Income And Educational Outcomes

The role of poverty in shaping educational outcomes is one of the most common debates going on today. It can also be one of the most shallow.

The debate tends to focus on income. For example (and I’m generalizing a bit here), one “side” argues that income and test scores are strongly correlated; the other “side” points to the fact that many low-income students do very well and cautions against making excuses for schools’ failure to help poor kids.

Both arguments have merit, but it bears quickly mentioning that the focus on the relationship between income and achievement is a rather crude conceptualization of the importance of family background (and non-schooling factors in general) for education outcomes. Income is probably among the best widely available proxies for these factors, insofar as it is correlated with many of the conditions that can hinder learning, especially during a child’s earliest years. This includes (but is not at all limited to): peer effects; parental education; access to print and background knowledge; parental involvement; family stressors; access to healthcare; and, of course, the quality of neighborhood schools and their teachers.

And that is why, when researchers try to examine school performance – while holding constant the effect of factors outside of schools’ control – income or some kind of income-based proxy (usually free/reduced price lunch) can be a useful variable. It is, however, quite limited.

This was apparent in a recent Brookings Institution analysis of the relationship between parents’ income/education and children’s readiness for school (as measured by academic skills, health and behavior).

The conclusion, following plenty of prior work, was that income during early childhood (and, to a lesser extent, maternal education) is a predictor of reading and math skills among five year olds, as well as behavioral traits such as the ability to concentrate and work independently (the latter characteristics were based on surveys of the children’s teachers).

But the estimated association between income and math/reading skills, though statistically discernible (i.e., not just random error), was actually rather more modest than many people might think. For instance, for a low-income family, a $1,000 increase in household income during early childhood was associated with a 0.015 standard deviation increase in math and reading scores once children reached age five (and the estimated association was even smaller for behavioral traits, as well as among higher-income families). To put this effect size in context, the average achievement gap between black and white students upon entry into schooling is generally characterized as roughly one standard deviation.*

These results – and those of prior research – suggest that there is much more to the role of student background in education than just income.

And the Brookings estimates are actually more precise than those typically used by education researchers, since the ubiquitous free/reduced-price lunch variable tells one nothing about the degree of poverty (or non-poverty), which may bias results (though see here). The same goes for other common variables, such as those measuring whether students are non-native English speakers or in special education plans.

So, while much of the variation in performance based on non-school factors will always remain unmeasured, there is a great deal of room for improvement in the quality of data. For instance, the work of Anthony Bryk and colleagues in Chicago shows how much can be gained by moving beyond the traditional variables used in most education research.

Bryk and his team focus not only on typically-ignored schooling conditions, such as curricular alignment and safety, but also on extremely important student and contextual characteristics, such as abuse/neglect, neighborhood crime, and social capital (students’ relationships with people outside of their neighborhoods). They find, put simply, that these variables add a good amount of power in explaining schools’ absolute performance and growth rates.

It’s certainly true that many people involved in education feel that we’ve reached a situation of “data overload." And that may very well be the case on the ground in many places, especially if states and districts don’t do a good job of training and presenting data in a manner that’s useful and accessible. From a research perspective, however, the standard set of educational variables measuring student characteristics are less than impressive.

Needless to say, gathering additional data on a larger scale would require significant resources and comparability between districts and states would be an issue, but, given the importance of separating student- from school-based variation in assessing and improving performance, such investments, even on a limited scale, would likely yield considerable benefits. After all, the process of “data-driven decision making” is only as good as the available data.

- Matt Di Carlo

*****

* It’s worth noting, as the authors point out, that this effect size is somewhat smaller than in some prior analyses (though still relatively small compared with reference points such as the traditionally-defined achievement gap). They speculate that methodological issues – most notably, their choice of covariates and the fact that income is imputed for much of their sample and does not include benefits such as food stamps  – may explain this discrepancy.