Advanced Cognitive Skill Deserts in the U.S.: Their Likely Causes and Implications

Caroline Hoxby, "Advanced Cognitive Skill Deserts in the U.S.: Their Likely Causes and Implications," Brookings Papers On Economic Activity, March 25, 2021, https://www.brookings.edu/wp-content/uploads/2021/03/BPEASP21_Hoxby_conf-draft_updated.pdf

Note the children’s scores aren’t based on

Her takeaway, “…What are we to make of these findings?They suggest that a child's cognitive skills in the third grade are not destiny. Even a child's skills in the fifth grade are not destiny. It is only as we look at the map for grade eight that we begin to see patterns that mostly resemble those for adults. By twelfth grade, the resemblance to the adult map is quite strong though of course imperfect…”

Where are advanced cognitive skills not prevalent?“…. The most obvious pattern is Appalachia, the area that follows the Appalachian mountain range through several states starting in northeastern Alabama and northwestern Georgia, running though eastern Tennessee and eastern Kentucky, encompassing all of West Virginia, continuing through southeastern Ohio, proceeding through Pennsylvania, and finally ending in New York.Although Appalachia is the most obvious desert, within it there are numerous counties, often in small clusters, in which advanced cognitive skills are prevalent. I show in the next figure that these are often counties in which a major university is located. Before examining the next figure, however,also note that advanced cognitive skills are not prevalent in the Ozarks (a mountainous plateau mainly in Missouri, Arkansas, and Oklahoma) and in some inland areas of the South—for instance, inland Louisiana, inland Mississippi, and an inland stretch that begins in north Florida, runs through northeastern Georgia, through inland South Carolina, and ends in inland North Carolina….”

Where are advanced cognitive skills prevalent? Where you would think, but also the Lutheran Belt ie the upper Midwest which was largely settled by Scandinavians ““…Figure 3 maps advanced cognitive skills by county for the U.S., based on the PIAAC numeracy data. The percentage of adults with advanced cognitive skills is divided in 10 deciles. In the bottom decile, only 0 to 17 percent of adults have advanced cognitive skills. In the top decile, 42.8 to 67.2 percent of adults have these skills... Observe the areas where advanced cognitive skills are most prevalent: most of southern New England but especially the area centered around Boston; scattered counties in New York and New Jersey but especially the area centered around New York City; coastal Washington and Oregon but especially the area centered around Seattle; coastal California but especially the areas centered around San Francisco, Sacramento, Los Angeles, and San Diego; the Salt Lake City area; and the Washington D.C. area including parts of Maryland and Virginia. None of these findings is surprising. A finding that may surprise some readers (although not those familiar with statistics on American education) is the overall dark shading of the so-called "Lutheran Belt" which occupies southern Minnesota and parts of Wisconsin, Iowa, and the Dakotas. The Lutheran Belt has long had a reputation for strong educational outcomes and explanations vary. While the area does not contain an advanced cognitive skills concentration like Boston's, say, most of its counties fall into the top four deciles of prevalence. (Remember to ignore some of its large, sparsely-populated counties.)…”

What are advanced cognitive skills for the purposes of the paper, "Advanced cognitive skills are generally defined to be those that require higher order reasoning. They require a capacity to solve problems through logic, think in the abstract, engage in critical thinking, and derive general principles from a set of facts. They are also often described as integrative or synthesizing. Advanced cognitive skills can be meaningfully differentiated from skills such as memorization, summarization, organization of facts, and other methods of acquiring concrete information...Long division, long multiplication, addition of fractions: these are all exercises that require a person to follow an algorithm that can even be fairly complex. They differ from algebra because algebra requires a person to translate a problem into equations, which are then solved to find the solution. Algebra thus requires abstract thinking. Proof-based mathematics is even more distinct because it requires logical reasoning that can extend over many steps. In short, algebra and the mathematics that typically follows it in the curriculum can be characterized as advanced cognitive skills….”

Core of the paper, “…attempts to test whether there is much variation among U.S. counties in the prevalence of adults with advanced cognitive skills. Being interested in areas where advanced cognitive skills are especially non-prevalent, I label such areas "deserts" as a shorthand. This exercise is not the same as testing whether there is much variation in the share of adults employed in skilled occupations. Such occupational variation could be caused by any number of supply and demand factors that vary among areas: regulations, tax policies, natural resources, place-based policies that favor certain firms. Rather, I am interested in variation that is a layer deeper: cognition. Understanding how the variation in the supply of advanced cognitive skills arises might reveal ways to improve such skills. It might also have consequences for economic policies such as trade and consequences for political economy. Second, this paper attempts to provide enough age-related evidence on the three explanations to help us parse them. Specifically, are advanced cognitive skill deserts associated with early-age factors that can be difficult to affect with policy? Are they associated with adolescent-age factors more likely to be influenced by policy? Finally, are they associated with individuals migrating towards areas where other people have advanced cognitive skills similar to their own?...,my measures of advanced cognitive skills depend only on the percentage of people with the skills, not the physical density of people with the skills. That is, I do not show advanced cognitive skills per square mile. This is a deliberate choice because I wish to avoid maps that mechanically predict that rural, low density areas are skill deserts. However, the consequence is that the viewer must consciously avoid focusing on sparsely populated counties with a large land mass: the skill measures might be based on relatively few people. The implications of a county’s having only 800 people, all of whom have advanced cognitive skills are very different from the implications of having 50,000 or 300,000 people with such skills. Generally speaking, the more sparsely populated a county is, the larger its land mass. It takes self-discipline on the viewer's part to ignore vast counties in Nevada, Arizona, Utah, New Mexico, south-eastern California, Montana, and the Dakotas…Third, the only nationally representative measure of cognitive skills among adults in the U.S. comes from the Program for the International Assessment of Adult Competencies (PIAAC). It measures 16 to 65 year olds’ numeracy (mathematics) and literacy. PIAAC has six levels of skill for each test, and its tests can identify very advanced skills. For instance, its top level in both numeracy and literacy contains only 1 percent of adults and its second-to-top numeracy level contains only 9 percent of adults. See Table 1. Also, PIAAC data can in theory be broken down by age which would be interesting even though it would not be longitudinal data.....For mapping, all of the adult ages are consolidated, only county level statistics are possible, and PIAAC aggregates the top three levels of skill for each test.16 Unfortunately, this makes only the numeracy test useful for mapping advanced cognitive skills. Its top three levels contain 37 percent of the adult population—already a coarser definition of advanced than I would prefer....PIAAC describes adults with the top three levels of numeracy skills as "proficient at working with math information and ideas. They have a range of numeracy skills from the ability to recognize math relationships and apply proportions to the ability to understand abstract representations of math concepts and engage in complex reasoning about quantities and data...”

Here’s her map, specifically the percentage of adults with advanced cognitive skills is divided in 10 deciles. In the bottom decile, only 0 to 17 percent of adults have advanced cognitive skills. In the top decile, 42.8 to 67.2 percent of adults have these skills.

Great data from Research being presented at BPEA today. Caroline Hoxby attempts to map what I’ll call the cognitive topography of the United States using the numeracy PIAAC data we’ve looked at. Her measure is the top 37% of the adult population which is the most limited run possible as PIAAC aggregates the top three levels of skill. Note she also does the same exercise for children at different age levels which ~ mirrors the adult population but uses SAT and State level test scores and does suggests some sorting, as does the clustering around college towns in the “Advanced Cognitive Skill Deserts”

Of course the underlying genetic topography isn't mentioned as a possible explanation.