- Slides: 15
Who Starts Ahead and Who Moves Ahead? Achievement, Social Adjustment and their Production in Children Randall J. Olsen Department of Economics
Why? • In 2002, NCES estimated the U. S. spent$500+ billion on primary & secondary education • We have long-running controversies on public vs private education, charter schools, testing and standards • Suppose, as some have thought, parental impacts on the production of child quality are very important, then we will incorrectly attribute to schools impacts that really belong to the household, chiefly the mother
Approach • Data – Children of the NLSY 79 plus maternal variables & limited school data • Outcomes are measured in terms of age-corrected percentile ranking • We use early childhood measures of achievement/ability before entering school and math and reading achievement for primary – middle school • We look at initial achievement (percentile) on entering school and change in percentile ranking from first measure through the last (always before 15)
From birth through entering school, the child development production process for child i evolves with post-natal inputs, starting at time 1, according to yi 1 = yi 0. f(Mi(1), Hi(1), Ci(1)) yit+1 = yit. yi 0. f(Mi(t+1), Hi(t+1), Ci(t+1)); t=1, …, k-1 so yit = yti 0. P f(Mi(1), Hi(1), Ci(1)). . . f(Mi(t), Hi(t), Ci(t)) Mi(t) are (possibly) time-varying maternal characteristics for child i, Hi(t) are likewise for household demographic characteristics, and Ci(t) are purchased inputs
Starting at the beginning of the primary grades, we use a different, more broad, measure cognition - the PIAT battery, and reflecting this change we indicate PIAT cognition at k+1 as y*ik+1. Following the above, the evolution of our measured PIAT achievement will be: y*ik+1 = yik. f*(Mik+1, Hik+1, Cik+1) y*ik+2 = y*ik+1. yik. P f*(Mik+2, Hik+2, Cik+2). ln(y*ik+r) – ln(y*ik+1) = (r-1). ln(yik) + ln[f*(Mij, Hij, Cij)]. If we let the general form of f*(Mij, Hij, Cij) be f*(Mij, Hij, Cij) = ebz(i)+ax(i, j) We estimate [ln(y*k+t) – ln(y*k+1)]/(t-1) = ln(yk) + bz +a. Sj=k+1 to j=k+t xj/(t-1)
The average yearly change in the percentile ranking becomes a function of time invariant characteristics of the child & mother and, for time-varying variables, the average value over time period for which we observe the percentile change in achievement. Finally, the specification allows for early childhood achievement (or ability) – that is, early childhood achievement not only puts the child ahead of their peers on entering school, but allows them to continue to accumulate skills at a faster pace. This is not a restriction but a testable hypothesis. We look at the accumulation of skills as measured by the three Peabody Individual Achievement Tests (PIAT) for academic skills and the Behavior Problems Index for social skills.
Tests – Early Childhood • Wechsler Intelligence Scales for Children – Revised Digit Span component (forward and reverse) – pure memory test; race & ethnicity explain little of its variance and is thought to have less cultural bias than most tests; first administered at 7 • Peabody Picture Vocabulary Test (PPVT); first administered at 4, but for some children as early as 3; child picks picture that matches spoken word. • These are our measures of early childhood achievement that augment learning during the primary and middle school grades.
Achievement Tests (Graduated to Match Age Spectrum) • PIAT reading recognition – a test of a child’s ability to read a series of words aloud. • PIAT reading comprehension – Child silently reads a sentence and then points to the picture that best conveys the meaning of the sentence • PIAT math – a multiple choice test; child picks one of four pictures or results that matches the mathematical concept or operation • Behavior Problems Index – Maternal report on a series of problem behaviors a child may have; taken as a measure of social development; problems appropriate to the child’s age
NLSY 79 - Sample • Nationally representative with oversamples of Blacks & Hispanics + originally Low Income non -Black, non-Hispanics and military members • N=12, 686 in 1979; dropped military oversample & “poor whites” – about 7200 completes + similar number in Children of the NLSY 79 • Data yearly on mothers 1979 -1994, every other year thereafter • Children assessed 1986 – present, tested in evennumbered years through 14 & then given an interview like the regular NLS every other year
Tables • Table 1 - early childhood ability on Digit Span memory test • Table 2 - Pre-school verbal ability (achievement) on PPVT • Table 3 – PIAT Reading Comprehension; initial percentile on entering school and yearly change in percentile ranking; use early childhood ability (achievement) as a continuing factor • Table 4 – PIAT Reading Recognition • Table 5 – PIAT Math • Table 6 – Same for Behavior Problems Index
Findings – Early Childhood • PPVT shows disadvantage for minorities, but Digit Span does not • Maternal Word Knowledge component of the AFQT (also in percentile points) has strong impact on PPVT and Digit Span; this maternal academic skill spills over more strongly to the academic PPVT • Maternal Arithmetic Reasoning has strong impact only on PPVT; popular psychology characterizes both skills as “left brain” activity • Maternal education has strong impact; mother’s age a birth also has a strong impact on the academic PPVT • Family income has a strong effect; maternal work time a strong negative effect • The cumulative effect of a bright, well-educated, stay-athome mother in a stable, affluent household in the preschool years is considerable
Achievement at School Entrance & How it Evolves • Better PPVT & Memory confer an advantage at entrance (if both are higher by one percentile, child enters. 4 points higher); that advantage will be about 25% larger after ten years in school. • Maternal education, net of early childhood achievement, lends a more modest advantage. When it is present, it tends to dissipate over time. Mother’s age at birth, however, has a continuing strong effect – older mothers are more savvy mothers? • Maternal achievement (AFQT components) has little impact net of child’s PPVT & Digit Span scores • Private school as little impact, along with school characteristics (not shown) • Father’s absence is not much of a factor except in Math, where children benefit by his presence – reflects the gain we see for males in primary and middle school? (next slide)
Gender & Ethnicity • Girls start out ahead of boys and lose that advantage over time • In verbal areas, controlling for other factors, Blacks and Hispanics start out ahead and then the Blacks lose ground; the Hispanics show little trend • In Math, Blacks and Hispanics start out behind; Blacks show insignificant loss of ground over time, Hispanice insignificant gain
Behavior Problems • Blacks and Hispanics start out doing better, adjusted, but trend is unremarkable • Older mothers and better educated mothers start out with an advantage in shaping behavior, but older mothers lose the edge as time passes; perhaps having the energy to “track” children is key – studies of delinquency point to the importance of tracking where the child is and who their friends are • Not having the father present has a very large effect
Takeaway • Upon entering school, children from advantaged backgrounds have a large headstart and certain characteristics (but not all) of advantaged households keep those children moving ahead • The weakness of schooling variables suggests counting on schools to change the initial ranking is not a winning bet • The mother is the 800 pound gorilla of child development if we don’t pay careful attention to her characteristics we will arrive at misleading conclusions