Cognitive Ability, Heterogeneity and Returns to Schooling in Chile: Outcomes of the 1981 Capitation Grant Scheme Harry Anthony Patrinos World Bank, Washington.

Cognitive Ability, Heterogeneity and Returns to Cognitive Ability, Heterogeneity and Returns to Schooling in Chile: Outcomes of the 1981 Capitation Schooling in Chile: Outcomes of the 1981 Capitation

Grant SchemeGrant Scheme

Harry Anthony PatrinosHarry Anthony PatrinosWorld Bank, Washington D.C.World Bank, Washington D.C.

Chris SakellariouChris SakellariouDepartment of Economics, Department of Economics,

Humanities, Arts and Social SciencesHumanities, Arts and Social Sciences

Nanyang Technological University, SingaporeNanyang Technological University, Singapore

IntroductionIntroduction

The main thread of this paper is the exploration of the role of cognitive ability in the The main thread of this paper is the exploration of the role of cognitive ability in the labor market in Chile and its interaction with schooling. labor market in Chile and its interaction with schooling.

To begin with, we hypothesize that quality of schooling contributes to the To begin with, we hypothesize that quality of schooling contributes to the development of better cognitive skills in the labor market.development of better cognitive skills in the labor market.

In the first part, we look at the bias in the estimate of the return to schooling when In the first part, we look at the bias in the estimate of the return to schooling when cognitive ability is not controlled for.cognitive ability is not controlled for.

In the second part, we examine the consequences of the 1981 vouchers policy, not In the second part, we examine the consequences of the 1981 vouchers policy, not on the average student in Chile, rather the group which acted upon the reform by on the average student in Chile, rather the group which acted upon the reform by switching to private subsidized schools or choosing to continue schooling (mainly at switching to private subsidized schools or choosing to continue schooling (mainly at the secondary level), as a response to the reform.the secondary level), as a response to the reform.

When estimating returns to schooling using a Mincer type earnings function, the When estimating returns to schooling using a Mincer type earnings function, the disturbance term will capture individual unobservable attributes which tend to disturbance term will capture individual unobservable attributes which tend to influence the schooling decisioninfluence the schooling decision→→ Correlation between schooling and the error term. Correlation between schooling and the error term.

Cognitive (as well as non-cognitive) ability is such an unobservable.Cognitive (as well as non-cognitive) ability is such an unobservable.

For example, Ingram and Neumann (2005) show that over the past decades, For example, Ingram and Neumann (2005) show that over the past decades, individuals with college education but without specific skills experienced the lowest individuals with college education but without specific skills experienced the lowest benefits from investing in education. benefits from investing in education.

Hence, individuals with low ability/skills may not benefit as much from investing in Hence, individuals with low ability/skills may not benefit as much from investing in

education, compared to individuals in the upper part of the ability distribution; for the education, compared to individuals in the upper part of the ability distribution; for the latter, ability is expected to interact positively with education resulting in higher latter, ability is expected to interact positively with education resulting in higher benefits from education investments.benefits from education investments.

Three possible empirical approaches:Three possible empirical approaches:

- Using twins- Using twins

- Using sources for exogenous variation in educational attainment, such as institutional - Using sources for exogenous variation in educational attainment, such as institutional changes in the schooling system (i.e, changes in compulsory schooling laws, or other changes in the schooling system (i.e, changes in compulsory schooling laws, or other “natural variations”). “natural variations”).

- Using achievement test scores measuring cognitive ability as additional controls - Using achievement test scores measuring cognitive ability as additional controls (keeping in mind that, both schooling and the test score may be generated by the (keeping in mind that, both schooling and the test score may be generated by the same latent ability - possible joint causality). same latent ability - possible joint causality).

DataData

International Adult Literacy Survey (IALS) data, which contain direct measures of International Adult Literacy Survey (IALS) data, which contain direct measures of cognitive ability. cognitive ability.

The International Adult Literacy Survey (IALS) was carried out in 20 countries between The International Adult Literacy Survey (IALS) was carried out in 20 countries between 1994 and 1998, a project undertaken by the governments of the countries and three 1994 and 1998, a project undertaken by the governments of the countries and three intergovernmental organizations, followed by an extensive quality review.intergovernmental organizations, followed by an extensive quality review.

Such data allows the researcher to identify features of earnings determination that are Such data allows the researcher to identify features of earnings determination that are

typically only indirectly observed.typically only indirectly observed. The data set includes three scales to measure individuals’ literacy levels - prose, The data set includes three scales to measure individuals’ literacy levels - prose,

document and quantitative literacy. document and quantitative literacy. ProseProse literacy tests the understanding and use of texts such as editorials, news literacy tests the understanding and use of texts such as editorials, news

stories, fiction and poems. stories, fiction and poems. DocumentDocument literacy tests skills required to locate and use literacy tests skills required to locate and use information contained in a variety of formats, such as job applications, payroll forms, information contained in a variety of formats, such as job applications, payroll forms, maps and tables. maps and tables. QuantitativeQuantitative literacy tests skills required in making calculations after literacy tests skills required in making calculations after locating numbers embedded in printed materials; examples of such calculations locating numbers embedded in printed materials; examples of such calculations include determining the interest on a loan, calculating a tip and balancing a include determining the interest on a loan, calculating a tip and balancing a checkbook.checkbook.

The IALS data have been used in several studies, with little evidence on Chile, the The IALS data have been used in several studies, with little evidence on Chile, the only country in the survey outside Europe and North America.only country in the survey outside Europe and North America.

Other studies for other countries using the IALS data found that cognitive skills Other studies for other countries using the IALS data found that cognitive skills contribute significantly to earnings and that their inclusion in earnings equations contribute significantly to earnings and that their inclusion in earnings equations reduces the measured impact of schooling. reduces the measured impact of schooling.

MethodologyMethodology

In the basic Mincerian human-capital model, schooling is assumed to be independent In the basic Mincerian human-capital model, schooling is assumed to be independent of ability, and that the return from schooling investments is equal for all individuals.of ability, and that the return from schooling investments is equal for all individuals.

One should allow ability to affect the rate of return to schooling investments.One should allow ability to affect the rate of return to schooling investments.

We use the IALS literacy score as a measure of cognitive ability to proxy for We use the IALS literacy score as a measure of cognitive ability to proxy for unobserved effects. unobserved effects.

2-pronged effect of introducing ability:2-pronged effect of introducing ability: Main effectMain effect: the more able individuals may be able to ‘convert’ schooling into human : the more able individuals may be able to ‘convert’ schooling into human

capital more efficiently than the less able, raising the return to schooling for the more capital more efficiently than the less able, raising the return to schooling for the more able. able. Other effectOther effect: the more able may have higher opportunity costs since they may : the more able may have higher opportunity costs since they may have been able to earn more in the labor market (if ability to progress in school is have been able to earn more in the labor market (if ability to progress in school is positively correlated with the ability to earn), and this reduces the rate of return to positively correlated with the ability to earn), and this reduces the rate of return to schooling. schooling.

Starting from: Ln wStarting from: Ln wii = α + βS = α + βSi i + γA+ γAii + ε + εii

Allowing the return to schooling depend on ability (A):Allowing the return to schooling depend on ability (A): Ln wLn wii = α + β(f(A = α + β(f(Aii))S))Sii + γA + γAii + ε + εii

using the IALS test score (T):using the IALS test score (T): Ln wLn wii = α + β(f(T = α + β(f(Tii))S))Sii + γT + γTii + ε + εi i

Assuming linear relationship between test score and ability and adding experience Assuming linear relationship between test score and ability and adding experience and its square:and its square:

Ln wLn wii = α + t = α + t00SSii + t + t11TTiiSSi i + γT+ γTii + exp + exp + exp + exp2 2 + X+ Xi i + ε+ εii

where X is any vector of covariates.where X is any vector of covariates.

After standardizing the test scores (mean 0, s.d. 1),After standardizing the test scores (mean 0, s.d. 1),

- the coefficient t- the coefficient t00 can be interpreted as can be interpreted as the return to schooling for an individual ofthe return to schooling for an individual of

mean ability. mean ability.

- the coefficient - the coefficient γγ measures measures the approximate percentage change in the hourly wage the approximate percentage change in the hourly wage arising from a one standard deviation increase in the score for an individual with no arising from a one standard deviation increase in the score for an individual with no schoolingschooling. However, . However, for an individual with mean years of schooling, the impact is for an individual with mean years of schooling, the impact is measured by: measured by: γγ + t + t11**mean years of schoolingmean years of schooling..

ResultsResults The working sample includes males employed for wages between the ages of 18 and The working sample includes males employed for wages between the ages of 18 and

65. The dependent variable in the earnings regressions is the logarithm of the hourly 65. The dependent variable in the earnings regressions is the logarithm of the hourly wage.wage.

As is the case for the rest of the IALS country data, in the Chilean data the three skills As is the case for the rest of the IALS country data, in the Chilean data the three skills are very highly correlated. Consequently, the average of the three scores will be used are very highly correlated. Consequently, the average of the three scores will be used in the analysis as an aggregate IALS score measure.in the analysis as an aggregate IALS score measure.

From Table A2, inclusion of the direct measure of cognitive ability in the traditional From Table A2, inclusion of the direct measure of cognitive ability in the traditional earnings function reduces the return to schooling by about 25 percent, while a one earnings function reduces the return to schooling by about 25 percent, while a one standard deviation increase in cognitive achievement increases earnings by a standard deviation increase in cognitive achievement increases earnings by a significant 15 percent. Once both the cognitive achievement and its interaction with significant 15 percent. Once both the cognitive achievement and its interaction with years of schooling are included (column 4), the interaction term is positive and years of schooling are included (column 4), the interaction term is positive and significant.significant.

Dependent Variable: log of hourly wageDependent Variable: log of hourly wage(1)(1) (2)(2) (3)(3) (4)(4)

Years of schoolingYears of schooling -- 0.1130.113(14.9)(14.9)

0.0870.087(8.9)(8.9)

0.0820.082(8.3)(8.3)

ExperienceExperience -- 0.0130.013(2.0)(2.0)

0.0150.015(2.2)(2.2)

0.0200.020(2.9)(2.9)

Exp. SquaredExp. Squared -- -0.0000-0.0000(0.1)(0.1)

-0.0000-0.0000(0.2)(0.2)

-0.0002-0.0002(1.2)(1.2)

Standardized ScoreStandardized ScoreIALSIALS 0.3420.342

(12.8)(12.8)-- 0.1530.153

(4.2)(4.2)0.004*0.004*(0.1)(0.1)

Standardized ScoreStandardized ScoreIALSIALS*years of schooling*years of schooling -- -- -- 0.0180.018

(3.0)(3.0)

ConstantConstant 6.456.45(243.9)(243.9)

5.115.11(41.0)(41.0)

5.335.33(39.7)(39.7)

5.305.30(39.6)(39.6)

RR22-adj.-adj. 0.1580.158 0.2150.215 0.2300.230 0.2370.237

NN 861861 861861 861861 861861

Table A2: Returns to Schooling - OLS (Male employees)

* for an individual with mean years of schooling, the impact isfor an individual with mean years of schooling, the impact is 0.171

Dependent Variable: log of Dependent Variable: log of hourly wage hourly wage

Q10Q10 Q25Q25 Q50Q50 Q75Q75 Q90Q90

Years of schoolingYears of schooling 0.1020.102(6.0)(6.0)

0.0750.075(9.0)(9.0)

0.1120.112(14.0)(14.0)

0.1140.114(8.7)(8.7)

0.1210.121(6.3)(6.3)

ExperienceExperience 0.0230.023(1.7)(1.7)

0.0170.017(2.1)(2.1)

0.0270.027(3.6)(3.6)

0.0020.002(0.2)(0.2)

0.0040.004(0.4)(0.4)

Exp. SquaredExp. Squared -0.0002-0.0002(0.9)(0.9)

-0.0002-0.0002(1.2)(1.2)

-0.0002-0.0002(1.6)(1.6)

0.00030.0003(1.6)(1.6)

0.00020.0002(1.2)(1.2)


5.125.12(36.3)(36.3)

5.015.01(37.4)(37.4)

5.655.65(27.8)(27.8)

5.865.86(20.4)(20.4)

PseudoRPseudoR22 0.0650.065 0.0780.078 0.1200.120 0.1670.167 0.2130.213

Table A4: Returns to Schooling Using Quantile Regressions – Standard Earnings Functions (Male employees)

The schooling coefficient for years of schooling at the 25th quantile is statistically different from other quantiles.

Table 1: Returns to Schooling and Cognitive Achievement – Quantile Regressions (Male employees)

Dependent Variable: log Dependent Variable: log of hourly wage of hourly wage

Q10Q10 Q25Q25 Q50Q50 Q75Q75 Q90Q90


0.0570.057(5.5)(5.5)

0.0690.069(4.8)(4.8)

0.0870.087(6.8)(6.8)

0.1180.118(4.6)(4.6)


0.0260.026(3.4)(3.4)

0.0220.022(2.2)(2.2)

0.0010.001(0.1)(0.1)

0.0030.003(0.3)(0.3)


-0.0003-0.0003(1.8)(1.8)

-0.0002-0.0002(1.1)(1.1)

0.00030.0003(1.8)(1.8)

0.00030.0003(1.2)(1.2)

Stand. ScoreStand. ScoreIALSIALS 0.2050.205

(4.0)(4.0)0.1970.197(4.9)(4.9)

0.2140.214(4.0)(4.0)

0.1740.174(4.0)(4.0)

0.0200.020(0.3)(0.3)


5.175.17(35.9)(35.9)

5.495.49(27.3)(27.3)

5.835.83(33.9)(33.9)

5.885.88(17.7)(17.7)

PseudoRPseudoR22 0.0820.082 0.0990.099 0.1390.139 0.1780.178 0.2130.213

- The schooling coefficient for years of schooling at the 10th quantile is statistically different than the 75th and 90th quantile. Likewise, 25th quantile estimate is different from the 75th and 90th quantile estimates.- The 90th quantile estimate for the standardized score is different from the lower quantile estimates.

Quantile regression results show that without controlling for cognitive achievement Quantile regression results show that without controlling for cognitive achievement Table A4), quantile returns to one additional year of schooling exhibit a U-shaped Table A4), quantile returns to one additional year of schooling exhibit a U-shaped pattern, increasing after the 10th quantile. The 90th-10th inter-quantile difference is pattern, increasing after the 10th quantile. The 90th-10th inter-quantile difference is about 2 percentage points, while the 90th-25th inter-quantile difference is about 4.5 about 2 percentage points, while the 90th-25th inter-quantile difference is about 4.5 percentage points.percentage points.

Adding cognitive scores in the equation (Table 1), returns by quantile now exhibit a Adding cognitive scores in the equation (Table 1), returns by quantile now exhibit a

sharp and strictly increasing pattern, with a 90th-10th inter-quantile difference of 7.5 sharp and strictly increasing pattern, with a 90th-10th inter-quantile difference of 7.5 percentage points. The independent effect of cognitive achievement is positive and percentage points. The independent effect of cognitive achievement is positive and highly significant up to the 75th quantile - a one standard deviation increase in highly significant up to the 75th quantile - a one standard deviation increase in achievement increases earnings by about 20 percent. However, this effect all but achievement increases earnings by about 20 percent. However, this effect all but disappears at the 90th quantile.disappears at the 90th quantile.

Schooling return estimates when cognitive skills is not controlled for are, therefore, Schooling return estimates when cognitive skills is not controlled for are, therefore, upward biased except for the 90th quantile of earnings. upward biased except for the 90th quantile of earnings.

Table 2: Returns to Schooling and Cognitive Achievement – Quantile Regressions (male employees)

Dependent Variable: log of Dependent Variable: log of hourly wage hourly wage

Q10Q10 Q25Q25 Q50Q50 Q75Q75 Q90Q90


0.0550.055(5.5)(5.5)

0.0650.065(4.5)(4.5)

0.0820.082(5.7)(5.7)

0.1000.100(5.1)(5.1)


0.0340.034(4.6)(4.6)

0.0300.030(2.9)(2.9)

0.0020.002(0.2)(0.2)

0.0140.014(1.3)(1.3)


-0.0005-0.0005(3.2)(3.2)

-0.0004-0.0004(2.0)(2.0)

0.00020.0002(1.1)(1.1)

-0.0000-0.0000(0.1)(0.1)

Standardized ScoreStandardized ScoreIALSIALS 0.370*0.370*

(4.1)(4.1)0.065*0.065*(1.0)(1.0)

-0.030*-0.030*(0.3)(0.3)

-0.119*-0.119*(1.5)(1.5)

-0.309*-0.309*(2.8)(2.8)

Stand. ScoreStand. ScoreIALSIALS*years of *years of

schoolingschooling-0.022-0.022(2.5)(2.5)

0.0150.015(2.5)(2.5)

0.0260.026(3.0)(3.0)

0.0260.026(3.5)(3.5)

0.0450.045(5.3)(5.3)


5.115.11(38.3)(38.3)

5.405.40(27.3)(27.3)

5.875.87(30.1)(30.1)

5.875.87(20.3)(20.3)

PseudoRPseudoR22 0.0860.086 0.1020.102 0.1490.149 0.1930.193 0.2470.247

- The schooling coefficient for years of schooling at the 10th quantile is statistically different than the 90th quantile. The 25th quantile estimate is different from the 75th and 90th quantile estimates. - The 10th quantile estimate for the standardized score is different from other quantile estimates.- The interaction term estimates are different between the 90th and other quantiles. ** For an individual with mean years of schooling, the impacts are: 0.218, 0.175, 0.188,For an individual with mean years of schooling, the impacts are: 0.218, 0.175, 0.188,

0.131 and 0.208.0.131 and 0.208.

When we control for both the independent effect of the cognitive ability measure and When we control for both the independent effect of the cognitive ability measure and its interaction with years of schooling (Table 2), at the lowest quantile the its interaction with years of schooling (Table 2), at the lowest quantile the independent effect of cognitive achievement is very strong. At higher quantiles the independent effect of cognitive achievement is very strong. At higher quantiles the picture changes: what matters is not the independent effect of cognitive ability, but picture changes: what matters is not the independent effect of cognitive ability, but the earnings-enhancing, complementary relationship between cognitive achievement the earnings-enhancing, complementary relationship between cognitive achievement and schooling. and schooling.

Summarizing, at the lowest earnings quantile, education qualifications at any level Summarizing, at the lowest earnings quantile, education qualifications at any level (quantity of schooling) have a rather small contribution to earnings; cognitive (quantity of schooling) have a rather small contribution to earnings; cognitive achievement (quality), along with experience are the contributors to higher earnings. achievement (quality), along with experience are the contributors to higher earnings. On the other hand, those in high quantiles (therefore, those of higher ability) benefit On the other hand, those in high quantiles (therefore, those of higher ability) benefit much more from acquiring more schooling, and from the interaction of additional much more from acquiring more schooling, and from the interaction of additional schooling with ability.schooling with ability.

An Evaluation of the 1981 “Vouchers” SchemeAn Evaluation of the 1981 “Vouchers” Scheme

We use the school reform of 1981 to identify a binary instrument and estimate returns We use the school reform of 1981 to identify a binary instrument and estimate returns to schooling and cognitive skills from IV. As in the first part of the paper the focus is to schooling and cognitive skills from IV. As in the first part of the paper the focus is on the role of cognitive skills and quality of schooling for heterogeneous individuals.on the role of cognitive skills and quality of schooling for heterogeneous individuals.

As a result of the opening of the education system to the private sector:As a result of the opening of the education system to the private sector:

- the dropout rate declined sharply - from 8 percent in 1981 to 2.7 percent in 1982 for - the dropout rate declined sharply - from 8 percent in 1981 to 2.7 percent in 1982 for basic education and from 8.3 percent to 6.2 percent for secondary education.basic education and from 8.3 percent to 6.2 percent for secondary education.

- Chile was able to absorb the pressure of secondary school expansion. - Chile was able to absorb the pressure of secondary school expansion.

- There was a large shift of students to the private subsidized schools (mainly in major - There was a large shift of students to the private subsidized schools (mainly in major urban areas), whose enrollment increased by 93 percent between 1980 and 1985, at urban areas), whose enrollment increased by 93 percent between 1980 and 1985, at the expense of municipal schools. the expense of municipal schools.

- Despite a decline in public spending for education (from 4 percentage of GDP in 1981 - Despite a decline in public spending for education (from 4 percentage of GDP in 1981 to 2.6 percent in 1990), student intake increased significantly, by about 15 percentage to 2.6 percent in 1990), student intake increased significantly, by about 15 percentage points in secondary education (and by 42 percent in higher education).points in secondary education (and by 42 percent in higher education).

- That the main effect of unrestricted school choice was an exodus of “middle class” - That the main effect of unrestricted school choice was an exodus of “middle class” students from public sector schooling (Hsieh and Urquiola 2006) and the practice of students from public sector schooling (Hsieh and Urquiola 2006) and the practice of screening by private subsidized schools (competing for better students). screening by private subsidized schools (competing for better students).

- It is possible that parents tended to select schools which provided good peer groups. - It is possible that parents tended to select schools which provided good peer groups.

- Overall, the reform improved the efficiency of the education system (number of - Overall, the reform improved the efficiency of the education system (number of students educated per unit of public spending increased significantly over the 1980s, students educated per unit of public spending increased significantly over the 1980s, without a significant change in average quality).without a significant change in average quality).

- However, this was achieved at the expense of equity, as subsidized private schools - However, this was achieved at the expense of equity, as subsidized private schools serving a better-informed, better-off population benefited from the opportunities serving a better-informed, better-off population benefited from the opportunities offered by the capitation grants system at the expense of municipal schools which offered by the capitation grants system at the expense of municipal schools which served the less well-off section of the population. served the less well-off section of the population.

What was the effect on quality (on average)?What was the effect on quality (on average)? International evidence is not conclusive.International evidence is not conclusive. Bellei (2005) outlines three principal reasons why it is difficult to make comparisons Bellei (2005) outlines three principal reasons why it is difficult to make comparisons

between public and private schools in Chile, hence the widely diverging results in between public and private schools in Chile, hence the widely diverging results in individual analyses, all stemming from the lack of random assignment of students to individual analyses, all stemming from the lack of random assignment of students to schools: (i) private schools tend to be located in urban areas and serve middle to schools: (i) private schools tend to be located in urban areas and serve middle to middle-high income students, (ii) there are wide differences in the level of resources middle-high income students, (ii) there are wide differences in the level of resources available to schools, even among the same types of schools, and (iii) there is very available to schools, even among the same types of schools, and (iii) there is very little information about how families select schools and how private schools select little information about how families select schools and how private schools select students. students.

This paper does not attempt to look at the impact of the reform in this light, but rather This paper does not attempt to look at the impact of the reform in this light, but rather use the reform itself as an instrument to evaluate the outcomes of the reform for the use the reform itself as an instrument to evaluate the outcomes of the reform for the group which acted upon the reform.group which acted upon the reform.

Chart 1: Average years of schoolin by cohort (age in 1981)

0

2

4

6

8

10

12

14

0 5 10 15 20 25 30 35

Age in 1981

Ave

rag

e ye

ars

of

sch

oo

ling

AverageYears ofschooling

The reform must have affected a large proportion of students. The affected group is The reform must have affected a large proportion of students. The affected group is expected to contain a large proportion of students from a better socio-economic expected to contain a large proportion of students from a better socio-economic background and from urban areas. This is because, firstly, in rural areas in Chile, low background and from urban areas. This is because, firstly, in rural areas in Chile, low population density does not permit a choice of schools; and secondly, there is population density does not permit a choice of schools; and secondly, there is evidence that private-subsidized schools tend to select students to a larger extent evidence that private-subsidized schools tend to select students to a larger extent than municipal schools. As a result, students who attend private-subsidized schools than municipal schools. As a result, students who attend private-subsidized schools come from higher-income and better educated families compared to municipal come from higher-income and better educated families compared to municipal schools (but not private non-subsidized schools).schools (but not private non-subsidized schools).

Prior expectations: Assuming that cognitive skills are largely determined by the Prior expectations: Assuming that cognitive skills are largely determined by the quality of schooling, and a group of students switched to better quality schools as a quality of schooling, and a group of students switched to better quality schools as a result of the reform, we would expect the affected group to have higher and less result of the reform, we would expect the affected group to have higher and less dispersed cognitive skills (especially those who switched to better quality schools at a dispersed cognitive skills (especially those who switched to better quality schools at a younger age).younger age).

For example, mainly in Scandinavian countries where IALS scores are high with low For example, mainly in Scandinavian countries where IALS scores are high with low

dispersion, the increase in earnings in response to one standard deviation increase in dispersion, the increase in earnings in response to one standard deviation increase in scores is generally low compared to Chile, which has the lowest scores among the scores is generally low compared to Chile, which has the lowest scores among the IALS group of countries. IALS group of countries.

Therefore, cognitive skill scores in the IV regressions (affected group) are expected to Therefore, cognitive skill scores in the IV regressions (affected group) are expected to be lower compared to the OLS regressions (average student).be lower compared to the OLS regressions (average student).

Country Sample Means: Years of Schooling and IALS Score, IALS DataCountry Sample Means: Years of Schooling and IALS Score, IALS Data

CountryCountry Mean Years of SchoolingMean Years of Schooling IALS/10IALS/10 Effect of IALS Effect of IALS score (rank)score (rank)

CanadaCanada 13.1 (3.5)13.1 (3.5) 2.92 (0.60)2.92 (0.60) 77

Chile Chile 9.6 (4.0)9.6 (4.0) 2.15 (0.58)2.15 (0.58) 11

Czech RepublicCzech Republic 13.0 (2.8)13.0 (2.8) 2.89 (0.47)2.89 (0.47) 1111

DenmarkDenmark 13.2 (3.3)13.2 (3.3) 2.96 (0.40)2.96 (0.40) 1313

Finland Finland 12.8 (3.5)12.8 (3.5) 2.99 (0.43)2.99 (0.43) 1212

GermanyGermany 11.8 (3.5)11.8 (3.5) 2.92 (0.47)2.92 (0.47) 66

HungaryHungary 12.3 (3.1)12.3 (3.1) 2.63 (0.48)2.63 (0.48) 22

ItalyItaly 11.1 (4.1)11.1 (4.1) 2.52 (0.62)2.52 (0.62) 1313

NetherlandsNetherlands 13.4 (4.2)13.4 (4.2) 2.94 (0.43)2.94 (0.43) 55

NorwayNorway 12.1 (2.8)12.1 (2.8) 3.00 (0.41)3.00 (0.41) 1010

PolandPoland 11.7 (3.0)11.7 (3.0) 2.43 (0.64)2.43 (0.64) 1515

SloveniaSlovenia 11.5 (3.0)11.5 (3.0) 2.40 (0.61)2.40 (0.61) 99

SwedenSweden 11.8 (3.7)11.8 (3.7) 3.12 (0.51)3.12 (0.51) 88

SwitzerlandSwitzerland 12.9 (3.4)12.9 (3.4) 2.83 (0.54)2.83 (0.54) 44

U.S.A.U.S.A. 13.9 (3.2)13.9 (3.2) 2.85 (0.65)2.85 (0.65) 33Source: Leuven et. al., 2004

A note on Identification:A note on Identification: Identification is based on variations across cohorts (similar but more problematic Identification is based on variations across cohorts (similar but more problematic

compared to - say - Oreopoulos' and Duflo's). The difference in our case is that we compared to - say - Oreopoulos' and Duflo's). The difference in our case is that we only have one cross-section. We cannot identify (unrestricted) age effects only have one cross-section. We cannot identify (unrestricted) age effects separately from cohort effects (they are collinear in one cross section). One separately from cohort effects (they are collinear in one cross section). One concern that might arise is that in reality we are using individuals at different points concern that might arise is that in reality we are using individuals at different points of the life cycle to identify different cohorts and hence different ages; age earnings of the life cycle to identify different cohorts and hence different ages; age earnings profiles might vary across age differently by education. This might potentially profiles might vary across age differently by education. This might potentially confound the effect of the instrument with different age-earnings profiles. confound the effect of the instrument with different age-earnings profiles.

However we have experience in the regressions, parametrically restricted (i.e., However we have experience in the regressions, parametrically restricted (i.e., quadratic) to still leave room for identification of the IV. It is worth emphasizing that quadratic) to still leave room for identification of the IV. It is worth emphasizing that we are conditioning on experience profiles and so we identify using variations we are conditioning on experience profiles and so we identify using variations across cohorts – around - these experience profilesacross cohorts – around - these experience profiles. .

Table 3: Returns to Schooling from IV – Male Employees 22-45 years(Instrument=1 if age in 1981 between 6 and 13 years)

VariableVariable OLSOLS OLSOLS IV: ReformIV: Reform IV: ReformIV: Reform IV: ReformIV: Reform


0.0580.058(4.8)(4.8)

0.1220.122(5.1)(5.1)

0.1020.102(2.7)(2.7)

0.0960.096(2.5)(2.5)


0.0010.001(0.1)(0.1)

0.0010.001(0.5)(0.5)

0.0090.009(0.5)(0.5)

0.0150.015(0.8)(0.8)

Experience squaredExperience squared 0.00020.0002(0.4)(0.4)

0.00020.0002(0.5)(0.5)

0.00020.0002(0.3)(0.3)

0.00020.0002(0.4)(0.4)

-0.0000-0.0000(0.1)(0.1)

Standardized IALS scoreStandardized IALS score -- 0.1740.174(4.2)(4.2)

-- 0.0860.086(1.0)(1.0)

0.0090.009(0.1)(0.1)

Stand. IALS score-years of Stand. IALS score-years of schooling interactionschooling interaction

-- -- -- -- 0.0110.011(1.0)(1.0)


5.755.75(30.0)(30.0)

4.984.98(13.3)(13.3)

5.195.19(10.2)(10.2)

5.195.19(10.5)(10.5)

Centered RCentered R22 0.1560.156 0.1790.179 0.1440.144 0.1670.167 0.1740.174

NN 586586 586586 586586 586586 586586

First Stage:First Stage:Shea’s partial RShea’s partial R22

F-valueF-value[p-value][p-value]

-- -- 0.2610.261205.5205.5

[0.000][0.000]

0.1390.13993.693.6

[0.000][0.000]

0.1300.13086.586.5

[0.000][0.000]

Endogeneity Test:Endogeneity Test:Durbin-Wu-Hausman Durbin-Wu-Hausman statisticstatistic[p-value][p-value]

-- -- 3.953.95[0.047][0.047]

2.092.09[0.148][0.148]

1.631.63[0.201][0.201]

t-values in parentheses

In the OLS regression, including the direct measure of cognitive ability reduces the In the OLS regression, including the direct measure of cognitive ability reduces the return to schooling by about 34 percent, while one standard deviation increase in the return to schooling by about 34 percent, while one standard deviation increase in the score increases earnings by 17 percent. score increases earnings by 17 percent.

In the IV regression, initially, the binary instrument is assigned the value of 1 for those In the IV regression, initially, the binary instrument is assigned the value of 1 for those who at the time of the reform where 6 years or older (22 years or older in 1997 – the who at the time of the reform where 6 years or older (22 years or older in 1997 – the year of the survey) and up to 13 years of age included (29 years old in 1997) and 0 year of the survey) and up to 13 years of age included (29 years old in 1997) and 0 otherwise, that is those in the 8 years of basic education. otherwise, that is those in the 8 years of basic education.

In the first stage, the correlation of years of schooling with the instrument is strong In the first stage, the correlation of years of schooling with the instrument is strong (Shea’s partial R(Shea’s partial R22 is high), indicating that the instrument is sufficiently relevant to is high), indicating that the instrument is sufficiently relevant to explain the endogenous regressor, while the instrument is uncorrelated with the explain the endogenous regressor, while the instrument is uncorrelated with the logarithm of hourly wage.logarithm of hourly wage.

In the standard Mincerian specification, the IV estimate of the return to schooling is In the standard Mincerian specification, the IV estimate of the return to schooling is about 37 percent higher that the OLS estimate, and the assumption that the schooling about 37 percent higher that the OLS estimate, and the assumption that the schooling variable is exogenous is rejected at the 5 percent level. variable is exogenous is rejected at the 5 percent level.

When we include the cognitive score in the equation, the estimate of the return to When we include the cognitive score in the equation, the estimate of the return to schooling decreases by about 16 percent (much less than the reduction one sees in schooling decreases by about 16 percent (much less than the reduction one sees in the OLS regressions), while the contribution of one standard deviation in the cognitive the OLS regressions), while the contribution of one standard deviation in the cognitive score to earnings is one-half the corresponding estimate from the OLS regressions score to earnings is one-half the corresponding estimate from the OLS regressions (8.5 compared to 17 percent). (8.5 compared to 17 percent).

Another instrument (in addition to the policy reform based instrument) is included in Another instrument (in addition to the policy reform based instrument) is included in

the specification presented in Table A6, namely if the pupil’s father has at least the specification presented in Table A6, namely if the pupil’s father has at least upper-secondary education. Here the objective is twofold: first, test the over-upper-secondary education. Here the objective is twofold: first, test the over-identifying restrictions and, second, test the hypothesis that if the treated group is an identifying restrictions and, second, test the hypothesis that if the treated group is an even more privileged one, its cognitive skills will be better and more homogeneous even more privileged one, its cognitive skills will be better and more homogeneous compared to the treated group in Table 3 (which in turn had better and more compared to the treated group in Table 3 (which in turn had better and more homogeneous cognitive skills compared to the average individual). Therefore, the homogeneous cognitive skills compared to the average individual). Therefore, the contribution of cognitive skills to earnings should be small, while returns to schooling contribution of cognitive skills to earnings should be small, while returns to schooling should increase.should increase.

Based on the value of Hansen’s J-statistic, the null hypothesis (that the instruments Based on the value of Hansen’s J-statistic, the null hypothesis (that the instruments appropriately uncorrelated with the disturbance process) is not rejected in all cases.appropriately uncorrelated with the disturbance process) is not rejected in all cases.

As hypothesized, the contribution of cognitive skills to earnings is nearly zero, while As hypothesized, the contribution of cognitive skills to earnings is nearly zero, while the schooling-cognitive skills interaction is much less than in Table 2 and statistically the schooling-cognitive skills interaction is much less than in Table 2 and statistically insignificant. At the same time, the estimate of the return to schooling increases by insignificant. At the same time, the estimate of the return to schooling increases by between 7-19 percent compared to Table 3.between 7-19 percent compared to Table 3.

VariableVariableIV: Reform +IV: Reform +

Educated fatherEducated fatherIV: Reform +IV: Reform +


Educated fatherEducated father


0.1200.120(3.1)(3.1)

0.1160.116(2.9)(2.9)


0.0240.024(1.2)(1.2)

0.0290.029(1.3)(1.3)

Experience squaredExperience squared -0.0004-0.0004(0.7)(0.7)

-0.0004-0.0004(0.8)(0.8)

-0.0006-0.0006(0.9)(0.9)


-0.022-0.022(0.2)(0.2)


-- 0.0060.006(0.5)(0.5)


4.914.91(9.5)(9.5)

4.924.92(9.7)(9.7)

Centered RCentered R22 0.1750.175 0.1850.185 0.1900.190

NN 490490 490490 490490



0.3080.308108.1108.1

[0.000][0.000]

0.1690.16949.049.0

[0.000][0.000]

0.1600.16046.246.2

[0.000][0.000]

Over-identification Test:Over-identification Test:Hansen’s J-statisticHansen’s J-statistic[p-value][p-value]

2.092.09[0.148][0.148]

1.961.96[0.161][0.161]

1.711.71[0.191][0.191]

Endogeneity Test:Endogeneity Test:Durbin-Wu-Hausman statisticDurbin-Wu-Hausman statistic[p-value][p-value]

5.385.38[0.020][0.020]

2.532.53[0.111][0.111]

1.931.93[0.164][0.164]

Table A6: Returns to Schooling from IV – Male Employees 22-45 years(Instrument=1 if age in 1981 between 6 and 13 years)

Table 3a: Returns to Schooling from IV – Male Employees 22-45 years(Instrument=1 if age in 1981 between 6 and 18 years)

VariableVariable OLSOLS OLSOLS IV: ReformIV: Reform IV: ReformIV: Reform IV: ReformIV: Reform


0.0580.058(4.8)(4.8)

0.0770.077(3.2)(3.2)

0.0430.043(1.3)(1.3)

0.0420.042(1.3)(1.3)


0.0010.001(0.1)(0.1)

0.0030.003(0.2)(0.2)

0.0020.002(0.1)(0.1)

0.0090.009(0.4)(0.4)


0.00020.0002(0.5)(0.5)

0.00020.0002(0.3)(0.3)

0.00020.0002(0.3)(0.3)

-0.0001-0.0001(0.3)(0.3)


-- 0.2040.204(2.9)(2.9)

0.0510.051(0.5)(0.5)


-- -- -- -- 0.0150.015(1.4)(1.4)


5.755.75(30.0)(30.0)

5.665.66(14.1)(14.1)

5.945.94(12.7)(12.7)

5.845.84(12.7)(12.7)

Centered RCentered R22 0.1560.156 0.1790.179 0.1580.158 0.1830.183 0.1880.188

NN 586586 586586 586586 586586 586586



-- --0.3000.300249.5249.5

[0.000][0.000]

0.2310.231174.4174.4

[0.000][0.000]

0.2330.233176.1176.1

[0.000][0.000]

Endogeneity Test:Endogeneity Test:Durbin-Wu-Hausman Durbin-Wu-Hausman statisticstatistic[p-value][p-value]

-- -- 0.640.64[0.42][0.42]

0.450.45[0.50][0.50]

0.330.33[0.56][0.56]

Note: t-values in parentheses

In Table 3a the binary instrument takes the value of 1 for those who at the time of the In Table 3a the binary instrument takes the value of 1 for those who at the time of the reform where 6 years or older (22 years or older in 1997 – the year of the survey) and reform where 6 years or older (22 years or older in 1997 – the year of the survey) and up to 18 years of age included (34 years old in 1997) and 0 otherwise; that is the up to 18 years of age included (34 years old in 1997) and 0 otherwise; that is the group includes all those who could have been affected by the reform, either by group includes all those who could have been affected by the reform, either by switching to private subsidized schools because of their perceived higher quality switching to private subsidized schools because of their perceived higher quality compared to municipal schools, or by choosing to continue schooling at the compared to municipal schools, or by choosing to continue schooling at the secondary level as a result of the increase in the supply of private schools associated secondary level as a result of the increase in the supply of private schools associated with the education reform, which led to a large increase in secondary school with the education reform, which led to a large increase in secondary school participation.participation.

Estimates of the return to schooling are now lower by 37 percent in the Mincerian Estimates of the return to schooling are now lower by 37 percent in the Mincerian specification and by more than 50 percent when we control for cognitive skills. At the specification and by more than 50 percent when we control for cognitive skills. At the same time, one standard deviation increase in the cognitive score increases earnings same time, one standard deviation increase in the cognitive score increases earnings by 20 percent, up from 8.5 percent in Table 3. The coefficient estimates are now by 20 percent, up from 8.5 percent in Table 3. The coefficient estimates are now close to those from OLS, as a result the endogeneity test does not reject the null close to those from OLS, as a result the endogeneity test does not reject the null hypothesis that the difference in coefficients between IV and OLS are not systematic.hypothesis that the difference in coefficients between IV and OLS are not systematic.

Table A6a presents the equivalent results of Table A6 for the 6-18 group.Table A6a presents the equivalent results of Table A6 for the 6-18 group.

VariableVariableIV: Reform +IV: Reform +



Educated fatherEducated father


0.0670.067(1.9)(1.9)

0.0630.063(1.7)(1.7)


0.0180.018(0.8)(0.8)

0.0250.025(1.1)(1.1)

Experience squaredExperience squared -0.0005-0.0005(1.0)(1.0)

-0.0005-0.0005(0.9)(0.9)

-0.0007-0.0007(1.1)(1.1)


0.0430.043(0.4)(0.4)


-- -- 0.0110.011(1.0)(1.0)


5.575.57(10.7)(10.7)

5.555.55(10.7)(10.7)

Centered RCentered R22 0.1900.190 0.2050.205 0.2100.210

NN 490490 490490 490490



0.3390.339124.1124.1

[0.000][0.000]

0.2420.24277.377.3

[0.000][0.000]

0.2410.24176.976.9

[0.000][0.000]

Over-identification Test:Over-identification Test:Hansen’s J-statisticHansen’s J-statistic[p-value][p-value]

3.293.29[0.069][0.069]

2.152.15[0.142][0.142]

1.701.70[0.193][0.193]

Endogeneity Test:Endogeneity Test:Durbin-Wu-Hausman statisticDurbin-Wu-Hausman statistic[p-value][p-value]

9.369.36[0.053][0.053]

7.147.14[0.210][0.210]

8.798.79[0.186][0.186]

Table A6a: Returns to Schooling from IV – Male Employees 22-45 years(Instrument=1 if age in 1981 between 6 and 18 years)

Finally, in the next 3 graphs we present results from progressively expanding the Finally, in the next 3 graphs we present results from progressively expanding the “treated” group one year at a time (starting with a group of very young children in “treated” group one year at a time (starting with a group of very young children in 1981), thus tracing the changes in coefficient estimates for schooling and cognitive 1981), thus tracing the changes in coefficient estimates for schooling and cognitive skills.skills.

Assuming that municipal schools on average provided lower quality schooling, Assuming that municipal schools on average provided lower quality schooling, switching to private-subsidized schools at a younger age should result in a higher and switching to private-subsidized schools at a younger age should result in a higher and more homogeneous cognitive skill endowment by the end of the schooling period, more homogeneous cognitive skill endowment by the end of the schooling period, compared to switching at a later age. We, therefore, would expect that as we compared to switching at a later age. We, therefore, would expect that as we progressively increase the instrument upper cutoff age, the treated group becomes progressively increase the instrument upper cutoff age, the treated group becomes less endowed in cognitive skills (because they have received most of their schooling less endowed in cognitive skills (because they have received most of their schooling before 1981) and less homogeneous in these skills. As a result, as we add the before 1981) and less homogeneous in these skills. As a result, as we add the marginal individual in the treated group, the return to cognitive skills will be increasing marginal individual in the treated group, the return to cognitive skills will be increasing and the return to schooling will be decreasing. and the return to schooling will be decreasing.

Chart 2: Returns to an additional year of schooling for different instrument cutoff points (Mincerian specification)

0

2

4

6

8

10

12

14

4 6 8 10 12 14 16 18 20 22

Instrument cutoff point (Age in 1981)

Retu

rn %

Chart 3: Return to an additional year of schooling and one standard deviation of IALS score for different instrument cutoff points

0

5

10

15

20

25

30

35

4 6 8 10 12 14 16 18 20 22


Re

turn

%

Return toschooling

Return toIALS score

Chart 4: Return to one additional year of schooling, one standard deviation of IALS score and having an educated father for different instrument cutoff points

0

5

10

15

20

25

30

4 6 8 10 12 14 16 18 20 22


Ret

urn

%

Schooling

IALS score

EducatedFather

In Chart 2, the Mincerian return is around 11-12 percent for the cohorts who in 1991 In Chart 2, the Mincerian return is around 11-12 percent for the cohorts who in 1991 were of primary school age or lower and subsequently (by the end of basic education) were of primary school age or lower and subsequently (by the end of basic education) declines sharply. declines sharply.

In Chart 3 (after adding the cognitive score), the return to schooling fluctuates at In Chart 3 (after adding the cognitive score), the return to schooling fluctuates at about 10 percent for basic schooling cohorts and subsequently declines sharply; the about 10 percent for basic schooling cohorts and subsequently declines sharply; the opposite pattern is observed for the return to the cognitive score. opposite pattern is observed for the return to the cognitive score.

We observe that the decline in the return to schooling estimate (increase in the return We observe that the decline in the return to schooling estimate (increase in the return to cognitive score) coincides with the end of the compulsory basic education in Chile.to cognitive score) coincides with the end of the compulsory basic education in Chile.

Concluding, the evidence seems to suggest that the students who took advantage of Concluding, the evidence seems to suggest that the students who took advantage of the 1981 “voucher” reform came from better socioeconomic backgrounds and mostly the 1981 “voucher” reform came from better socioeconomic backgrounds and mostly urban areas. Those who benefited most were students who in 1981 were enrolled in urban areas. Those who benefited most were students who in 1981 were enrolled in the early stages of schooling. However, the story seems to be different for the older the early stages of schooling. However, the story seems to be different for the older cohorts. cohorts.

The reform (capitation grants paid to schools and the opening of the education The reform (capitation grants paid to schools and the opening of the education

system to the private sector) facilitated the absorption of the large secondary school system to the private sector) facilitated the absorption of the large secondary school expansion, despite the decline in public spending for education. However, because expansion, despite the decline in public spending for education. However, because of the secondary school expansion, lower “quality” students (from middle class of the secondary school expansion, lower “quality” students (from middle class families or otherwise) may have been admitted into the system, students who had families or otherwise) may have been admitted into the system, students who had earlier received their basic education in public schools. earlier received their basic education in public schools.

Table A1: Means of Variables by Earnings Quantile, Male Employees

VariableVariableOverallOverallmeanmean

10th10th 25th25th 5050thth 75th75th 90th90th 9999thth

Log (hourly wage)Log (hourly wage) 6.416.41(0.88)(0.88)

5.565.56(0.63)(0.63)

5.955.95(0.12)(0.12)

6.176.17(0.13)(0.13)

6.876.87(0.16)(0.16)

7.467.46(0.16)(0.16)

8.848.84(0.36)(0.36)


6.96.9(3.9)(3.9)

7.37.3(3.9)(3.9)

8.48.4(3.7)(3.7)

9.69.6(3.8)(3.8)

11.511.5(3.6)(3.6)

13.413.4(4.4)(4.4)

IALS scoreIALS score 209.3209.3(62.8)(62.8)

167.3167.3(65.1)(65.1)

181.5181.5(61.4)(61.4)

198.8198.8(54.5)(54.5)

221.1221.1(55.7)(55.7)

237.3237.3(51.7)(51.7)

259.3259.3(54.7)(54.7)

Years of experienceYears of experience 21.521.5(13.4)(13.4)

25.725.7(14.5)(14.5)

22.922.9(14.8)(14.8)

21.921.9(13.5)(13.5)

20.420.4(11.9)(11.9)

20.220.2(12.7)(12.7)

17.917.9(12.5)(12.5)

NN 892*892* 8888 159159 217217 211211 125125 8282

Note: standard deviation in parentheses.*Includes ten observations for those with earnings which exceed the 99 th percentile.

APENDIX

Table A3: Returns to Cognitive Achievement Using Quantile Regressions (Male employees)

Dependent Variable: Dependent Variable: log of hourly wage log of hourly wage

Q10Q10 Q25Q25 Q50Q50 Q75Q75 Q90Q90

Stad. ScoreStad. ScoreIALSIALS 0.2550.255

(5.2)(5.2)0.2500.250(8.0)(8.0)

0.3300.330(11.9)(11.9)

0.3540.354(7.6)(7.6)

0.3280.328(7.1)(7.1)


6.056.05(182.7)(182.7)

6.456.45(221.1)(221.1)

6.866.86(150.3)(150.3)

7.337.33(171.4)(171.4)

PseudoRPseudoR22 0.0610.061 0.0680.068 0.1010.101 0.0980.098 0.1270.127

Table A5: Returns to Schooling from OLS – Male Employees 22-45 years

VariableVariable (1)(1) (2)(2) (3)(3) (4)(4)


0.0580.058(4.8)(4.8)

0.0550.055(4.5)(4.5)

0.0680.068(4.9)(4.9)


0.0010.001(0.1)(0.1)

0.0110.011(0.7)(0.7)

0.0110.011(0.7)(0.7)


0.00020.0002(0.5)(0.5)

-0.0001-0.0001(0.3)(0.3)

-0.0002-0.0002(0.4)(0.4)


0.037*0.037*(0.4)(0.4)

0.1280.128(2.7)(2.7)

Stand. IALS score-Years of Stand. IALS score-Years of schooling interactionschooling interaction

-- -- 0.0140.014(1.9)(1.9)

--

Father at least Upper Sec. Father at least Upper Sec. educationeducation

-- -- -- 0.1940.194(2.3)(2.3)


5.755.75(30.0)(30.0)

5.695.69(29.2)(29.2)

5.565.56(25.8)(25.8)

Adj. RAdj. R22 0.1560.156 0.1790.179 0.1820.182 0.2070.207

NN 586586 586586 586586 490490

* for an individual with mean years of schooling, the impact isfor an individual with mean years of schooling, the impact is 0.167

Cognitive Ability, Heterogeneity and Returns to Schooling in Chile: Outcomes of the 1981 Capitation Grant Scheme Harry Anthony Patrinos World Bank, Washington.

Documents

quality of schooling

role of cognitive ability

noncognitive ability

ability distribution

prose literacy

schooling system i

compulsory schooling

schooling decision correlation