Inputs in the Production of Early Childhood Human Capital: Evidence from Head Start

School Reform, October 2015

A study con­duct­ed by Christopher Walters finds that Head Start cen­ters offer­ing full-day ser­vice boost cog­ni­tive skills more than oth­er cen­ters, while Head Start cen­ters offer­ing fre­quent home vis­it­ing are espe­cial­ly effec­tive at rais­ing non-cog­ni­tive skills. Other key inputs, includ­ing the High/Scope cur­ricu­lum, teacher edu­ca­tion and cer­ti­fi­ca­tion, and class size, are not asso­ci­at­ed with increased effec­tive­ness in Head Start. Together, observed inputs explain a small share of the vari­a­tion in Head Start effec­tive­ness.

Studies of small-scale “mod­el” ear­ly-child­hood pro­grams show that high-qual­i­ty preschool can have trans­for­ma­tive effects on human cap­i­tal and eco­nom­ic out­comes. Evidence on the Head Start pro­gram is more mixed. Inputs and prac­tices vary wide­ly across Head Start cen­ters, how­ev­er, and lit­tle is known about vari­a­tion in effec­tive­ness with­in Head Start. This paper uses data from a mul­ti-site ran­dom­ized eval­u­a­tion to quan­ti­fy and explain vari­a­tion in effec­tive­ness across Head Start child­care cen­ters. Walters answer two ques­tions: (1) How much do short-run effects vary across Head Start cen­ters? and (2) To what extent do inputs and prac­tices explain this vari­a­tion? To answer the first ques­tion, the author devel­ops a ran­dom coef­fi­cients sam­ple selec­tion mod­el that quan­ti­fies het­ero­gene­ity in Head Start effects, account­ing for non-com­pli­ance with exper­i­men­tal assign­ments. Empirical Bayes esti­mates of the mod­el show that the cross-cen­ter stan­dard devi­a­tion of cog­ni­tive effects is 0.18 test score stan­dard devi­a­tions, which is larg­er than typ­i­cal esti­mates of vari­a­tion in teacher or school effec­tive­ness. Next, the author assess the role of inputs in gen­er­at­ing this vari­a­tion, focus­ing on inputs com­mon­ly cit­ed as cen­tral to the suc­cess of mod­el pro­grams.