No evidence for cumulating socioeconomic advantage. Ability explains increasing SES effects with age on children's domain test scores

Steady as she goes, "Studies that investigate the effects of socioeconomic background (SES) on student achievement tend to find stronger SES effects with age, although there is much inconsistency between studies. There is also a large academic literature on cumulative advantage arguing that SES inequalities increase as children age, a type of Matthew Effect. This study analysing data from the children of NLSY79 mothers (N ≈ 9000, Obs ≈ 27,000) investigates the relationship of SES by children's age for two cognitive domains (Peabody Picture Vocabulary test and digit span memory) and three achievement domains (reading comprehension, reading recognition and math). There are small increases in the SES-test score correlations for several domains, but there are more substantial increases in the test score correlations with mother's ability and prior ability. Regression analyses found linear increases in SES effects for all domains except digit memory. However, when considering mother's ability, the substantially reduced SES effects did not increase with children's age. Much of the effects of SES on children's domain scores are accounted for by mother's ability. The effects of prior ability also increase with age and SES effects are small. Therefore, there is no evidence for cumulative socioeconomic advantage for these domains. Generally, increases in SES effects on children's cognitive development and student achievement are likely to be spurious because of the importance of parents' abilities and their transmission from parents to children..."
Takeaway, "...The main conclusion from this study is that the increase in the SES effects on cognitive development and student achievement as children grow older observed in these data is spurious, confounded by parental and child's abilities. Increases in SES effects with age for some domains disappear with the inclusion of mother's ability and mother's ability age interaction terms, or with analogous measures of prior ability. Net of mother's and child's abilities, SES effects on children's test scores are small, so changes in its effects with age are unimportant..."
Evidence

“…Five childhood outcomes are investigated. Two are cognitive: the Peabody Picture Vocabulary Test (PPVT) and the Wechsler digit memory span scores; and three are achievement measures: the Peabody Individual Achievement Tests (PIAT) for reading comprehension, reading recognition, and math. At ages 3 and 4, children were tested only by theppVT, at age 5 only by theppVT and PIAT math, and at age 6 in all domains but digit memory. From ages 7 to 12, children were tested in all domains, and from age 13 to 15 only in the two PIAT reading domains and PIAT math…Mother's cognitive ability is measured by AFQT score from parts of the Armed Services Vocational Aptitude Battery; a special survey administered in 1980 to NLSY79 respondents. The raw AFQT score in the NLSY79 data is the sum of scores in the arithmetic reasoning, word knowledge and paragraph comprehension subtests and one-half of the score in the numeric operations subtest…At ages 3 and 4 there were onlyppVT items, so the multiple-domain factor was the same as the single-domain factor. Between ages 7 and 12 when students were tested in all 5 domains, the variation accounted for by g increased from 25% to 37% consistent with the increasing heritability of cognitive ability with age. If theppVT and digit memory items are excluded, the percentage of variance explained by the principal factor increases by about 10 percentage points. Pooling all the data, mother's ability correlates at 0.26 with digit memory and between 0.4 and 0.5 with the other four domains. Cognitive ability, g, correlates most strongly withppVT scores (r = 0.78), most weakly with digit memory (r = 0.47) and between 0.66 and 0.73 with the other domains. Mother's ability and g correlate at 0.46 consistent with studies cited earlier that have reported intergenerational correlations for ability.Family SES correlates at 0.62 with mother's ability which is higher than Torres's estimate (Torres, 2013, p. 166). SES and g correlate at 0.41. The average correlation between ability and prior ability is 0.6 (Table A4). The correlations of ability and prior ability were generally higher at older ages (Table A5) again consistent with the increasing heritability of cognitive ability….”

Results

“…children's age for the five domains. TheppVT (blue diamond, solid line) and math (purple triangle, long-dash short-dash line) show the highest correlations and digit memory (red squares, dashed line), the lowest correlations. For reading comprehension (green star, dotted line) there is dramatic increase in the SES-achievement correlation with age. There are shallower increases in the correlations with age for theppVT and math. For reading recognition (brown circle, short-dash dot short-dash line) there is an anomalous correlation (0.44) at age 5 but for older ages the correlations are stable at around 0.35. The correlations between SES and digit memory decline from ages 7 to 9 and then increase. So even without considering mother's or child's ability, the evidence for cumulating socioeconomic inequalities is not unequivocable. Fig. 2 presents the correlations between mother's ability (AFQT) score and test scores. There is a substantial increase in the correlation between mother's ability and reading comprehension with children's age. The AFQT correlations with theppVT, math and reading recognition show clear but smaller increases. For digit memory, the correlations increase between ages 8 and 12. The LOESS curves show largely linear or curvilinear relationships. The correlations between prior ability and children's test scores show clear increases with age for all five domains (Fig. 3). For theppVT, there is an anomalous very large correlation at age 6 (≈0.7) and very low rapidly until age 10 and then stabilizes. The correlations between prior ability and digit memory are weaker but also show an increase with age…”

Gary Marks and Michael O'Connell, "No evidence for cumulating socioeconomic advantage. Ability explains increasing SES effects with age on children's domain test scores,"Intelligence, September/October 2021, https://www.sciencedirect.com/science/article/abs/pii/S0160289621000660