INEQUALITIES IN CHILDREN’S OUTCOMES IN DEVELOPING COUNTRIES Intergenerational Associations between Parents and Children, Inequality and Poverty Jere Behrman, University of Pennsylvania Discussant: Karen Macours (Paris School of Economics) University of Oxford, 8–9 July 2013 Intergenerational Transmission of Poverty and Inequality: Young Lives Jere Behrman, Benjamin Crookston, Kirk Dearden, Le Thuc Duc, Subha Mani, Whitney Schott, Aryeh Stein, and the Young Lives Determinants and Consequences of Child Growth Project Team Research support acknowledged from BMGF Global Health Grant OPP10327313, NICHD Grant R01 HD070993, GCC Grant 0072-03 and, through Young Lives, DFID and the Netherlands Ministry of Foreign Affairs Introduction • Considerable literature on how family and community background affect investments in children • Empirical estimates generally show significant associations between family and community background and child outcomes • However, not much evidence on how much improving family and community background would reduce the “poor” (those with the lowest levels) and inequality of child outcomes Introduction • We provide estimates for Young Lives countries for parental resources/human capital and child outcomes, and implied future child household consuption (as adults) under the assumption that intergenerational familial and community associations reflect causality • Likely to be upper-bound estimates because of unobservables and tendency for reinforcement of endowment differentials, though random measurement errors or compensation for endowment differences work in the other direction Digression on Estimated Impact of Parents’ Schooling on Child Schooling in Minnesota (Behrman & Rosenzweig AER 2002) Strong emphasis on critical role of parents’, particularly mother’s, schooling But interpretation difficult because: • Intergenerationally correlated endowments E • Assortative mating on S and E Child schooling determined by parents’ S & E (1) Sc = aSm + bEm + cSf + dEf + u Assortative mating relation: (2) Sf = mSm + nEm + v Control for E: adult within-identical twins Assortative Mating: Impact of Additional Year of Own Schooling on Additional Year of Spouse’s Schooling (Relation 2) – Association Because Spouse More Schooled (though less so in standard estimates) AND Has More Endowments 0.7 0.6 0.5 0.4 Own Sch 0.3 0.2 0.1 OLS Twins Impact of Additional Year of Father’s Schooling (Sf) on Child Schooling (w/o and w Sm) 0.5 0.4 0.3 0.2 Effect Sf 0.1 OLS OLS w Sm Twins Twins w Sm Impact of Additional Year of Mother’s Schooling (Sm) on Child Schooling (w/o and w Father’s Schooling Sf) (Note: Controlling for Endowments, More Schooled Women Spend More Time in Labor Market) 0.4 0.3 0.2 0.1 Effect Sm -0.1 -0.2 -0.3 OLS OLS w Sf Twins Twn w Sf Data for Present Paper • Longitudinal study of poverty led by Department of International Development at the University of Oxford, with research and policy partners in Ethiopia, India, Peru and Vietnam • Fairly representative of the population, except the highest part of the income distribution • Involves roughly 12,000 children (8,000 enrolled at ages to 18 months, 4,000 at age years) • We use 5,763 children from younger cohort, collected at ages ~1, 5, and in 2002, 2006, 2009 Child outcomes • Scores on cognitive tests at age – PPVT: Peabody Picture Vocabulary Test, measure of cognitive performance and arguably ability – Math: early math skills – EGRA: reading test • Height-for-age z-score at age 8, HAZ(8) Father’s schooling Vietnam 10 15 10 15 India Peru father's completed schooling, reported in R2 Ethiopia Graphs by country India Peru Vietnam 10 15 10 15 Ethiopia Average over R2 and R3 total monthly consumption * 12/365 in USD Household per Capita Consumption per Day (average over R2 and R3) Graphs by country Estimated Inequality and Poverty for EGRA: Gini none Ethiopia India Peru Vietnam MS=P MS=9y MC=20p MS=9 &MC=40p, CW40p 0.308 0.006 0.344 0.005 0.208 0.004 0.134 0.003 0.298 0.005 0.322 0.006 0.200 0.004 0.130 0.003 0.274 0.005 0.301 0.005 0.188 0.004 0.122 0.003 0.306 0.006 0.344 0.006 0.200 0.004 0.132 0.003 0.264 0.005 0.292 0.005 0.170 0.004 0.120 0.003 Notes: Standard errors below coefficient estimates; zeros coded to 0.4 MS= minimum schooling, MC=minimum consumption, MCW = minimum community wealth, P=primary, 5y=5 years, 9y=9 years, 20p= 20th percentile, 40p=40th percentile Estimated Inequality and Poverty for Math: Gini none Ethiopia India Peru Vietnam MS=P MS=9y MC=20p MS=9 &MC=40p, CW40p 0.432 0.007 0.302 0.005 0.212 0.005 0.171 0.003 0.413 0.006 0.264 0.004 0.204 0.004 0.162 0.003 0.363 0.006 0.237 0.004 0.190 0.004 0.146 0.003 0.427 0.007 0.297 0.006 0.203 0.004 0.168 0.003 0.344 0.005 0.224 0.003 0.175 0.004 0.138 0.003 Notes: Standard errors below coefficient estimates; zeros coded to 0.4 MS= minimum schooling, MC=minimum consumption, MCW = minimum community wealth, P=primary, 5y=5 years, 9y=9 years, 20p= 20th percentile, 40p=40th percentile Estimated Inequality and Poverty, Math: PH none Ethiopia India Peru Vietnam MS=P MS=9y MC=20p MS=9 &MC=40p, CW40p 0.131 0.010 0.164 0.009 0.182 0.010 0.151 0.009 0.131 0.010 0.115 0.008 0.171 0.010 0.137 0.009 0.062 0.007 0.053 0.006 0.146 0.009 0.091 0.007 0.130 0.010 0.163 0.009 0.169 0.010 0.149 0.009 0.033 0.005 0.042 0.005 0.104 0.008 0.063 0.006 Notes: Standard errors below coefficient estimates; zeros coded to 0.4 MS= minimum schooling, MC=minimum consumption, MCW = minimum community wealth, P=primary, 5y=5 years, 9y=9 years, 20p= 20th percentile, 40p=40th percentile Estimated Inequality and Poverty, EGRA: PH none Ethiopia India Peru Vietnam MS=P MS=9y MC=20p MS=9 &MC=40p, CW40p 0.149 0.011 0.192 0.010 0.154 0.009 0.147 0.009 0.149 0.011 0.192 0.010 0.148 0.009 0.143 0.009 0.130 0.010 0.131 0.009 0.125 0.009 0.130 0.008 0.149 0.011 0.192 0.010 0.148 0.009 0.147 0.009 0.116 0.010 0.128 0.008 0.074 0.007 0.119 0.008 Notes: Standard errors below coefficient estimates; zeros coded to 0.4 MS= minimum schooling, MC=minimum consumption, MCW = minimum community wealth, P=primary, 5y=5 years, 9y=9 years, 20p= 20th percentile, 40p=40th percentile Inequality and Poverty, Children’s Generation Table Gini coefficient and poverty headcount (PH), next generation Ethiopia Gini PH 0.350 0.186 0.008 0.012 India Gini 0.252 0.005 Estimated mother's schooling 0.370 0.624 0.005 0.015 Estimated father's schooling Estimated mother's height Estimated consumption PH 0.126 0.008 Peru Gini 0.324 0.008 PH 0.119 0.008 Vietnam Gini PH 0.307 0.107 0.008 0.008 0.393 0.003 0.604 0.012 0.191 0.004 0.068 0.007 0.160 0.004 0.059 0.006 0.313 0.295 0.006 0.014 0.356 0.005 0.377 0.012 0.144 0.003 0.016 0.003 0.153 0.004 0.039 0.005 0.020 0.020 0.020 0.020 0.000 0.000 0.000 0.000 Notes: Poverty line is 20th percentile of parents' distribution for consumption per capita, and is grades for mother's and father's schooling attainment Inequality and Poverty, Children’s Generation Table Gini coefficient and poverty headcount (PH), next generation Ethiopia Gini PH 0.350 0.186 0.008 0.012 India Gini 0.252 0.005 Estimated mother's schooling 0.370 0.624 0.005 0.015 Estimated father's schooling Estimated mother's height Estimated consumption PH 0.126 0.008 Peru Gini 0.324 0.008 PH 0.119 0.008 Vietnam Gini PH 0.307 0.107 0.008 0.008 0.393 0.003 0.604 0.012 0.191 0.004 0.068 0.007 0.160 0.004 0.059 0.006 0.313 0.295 0.006 0.014 0.356 0.005 0.377 0.012 0.144 0.003 0.016 0.003 0.153 0.004 0.039 0.005 0.020 0.020 0.020 0.020 0.000 0.000 0.000 0.000 Notes: Poverty line is 20th percentile of parents' distribution for consumption per capita, and is grades for mother's and father's schooling attainment Inequality and Poverty, Children’s Generation Table Gini coefficient and poverty headcount (PH), next generation Ethiopia Gini PH 0.350 0.186 0.008 0.012 India Gini 0.252 0.005 Estimated mother's schooling 0.370 0.624 0.005 0.015 Estimated father's schooling Estimated mother's height Estimated consumption PH 0.126 0.008 Peru Gini 0.324 0.008 PH 0.119 0.008 Vietnam Gini PH 0.307 0.107 0.008 0.008 0.393 0.003 0.604 0.012 0.191 0.004 0.068 0.007 0.160 0.004 0.059 0.006 0.313 0.295 0.006 0.014 0.356 0.005 0.377 0.012 0.144 0.003 0.016 0.003 0.153 0.004 0.039 0.005 0.020 0.020 0.020 0.020 0.000 0.000 0.000 0.000 Notes: Poverty line is 20th percentile of parents' distribution for consumption per capita, and is grades for mother's and father's schooling attainment Intergenera(onal  Transmission  of   Poverty  and  Inequality:  Young  Lives   Discussion:  Karen  Macours   Paris  School  of  Economics  -­‐  INRA     Summary   •  How  large  is  the  intergenera(onal  transmission  of  poverty  and   inequality?   –  This  paper:  through  rela(onship  between  parental  human  capital  and   investments  and  children’s  HK   –  Es(mate     •  ln(C)  =  f(parents  schooling,  mothers  age  and  height,  unobserved  family  facto   =>  This  rela(onship  is  assumed  to  carry  over  to  child’s  genera(on   •  Child’s  human  capital  =  f(C,  parents  schooling,  family  and  community   characteris(cs)   –  Simulate  what  happens  if  one  increases  parental  schooling,  consump(on,   and  community  wealth  to  certain  thresholds   –  Find  large  change  in  parental  HK/investments  leads  to  only  modest   changes  in  poverty,  and  almost  no  changes  in  inequality   ⇒ Given  findings:  are  we  all  focusing  on  a  “second  order”  ques(on  (or  at   least  should  we  find  a  beWer  mo(va(on  for  our  work)?     Big  issue   •  Possible  intergenera(onal  transmission  implicitly   or  explicitly  mo(vates  a  lot  of  work  on  ECD,  health,   educa(on,  …     •  But  how  important  is  it?   –  Ideally  we  have  very  long  term  panel  data  over  mul(ple   genera(ons  to  look  at  this   –  YLS  panel  can  possibly  proxy  for  even  longer  term  data     –  Focus  on  transfer  through  HK,  as  opposed  to  physical   capital  (assets,  land,  …)  helps  understand  part  of   puzzle  ?   •  But  are  we  sure  HK  is  the  first  order  issue?  And  what  about   complementari(es?       Model,  outcomes,  bias   •  Other  factors  determining  returns  to  HK?   –  General  equilibrium  effects?  E.g  higher  educa(on  for  all   could  decrease  returns  to  1  year  of  educa(on?   •  Outcome  measures:     –  rela(ve  poverty  (boWom  20%)  :  how  does  it  look  like   with  absolute  poverty  line?   –  GINI:  possibly  some  type  of  rank  order  measure?   •  Argument  of  upward  endogeneity  bias?     –  AWenua(on  bias?  measurement  error  ~  “rough”  proxy   for  parental  HK  (schooling)   Assump(ons   •  How  important  are  some  of  the  key  assump(ons?   –  Some  seem  directly  related  to  outcomes  of  interest?   •  Human  capital  distribu(on  for  children  is  the  same  as  for  their  parents   –  If  nobody  is  changing  ranks,  not  surprising  there  is  no  large  impact  on   inequality?   •  Child’s  percen(le  distribu(on  in  HK  at  age  8  (math,  ppvt,  egra)  persists   and  determines  rank  in  adult  schooling  aWainment   –  To  the  extent  that  in  many  countries  decisions  afer  8  might  differen(ate   adult  HK  more,  this  may  lead  to  underes(ma(on?   –  Can  you  support  some  of  these  assump(ons  with  the  data?   •  E.g  rela(onship  test  scores  kids  at  8  and  schooling  at  12  and  15?   •  Sensi(vity  analysis  on  assump(ons  secular  trends  (e.g  what  happens   in  case  of  secular  declines),  on  modifica(ons  to  distribu(ons,  …  ?   INEQUALITIES IN CHILDREN’S OUTCOMES IN DEVELOPING COUNTRIES This presentation was given during the conference on Inequalities in Children’s Outcomes in Developing Countries hosted by Young Lives in Oxford, 8–9 July 2013 http://www.younglives.org.uk/what-we-do/news/children-inequalities-younglives-conference-2013/overview ... Vietnam Parents (actual) Children (expected) 1.00 1.20 0.77 0.83 0.15 0.17 13.69 12.93 1,602 1,602 Inequality and Poverty, Parents? ?? Generation Gini coefficient and poverty headcount (PH), parents'' ... and is years of schooling for mother''s and father''s schooling Inequality and Poverty, Children? ??s Generation Gini coefficient and poverty headcount (PH), next generation Ethiopia Gini PH 0.350.. .Intergenerational Transmission of Poverty and Inequality: Young Lives Jere Behrman, Benjamin Crookston, Kirk Dearden, Le Thuc Duc, Subha Mani, Whitney Schott, Aryeh Stein, and the Young