Human Capital, Age Structure and Economic Growth

Jesús Crespo-Cuaresma and Wolfgang Lutz World Population Program, International Institute for Applied Systems Analysis

(IIASA)

2

Education Matters

• For individual income (Micro)• For economic development (Macro)

“Whereas at the micro case ... it is established beyond any reasonable doubt that there are tangible and measurable returns to investment in education, such evidence is not as consistent and forthcoming in the macro literature “ (Psacharopoulos and Patrinos 2002)

Findings are dependent on education indicators chosen (mean years of schooling, for which age groups, distribution by attainment level) and quality (consistency across countries and time)

• For health/mortality and fertility • Indirect effects of education on institutions and good

governance

3

Measuring Formal Education

• Education Flows – Policy variable (Gross and Net Enrolment by Age, Repetition Rates)

• Education Stocks - Change very slowly due to great momentum– Mean years of schooling– Distribution by highest educational attainment– Functional literacy (IALS, LAMP)

4

5

6

7

Purpose of the Reconstruction1. Assessments of the returns to education at the aggregate

(national) level require empirical information about the educational status of the adult population over some time period.

2. This information needs to be consistent in terms of the definition

of educational categories across countries and over time.

3. Since the effects of educational attainment can be expected to differ by age (e.g. the education of 20-30 year olds being more important for economic growth than that of 65-75 year olds) as well as by sex, having full age detail for men and women can be considered a great asset.

4. Only the explicit consideration of distinct levels of educational attainment allows for the analysis of the relative importance of primary versus secondary or tertiary education (and different mixes of the three) which should be key to all national and international education policy plans.

8

Previous Work• Official data from censuses (and some sample surveys) as collected by

UNESCO are only very fragmentary and scattered over time and countries and have the problem of changing definitions.

• Barro and Lee (1993, 1993 and 2000) undertook the ambitious effort to complement these data with the somewhat more consistent time series of national school enrollment data at different levels using perpetual inventory methods. This resulted in a widely used data set that gives mean years of schooling of the entire adult population 15+ and 25+ (by sex but without age detail) for 107 countries for 1960 to 2000.

• Similar independent efforts have since been made by Kyriacou (1991), Lau, Jamison and Louat (1991), Lau, Bhalla and Louat (1991), Nehru, Swanson and Dubey (1995), De la Fuente and Domenech (2001) and by Cohen and Soto (2001) which in many cases result in quite different estimates of mean years of schooling.

• None of these reconstruction efforts gives the desirable age detail combined with the distribution over different educational attainment groups and accounts for the fact that mortality differs by education.

9

Our method: Demographic multi-state back projections from 2000

• We need only one reliable data point for each country (typically around the year 2000) which gives the total population by sex, 5-year age groups and the four attainment categories based on ISCED: no education, primary, secondary and tertiary.

• We use population age structures as estimated by the UN Population Division in five year intervals for all countries in the world since 1950.

• Taking these UN estimates as a basis, our reconstruction effort is reduced to estimating the proportions in each of the four education categories for each 5-year age group of men and women.

10

Our four education categories for 2000based on ISCED, from censuses and surveys (mostly DHS)

• No education Never been to school

• Primary Some primary, complete primary, incomplete lower secondary

• Secondary Completed lower secondary to incomplete first level of tertiary

• Tertiary Completed first level of tertiary or higher.

11

2000 (India)

60000 40000 20000 0 20000 40000 60000

15-19

20-24

25-29

30-34

35-39

40-44

45-49

50-54

55-59

60-64

Population in thousands

no edu

primary

secondary

tertiary

12

1990 (India)

60000 40000 20000 0 20000 40000 60000

15-19

20-24

25-29

30-34

35-39

40-44

45-49

50-54

55-59

60-64

Population in thousands

no edu

primary

secondary

tertiary

13

1980 (India)

60000 40000 20000 0 20000 40000 60000

15-19

20-24

25-29

30-34

35-39

40-44

45-49

50-54

55-59

60-64

Population in thousands

no edu

primary

secondary

tertiary

14

1970 (India)

60000 40000 20000 0 20000 40000 60000

15-19

20-24

25-29

30-34

35-39

40-44

45-49

50-54

55-59

60-64

Population in thousands

no edu

primary

secondary

tertiary

15

Sample TableIndia

Population distribution ('000) by age, sex and level of education and mean years of schooling (MYS)

MalesNo Edu. Primary Secondary Tertiary MYS

2000 15-19 8867.9 14274.4 29681.6 0.0 6.620-24 8917.9 10326.9 23088.3 3906.1 7.225-29 10535.7 10117.4 18153.7 5191.1 6.930-34 11251.9 9467.0 14566.2 4328.1 6.435-39 11258.2 9195.7 11243.9 3274.6 5.740-44 10281.5 8067.7 9242.1 2785.4 5.645-49 8523.5 6780.7 7453.6 2298.9 5.550-54 7591.1 5223.6 5313.0 1683.8 5.155-59 6255.8 4633.0 3937.7 1157.7 4.760-64 6389.9 3614.5 2346.7 619.9 3.665+ 12124.9 7331.6 3515.5 855.0 3.215+ 101998.4 89032.5 128542.2 26100.7 5.925+ 84212.5 64431.2 75772.3 22194.6 5.5

16

2000 (Singapore)

250 200 150 100 50 0 50 100 150 200 250

15-19

20-24

25-29

30-34

35-39

40-44

45-49

50-54

55-59

60-64

Population in thousands

no edu

primary

secondary

tertiary

17

1990 (Singapore)

250 200 150 100 50 0 50 100 150 200 250

15-19

20-24

25-29

30-34

35-39

40-44

45-49

50-54

55-59

60-64

Population in thousands

no edu

primary

secondary

tertiary

18

1980 (Singapore)

250 200 150 100 50 0 50 100 150 200 250

15-19

20-24

25-29

30-34

35-39

40-44

45-49

50-54

55-59

60-64

Population in thousands

no edu

primary

secondary

tertiary

19

1970 (Singapore)

250 200 150 100 50 0 50 100 150 200 250

15-19

20-24

25-29

30-34

35-39

40-44

45-49

50-54

55-59

60-64

Population in thousands

no edu

primary

secondary

tertiary

20

When moving back along cohort lines in 5-year intervals we take advantage of the demographic identity:

P(x,t,) = P(x+5,t+5) + Deaths +/- Migrants

P(x,t,e) = P(x+5,t+5,e) + Deaths (,,e) +/- Migrants (,,e) +/- EducTransitions (x to x+5, t to t+5,e)

Necessary Assumptions:

1. Education differentials in mortality and migration

2. Dealing with the open interval in highest age group

3. Age of transition to higher educational categories

21

Differential Mortality by Education

• We defined mortality differentials in terms of life expectancy at age 15 (child mortality is more a function of parents’ education).

• We carried out numerous studies for countries for which two reliable subsequent censuses (or public use samples of them) were available to assess differential mortality (combined with migration) along cohort lines.

• Despite all differentials among countries a rather general pattern appeared: The differentials between the lower categories are smaller than to the higher.

• Our choice: No education – Primary: 1 year difference

Primary – Secondary: 2 years difference

Secondary – Tertiary: 2 years difference

22

Differential Mortality by EducationRelative Difference in Proportions (Number with Differential - Number

without Differential)/Number with Differential India 1970

-90%

-80%

-70%

-60%

-50%

-40%

-30%

-20%

-10%

0%

10%

20%

15 - 19 20 - 24 25 - 29 30 - 34 35 - 39 40 - 44 45 - 49 50 - 54 55 - 59 60 - 64

Age Group

Rel

ativ

e D

iffe

ren

ce i

n P

rop

ort

ion

s

male NoEdu

male Prim

male Sec

male Ter

female NoEdu

female Prim

female Sec

female Ter

23

Validation of Results(April – October 2006)

The first reconstruction results have been checked against much available historical evidence on educational attainment distributions, in particular:

• UIS Attainment Data Set• Earlier DHS, WFS or other surveys• OECD and Eurostat education data base• Numerous individual national statistical agencies (in personal

communication)

Whenever there was a significant discrepancy, we tried to identify the reason: Was it a problem with our assumptions (mostly migration) or a problem with the other data (mostly definition of categories or representativeness of survey) and made appropriate adjustments in our data, if necessary.

24

Exploiting the age dimension of human capital

1) How does the new data compare to the Barro-Lee dataset in terms of correlation with growth? (Sala-i-Martin, 1997)

2) Do differences in cohort education help us better explain differences in income per capita? (Hall and Jones, 1998)

3) What is the role of age-specific human capital in technology adoption and growth? (Benhabib and Spiegel, 1994)

25

Growth and education: a robustness exercise

Are changes in educational attainment a robust determinant of growth?

A stylised growth regression:

How does the sign, size and significance of depend on the set of conditioning variables?

J

jiijjii zEDy

1,

26

Growth and education: a robustness exercise

- IIASA/VID dataset versus Barro-Lee- 59 other variables in the pool of Z-variables- Initial GDP per capita, primary school enrolment and life expectancy as fixed variables

J

jiijjii zEDy

1,

27

Unweighted results

Variable Average parameter t-statistic Empirical % % significantBarro-Lee, 25+ 0.0024 1.114 95.3% 13.1%

IIASA/VID 15 -19 0.0042 1.793 99.2% 64.3%IIASA/VID 20 -24 0.0038 1.894 99.8% 66.1%IIASA/VID 25 -29 0.0035 1.956 99.8% 66.4%IIASA/VID 30 -34 0.0033 1.992 99.4% 67.8%IIASA/VID 35 -39 0.0033 2.043 99.6% 70.6%IIASA/VID 40 -44 0.0037 2.155 99.5% 77.0%IIASA/VID 45 -49 0.0038 2.258 99.7% 84.5%IIASA/VID 50 -54 0.0040 2.248 99.9% 84.3%IIASA/VID 55 -59 0.0038 1.892 99.9% 67.0%IIASA/VID 60 -64 0.0038 1.618 99.6% 53.8%

IIASA/VID 65+ 0.0014 0.435 83.6% 0.1%IIASA/VID 15+ 0.0061 2.415 99.9% 84.1%IIASA/VID 25+ 0.0056 2.596 99.9% 89.8%

28

Weighted results

Variable Average parameter t-statistic Empirical % % significantBarro-Lee, 25+ -0.0013 -0.743 95.3% 13.1%

IIASA/VID 15 -19 0.0034 2.089 99.2% 64.3%IIASA/VID 20 -24 0.0027 1.798 99.8% 66.1%IIASA/VID 25 -29 0.0020 1.395 99.8% 66.4%IIASA/VID 30 -34 0.0016 1.211 99.4% 67.8%IIASA/VID 35 -39 0.0014 0.993 99.6% 70.6%IIASA/VID 40 -44 0.0010 0.659 99.5% 77.0%IIASA/VID 45 -49 0.0010 0.659 99.7% 84.5%IIASA/VID 50 -54 0.0008 0.506 99.9% 84.3%IIASA/VID 55 -59 0.0009 0.516 99.9% 67.0%IIASA/VID 60 -64 0.0008 0.417 99.6% 53.8%

IIASA/VID 65+ -0.0017 -0.661 83.6% 0.1%IIASA/VID 15+ 0.0035 1.669 99.9% 84.1%IIASA/VID 25+ 0.0031 1.656 99.9% 89.8%

29

The contribution of human capital to income differences

where () is a piecewise linear function based on average returns to schooling worldwide.

,

,

,)(

)(

)1/(

1

iE

i

ii

i

i

ii

iiii

LeH

AL

H

Y

Ky

HAKY

i

30

Income decomposition

31

Human capital and technology adoption

iiii

iii lky

yHHy

43

max210

32

Human capital and technology adoption

33

Human capital and technology adoption

34

Conclusions

Exploiting the age structure of human capital:

- we obtain variables that correlate robustly with economic growth,

- we are better able to explain differences in income per capita across countries,

- we gain new insights to the nature of the interaction between education and technology adoption.

35

Paths of ongoing and further research

- Exploiting the age structure of human capital in the time dimension: panel data analysis.

- Interaction between gender and education (“If you educate a man, you educate a person, if you educate a woman you educate a family, and a whole nation”)

- Economic growth projections: using age-structure human capital to improve reliability of forecasts.