: Pooled Quantile Regressions with Cohort Effects, By Education Group HighSchoolor LessUniversity Variable10th QuantileMedian90th Quantile10th QuantileMedian90th Quantile Female0.064*** **-0.021** (0.013)(0.0049)(0.0047)(0.010)(0.0069)(0.0084) Age Group ***-0.027**-.064** * (0.021)(0.01)(0.012)(0.027)(0.012)(0.050) ***-.074* *** (0.035)(0.011)(0.015)(0.042)(0.015)(0.022) ***-0.12***-0.18***-0.092***-0.15*** (0.068)(0.016)(0.024)(0.045)(0.017)(0.026) ***-0.24***-0.31***-0.16***-0.15*** (0.12)(0.024)(0.027)(0.093)(0.032)(0.038) Cohort ***0.061*** (0.013)(0.010)(0.0094)(0.024)(0.017)(0.009) Cohort ** (0.053)(0.017)(0.016)(0.032)(0.021)(0.018) Cohort *** * (0.068)(0.021)(0.025)(0.085)(0.039)(0.023) Cohort *** ** *** (0.11)(0.027)(0.024)(0.11)(0.041)(0.035) Constant5.45***5.76***5.87***5.73***5.80***5.99*** (0.027)(0.011)(0.014)(0.032)(0.017)(0.015) Observations

31 Implications Cognitive skills decline with age Less evidence of decline at bottom of skill distribution, but strong negative age effects at the 90th percentile Literacy declines across successive cohorts, particularly at top of distribution Absence of relationship between cognitive skills and age in cross-section results from offsetting age and cohort effects

32 Skills and earnings: interpretive framework Simple framework builds on Green and Riddell (2003) who use the 1994 IALS to examine the role of cognitive skills in Canadian earnings patterns. Hedonic model in which earnings are determined by the skills an individual possesses and the prices of those skills.

33 Abstracting from other influences on earnings, individual earnings are a function of the skills an individual possesses and puts into use: E i = f (G 1i, G 2i, G 3i ) + e i (1) where E i are earnings for individual i, G ki is the amount of skill k that person i sells in the market, and e i is a disturbance term that is independent of the skills.

34 Typically, we do not observe the skills but we do observe some of the inputs that generate them. They enter via skill production functions: G ki = h k (edn i, exp i, v ki ) (2) where k indexes the skill type, edn corresponds to education, exp is years of work experience and v k is an ability specific to the production of the kth skill.

35 If we do not observe the G ki 's directly, we can obtain an estimating equation by substituting equation 2) into 1). This yields a reduced form specification for earnings as a function of education and experience. The coefficient on a covariate such as education reflects the combination of how that input contributes to production of each skill and how those skills contribute to earnings.

36 What can be learned when only some skills are observed? If we observe the cognitive skills G 1, we get a quasi- reduced form earnings regression that includes G 1 (the vector of cognitive skills), experience and education. The quasi-reduced form parameters on edn and exp now reflect the impact of education and experience on the production of skills other than the observed cognitive skills.

37 Table 7: Earnings Regressions OLS 1OLS 2IV 1IV 2 Female-0.41***-0.40*** (0.024) Years of School0.087***0.069***0.070***0.050*** (0.0042)(0.0047)(0.013)(0.017) Experience0.067*** 0.068*** (0.0035) (0.0041)(0.004) Experience Squared *** *** (0.0001) Average Skill Score *** *** -(0.0003)-(0.0007) Constant4.80***4.28***5.04***4.34*** (0.077)(0.10)(0.16)(0.10) Observations7768 R-squared

38 Cognitive skills and earnings Introducing cognitive skills reduces schooling coefficient by 20% (OLS) to 28% (IV). Earnings increase with age and experience. Since cognitive skills decline with age, non-cognitive skills must increase at a rapid rate Impact of cognitive skills on earnings is substantial. A 25 point increase in skills (half a standard deviation) is associated with an earnings increase equivalent to 1 extra year of schooling

Quantile regressions Table 8 shows estimates at 10 th, 25 th, 50 th, 75 th, 90 th quantiles Returns to schooling and experience decline across quantiles Coefficient on skills approximately constant across quantiles Implies cognitive skills do not interact with other (unobs) skills or attributes in generating earnings 39

40 Conclusions Conventional estimates of returns to education and experience confound two effects: –Skill production effect –Market valuation of skills effect Cognitive skills increase strongly with schooling Parental education has an impact, but not as large as might be expected Literacy & numeracy skills decline with age after leaving school, particularly at the top of the distribution Cognitive skills have been declining across cohorts, particularly at the top of the distribution

Conclusions Absence of relationship between age and skills in cross section appears due to offsetting age and cohort effects Cognitive skills account for about 20% of the returns to schooling (IV estimates suggest more) Literacy and numeracy skills have substantial effects on earnings 41