Silke Schmidt

Black-White Achievement Gaps in Our States: An Analysis of Trends over Time

By Silke Schmidt

What factors contribute to persistent differences in academic achievement between black and white students? Can we explain some of the observed variation in achievement gaps from state to state? For example, do states with persistently high gaps share any demographic characteristics? And are there lessons we can learn from demographically similar states that have been able to reduce their gaps?

Data from the National Assessment of Educational Progress (NAEP), available from 1990 to 2015, combined with demographic information from the 1990 U.S. census and 2010-2014 American Community Survey can be used to address these types of questions. After excluding 13 states with small black populations and/or incomplete data, and the District of Columbia, trends in 8th grade mathematics test scores were evaluated in 37 states. The achievement gap for a given year was defined by subtracting average scores of black students from average scores of white students. To evaluate the change in gap, the 2015 gap was subtracted from the earliest available gap (1990 for 29 states and between 1992 and 2000 for the other 8 states). Therefore, negative numbers represent a widening gap and positive numbers a narrowing gap.

Descriptive Summary

The state-specific achievement gaps ranged from a minimum of 7.3 (West Virginia, 2009) to a maximum of 50.9 (Nebraska, 2007). The U.S. average decreased only slightly from 32.9 in 1990 to 31.6 in 2015. Even in states that had high average scores for both white (~300) and black (~270) students in 2015, the gap between them was on the order of 30 (Massachusetts, New Jersey, Texas).

The average standard error (SE) of the states’ mean scores for the larger of the two samples (white students) was 1.1 in 1990 and 1.3 in 2015. By rounding up twice the average SE in 2015, states whose gaps changed by an absolute value of 3 or less (from the first available year to 2015) were classified as “gap similar” (14 states). All other states were classified as either “gap narrowed” (17 states) or “gap widened” (6 states). A map of the three groups of states and the temporal change in black-white achievement gaps is shown in Figure 1.

Question 1: Do the 6 states with increased achievement gaps share basic demographic characteristics and are those also found in states that narrowed gaps?

The 6 “gap widened” states include three in the Midwest (Iowa, Indiana, Wisconsin) and one each in the Northeast (Pennsylvania), Central South (Kentucky) and West (Nevada).

When population size is plotted against the percentage of blacks in a state’s population (averaged across 1990 and 2014; Figure 2), with the color indicating the gap change, all 6 “gap widened” states (shown in red) are clustered in the bottom left corner. These states have fewer than 10% of African Americans and a population size at or below 6 million, with only PA falling outside that range (10.1% and 12.3 million). The same bottom left corner of the graph includes four states with similar demographic characteristics, but with decreasing gaps from 1990 to 2015 (shown in green): Rhode Island, Nebraska, West Virginia and Oklahoma. The only other green state with fewer than 10% African Americans is California (6.7%), an extreme outlier for its population size (33.9 million).

The percentage of blacks also correlates well with the numerical change in the gap: All states with higher percentages either maintained or narrowed their gaps over time. This includes most of the southern states where the black middle class is larger than in the Midwest.

Question 2: What can we learn from demographically similar states in which the gap has decreased over time and is similar to or below the national average in 2015?

Of the 4 demographically similar states identified above, Nebraska’s actual gap was 40.9 in 2015 (Figure 1), almost 10 points higher than the national average of 31.6, while OK, RI and WV had gaps similar to or below the national average. Thus, the following analysis excludes Nebraska.

While a student’s performance on standardized tests is influenced by a complex interplay of multiple factors, some researchers argue that early childhood education is critical in preventing achievement gaps at older ages. The contrast between the two demographically similar states of WV and KY may provide some support for this hypothesis. The 2015 report “Education Week’s Quality Counts” calculated quality indices for each state’s early childhood education programs, including its preschool poverty gap: the difference in preschool attendance rates of children who do or do not live in poverty. This gap was much lower in WV (4.6 percentage points in 2014) than in KY (15.6 percentage points), which may partially explain why Kentucky’s black students showed less improvement from 1990 to 2015 than West Virginia’s (Figure 3). The other reason for Kentucky’s widening gap was that its white students showed greater improvement than West Virginia’s during the time period of interest, though both continued to score well below the national average.

Oklahoma’s white students also scored below the national average in 1990 and 2015. This is consistent with all three states (OK, KY and WV) having black and white poverty rates above the respective national average. Oklahoma’s gap narrowed over time, however, because its black students surpassed those in KY and WV to reach the black national average in 2015. The state’s preschool poverty gap was 7.4 percentage points, less than half the national average of 15.8. Interestingly, OK was highlighted by President Obama in his 2013 State of the Union address as one of the first states to have a universal state-funded preschool program with consistently high attendance rates.

It has also been demonstrated that exposure to the heavy metal lead at young ages has long-lasting effects on cognitive and behavioral outcomes. Rhode Island is a particularly compelling case in point since a recent paper suggested that state policies for reducing environmental lead exposure had a measurable impact on test scores in children born between 1997 and 2004 (Figure 4).

Lead is a powerful neurotoxin, especially for the developing brains of children under 6 years of age. Poor minority families are much more likely than wealthy white families to live in old homes with deteriorating lead paint and/or lead pipes that deliver drinking water. Starting in 1997, the state of Rhode Island strongly encouraged landlords to reduce lead hazards in older homes they rented out, to the point that they “could (and were) sued in civil court when children living in their homes were found to have elevated lead levels if the home did not have a lead-safe certificate.” By 2010, the number of these certificates had increased from 333 to more than 41,000.

By analyzing the relationship between blood lead levels in preschool and standardized test scores later in life in ~57,000 Rhode Island children, the study concluded that “the decline in racial disparities in lead explains between 37 and 76% of the decline in racial disparities in test scores witnessed over the past decade in RI.” This remarkably strong relationship suggests that public funds would be well spent on reducing lead exposure hazards in large urban areas with a high proportion of poor minority families. A recent lead risk map produced by Vox that is based on the average age of homes and poverty rates for each census tract could prove very useful in identifying priority areas for such state policies.



Question 3: Are racial achievement gaps correlated with demographic indices of racial disparities?

In addition to analyzing trends over time, it is important to relate the size of the observed gap to demographic differences between states. Two particular measures were considered here.

The first is a statewide measure of spatial segregation that was presented in the 2015 “Poverty and Inequality Report”. The index of dissimilarity (D) is calculated from census tract data and “indicates the percentage of [blacks] that would have to move to other neighborhoods (within their state) in order to achieve parity between [blacks] and whites in their percentage distributions across all neighborhoods.” The report notes a general trend for states with small black populations to be more segregated. This is consistent with the data analyzed here, although the correlation is moderate. The left panel of Figure 5 illustrates a non-significant positive correlation between D and the size of the achievement gap (in 2015). Interestingly, the slope is steeper and the regression R2 increases from 0.05 (p=0.20) to 0.10 (p=0.08) when western states are excluded.

The second measure quantifies socioeconomic disparities between black and white families, using the official measure of poverty (OMP). It takes into account family size, family income and inflation-adjusted cost of living estimates (but not geographic residence) to derive a “ratio of income to poverty level.” If the ratio is 1.0 or smaller, all adults and children that make up a family are classified as “doing poorly;” if the ratio is 2.0 or smaller, the family is considered “poor or struggling.” The difference in the percentage of black and white families living in poverty (using the 1.0 cutoff) is plotted against the size of the achievement gap (in 2015) in the right panel of Figure 5. While the correlation is stronger (R2=0.21, p=0.004) than for the segregation index D, the differences in poverty alone do not explain all of the observed educational gaps. This suggests that a more comprehensive socioeconomic disparity measure may be useful. The 2015 “Poverty and Inequality Report” (“Education” chapter authored by Sean Reardon) proposes such a measure as the “weighted average of racial differences in income, educational attainment, poverty rates, and unemployment rates, each among parents of school-age children in the state,” with a value of 0 indicating equal rates in black and white parents.  Using this measure, Reardon reported a high correlation (0.68; p. 48 of the report) between each state’s black-white achievement gaps and socioeconomic disparities.

Poverty rates and levels of segregation are frequently evaluated at the level of cities, instead of states. In the present dataset, an analysis of black-white differences in poverty for the two largest cities in each state confirmed that cities indeed have even greater racial disparities (data not shown). It is well known that blacks tended to settle in urban areas during the “Great Migration,” the relocation of millions of African Americans from the South to the North and Midwest between 1916 and 1970. Larger cities typically provided more job opportunities, especially in the manufacturing industry; therefore, the percentage of blacks in a large city’s population is usually much higher than the statewide average. The decline of the manufacturing industry during the 1970s created areas of concentrated poverty in many large cities. When wealthier white families reacted to these problem areas by moving to city suburbs (“white flight”), this had the potential to induce a strong degree of spatial segregation between blacks and whites and exacerbate socioeconomic and educational gaps. Philadelphia and Milwaukee have been identified as some of the most segregated cities in the nation; 48% and 66% of each state’s black population lived in these two cities in 2014, and much of the achievement gap observed in the two states is likely due to the cities’ pronounced racial disparities. However, since NAEP scores aren’t available for cities, a more detailed analysis is beyond the scope of this project.

Question 4: Taking into account the geographic and historical context, which regions are making the most (or least) progress in narrowing achievement gaps?

Analyzing only the numerical change in gap over time misses important information about the nature of the gap: Was it large to begin with and did it widen because black students improved more slowly than white students? Or was the gap small, relative to the nation’s average, at the first available time point and did it “catch up” to the national pattern as it widened over time? Are there states in which the improvement in black student scores was particularly notable, suggesting that other states may adopt some of their educational policies?

To evaluate these questions, the 37 states were grouped by geography into 5 regions. Since the 10% cutoff for the percentage of blacks was particularly relevant for the 10 Midwestern states, this group was further subdivided, resulting in 6 groups of 4 to 8 states each. Boxplots of black and white scores and the gap between them (Figure 6) as well as slope graphs (Figure 7) were used to summarize the nature of the gap and its change over time for each region.

The following patterns emerged from these visualizations:

  • While almost all of the South Central states narrowed their gaps over time, these gaps were generally low because white students performed substantially below the national average, especially in 2015. A notable exception from this trend is Texas. Oklahoma’s black students had the second-highest growth after Texas, which is noteworthy since the state has the lowest proportion of blacks within the region. The gap widened in Kentucky because its black students improved more slowly than those in the rest of the region.
  • As a group, the Midwestern states with fewer than 10% blacks experienced the worst changes in achievement gaps. Here, the median scores of black and white students were above the national average in 1990, but 25 years later, white students performed closer to the national average and black students scored well below the national average, especially in Wisconsin. Iowa was the only state in the country where black students performed slightly worse in 2015 than 1990. Even though black students in MN improved the most within the region (Fig. 7) and the gap narrowed in NE, the median gap for the region was substantially larger than the national average in 2015 (Fig. 6).
  • Most of the states in the Northeast and Midwest with >=10% blacks narrowed their gaps, but the main reason is that the scores of white students decreased over time, moving either closer to the national average (Northeast) or dropping below it (Midwest). NY narrowed its gap the most (Fig. 1) because white students improved at a much slower rate than black students. Within the Northeast, black students improved the least in Pennsylvania.
  • Western states mostly maintained or narrowed their gaps (except for Nevada), although black students performed above average in 1990 and very similar to the average in 2015 (Fig. 6).
  • The South Atlantic states represent the best-case scenario: They narrowed their gaps because black students improved over time while white students remained close to the national average.


The data analyzed here support a strong association between black-white achievement gaps and the number and proportion of African Americans in a state, especially in the Midwest. The influence of large cities with concentrated areas of black poverty is perhaps best illustrated with the example of Milwaukee, one of the most segregated and politically polarized cities in the nation. It has been noted that blacks began to settle in Milwaukee much later than in Chicago, Detroit and Cleveland and suffered from the collapse of Milwaukee’s industrial base and the massive elimination of manufacturing jobs before a black middle class had developed.

The observation of worsening achievement gaps and higher levels of spatial segregation in Midwestern states with small black populations is worrisome, although a few states show promising trends. In Minnesota, which prides itself on having a high-quality public education system, and to a lesser extent in Kansas and Nebraska, black scores have improved over time. However, Minnesota’s gap is still much higher than the national average (~40) and it has yet to be seen whether a trend in the right direction can be maintained. Real estate businesses and affordable housing programs may have an important role to play in reducing the kind of spatial segregation at the neighborhood level that appears to manifest itself in racial achievement gaps.

The South Atlantic states, which generally have a larger black middle class than Midwestern states, illustrate a best-case scenario of black scores improving over time and white scores remaining at or above the national average. Despite this promising trend during the last 25 years, however, the median gap in the region (28.4, below the national average of 31.6) remains stubbornly large.

In terms of state policies, early childhood education facilities, including the federally funded Head Start program for low-income families, should continue to be supported. In addition, a much greater effort should be directed at reducing the lead exposure of young children living in old homes. The example of Rhode Island is powerful, both in terms of its success with lead mitigation policies and the resulting improvement in the academic performance of its minority students.

Tableau Story: