Ability in Pre-Schoolers, Earnings, and Home-Environment SWP322 World Bank Staff Working Paper No. 322 WASUMGTON DC. N431 April 1979 0 Prepared by. Roger Grawe Development Economics Department Copyright © 1979 The World Bank 1818 H Street, N.W. Washington, D C 20433, U S.A The views and interpretations in this document are those o and should not be attributed to the World Bank, to its affilia organizations, or to any individual acting in their behalf FILE CPY The views and interpretations in this document are those of the authors and should not be attributed to the World Bank, to its affiliated organizations, or to any individual acting in their behalf. WORLD BANK Staff Working Paper No. 322 April 1979 ABILITY IN PRE-SCHOOLERS, EARNINGS, AND HOME-ENVIRONMENT This paper explores on the basis of available data the role of various family characteristics in fostering abilities in pre-school children and the subsequent effect of ability on earnings. The paper is part of the Bank's effort to improve the understanding of the process of human capital development as a basic determinant of earnings. The results suggest that household income and the status of the mother are key predictors of abilities of pre-school children from disadvantaged backgrounds. Prepared by: Roger Grawe Development Economics Department Copyright ( 1979 The World Bank 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. TABLE OF CONT:ENTS PREFACE Sutmary, Conclusions and Recommendations 1 I. Ability as a Determinant of Earnings 5 II. Fanmily-related Deteiminants of Ability in Young Children 22 III. Pre-school Iutervention 37 Implications for the Bank 51 Appendix 53 LIST OF TABLES Table 1. Outline of Models of Economic Success 10 Table 2. Environmental Determinants of Childhood Abilities 26 Table 3. Bernard van Leer Foundation Projects Affecting 48 Pre-school Children Table A.l. Pre-school Ability Measures 76 Table A.2. Explanatory Variables - Description & Distribution 77 Table A.3. Significant Coefficients on Measures Derived from 78 the Seguin Form Board Test Table A.4. Significant Coefficients in Four Preschool Outcome 83 Measures Table A.5. Significant Coefficienits on Sub-test of "Wechsler 86 Preschool and Primary Scale of Intelligence" PREFACE This paper and an accompanying paper by Moshe Smilansky have been produced as part of ongoing work in Population and Human Resources Division of the Developmient Economics Department and the Education Department to evaluate relative educational investment priorities with respect to different age groups. In view of the increased priority attached to pre-school intervention in recent years, an informal working group within the 11orld Bank was formed to consider whether the pre-school period deserved greater emphasis in its activities in education, nutrition, and health. Foremost among the questions to be considered was the relationship between the timing of the intervention on investment and subsequent productivity. Though little direct longitudinal evidence could be brought to bear, the hypothesis that investments during pre- school years generated outcomes in skills and abilities which themselves subscquently enhanced productivity was considered worthy of investigation. As the Batik is increasingly involved in assisting less-developed countries restructure their educational priorities, an important question was, simply, should maore priority be given to the education of disadvantaged pre-school children in an effort to generate a more equitable distribution of outcomes. To review the evidence on pre-school intervention the Bank invited Professor Moshe Smilansky pf Tel Aviv University to participate in a staff seminar in November, 1975. Smilansky contended that the results of pre-school intervention experiments and programs, at least as measured by cognitive tests, were inconclusive at best and did not establish a case for investments priority to the pre-school age category. However, Smilansky noted that the more successful experiments had been those involving the participation of parents. Extrapolating from his experience as an educational policy-maker in lsrael, Smilansky proposed that the need, particularly in modernizing societies, is for greater support to the family unit. The most effective target age for such support, he argued, was adolescence. Subsequently Mr. Smilansky was invited by the Education Department and the Population and Human Resources Division to prepare a paper setting out his views. Priorities in Education: Pre- school, Evidence and Conclusions reviews in detail the evidence on U.S. and Israeli experiments in pre-school intervention, and contains an annex which sketches out in a preliminary fashion ways in which greater priority to the needs o, adolescents could provide family support. The paper bly Roger Grawe reviews the role of childhood ability as a determinant of subsequient earnings. Grawe then develops some new emnirica). evidence on the effects of various family characteristics in fostering abilities in pre-school children. This research reinforces, through a quite different methodological route, Smilansky's conclusion that pre-school outcomes can best be enchanced by support to the existing farmily structure; in particular, Grawe identifies the status of the mother and income as consistent predictors of ability for children from disadvan- taged backgrounds. A significant litLitation to the research reviewed and reported by Sniilansky and Grawe is its reliance on experience and data from developed economJ.cs. This is tempered E;omewhat by the focus on disadvan- ta-,ed rtoups withini Ohcsc oconomic.3 but goncralizations stil'l caniiot be acccptcJ. Thc p.Ipers serve to indicate that too little is yet known i concerning the development of intelligence, and its relevance as conven-- tionally measured9 to formulate new educational priorities in developing countries. But the weight of the evidence is now that intervention through support of the existing family structure offers the best clhance of augmenting the abilities of disadvantaged pre-school children and increasing the likelihood of greater subsequent achievement. The author would like to express their appreciation to Dov Chernichovsky and Martha Grosse for their detailed comments and assistanice in the production of this paper; Susan Cochrane, Mats Hultin, Timothy Kinig, and John Simmons also contributed valuable comments. Of course, none of these individuals are responsible for remaining deficiencies. ABILITIES IN PRE-SCHOOL CHILDREN: PRODUCTIVITY EFFECTS AND FAMILY ENVIRONMENTAL DETERMINANTS Summary, Conclusions and Recommendations A primary objective of development is to raise the welfare of in- dividuals, by increasing their economic contribution and earning power (and likewise improving non-pecuniary outcomes). Obviously, the expansion of income earning opportunities, by increasing the demand for entrepreneurial and labor services through policies and projects which contribute to general economic growth, is the primary means of tackling this objective. However, individuals themselves have traits and characteristics which also affect their ability to take advantage of and expand the opportunities offered in any economic environment. In order to understand how better to increase and broaden the distribution of welfare-generating outcomes for individuals, it is important to know how individual traits such as schooling and ability affect outcomes, how such traits interact, and how they may be affected by manipulable variables (educational inputs, day care, mother and child health, other forms of family support), and by variables endogenous to the system (income, family structure, labor force participation). This general formulation leads to three specific questions that are the focus of this paper: 1. What are the characteristics of individuals that sig- nificantly affect earnings? In particular, what is the particular role of ability in generating economic returns? 2. What factors influence the development of abilities in young children? 3. What has been international experience with interven- tion programs intended to increase the opportunities for children with disadvantaged backgrounds by enhanc- ing their abilities? - 2 Section I, which concerns the first question, begins with a descrip- tion of models of earnings determination and considers some of the statistical problems that complicate empirical verification of the models. An appealing formulation of earnings generation is a recursive model in which the different stages of the model correspond to a life-cycle sequence. However, it is shown that the introduction of greater complexity, for example, the reciprocal effect of child ability on home environment, creates difficult estimation problems in an expanded model. Nevertheless, a review of the evidence indicates that we can identify and measure the impact of ability on earnings. Section II expands on the particular aspect of the general model which addresses the second question, and is the main focus of this paper. This analysis of the first stage in the recursive earnings model, which relates childhood abilities to family and environmental variables, begins with a discussion of the notion of ability and its measurement. Ability emerges as an operational concept which must be recognized as closely tied to the specific measuring instruments0 From this perspective, focussing on a single ability dimension, such as IQ, probably does little to facilitate a better understanding of the issues. Utilizing data from.a sample of low income black families in the United States, the relationship between several different pre-school ability measures and selected family background vari- ables is explored in some detail0 The results of this analysis show several interesting, though tentative, conclusions. Family income plays a generally significant, and occasionally large, role in the generation of children's abilities. Also important is the effect of mother's status represented here, either singly or in combination, by marital status, mother's education and employment. The strength of these mother's status effects on a wide range of ability measures suggests that a development activity which en- hances the status of women is more likely to improve the capacity and performance of the next generation. Conversely, a development project which has an adverse effect on the status of women is likely, ceteris paribus, to set back the capacity of the subsequent generation. These conclusions are offered with significant caveats: (i) they assume cross-cultural validity between low-income US and LDC environments; (ii) the notion of status is an intervening variable, not directly quantified, and it is, therefore, rather difficult to predict how much status, gained or lost, would result from specific policy measures or would lead to the postulated intergenerational effects. To begin Section III, some comments are offered concerning the gen- erally negative evaluation of pre-school intervention programs. Methodologi- cal and measurement problems are serious enough to cast into doubt broad conclusions from most evaluation efforts thus far. Some recently available longer term evaluations are cited as evidence that the jury must be considered still out on the question of general success or failure of pre-school inter- vention. In the LDC's, few countries have begun to articulate a policy of support to young children in a family context. However, many countries do have day care programs which could be considerably upgraded if child devel- opment objectives were more widely recognized. A review of UNICEF and privately assisted projects indicates some of the potential for effective intervention in LDC's. - 4 - On the basis of the issues and evidence diccuosed in t;his pap,2ZT the following suggestions are offered as directions for future 3ank activity in the pre-school area: lo The Bank should be instrumental in undertaking apopZTiste studies that will relate abillties to earnings Anci home environment in LDCvs8 and thereby enhance our und2rotanding of the process of human capital formation in thoea countries. 2. Particular consideration should be given to possible ressazch inputs to existing pre-school projects or programs perhaps on a collaborative basis with UNICCEFo This could signifl cantly increase our understanding of the offectGs f different pre-school environments in the LDC context. A better understanding the curriculums and institutional oettings that seem to work for pre-school and early primary childrea could increase the efficiency of primary education programe. 3. Any effort to monitor the effect of Bank projects on opportunities for women should include the effects on the envirorment of pre-school children. This T^fould provide a basis for accumulating evidence on the need for aad feasi- bility of specific project components designed to eubEnt the abilities of young children. - 5 - SECTION I - Ability as a Determinant of Earnings This section focusses on the role of cognitive ability in earnings determination. Although there are diverse reasons why social policy might be directed toward augmenting individual cognitive abilities, the primary justification considered here is the effect on the economic well-being and earning power of individuals. 1/ On the basis of evidence from numerous data sets to be reviewed below the following conclusions seem warranted: 1. Developmental or recursive earnings models imply clearly that enhancing the abilities of children will result in higher incojmes subse- quently. However, empirical efforts to confirm or reject this hypothesis must confront difficult problems of omitted variables, errors in variables, and statistical identification. Nevertheless, the impact of ability on earnings can be demonstrated statistically. 2. The size of the ability effect on earnings varies among data sources, but ranges from small to moderate. There seems to be some evi- dence that post-school ability has greater direct explanatory power in earnings equations than does pre-school ability, though this is not uni- versally true. 1/ Other reasons for advocating increased cognitive abilities tend to parallel the external effects alleged for education, though I would argue that augmented abilities may have even broader effects. They relate to the technological adaptiveness of populations, and'to the level and quality of inter-societal communication (for an economic analysis of the importance of such communication in modern society, see A. Hirschman, Exit, Voice and Loyalty). Non-pecunia,ry outcomes for individuals due to enhanced abilities (e.g., greater prbductivity in leisure) and positive cultural effects that a family-oriented abil- ity enhancing program might have, may be quite impo8rtantin the- calculus of individual decision-making though quantification and aggregation problems may reduce their appeal to the policy-maker. 3. The relationship of ability with other determinants of earnings is complex. There seem to be significant interactions between ability and education and between ability and experience, in their joint effect on earnings. Investigations which have allowed non-linear ability effects have been modestly successful. 4. There appears little doubt that childhood ability has consider- able indirect effect on earnings through its consistently significant effect: on schooling. The identification of individual characteristics which significantly affect the growth and interpersonal distribution of earnings has only become an important focus of economic research in the past decade0 The human capital paradigm has focused attention on a particular formulation of the general issues, an earnings function: lol E = FI(S,T) in which individual earnings are determined by the human capital investments , schooling S, and other forms of training T. Ignoring for the moment specifi- cation and estimation problems, the coefficient on S in econometric versions of 101 can be interpreted as the rate of return to investment in education. Early objections to this procedure emphasized the likely exaggerated effect estimated for schooling as a result of omitting crucial correlates, particu- larly ability, from the "independent" variables0 The time dimension over which earnings are measured also signif- icantly influences the relative importance of individual characteristics0 Ideally the human capital earnings function should evaluate the determinants of lifetime earnings, since different patterns of human capital investment-- -7- vocational education, university education, on-the-job training--give rise to different earning-age profiles. 1/ These modifications have resulted in a modified earnings function 1.2 El= F (S,A,Z) where lifetime earnings, E, are described as a function of schooling S, ability A, and other factors, Z, which may include training and labor-force experience. At this point, the pure investment analogy of human capital breaks down, as it is clear that school and ability have their own deter- minants which must be recognized and modeled if the earnings function is to be correctly estimated. 2/ One model that has a considerable statistical and theoretical appeal is a recursive system which builds on a life-cycle approach charac- terized by the initial development of ability, followed by a period of schooling, and finally, entry into the labor force. The three stages of this model may be characterized as follows: 3/ 1.3a A = F (X,G) 1.3b S = F 2(X,A) 1.3c E = F 3(X,S,A) 1/ Some empirical implications affecting ability will be discussed below. The importance of the earnings-age profile is emphasized in J. Mincer, who has developed the notion of the overtaking point, which occurs when the age earnings profiles of individuals of different schooling cross each other. The profiles differ due to varying rates of personal time preference. The effects of schooling on earnings are greatest at this point. 2/ See Mark Blaug, "Human Capital Theory, A Slightly Jaundiced Survey," Journal of Economic Literature, September 1976, for a discussion of the "investment" model in this context. 3/ Path coefficient models (notably, 0. Duncan, "Ability and Achievement," Eugenics Quarterly, 1968) are necessarily based on a similar recursive system. where A represents early childhood ability Ss schooling and post-school formal training E9 some measure of lifetim2 earnings X, a set of family background characteristics G9 genetic endowment. As a structural models, these relationships capture some important features of reality9 namely that there are distinct developmental stages each focussed on different outcomes which precede chronologically the earnings function. Of course the underlying causal basis of this recursive model is found in the exogenous variables, X and G, of the first equation. The creation of abilities and schooling outcomes are represented only as important processes through which genetic and early environmental characteristics ultimately influence earnings and productivity. In this respect this recursive model presents an extreme case; but its structure offers a useful perspective from which to evaluate the empirical work in this area. An important implication of the recursive framework is that each of the equations can be estimated singly by ordinary least squares techniques and still yield best, unbiased estimates of the coefficients. In the earnings equation, this has meant that the inclusion of an ability measure has been widely interpreted as allowing the estimation of an unbiased rate of return to education, so that several data sets which include an ability measure have been developed. In a recent review of this literature, A. Leibowitz has presented the salient features of those investigations which included one or more of the relationships in the recursive model. 1/ This information is reproduced here as Table 1. The variety of data sets and estimation techniques represented in these studies has resulted in a rather heterogenous pattern of results. Of the nine studies reported which include an ability measure, seven find it to be significant predictor of income and the other two find its effects confined to schooling. In all reported studies which investigated schooling attainments, ability figured as a significant determinant. It is important to emphasize that this effect is independent of family background; furthermore only a few of these studies distinguish pre- school ability from post- or mid-school measures. If, as discussed above, ability is viewed as a developmental concept, then ability and education might be expected to have interactive effects; however, ordinary least squares estimation of the simple recursive model assumes linear independence and additivity between the incremental effects of education and ability. More important the recursive formulation does not permit any feedback of schooling on ability, though such feedback is a clear implication of the developmental approach to ability generation. In more formal terms, assume for the moment that schooling and ability are considered to be proxies for an unobserved variable "productive capacity," P, so that 1.3c becomes, E = F (P,X). Only if this relation is itself linear, i.e., 1.4 P = a A + a S 1 2 1/ A. Leibowitz, "Family Background and Economic Success: A Review of the Evidence," Law and Economics Center and Department of Family Medicine, University of Miami, mimeo. 1976. t la - 'rA' I..: I OU J'n n r mioir. of . CnrlOmit i SUUcc' s Va, v-,ck.' Author Sample lep.rndant Va i.laitu Rulotod to Not Rvklan'd to !L%p Laillf.\ Sao'lle And 1962 1962 Income Schooling 1962 15% Nelson .pccup.atiunal ages 25-34 rather's Occupation 26% changes in 35-44 EducaLiori, IncoMn Generation 45-54 Childhood I.Q. 35-64 1962 Occupation Schooling 1962 4a ages 25-34 Father's OccupaLion S8t 35-44 Eduication, Incomo 45-54 Chtildhoocd I.Q. 55-64 (except age 25-34) 1962 Schooling Father's Occupation 41% ages 25-34 Education, Incomne 47% 35-44 Childhood I.Q. 45-54 55-64 Chamberlin Gorselina 192Q Incowm Schooling and 156 pairo in Indiana "Ability" Grilichle of brothers Age 1928 Occupation Schooling 'Ability' Age 1928 Schooling Age "Ability" Conlisk DarkIl^y 1960 Earnin.s Schcoling (nste '0) I.Q. (i'nq 1-lR) 43t Guidance (aee c.30 ) Parents' Income Parents Schooling Study (ent from occu- (192S-29) Number of own 75 men pation data from children 1960 Census) Schooling (ago 30) I.Q. (ageo 1-18) Parents' Schooling 44% Parents' Income de Wolff, 1938 Students 1963 Taxed Schooling . 33% Van Slijpe in Malmo, Income Social Class Sweden Childhoodl I.Q. 545 man (I.O. and Social Class Interaction) Duncan. 1962 1964 Earninga 1964 Occupation Ntmbnr of Siblings 11t Occripational aten aqed 25-34 Adult I.q. Father's Occupation Changes EducuLion in a Fatlhor's Education Generation; 1964 CPS __ __ 1964 OccupatLon Education Father's OccupaLion Adult 1.Q. t'athor's Education Nuimher of Sil,linvs Lducatior. Chit0hood I.Q. 42t Father's Occupation Fnvher' Education Source: A. LeLbowtIz, "ft.litly Nuu0or of S1blInqs background Jilll 'e¢)m'nic Succews: A Kt',t.w of tl. Evidence," Ln4 6 Economice Center a-id Piepartl.:ntit 11 t Vu.tt _ a -d_ ixi_ i. _.... 1....L , "ou 4 Table 1 cont. vn,i ANtlior Saap le Depondenr Variable . Related to Not 1velaLed to . I:x' niIm Teibowitz Terinan Sample 1940, 1950, 1960 Schooling Mother's Schooling 6% 700-784 obs. I.arnings Experience Father's Schoollng 10% males (age 29, 39, 49) Family Income Childhood 1.O. 615 females (1940 and 1950J Family Income I (1960) Schooling 1940, Mother's Schooling Time Inputs in 8% 1950 males Father's Schooling Early Chiildhood 10% Childhood I.Q. Siblings (1950) Sihlings (in 1940) Family Income Schooling 1940, Mother's Schooling Childhood I.Q. 17% 1950 females Father's Schooling Family Income 15% Siblings Birth Order Childhood Ability Mother's Schooling Father's Schooling 19% I.Q. at c. age 11 Birth order males Time inputs to child Fanmily income Childhood Ability Family Income MotheL's Schooling 13% I.Q. at c. age 11 Birth Order females Time inL)uts to Child Father's Schooling Lindert 312 New Jersey 1963 Schooling Mother's Schooling Age 32% executives age 31-81 Father's Schooling Sex across 1008 Father's Occupation siblings *Pb3ervationsL Broken home 8% Birth order between No. of siblings siblings Estimated Time Inputs 1963 Occupational Father's Schooling Mother's Schooling 27% Status of 659 Father's Occupation Miales Broken home Age Birth Order No. of siblings Estimated Time Inputs Morgenstern 19G8 Urban 1968 Average Schooling School Quality 24% Problcm yearly Experiepce for whites 27% Survey Wage Rate Father's Schooling 842 blacks Sex 782 whites,\ School Quality for men and women Blacks Schooling 1968 School Quality 21t Region 30% MothUr's Schooling Source: A. 1cibowLtz, "Faimily Father's Sehooling Nnckgroutnd and .conorntc Sex Succcis: A Review uC ttie Age Evicdrioce,"l Lr> & I:conloullc I Ce:nter and Dcpartmt-itt of Fa.nily liediriric, Untvetnity of Miami, mirnlfo 1976. - 12 - Table 1 cont. Auithor S.imple Dependentk Variable Related to Not Related to- ria.lac. Sewcil and Wisconsin 1967 Earnings Schooling .Father's Schoolinq 7% Hauser Study (age c.29) Ability MoLhcr's Schoolinq 2069 men (from Social Security) Occupation 1964 Father's Occupation Parental Income (1957-1960) 1964 Occupation Schooling Father's Schoolinq 41% measured by Ability Mother's Schooling Duncan SEI Father's Occu.pation Parcnt's Income Years of Schooling bleasured Ability 28% Father's Schooling Mother's Schooling Father's Occupation Parents' Income Measured Ability Father's Schooling 9% Mother's Schooling Father's OccupaLion Parents' Average Income Taubman NAS-NRC 1973 Average Av. Schooling Mother's Schooling Income of twins Father's Schooling (MZ) (MZ) Father's Schooling Isother's Schooling (DZ) (DZ) Age (MZ) llours Government Employee Disability Status Other Religion Self Employment Status Jewish Religion 37- Catholic Religion (MZ) (DZ twins only) 38% Age (DZ) (DZ) 1973 Income Difference in Government 5% Difference Schooling Employee (MZ) between twins Self Employment Disability 124 Status Status Government Employee (DZ) Disability Status (MZ) --hours (if self employmenL, government status not included) Wachtel NBER - Thorndike 1969 Income Schooling Siblings 19% 1812 men Log of Hours Mother's Educdtion aged 44-52 Ability College Quality School Quality Experience Source: A. lItbowitz, "Family Father's Etucation Background and Econoirc - Succeso: A Kevicv of tIle Evidence of Famtty Schoolillg Ability Sclhool QualiLy 1o% Medicitc, lUtitversity of Siblings Miumi, isieo, 1976. Father's Lklucation Mother's Elducation Agt' AidliLty 1913 Motliny's 1'tI.'wti on 3% -rather's rldtction School Qunlity - 13 - will the linear form of the recursive system (1.3) adequately account for the education-ability interaction. 1/ Recognizing that pre- and post-school abilities may vary, captures one aspect of this interaction. For example, let A1 and A2 refer to pre- and post-school abilities respectively, and consider the model: 1.5 A = g1(X,G) 1.5b S = g2(X,A1) 1.5c A2 = g3(A1,S) 1.5d E = g4(A 29S,X) If this model correctly described reality but were specified incorrectly by an econometric version of the simple recursive model (equations, 1.3, above) it can be readily demonstrated that the estimated effect of schooling would reflect the product of the marginal effect of post-school ability on earnings and of schooling on post-school ability, while the real effects of schooling 1/ F. Welch, op. cit. demonstrates that with such a linear specification and plausible parameter estimates for the first order partial correla- tion coefficients r and r result in an interpretation of the schooling coefficient not as a rate of return but as itself a much better estimate of the incremental importance of "true" ability or- productive capacity. - 14 - and post-school ability on earnings would be weighted components of the estimated coefficient on pre-school ability. 1/ An alternative form of 15d still consistent with a recursive for- mulation is 1.6 E = g'(X,AlDS,A) An appealing transformation of this general form would be 101 E =h 1 (1A)h 2(X9A 2S) in which early childhood abilities could have the effect of an efficiency parameter augmenting the effects of schooling, post-school abilities, and other background factors. This particular version of the earnings function might be particularly appropriate if A1 were considered to be a vector of pre-school outcomes which could not be subsequently made up. This might 1/ Consider economic formulations of 13 and 1l5: lo3a' A = a X + a2 G + U1 103bW S = a 21X + a22A + U2 13c' E =a31X +a32 S + a33A 3 1.5a' A1 =b 1X +b12G + eI 1.5b' S =b21 +b22AI + e2 15c' A2 =b31S S+b32 A e3 1.5d0 E =b41X +b42 S + b 43A 2+ e4 Suppose that 1.3c' is estimated even though model 105 better reflects reality. In this case substitution illustrates that the esti- mated coefficients a32 and a33 will actually equal a32 b43 31 a33 -b42b22 + b43b32 Model 13 in effect, a semi-reduced form version of the structural model, 105; as such it must be recognized that the coefficients have no direct interpretation. - 15 - be particularly appropriate for children in multiply disadvantaged envi- ronments, in which brain development might be permanently set back through malnutrition or mental growth inhibited through the lack of environmental stimulation at appropriate times in terms of brain development. 1/ In such cases the function, h1, might be of the form, hI(A1) = (1,0) where Q < 1 obtains only in the case of compounded or severe deprivatsions. In the normal case, the effects of early childhood ability work themselves out indirectly through earlier stages in the recursive framework (i.e., h1(A1) 1). The formulation of 1.7 is merely one specification of many in which interac- tions between ability and other variables contribute to economic performance; but it does summarize usefully the developmental plausibility of these sorts of interactions. Turning again to the empirical studies summarized in Table 1, one notes that few have had the required data or were organized so as to explore fully the types of interactions that might be expected. In general the inclusion of ability measures in earnings functions has been found to reduce schooling coefficients in the range of 4 to 35% with a central tendency around 10%. 2/ Post-school ability measures seem to be associated on average 1/ Herman Epstein, a physiological psychologist at Brandeis University, has presented some very suggestive evidence on these points in an informal seminar at the World Bank. His research indicates the existence of devel- opmentally significant growth spurts in brain size which seem to be highly correlated with mental growth. These spurts occur at approximately two- year intervals until adulthood; but deficits caused by early childhood malnutrition do not appear to be made up subsequently. 2/ H. Gintis, "Education, Technology and the Characteristics of Worker Pro- ductivity," American Economic Review, 61 (May, 1971), reviews nine studies which yield results within these bounds. Subsequent studies seem to narrow this range: Griliches and Mason (7-10%), Taubman and Wales (9%). Negligible effects of pre-school ability on the education-income rela- tionship were found by both Conlisk and Leibowitz, however. 16 - with larger reductions in the schooling coefficient, between 15 and 20%, providing some indirect evidence for the specification of a development,al theory in which abilities at different stages are treated differently. 1/ Explicit evidence of interaction effects between education, expe- rience, and ability is provided by J0 Hause in a reanalysis of several data sets stratified by labor force experience and educational level. The data 0o0 imply that measures of cognitive ability are associated with an empirically significant but modest in- crease in annual earnings for those with high levels of schooling 000 In the United States samples, there appears to be a distinct tendency for the ability coefficient to increase with labor force experience. 2/ In the four data sets reanalyzed by Hause the ability coefficients at the highest educational levels (either one or more college degrees) range from 1.65 to 7.7 times as great as the ability coefficients on the lowest educa- tion levels (high school dropouts) with the variation in the range related to both the type of ability measure and the age of the cohort. The greater the age (experience) the larger the ability coefficients in all educational strata; but those in the highest educational levels increased relatively moreo An interesting dimension of Hause's results that he does not elaborate 1/ Griliches and Mason established a post-school reduction in the school- ing coefficient of 13-17%o Hanushek in a study not reported in Table 1 ("Teacher Characteristics and Gains in Student Achievement: Estimation Using Micro Data," American Economic Review, 61 (May 1971)), found a bias of 15% in the schooling coefficient on earnings when post-school ability was uncontrolled0 Taubman and Wales' data and estimation tech- niques yield a range between 10% and 19%. 2/ J. Hause, "Ability and Schooling as Determinants of Lifetime Earnings," in F0 Juster, ed0, Education, Income and Human Behavior, Carnegie Foun- dation For the Advancement of Teaching and National Bureau of Economic Research, 1975. - 17 - is the tendency for ability coefficients for both high-school and college dropouts to be significantly lower than for those for high-school and college graduates. This presumably indicates the interaction of ability with some sort of motivational factor. In terms of the model suggested in 1.7 this motivational dimension may figure as a component of the background factors, X; or it may be that the ability outcomes themselves include non-cognitive components. Further evidence on interaction is provided in another data set (the National Bureau of Economic Research--Thorndike data) extensively analyzed by P. Taubman. 1/ Ability tests were administered during World War II as part of the qualifying process for pilot selection in the Air Force. Earnings profiles were subsequently obtained in 1955 and again 1969. Taubman finds strong ability effects: the increment to earnings in both years as- sociated with the difference between the top and bottom fifths in the abil- ity distribution exceeds the earnings increment associated with a bachelor's degree. 2/ Taubman tested for interactions in the most general functional form by computing separate earnings equations for each ability fifth. He concludes "using analysis of covariance, we could not reject the hypothesis that the effects of all variables including education, were independent of \ 1/ P. Taubman, Sources of Inequality in Earnings, North Holland Publish- ing Co., Ltd.; Amsterdam, Oxford; 1976. 2/ Taubman uses ability fifths from one factor of the raw scores as his ability measure. As the test population had already scored above the median on a general intelligence test, Taubman notes that his fifths correspond more closely to tenths in the general population. - 18 - the level of ability in each year0" I/ However, Taubman does find signifi- cant differentials between the bottom ability fifth and the top four fifths with successively greater effects even within the top four fifths. Thiese non-linear effects of ability are more pronounced in 1969 than in 1955, and supplement across-the-board increase in ability effects in the later year0 One may conclude from Taubman's empirical analysis that ability interacts non-linearly with age or experience in affecting earnings (though, at 'Least as measured in the NBER--Thorndike data, ability does not interact with education and other determinants of earnings). Finally, Leibowitz reports two additional pieces of evidence on ability interactions. One, an extensive study by Sewell and Hauser, 2/ analyzes an early career profile of earnings in a large sample of the cohort of 1957 high school seniors in Wisconsin. Both ability-education and ability° experience interactions were well documented0 The second is the Swedish study by de Wolfff and Van Slipje who "found that intelligence and social clatss reinforce each other in generating earnings, and their combined effect, particularly at high levels of education, is much greater than implied by an additivity of effect0" 3/ The Swedish data on social class and ability were 1/ Taubman, gop cit., po 34. Note that this finding does not preclucle that a continuous scaling of ability with the stronger assumption of a functional form such as 3.3 might not yield significant interactions coefficients0 Furthermore, Taubman's ability is a post-schooling measure which might itself be thought of as the outcome of a school- ing--pre-school interaction0 Under this interpretation, the inde-- pendence of education and ability lends some support to the Gintis- Bowles hypothesis on education as a proxy for non-cognitive outcomes (cf. S0 Bolwes, H0 Gintis, and J. Simmons) 2/ Sewall, W. H0 and Hauser, R., Education, Occupation and Earnings: Achievements in the Early Career0 New York: Academic Press, 1975' 3/ A0 Leibowitz, opo cit*, po 20O - 19 - collected in 1938 and income, 25 years later in 1963. One might rationalize ex-post that the behavioral correlates of social class would have been much more homogenously grouped in Sweden of 1938 than today; also an investigation of ability-education interactions stratified by SES levels would have been informative. Another important' aspect of the relationship between ability and earnings is the indirect effect through the influence of ability on 5chooling. In terms of the evidence presented in Table 1, all of the studies which had a direct ability measure among the determinants of schooling found the rela- tionship to be significant. 1/ It is also generally true that the variance explained in the schooling relationships is greater than in the earnings function, so that the indirect effects of ability may often exceed the direct effects in actual dollar impact on earnings (this obviously depends specifi- cally on the relative sizes of the various coefficients involved; but, in general, the effects of ability on schooling and schooling on earnings are both rather large while the direct effect of ability on earnings though often sigificant is not usually large). Some of the problems in sorting out the effects of schooling and ability have already been mentioned above. These became further compounded as one takes a more generalized view of schooling and ability outcomes. If abilities are defined to include non-cognitive outcomes as well as a broad range of cognitive skills, and schooling outcomes consist of learned knowledge 1/ The single exception, the study by Chamberlain and Griliches, infers an ability measure from a components of variance model estimated on a sample of brothers. Whatever skills are represented by this ability measure are a significant determinant of adult earnings and occupation but not schooling. - 20 - as well as work habits, attitudes, etc.D then a system with feedback is needed to portray the types of interaction that presumably occur: 108a A = hI(X 'S G) 108b S - h2(X A) 18c Et h3(At St 9) where the subscript t refers to time. In this variant schooling and ability interact as determinants of each other at any given point in time. Onc:e the interaction is admitted the practical benefits of the recursive system are lost and it is no longer possible to estimate unique coefficients for the variables in the various equations. In terms of an econometric models, only the schooling relationship can be estimated without further restrictior..s Neither the ability nor the earnings equations can be identified. This raises serious doubts about the consistency of single equation estimates of the earnings function (108c) which do not explicitly deal with the econo= metric effect of the schooling-ability model. Alternatively one may specify a model which combines the simulta- neity features of "l.8" and distinguishes pre° and post-school ability as "105" above. In this case the pre-school ability equation can be identi- fied and estimated with standard techniques even though the simultaneity - 21 - difficulties remain in the the rest of the model. 1/ The next section of this paper deals specifically with the determinants of pre-school ability. 2/ 1/ This model may be specified as follows: 1.8a' A'=h G 1 h 1(Xt') l.8a A = h (X 'S t) 1.8b St = h 3(Xt ,At) 1.8c Et = h4(Xt At' t In this formulation the pre-school ability equation combines the struc- tural and reduced forms and can be estimated using standard regression techniques. It should be noted that were adequate longitudinal data available, the inclusion of "previous period" ability and schooling outcomes in the ability and schooling equations would result in an indentifiable structure. 2/ Of course\, the ability-earnings relationship is not even a necessary condition for policies to augment the abilities of young children if other than economic goals are considered as policy objectives. In Tanzania and Peru, for example, day care programs for young children are justified as an important element in the restructuring of rural society: this as much for their role in socializing future adults as for their cognitive value (though the latter is more explicitly recog- nized in Peru). Or intervention may be reasonable widespread as in the provision of day care in Kenya but remain a privately demanded and supplied service with little government direction or support. 22 - SECTION II - Family-related Determinants of Ability in Young Children In Section I a general model of the development of individual economic performance was developed in which ability plays an importani: role first as an outcome in its own right and then as an important determinant of schooling and earnings. In the present section, attention is focussed on *the determinants of ability and, in particular, of pre-school ability. The general model for the discussion is 2.1 At = F(X itG) where At represents a vector of ability outcomes at time t, Xit9 a set: of i environmental characteristics, and G, genetic endowment0 The discussion will first take up some issues in the definition and measurement of ability0 A brief discussion of heritability will follow. The main part of the section will present and evaluate empirical evidence on the environmenltal determinants of ability0 The Measurement and Meaning of Abilit Ability is an operational concept: in the context of the present model, ability is the intervening variable that relates a set of famil.y background characteristics to subsequent individual performance in school and work0 The practical validity of this concept depends crucially on the speci- fic instrument that is to measure ability0 Traditionally cognitive ability measures were designed to predict school performance, usually within Et short to medium time horizon0 The instruments that were to be designed for this objective were selected on certain characteristics. As school performance is, by design, a relative concept with a near normal distribution, the sta.ndard w 23 - cognitive ability measure should generate over similar populations a similarly normal distribution. Furthermore as school performance does not, on average, fluctuate dramatically from year to year, a predictive measure was desired which demonstrated similar stability. Finis Welch has pointed out that these objectives will logically result in the development of ability measures which are heavily influenced by the more stable set of underlying factors, such as home environment. 1/ Furthermore a test which reflects predominantly charac- teristics of middle-class home environment will also tend to the desired normal distribution. It is natural then that IQ tests will tend to be less reliable over longer periods of time as the background characteristics them- selves will tend to vary more. The phenomenon reported by Bloom in Stability and Change in Human Characteristics, namely that "in terms of intelligence measured at age 17, at least 20% is developed by age 1, 50% by age 4, 80% by about age 8, and 92% by age 13," 2/ is attributable to (i) the increasing sensitivity of the intelligence test to the more stable elements in the background of the individual, and (ii) to the increased probability of stabil- ity in the determinants of test response. The latter is itself due to simply the shorter span of years involved and the resulting lower likelihood of significant changes in the environmental determinants of ability and, I would speculate, to the increasing feedback effect of the individual on his own environment, thereby increasing the stability of that environment. 1/ F. Welch, "Relationships between Income and Schooling" in Review of Research in Education, 2. 2/ This statement forms the basis for an illuminating discussion by M. Selowsky, "A Note on Pre-School Age Investment in Human Capital in Developing Countries," in Economic Development and Cultural Change, Vol. 24, No. 4, July 1976. 24 In the following discussion, the term ability will refer to a variety of specific instruments, though most are variants of "IQO" Each of these tests was designed for a specific evaluation or predictive purpose which was not directly related to the general earnings model. In the case of cognitive measures, such as IQ, we know that they do relate to subsequent outcomes, particularly school performance, but also earnings. In the case of non-cognitive, or affective, measures, the primary objecives of the tests have been short-term evaluations for counseling purposes. Given such objec- tives it is not surprising that affective tests have shown little stability in the context of long-term models0 Few would dispute that affective tra:Lts play a crucial role in determining economic performance, but little progress has been made in devising the measures with which such traits could be iden- tified. In the empirical treatment of outcomes for pre-school children to b- presented below, affective outcomes will be used to supplement informatiorL on cognitive outcomes. One disclaimer to this discussion of ability measure should be noted0 This review, and the subsequent discussion of empirical results do not discuss the issues posed by the ongoing but seemingly sterile debate on the heritability of IQ0 There now exists a comprehensive literature on this matter which has demonstrated conclusively the inadequacy of the data (some of which has now been shown to have been fabricated), and of the statistical procedures that have been utilized to demonstrate heritability coefficients0 1/ Even more to the point is the fact that heritability estimates of any trait 1/ See L. Kamin, The Science and Politics of IQ, Halstead Press, iea Yorlc, 1974, and A0 Goldberger, Discussion Papers Nos. 225-74, 242-74, 30175, 340-76, 341=76, Institute for Research on Poverty: University of Wisconsin, Madison. - 25 - are population specific. Such estimates do not establish relative herita- bility characteristics between distinct genetic pools; nor do they establish anything concerning the relative heritability of traits within populations having quite different environmental characteristics. As the objective of this exercise is to identify policy issues which may relate to the pre-school populations of disadvantaged groups in developing countries, the heritability of IQ within samples of relatively affluent American and British populations can only establish an upper bound which is likely to be of little relevance in rapidly changing LDC environments. The dominance of gross changes in environment over heritability has been effectively demonstrated in the case of anthropometric measures such as weight and height. 1/ However, in the discussion of empirical results it must be borne in mind that without an explicit control for the genetic component of the ability measures, the coefficients on other determinants will be biased to the extent they are correlated with genetic factors. The empirical results of this section focus on the determinants of abilities in children, particularly pre-school children. Three such studies are summarized in Table 2. What conclusions may be drawn from these investigations? 1. The first and most basic point is that systematic effects of family characteristics occur both in samples that range widely over socio-economic strata and in a sample that is confined to poor and lower middle-income blacks. In particular mother's education and family demographic char- acteristics appear as significant,determinants. 2. New information emerges from these studies on the relation of mother's education to other significant 1/ See, for example, G. Graham and B. Adrianzen, "Growth, Inheritance, and Environment," Pediatric Research 5 (1971); and J. Halricht et. al "Height and Weight Standards for Preschool Children - How Relevant Are Ethnic Differences in Growth Potential?" The Lancet April 6, 1974. -26 - Table 2: ENVIRONMENTAL DETERMINANTS OF C.9ILDIIOOD ABILITIES Sample Study Ability Measures Characteristics Type of Test Significant Determiniants (1) (2) 3(4) (5) Taubsan Pilot qualifying Post-high school-- Cognitive/ (1) Mother's education, test upper half of Achievement (2) Biography variable (a composite of scaled sub-test ability distri- education and family socio-economic bution nationwide status, (3) Sociological factors (Jewish upbringing, raised by grandparents-negative) (4) Affective or motivational elements: (financial motives in occupational choice, date marriage, self-employment, perceived financial pressure) Leibowitz Peabody picture Preschool visions Cognitive (1) Family size, vocabulary of "Sesame (2) Mother's education, Street" (3) Labor-saving consumer durables, Rural & Urban (4) Presence of encyclopedia samples of all (5) Joint mother-child activities--e.g., SES reading together, (6) Mother's TV watching - negative. Grave Setuin form Preschool parti- Visual/Motor (1) Child spacing, board cipating families skills (2) Female-headed household, in Gary, Indiana (3) Mother's education, income mainte- (4) Labor-force participation of family heaci, nance experiment (5) Income and charges in inccme. Preschool - Cognitive/ (1) Earned income - PSI, ROL inventory (PSI) Socio-emotional (2) Labor-force participation - GMP, BRN Role Identifica- (3) Family size negative - GMF, BRN tion (ROL) (4) Child spacing - C0T, BRN (negative) Brown referent (5) Increased transfer payments - negative. scale (Self- concept) (BRN) School readiness ("Gumpgookies") (GNP) Wechsler Preschool - Cognitive (1) Birth order, & Primary (IQ) (2) Child spacing, Scale of intelli- (3) Change in earned income, gence (4) Fema.e-hp.id,t (E) household Subscales: (5) Level of monthly income (b, C) Arithmetic (A) Block Design (B) Comprehension (C) Picture Com- pletion (P) - 27 - Table 2 cont. MNEY $/100 2.50 (3.07) Average monthly income earned by all family workers in 1972. MNEYSQ ($/100)2 15.65 (25.91) The square of MNFY MNUY $/100 2.80 (1.79) The average of all unearned income received monthly by the faiily inclu- ding all forms of welfare .payments, the net value of food stamps, social security payments, etc. during 1972. MNTPY $/100 1.08 (0.98) Average 1972 monthly transfer payments received by families in the treatment groups of the Income Maintenance Experiment. DMNEY $/100 0.11 (1.58) The change in average monthly earned family income between 1972 and 1973. DMNUY $1100 -1.74 (1.44) The change in average monthly unearned income between 1972 and 1973. IMNTP $/100 0.51 (0.88) The change in average monthly experi- mental transfer payments between 1972 and 1973. - 28 - determinants. From the Leibowitz data9 there appear to be strong indications that mother's education matters because of the activities that educated mothers undertake with their children (reading9 not watching TV, etc.). The Gary Income Maintenance data broaden the picture to include factors other than education which may serve to increase the status of the mother, such as heading the household and participating in the labor force. Each of these factors was found to contribute to specific types of childhood abilities, giving a somewhat broader picture of the importance of the mother-child relationship. 3. Controlling for the status variables, one may generalize that income has a significant impact on abilities in a sample of predominantly low income black American children. There is some reason to believe that for some abilities (including both cognitive and affective traits) the signifi-- cance of income or changes in income dominates the effect olf mother's education at least within the ranges of income less than $159000 a year and education of high school or less as covered by the Gary sample. There is some evidence that income plays a more significant role in families in which the mother is less well-educated0 4. Demographic effects are mixed: there does seem to be a general detrimental effect on abilities from close child- spacing, particularly in two-parent families or families having a relatively more educated mother0 There does not, however, appear to be a consistent pattern to the effects of birth order0 In the Leibowitz sample, family size does consistently have the predicted negative effects on ability0 The fact that this is not the case in the Gary sample suggests that family size effects might be related to broader socio-economic distinctions0 Child-specific variables (birth order, spacing, age) do interact with mother's education and family status. Thus test scores standardized by age, for example, may fail to reflect accurately environmental influences. 5. Finally, the Taubman post-high-school data suggest, through the significance of the general "background" variable, important feedback effects of schooling on ability con- firming the simultaneity problems discussed in Section I. The Taubman data also indicate the importance of socio- logical and motivational factors though the direction of causality for the latter is far from clear. The following discussion will summarize the most important results across the various ability measures first in the Taubman and Leibowitz - 29 - studies and then in the Gary sample, leaving a detailed description of the outcome measures and the specific explanatory variables and a test-by-test discussion of results to the Appendix. In a study that involved mid-school (as opposed to pre-school) abilities, Taubman estimated a variant of the ability function 2.1. The ability measure was a post high school qualifying test given to candidates for pilot training during World War II. Of the many sub-tests, Taubman constructed a scaled score predominantly reflecting skills in numerical operations, mathematics, and reading. Although the sample size allowed quite detailed estimation using nearly fifty independent variables, the explained variance in the ability scores barely exceeded 8%. Taubman's results indi- cate strong effects of parental education; but a single interpretation does not readily follow as there is no family income variable among the background measures. The effects of parental education may be interpreted as reflect- ing (i) the genetic endowment of the parent, (ii) actual earnings, or (iii) non-income related and otherwise unmeasured aspects of the home environment. On the basis of the low labor force participation rates reported for mothers in his sample, Taubman argues for an interpretation based on (i) and (iii); but this reasoning cannot exclude a family income-related interpretation as assortive mating could lead directly to the interaction of (unmeasured) family income and mother's education. A. Leibowitz has analyzed a set of data for which the ability mea- sure relates to the pre-school period. Using scores on the Peabody Picture Vocabulary Test as the outcome measure, At, she estimates the effect of a reasonably rich set of home environment variables. In particular she finds significant effects from family size, mother's education, the presence of - 30 labor-saving consumer durables (dryer and dishwasher)1, the presence of an encyclopedia, and joint mother-child activities,, such as readingo However, the work status of neither the mother nor father was found to have a sig- nificant effect on pre-school verbal ability, nor did the presence of con- sumption-oriented durable goods (TV, HiFi)1, nor attendance at day care, Head Start, or kindergarten. Children of mothers who watched TV regularly when the child was present did score significantly lower0 The data analyzed by Leibowitz allowed a breakdown into urban, suburban, and rural subsamples. Among the urban families1, the effect of mother's education was considerably weaker, as were the various activity effectso However1, the presence of labor-saving or education-related goods had considerably greater effects0 In contrast, only the reading activity variable was estimated to have a significant effect among suburban families; while in the rural sample,, mother0s labor force participation had a strong negative effect (which, however, was halved by accounting for TV watching)0 Family size was not related to the ability measure in either the suburban or rural samples0 These differential effects seem to illustrate that the specific activities which mothers engage in affect the human capital acqui- sition of their children as much or more than gross measures of socio-eco- nomic status or time spent with children0 The most important class of variables that emerge from analysis of the Gary Income Naintenance data concerns the effect of mother's status on child performance. Nother's status in the context of these families is reflected in education, heading the household, and participatirig in the labor forceo The effects of thib class of variables is strongest on visual-motor skills as measured by the Seguin Form Board and the Block - 31 - Design sub-test of the Wechsler Primary and Pre-school Scale of Intelligence. Since a major developmental focus of the pre-school period is physiological maturation, the coordination of motor skills with cognitive ability assumes considerable importance (arguably greater than pure cognitive effects). The interaction of mother's status with other variables indicates more how the status variables operate. This analysis suggests a degree of substitutabil- ity between mother's education and single parenthood in their effects and ability whereas labor force participation and single-parenthood (synonymous with female headed households in this sample) appear to have complementary effects. 1/ There is also some indication in the covariance analysis that earned income and regular employment may substitute for mother's education. On a limited scope of outcomes, it appears that labor force participation among less educated mothers hinders performance. In the affective domain, the major differential effect is that of labor force participation in fe- male-headed households--a result consistent with the importance attached here to the status of the mother. There is some indication that child-spe- cific characteristics as age, birth order, and spacing lose their normal importance as ability determinants in female-headed households. The second major category of variables which were included in the investigation of the Gary data were income-related. In the context of the income maintenance experiment it was possible to distinguish between sources of income (earned, unearned, experimental transfer payments) 2/ and 1/ In the case of two sub-tests on the IQ measure some puzzling reversals of the general propositions on mother's status occurred. Single parent- hood was found to enhance the effect of the better-educated mother. 2/ Unearned income generally consisted of governmental transfer payments other than the experimental payments. - 32 - to specify changes in average monthly income between the two years during which the study was conducted. Also related to this category were measures of the frequency and duration of employment. In general, earned income significantly affects a broader range of abilities than income from other sources. In particular, higher levels of earned income enhance the performance of children on the visual-motor sk1ills and on cognitive measures which have a significant informational component. In the case of the latter measures there does seem to be a significant topping-off effect to earned income, with the overall income effect negative at very high levels (relative, of course, to the sample mean)0 1/ Increases in earned income have consistently positive effects over the all IQ sub-tests that were used in the sample as well as on a "self-concept"' measure. The effects of increments in earned income are significant but much weaker in the case of visual-motor skills. The policy significance of direct income effects in this sample of low income urban blacks is great, for it indicates that within a disadvantaged socio-economic stratum, controlling for education and family structure, better performance in the economic system generates better outcomes for young children. The results presented here do not allow a precise specification of the means by which income affects performance; the major alternative inter- pretations would seem to be related to (i) increased allocation of goods 1/ This effect can best be interpreted as sample-specific, i.eo within the restricted range of income-earning opportunities available to families in the Gary sample, those at the high range had made in some sense, extraordinary efforts with sacrifices in other dimensions in order to attain incomes which for the population as a whole would not be considered high. - 33 - to young children or simply the presence of a materially more stimulating environment, (ii) a more achievement-oriented environment, stemming from the greater motivation of the income earner, (iii) the greater security and stability implicit in a wage-earning environment, though this effect should be controlled by the labor force participation status of the household, or (iv) residual effects of genetic endowment, parental intelligence, not captured elsewhere (in a sample for whom educational opportunities may have been arbitrarily limited, these effects may be plausibly proxied, at least in part, by income). Earned income also interacts systematically with other determinants of ability. In particular, there is evidence from several ability measures that earned income effects are greater for children with less educated mothers. This lends support to the productivity rationale for income effects or suggests that whatever income buys may act as a substitute for mother's education. In one case, income effects are stronger among children who respond more readily to the test situation. 1/ This sort of interaction is indicative of the basic identification problems that result from the feedback between the child's own development and the environment that further stimulates that development. The interaction between single-parenthood and earned income is inconclusive. On some outcome measures performance is sig- nificantly reduced by increases in earned income in female-headed families while on other measures income has positive effects. 1/ In all the various tests of pre-school children in the Gary sample, a minority of 10-15% refused to respond at all to the test. This state- ment is based on separate regressions which were computed for the sub- group of responders. - 34 - The role of unearned income and experimental transfer payments is much spottier. In the few cases in which experimental payments emerge as significant determinants they seem to behave in a complementary fashion to earned income and in the direction opposite to other unearned income. The implications of this are provocative in the context of redistributive policy; but the actual results are too thin to support more than speculation. The major implication relevant to LDCs is the importance of earned versus unearned income. A final set of explanatory factors that bear importantly on child- hood ability includes demographic factors such as birth order and child spacing0 1/ Within the rather narrowly defined socio-economic group that constitutes the Gary Income Maintenance sample, there emerge few consistent patterns to demographic effects across tests0 This heterogeneity seemE mainly due to differential demographic effects within different family types or different education strata0 The negative effects of birth order on performance were, with one notable exception, concentrated in two-parent families0 The magnitude of significant birth-order effects was not exceedingly large, however: the differential due to a single position later in the sibling order varied from the equivalent of three to seven months in maturation0 In some cases a sig- nificant, positive birth order effect occurred within single-parent families indicating, perhaps, some substitution of older siblings for the second parent in fostering certain skills (visual/motor skills in particular)0 In 1/ In the Gary sample, preliminary investigations determined that the number of children in the family had no explanatory power, except as a proxy for birth-order0 - 35 - all sub-scales of the WPPSI-IQ test the positive effects of birth order within the sub-group having high-school or better educated mothers dominated the full sample. One can only speculate as to why this effect should occur in the restricted domain of the IQ tests. A plausible but by no means certain explanation is that older children attending school particularly enhance the abilities of younger siblings in families which value and support education as proxied by the mother's education. The second demographic dimension that is investigated in some detail in the Gary sample is the effect of close child-spacing. On balance, a narrow gap between the subject child and his next oldest sibling retarded performance; if the gap were less than one year the decline in performance was often substantial, exceeding a standard deviation for some subgroups. These effects are again most consistent on the WPPSI-IQ measure for children whose mothers attained at least a high school education. This is at least consistent with the rationale offered above for the positive birth order effects in such families since close child spacing would vitiate the hypothe- sized cognitive stimulation of the older sibling who would not be sufficiently advanced in school. Negative child-spacing effects are also found in the non-cognitive school-readiness measure particularly for children in two-parent families in which an interval of less than 1 year with the next oldest sib- ling retarded performance by over 1.3 times the sample standard deviation of the test score. It should be noted that, anomalously, in the case of some visual skills and the self-image measure, close child-spacing actually facil- itated performance in two-parent families. On the whole, demographic effects on child performance in this sample do not reveal clear-cut patterns. One may conclude only that the - 36 - standard negative effects associated uith child order and opecing canot be assumed to hold for all disadvantaged groups. It atlo esmo clsasg thet effect of family size or spacing can be expected to viyy with pareicular outcome or ability measures. - 37 - SECTION III - Pre-school Intervention The previous two sections have examined in detail the role of ability in the modeling of earnings determination and the factors which themselves contribute to different childhood abilities. In the present section some comments are offered on the evaluation of intervention programs designed to improve pre-school abilities and on the prospects for organizing operationally-oriented research on these issues in LDCs. Numerous detailed reviews of the literature on compensatory educa- tion for pre-school children have been completed in the past four years. The accompanying paper by M. Smilansky 1/ reviews in detail most of the experi- ments and special projects in compensatory education conducted during the late sixties and early seventies. Smilansky's conclusions are similar to those of other reviewers, such as U. Bronfenbrenner 2/ in stressing the ephemeral gains produced by most well-designed intervention projects (ill-designed projects do not generally produce even ephemeral gains, though many different curricula have been found effective so that the constraint, "well-designed" encompasses a wide range of "treatments"). Smilansky's discussion of parental involvement is particularly valuable as this quite promising aspect of intervention programs seemed to be in danger of overselling as a result of the Bronfenbrenner review. 3/ There is no question that pessimism concerning the medium- term effects of pre-school intervention in IQ is well-founded. But 1/ M. Smilansky,"Priorities in Education-Preschool Evidence and Conclusions, World Bank, 1977. 2/ U. Bronfenbrenner, "Is Early Intervention Effective?", mimeo, 1972. 3 On this point, see also D. Weikart, "Parental Involvement Through Hlome Teaching," in Report 1974-75, High/Scope Educational Research Foundation, Ypsilanti, 1975. - 38 - before writing off any possibility of significantly augmenting the intel- lectual capacity of environmentally disadvantaged children, the following mitigating factors might be considered (as the thoughts of an economist wandering somewhat afield these observations should be appropriately dis- counted): 10 A critical consideration of the evaluation tools and the determination of appropriate criteria for success; 2. The adequacy of evidence concerning the appropriate period for intervention; 3. Two counter-instances of long-term intervention success. The evaluation pattern for most intervention projects has followed standard experimental procedure. An experimental group(s) and a more or less carefully matched control group are followed through the treatment period and beyond, with longitudinal measurement of the selected outcome measures. The treatment group and controls are presumed to have an identical distribution of environments outside the intervention program and subsequently (in primary school). The standard pattern of results reviewed by Smilansky and others is shown in Figure 1l Figure I IQ 120 115 ,' 110 - ' 100 _ 95 3 4 5 6 7 8 9 Age (years) Treatment Control Typical Pre-school Intervention Experiment Results (IQ units assumed initially at 100 for illustrative purposes) - 39 - Although formally similar, the methodology which has consistently generated this pattern of results may yield quite different implications than econometric methodology. The two approaches are illustrated in Figure 2: 1/ Figure 2 Experimental Econometric -ZI T > GI 2 Z z3 xI X0 .5 oP d S- 0) Cd CSo-ri HO CU Pe Po00, z~~~~~~~~~~~~~~~O a 0o 0 oz ~~~~~~~~CM ul ri r-i 0% oo H0 coi 0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ e o o o o c 0 0 0 0 g " 4 B CHO 0 . H (S '0 S'~ CMC'- C H z'd0 No co ~ (S 4f 'IO0'. C'\j _: -IA \0 0 Ed 0.- - I ', 'f'n( r- 'A %ON.' coV o 0 0 0 0 0 00 0 0 0 HO :r-lo\ HO HO H 0 000 0 -g t'.o H '- '- '0-H sm\ co 430~~~~~4 w n n wso owe w%s w 00 -4 HC03 1-H .,-41 ~~~~~~~ ~ ~ ~~.0 C - 77 - TABLEA._2 EXPLANATORY VARIABLES - DESCRIPTION & DISTRIBUTION VARIABLE UNITS MEAN (Standard DESCRIPTION OR RANGE Deviation) . AGE Months 38.2 (27.64) Age means on various tests obviously differ. TESTORD 1 - 5 - 2.78 (1.48) Position of particular test in the sequence given. Range from 1 to 5. TESTADIM 0,1 .37 Tests were administered by two sets of administrators; one was extensively trained, the other was not (1 = less well trained). KDORDER 1 - 9 - 2.81 (1.94) Position of child in family. NKDS 1 - 14 4.53 (2.71) Number of children in the family, calculated as the remainder after the number of parents is subtracted from resident family size, so includes other permanently resident non-family children and excludes non-resident children. NDC 1 - 3 0.27 (0.61) Number of "disabled" children reported by mother. KDSEX 0,1 o.48 Sex of child (1 = boy). KDGAP 0 - 1 0.26 (0.27) A measure of the closeness of spacing between the subject and his next oldest sibling; 0 - only child, 1 gap of 10 months. SGLPAR 0,1 0.65 1 = Single parent family, all are female headed. LFPR 0,1 0.39 1 = Family head participated in labor force at some time in two years during which testing occurred. UNENP 0 - 8 0.65 (0.97) Measures the incidence of unemployment of either the primary or secondary worker within the family during the two years in which testing occurred (1972-73). MOMHS 0,1 0.40 1 = mother attained twelve or more years of schooling. -78 - a) H 0Dr 0 C- o co -P~~~~~~~~~~~~~~ -P1- 0 H H o 0 0 0~~~~~~~~~~~~~~~~~~~~~ M r H *d. 0) .0E-40+)- 0 0 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~000 '0'O 0 c~~~~~~~~~~~~~~~~~~~~~ o'~~~~~~~~~~ ~ ~ ~ ~ ~ "~C -. 1.40 rn 0 (D~~~~~~~~~~~~~~~~ 0 (~ o 0( C\l 0~~~~~~~~~~~~~ IA Cd m~~~~~~~~~~~~~~~~~~r l \0 cl- C) 0 ON cq n o00 0 0' \4 4~o. 1' 0 00 0~~ ~~~~ ~~0 000 IH 0% o IV C.~4 co ) - Uo IH H 0 ch Xf\~~~~~~ 4 -4 W I a) 44~~~~~~~~~~~~~~~~00 l 40 4'' 0 Cl) Ci)oCO '0 ~ - 0MH C.) 4 4'0 0: co E-40 0l 0\ 0 -OH D '00 C8 Crof ~~~~~~~0\ 0\0 0 0 go W. (' 0%\' CO CO co 44 0 CO ) 0 a ('4 H ~~~~~~~~~~~~~~~~~~H C'J (0n c H 4$00 d5 - o 0 ~-14 U) co ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ON 0 ~~~~~~~~~~~~~~~0 004'0 0 0D N C.W4) 0T 0 -HC - E4- E'.0CHg 0 IA \0 a% UN u' 0 u'0co 0 0 Ck 04 U'~~~~ 0~~1 ON cn CY \'. -,0r4 ' (go p~~~~~~~ ~~0 C,,CO a,01 11 cm, _4 cz 0 0 0 0'a~~~~'dl to 0 coI-I rI Ci) ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~c,j~~~~~~~~~~~~~~go~~a)4~ E-4 o.4(-H OC'j'0C'J H -:?". ~~~~~~~~~~~~~~~~~~~~~~~~ 0 4'~Q r:,.4020 o \oo 000 0 -C00(C 1.la r C".J-- '.JCO CO (\' '0 14' 0d V - 02.-4 a ~~~~~~~~~~SHH I t~~~~~~~~~H 0 P 00.1. 0 r~~~~(go I I B B I *r-I .00~~~~~~~~~~~~~~~~~~~~~~~~~~- -OH go~~~~~~c 40 vo~~~~~~~~~0 HO 4 1.4 E~~~ 0) 00 -g 0 go~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~4 0 -4140 C 0- : -r40 C \ \ -- n I 1.4 41.. 0 :j, 0 0o -H v.4 0~ 0o 0n I *.-4- Is- 0 04 '04' i-~~~~~~~4 . 0 Ca O~~~~~~~~~ B C ' 4 ~~~~~~~~~~~~~~0 ~~~~~~~~ 4'.0~~~4 - -I 4 01.. '04'0 bD~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ZIr-4 (1) -4) 0 0 W Oo~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0 m c n5.4 0r4 0 C-Ia) O'i ~~~~~~0 0 43--- 02Hr-4 ~ ~ ~ ~ ~ ~ ~ ~~~~2 02 o 0 HA 4 - 79- > H r4 | S - o wm £ > £ £ > £ H. o 4 4a) co \0 * \He- H-1CC \4 C't '0 \ H. Cn CC) 0 Ir\ Cm H4 C.) c\J~~~~~~~~~~~~~~~~~c MCIH HO 00 H A A Ca * C"JCY.~~~ 0% 0 0)\ O H * . *O * * o fi W_ £ gm ° ii H H '.0>> COO S y , X, H CO f H > £ Ho 4 oC CO) r-i r C) 0 aO 0 = 0 4li U11 I' CC) El I A>^ S CO '.0.-~~~~ 0c Co 0 r'- 0%C'.J Jo Hu 0 j ^1 u aO h- -- E0- O(9 '-0 \.0 . H *~ * *rI0 * e -- P4 C) ON 0CmJ'AC~ \- *43 C)~ ~ ~ h:Cj OC OH HC) LI- °$ Is\ r-i L-~~% 'U r4 'IO~~~- a -"i C'o' 0I CO m '.O0 E-4 r-i cli ~~~co (\j Oe- CL\C) ~ H 0% I C~~I H\ . a) O* 4' Cs-tl - C'J H Ca ) 0 'U~~~~~~~~~~~~~~~~~~~~~0c CO ~~~~~~~~~~~ ~ ~~~~~~~~~~~co C- lX) 0I 0 C ) 04) IA~~~4 A c~~~OO'0 .4 'LA 0 0 ~~~~~~~~~~~~~~~~4 r- Cj CIJ co ~ ~ ~ ~ ~ ~ ~ ~ - 4 0 0 0~~~~~~~~~~~~~~~~~~~~l - 80 - 0~~~~~~~~ C ~ ~ ~ ~ ~ ~ 0 xi~~~~~~~~~r C% :: X Q~ w cm - :2 C)l a 0 0 0 00 0 0 C4 .hco4' - _~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~*- cn a: cn On r 0 9l CC)o 0 0 0 0 0 0 0~~ 0 r0 '%O (Y)~ ~ ~ ~~~~~~~~~~~~~~~~~~~~~~4 r- C'J C\c to 8O-:Tn l FY4 0 0 0 - 81 - 4)3 0 * 0% ~~~~~~~ M0~ co cli ~ ~ ~ c ho r- * . . 0~~~~~~~~~~~~~~ co _ - r1i C\JO a~~~~~~~~\ a 4 c C\ ¢ a £ x SH o S > - , 8 i tf < \0 t t_ h~~~~c l Ll E-4~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~- C'J ('~~~P 04 0 CN , , f C~ ~ ~ ~ ~ ~ ~ ~ ~~n @~~~~~~~~~~~~~~~~~~~~ r C \ N \ P 4 N00 CO CmHO l~~~~~ H > aa N > : N < N @ O < O O h X~~~* a | v | o @ N > t > ^ O > > > ca CJ 2 .0 N _ I < N ^ I < I n CO z 4 -4~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~- ON c N3 f C\ CMN H U N 0 C 0% C'J 0% C'~~~~~~~~~~~~~~~~~~coC CC)%\ 0N -4\j1 0ri * H~~~~ 'h\ 0 -- IS\ 0% - 4 tXCI A ZA "A c 4) U)- CZ~~~~~~~. C* * * CZ 04 J\l lj HI H aC U) ccZ) 0%r c1-iH - 82 4J 0 ca Cc 0 LO 2 o sS n Q o o C4o cn-flx~ '0 HO u4 a Co o 8 'Lrsr~ H1 H Q o o I H \02 0X D '0(l D C 4 sO'0 ~ - r- c'J ~ 002\0 C N a ~~- C'- \O ~~:3 ~- r- 0 tr f\ \10 H dC 0 0 0 0 0 0 0s 0: £ II ce o0 H o 0 ° °oN4 <~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~b r D N C,, 0 C~~~~~~~~a 0 0 0 C ka ao 03~ 0202, "0~~~~~~~- '-.0 o~~~~~~~~ %O t- co \~~~~~~~~~~~~~~~'0 --Tci 0 ~~~0 0 0 0 0 CM 4J r~~~~~ if ~~~~~~~COHr- 0". XA 0 C1) <~c"J C\ 8 CM IA \ - 83 - 'I)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\ CHO 73 $ q) n co cUn 0 H O 0 n co \0 i< ~~~~~I '0 10 Nve 1. 11 . . . . .. 0% in c% C\JHco C m-J HU, | HC\J 0 0 _H H'- 0 - H N \ 0~~~~~~~~~\ \ CD~~ ~~~ m H O c - CO C\ C\ $ AlN S n u 9Co A'II'C-C -C - -i 0 v) *H 4 \ 0co \j U 0i h CC) H 8 ~ ~ ~ ~ ~ " H C- r4 -t, IF U3 3A n -3 co I ^r4 U N 9 2D * - 0 0\CN c \o C 0 R 02 @ (C 0 H H O C I $\ C CiH o 0n (I.(O-O ,-4 C\ 0 O N c-a C\0j 0 \ C _ H *24011 0 'IA .40%rY' 4 0~'I 'It\ J"- o c0 0 0~~~~~~~~~~~~~~~~~~~~~~~~~~~~0 0 VI~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0 En 2 HC O -* ° H H , * 0 * rn ~ O0 -%O~$% - c00 C,L. F-* . .. . Xl Dal 0a 0 ~~~~~~~~~C~0 r f' %0 0~~~~~~~~~~~~~~~~~~~~ rl)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0 0 ~~~~~~~~~~~~~C 10c 4 \ciU H 4 C1 . ' C' "\0 r¶'' % " 0 ~ ~ U ("*1 '0 C"J H ~~~ ~~~~~~~~~~~~0% CNJ 'IJ rjII 0 HO (~~~~~~'I ~~4~~t\\O O 02~~~~U U) rl~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0 044 m o~~~~~~~~~~~~~~0 CO U) 0 U) U ~ ~ ~ ~ ~ ~ ~~ 0)0 " E-i, 'C\ C ** 0 g~~~ c ON co co C bem G~~~~~~~~~~~~~~~~~~~~~~ 0 ~~~~~~~~~~~~~~~~~~~~o 1 \0~~~~~~~~ ~~~~~~~~~~~~ 0 C), Qa4 7325Z It C 1~ ~ ~ ~ ~ ~~~~~~~~~~~~~~~C Ei Hp 0s ~~~~ ~ ~~N C\j (I E> co 1 Crl 0$ 0\ 10 t( ,x IC\ co0 ': i cn A .e ~~~~~~~~~~~~~~~~~~n O v) NO l~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ @f c 0- - 85 - C') 0i 1 CM CO 0 0 'IOH 0 H C 0 0 -o --I tn \ \ r- (N :c q ) coH cJi CDo . ... .... 0~~~~~~~~~~~~~~~~1 C\j 0 cm CN ~ ~CN Cj en'i0 '.0 0 r 'IO 0\'.\.r t-'.*f \C O ' \O \- CO H I I I HI I--HI co 0 co~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. CP\~~~~~~~~~~~~~~~~~~~~C 01%) ~ ~ - - ---I- 0 tQ O r S S O . . O g 8co C)0)11 H C'- (Ni en 0 ; H -- CC) G =~~~~~~<(1 S * SX * CO ~ ~~ C\ \ HHi cv-i r0 CO C\j ()C * H 0 i-l- -x '. ' CO 0 :11 \ N b O H N\ C t H S ~ ~ ~ IS I 0 H r- it-. 0. -i 0 U) 0 E~~~C-C 0o 0 (CM Al t- r en 1>\Nci J MO -o H N t- 00 'i C ( C" O. UO0 0'. m ~ ~ H . . * . . . HH _ o2 . O C\J '' H CM P4 00U I . . . * * * 0 m~~~~~~~~~~~~I 1<- 0.S 86 miW~~~~~~~~~~~~~~~~~~~~~l ri vo Cao ts~~~~~c co 8 cm gl a: SE N F~~~~~~~~~~~~I c%j co 1 g S O X y c_ 8 _ F H F z9 0 Si r = U] B ' o o 9 H o 9 o i - 1 X S< § 8 t < | E; g~~~~~( 0 c- '51 ~ ~~~~~~ ci '10 0 a- 0L- 1 1 1 02 > I ~4 F ha,< Ut :2 U) | <; ~~~~~~~~~~~~~~~~~~~~r-i lclg 4-2 C> 1 fi | ~ ~~~~~~~~~~~~~~~~~~~~~~~~ '0 ' n H O ~~~~~~~~~~~~~~~~1 Eo 0 co C M8 S X x o o o ~~VCm o o ;~~~~~~~~~ H 0~~~~~~~~~~~~ O E: (V c t Co _0 o 0 0 0 0 0 0'. -0" o '.-) H \ - CV H- C\j IA H0 0 C\V '.0 i - cje 0 0 0 0 0 0 0 0 0 0 0 0 a a~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ C0 NH H c 0 a\a 0 o oH4 n 0 0 0 ~~~~~~~~ 000 0 0 Cf) 0 0 0 0 0 CV 0 0 C\V C/3 c 8- 0 0 0 0 0 lo o o o o o o o o o - 89- co cli~~~~~~~~~~~~~~~~~~~~~~~~~~: cli OC) r- co co ~ ~ ~ 4. ~ - P '.0 e H sD rt- l 'I O. 0 U\ H H > -b X o rQ ~~~~n . 4 n 0 0 ~H '0 C \i'. U3 090 0 9 0 - I UI S IA \r 0 £0 0 ~~~4 0 n~~~~~~~ 0 0 -: CC)S > 3z C O9 0co iSt~~c *0 mses i o 0. .? . r c) X > n Cq £ cq r n o~~~~~~c co r- O \O V\ tOa0 F \ 0s% .0 0'. ct) ~ ~ ~ ~ ~ ~ ~ ~ c 22'. HH H O Z 1 00 1 *.C *' *~ . ~ * * H % :> H O ^ CM H -X - .-4 c cv~~~~~~~~~~- 0 m \0 04i-I c C'-0 0 0 00 If\-4r-0 o CP\' C/)cr\c.c.ifH- L $ 0~~~~~~ CC/~~~~. 0 0\ co ij.~~~~~~~~ - ~ c C-) cf) 0 0 Q\9 29 co 1 13Lj C\j co COH c o 0 LA~~~ H ON r-i 00 r-i C\j 00 Q - 90 - a% a% 'IO cn 0 \ NM n cl as o g --:ON 1 0 0% oJ vaJ (\ 030 ) C) ' 1 \S ON 0\A H N\ CM rO z _:I o 0 o 0 0 0 0 o n P.* 0 0 % co V% $ OE; Si 0\ CO \ r-i t-co ON E 1Jn ON co HO 0 0 0 0 0 0 0 0 0 0 in cl; in r- CM cwOO C 0% E-4~~~~~~~~~~~~c0 ~~~~~~~C\j c' 0O ID0 r0 O H c H CC ' - (¶ c0 'N ° o-4 9A o 0 0 0 0 0 0 C) U) o I C8d J ° UE UE N N :t~~I^-4 V II\ C\J lil U) o n a:>~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~l 0% 0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ u H z l % H : - 5 O n XA cj 0 O~0 cm,\ O0 o ON CM UN i9 rU H ~~ ~~~~0 0 0 0 0 0 0 0 Lr- C\j c( 0 _ o a:N a 0 H \O O C/) c --I 0 I.l\1\ C1 cmJ in o 0 0 0 0 0 0 o