WIXLIB

—7—The mean family incomeoftional average of 12,500 and 9,280 is considerably below the 1972 na-isindicative of the oversampling of lowprotein intakes of 1516 and 56and calorie standards of roughlyin-come families. The mean calorie andgramsare considerably above the protein1330and26, grams for children of the age and weight corresponding to the samplemeans.5 This finding of higher average protein and calorie intakes thandietary standards is not surprising given the similar results from theTSNS, a sample characterized by significantly lower family incomes. Children's intakes of calcium, vitamin A and vitamin C average two to threetimes recommended dietary standards in the MANES working sample. In thecase of iron, the average intake is two—thirds of dietary standards.In column 1 of Table 2, levels of growth, health, and nutrient intakesare presented for children in households falling into the upper and lowerthirty percentiles of thesignificantgroups,differencesin height, weight, or head growth between thesenor in protein, calorie, vitamin A,of vitamin C andemergespoverty index (PIR) distribution.6 There are nocalcium intakesor iron intakes. In the casesa statistically significant differencein favor of the higher PIR group. For both groupsmeannutrientintake levels consistently exceed dietary standards, with the exception ofiron. There are also no significant differences in hospitalizations although the lower PIR group had a significantly greater number of colds inthe six months prior to the medical history. The average family income andhouseholduressizefor the higher PIR group are 14,766 and 4.2. The same fig-for the lower PIR group are 3,673 and 5.7.Inthe remaining portion of Table 2 similar high and low PIR compari-sons are made for blacks and for whites in the working sample. The pat-terns of statistical significance within these stratifications areto that for the sample as a whole.similar

15.5097.9949.249358.2787.94151456.26White 1.01t n—White the Upper and Lower Thirty Percentilesbvalues are for significance test of difference between —.10t—aThe sample sizes for the lower and upper thirty percentiles are, for the entire sample, 1036 and 918; forwhites,521 and 831; for non—whites 515 and 87. The Imbalance in the white and non—white categories results from the use ofthe entire sample thirty percentile cut—off values for these sub—sample stratifications.Daily caloriesDaily protein (gm)Daily Vitamin C (mg)Daily iron (mg)Daily calcium (mg)Daily Vitamin A (IU)Weight (kg)height (cm)Head (cm)HospitalizationsColds (last six months)All eMean Levels of Growth, Nutrient Intakes, and Health for Familiesof the Poverty Index DistributionTable

—9—Further information on the nutritional status of young American childrencan be obtained by examining the diets of children light for their age andsex. In our HANES working sample the calorie and protein intakes of childrenbelow the 10th percentile in weight for their age and sex are 1440 and 53 grams,not significantly different from the working sample means and indicative ofmore than adequate intakes of these nutrients according to dietary standards.The mean family income for this group of light children is 8470. Unlesspresent and past nutrient intakes are not correlated, these numbers imi,ly thatinfluences other than diet may be responsible for producingassociatedthe condition usuallywith undernutrition. The consideration of the emprical results in thenext section will, enable us to come to firmer conclusions about the role ofsocioeconomic variables in the choice of diet by arents for their children andabout the subsequent effect of nutrient intakes on children's growth.IV. EmpiricalResultsAt the empirical level we have estimated several variations of our modelof children's diet, health, and growth. With the excei,tion of calories andprotein, the nutrient intake variables did not ai'proach statistical significanceon the growth equations, either because of their high correlations with roteinand calorie intakes or because they have very small impacts on growth at themargin. The health variables also oerformed poorly in the growth equationsin the statistical sense, apparently because these conditions have minor orvery short term growth effects that are rapidly overcome.In the presentation of the empirical results, therefore, we etmhasize amodel with the following endogenous variables: height, weight, head circumference,protein intake, andcalorieand for vitamin C intake.intake.We also report results for the colds variable

uauo uao.idIs-suptdx iaq o suzs uatuq itoi8 qatpaiadd os uaod uaq ua4a suoinba asq u.aqurnu ouIdxa ue s ppnppaqoidd '1aAatoi 's3poj;—u osie 1atbsps s asnaq uoTnbauE ;u2su-r 1-aqiuoi; ppnpxa uaaq sq aqw.1eA uTao1d auiaisapoi josuoTnba qo1 suap'[pp aq ;o sa1ursa uoTnba snou peuasa.idirnrs au4tO,I sua1pp43 iu31ppp unot( .o; uospap ap aqS[EA asaq ;o .r&.iodur aq uopsbUTuoza;o uoT oaodtwsqns u1dxa oui no UT saqpeA snouoxpaaxaaqo supq s&u-j uapi-iu utjti.y.i;o uou 'suo1nb aeu' uinusnsaz wo panpai atp ;o 3ads ups.iau iaouysaIqT1uapudappauTdau'—tin at sanpiuaiu pu aua aq apn-UT 'uo SaqB1eAApaxt eq suosso.zez qoib s,U31p43 o azts ptottasno pu saq'ç aseq ;ouopTpp aip pu sanT—T'tLIaua2uopnpa s4iaqom puU1O3UT 4ITw; 4Joi2 suaipap UTUtTdXa uT U3TUS pauv o aiesaq&A o as e- aqtoqs STTtSa. aqjioi s1ua1p-p3 ;o uopupuaq u S.ZO3TpUT e1OAeqaq UB 3pUOUO3aOpoS a4 SflSIaA Sa[q1ATuad pu aua aq ;o 93UaSfl3q suuo;TTUS aq uoapTAo.d iaq uoTu1.1o;uT aqpanpa.z qo,z s1uaiptp aq u pasa.IauT AITnc1?d aiaq UT pauasaid ai s-nsa1 uuo;panpai auiuia; T paq aqpu 'pq ptoasnoq aq ;o uonp aqptoastoq'awoQutpuiwu1np UB 'azsptotasnoq Aq pe.znsvaux saouant;trr -mouooaoos pu .xapxo qTq'qa siaioui 'sçq suaid 'ai 'xas 'ae iq panseaiuIuaid pu paua o sau uepnu pu 'aç 'qoi sua.ipppwai(8)—(9) suotnba u1o.; pa4-pap sdTqsuoTTa uuopanpaisU1o.—ot —aqjpa3npa'y

— 11—Table 3Reduced Form tSexBirth orderMother's weightMother's heightRaceIncomeHousehold 80.05(2.41)(1.55)(1.67)(0.79)0.378.380.09—0.37(1 0.02(—1.41)—0.04(—1.33)Schooling 2Colds(34.61)0.13(4.54)Father's heightProtein Vitamin C4.71(5.21)Birth weightHeight Calories(3.59)(U.16)Age squaredCircumferenceDevenden t 9(3.92)—0.12(—0.11)(0.06)Schooling chooling 0.08(0.36)(—3.84)(1.09)(1.78)(1.11)(0.66)Sex of 78.92515a(0.81)statistics in 2515.035.22515

— 12 —Table 4aStructural Equation Estimates for Chi1dren' Growth and Health,Three Stage Least S4uaresDeendent 1(1.49).002(5.02)Vitamin C.011(.10).84AgeAge 7.10)(—9.45).48SexBirthweightBirth orderFather'sheightMother's .06)(.72)— .19(—3.82)Mother's height.24(31.52)— 001(—6.44)Household size.06(1.86)Schooling 2.72(—4.90)Schooling 3—1.40(—6.83)Schooling4—. 79(—2.84)aN2515b.indicates endogenous variable

—13—TABLE4bElasticities of Selected Variables aDependent ht.05CaloriesCircumference.25.20Age.38.42.21Birth weight.004.16.005Mother's weight.004.07.01Mother's height.33.46.54Father's height.27.27Race.003.001—.005Sex.003.004.003at mean values of dependent and independent variables.bindicates endogenous variable.

— 14 —ofthe diet. Protein and calories are highly colinear (r .82) so a good portionof the protein influence is captured by the calorie variable. An argument with asimilarframework explains why protein appears in the height and headgrowth equa-tions while calories does not.9In elasticity terms the most important variables in the growth equations arechildren' sage and height of the mother andwere expected anddemonstrateheightagain the importanceof the household decision makerinof the father. These resultsofvariables beyond the influencethe children's growth process. A result that issurprising is the rather substantial elasticities of children's growth with respectto nutrient intakes. The elasticities (at the means) of height and head circuxu—ference with respect to protein are .05 and .25, respectively, and the elasticityofweight with respect to calories is .20. These results imply that an increaseindaily protein consumption of ten percent or about five and one—half grams wouldincrease height by an average of one—fifth of one inch andheadcircumference byan average of one—half of one inch. A ten percent increase in calorie intakeswould increase children's weight by an average of seven—tenths of one pound. Theprotein elasticities in the height and head circumference equations seam particu-larly large in light of the fact that protein intakes average more than twicedietary standards. The protein effects on growth seem to be linear throughout therange ofappearintakes characterizing the HANES working sample: that is, we do notto be approximating a non—linear protein effect with a very small impacton growth at the margin. 10These substantial elasticities of growth with respect to nutrients thatare consumed in excess of dietary standards is consistent with the findings ofthe evaluation of the Special Supplemental Food Program for Women, Inf ants andChildren (WIC). This analysis showed that although children in poor Americanhouseholds generally consumed nutrients well in excess of dietary standards,their growth could be accelerated by increasing nutrient intakes. In light of

— 15 —these findings the WIC evaluatorsrecor!,merlded a reassessment of dietarystandards and singled out protein in particular (Edozien, et.al., 1976).C. Nutrient Intake EquationsThe simultaneous equation estimates of the protein, calorie, and vitaminequationsCare presented in Table 5. The results are similar for each of thenutrients. Simply stated they indicate, within the context of our.model, thatchildren get the amount of these nutrients that they "ask for." The child'sfor nutrients, represented by weight of the child, is a very ir!rnortantdemanddeterminant of intakes. The nutrient—weight elasticities are about one ineach case and the t—values of the weight coefficients are substantial. Thefamily income coefficients approach statistical significance but imolv veryelasticities (about .02 in each case).smallEducationnutrientof the household head has apositive but nonlinear effect onintakes. Children in families where the head has 12 years of schoolingreceive about five iercent more of these nutrients relative to children in familieswhere the head has less than 12 years of schooling. However, this education differential falls when children in families where the head has college or graduate edu-are compared to children in families where the head has less than 12 yearscationofschooling.V.primary purpose of this paper was to investigate the extent to whichAfamilyyoungSunaryincome and education are obstacles to the provision of adequate diets forchildren in the United States. Based onour examination of the HANESdatawe have, found that:1. Average nutrient intakes of youngdietarychildrenstandards, with the exception of iron.are well above recommended

— 16—TABLEStructural5aEquation Estimates for Children's Nutrient ProteinIncomeHouseholdsizeSchooling 2Schooling 3Schooling 4Sex of 3)(.70)(—1.16)aN 2515.b.Vitamin C4.03(—1.9)SexCalories(12.78)(—1.06)Age squaredVariablesindicates endogenous variable.

— 17 —Table5bElasticities of Selected VariablesaIndependentVariablesWeightDependent ncome.02.02Household Size.05.05aCouted at mean values of dependent anddent variables.bindicatesendogenous variable.indepen-

— 18 —2.nomicAverage nutrient intakes for children in households of lower eco-status arevery similar to intakes of children in householdsof highereconomic status. Rates of children's growth are also similar in thesehouseholds.3.Family income and education of the household head have statisticallysignificant but very small positive effects on the nutrient intake levels ofyoung children in the model of children's diet, growth, and health estimatedinthis paper.These findings are very consistent with those from a similar analysiswe performed with the Ten State Nutrition Survey.oftheA most interesting resultpresent study is the rather substantial estimated effects of proteinintakes on children's height and head growth, even though protein is con-sumed well in excess of dietarystandards. Thisfinding and the apparentcorrelation between children's growth and their intellectual developmentbrings to question the adequacy of present protein standards. Couldamericaninmothers, who provide very high protein diets for their childrenhouseholds at a].]. levels of socioeconomic status know more about whatconstitutes an adequate diet for their children than the experts do?

.F—lFootnotes'For examples of research into the problem of undernutrition in Americanschool and pre—school.children in the U.S., seeSims and norris (1974), andOwenChristakis(1968), Owen(1969),(1974).2Owen (1977), in his review of the effects of nutrition on growth anddevelopment, concludes that the "evidence, which still should be considered preliminary in nature, . [indicates] that bigger is srtarter, at leastcognitiveamongpre—school children."3More formally,it couldbe argued that rates of children's growth enterthe utility function in a non—linear fashion and that(e.g.excessive rates of growthobesity) are negatively related to parent's utility.4it is often pointed out that inagricultural societies parents are veryconcernedabout the size of their children because physical strength is animportantcorrelate of individual output. Although a desire for larger childrenin modern societies may not be based on a similar observation, there is evidencethatthe height or weight of children at younger ages correlate with their in-tellectual develomnent and health in later years, and thus withtheirfutureearnings.5The dietary standards cited in the text are those of the HANES dietarystandards committee for children 24—47 months weighing the sample mean of 15.5kilograms.computed in HANES, the poverty index ratio takesaccountof householdincome, household size, and household diet requirements as reflected by the agedistribution of the household members.

-7F—2Adjusted R2,s increased by less than .01 when the socioeconomic variableswere added to either height, weight, or head size regressions that already contained age, the square of age, sex, parent's height, birthweight and birthorder.It should also be pointed out that the limited significance of the socioeconomicvariables does not appear to be due to colinearity with the genetic and parentaltrait variables. The t—values of the socioeconomic variables do not increasemarkedly even when the genetic and parental trait variables are excluded fromthe children's growth equations.8The results for the colds variable are also presented in Table 4 but arenot discussed in the text. Household income and education of the householdhead are inversely related to the number of children's colds while protein andvitamin C intakes do not have statistically significant impacts.9Because the growth equations formed part of a simultaneous system traditional F tests could not be employed to test the individual and joint contribu-tions of the protein and calorie variables. Results from OLS regressions indicate that protein makes a significant incremental contribution to explainingthe variance in height and head growth when added to regressions containingthe other independent variables, while calories does not. When both dietvariables are added jointly to height and head growth regressions the incre—mental contribution is insignificant. For the weight equation, the incrementalcontribution to explained variance is significant when the trotein and calorievariables are entered individually or jointly to regressions contaning theother independent variables.

F—310Predicted protein and the square of predicted protein were enteredas independent variables in the height and head circumference equations inthe final stage of a two—stage least squares process. No evidence of a non—linear orotein effect was uncovered.

R-].REFERENCESChase, A. "Poverty andLow IQ: TheVicious Circle," NewYork Times BookReview, July 17, 1977, p. 9.Chernichovsky,Dov, andchildren a

052—26556 Newark, NJ 07102 (201) 648—5940 National Bureau of Economic Research 15—19 West 4th Street, 8th Fir. Washington Square New York, NY 10012 (212)598—7045. An Economic Analysis of the Diet, Growth, and Health of Young Children in the United States . The child's health status is a function of his diet, growth and otherinuts which