The Geography of Rural Educational Opportunity

GEOGRAPHIC VARIATION IN EDUCATIONAL OPPORTUNITY

Research conducted on geographic variation in educational outcomes has often focused on achievement disparities within and across urban and suburban spaces, generally finding that achievement gaps are larger in cities or metropolitan areas with higher levels of racial-ethnic and socioeconomic inequality and segregation (Owens 2018; Owens and Candipan 2019; Reardon, Kalogrides, and Shores 2019). In addition, the literature on neighborhood effects on educational outcomes in densely populated areas is robust (Chetty, Hendren, and Katz 2016; Sharkey and Faber 2014). These analyses provide important insights on the distribution of educational opportunity as well as racial-ethnic and socioeconomic gaps in opportunity across the country’s largest metropolitan areas, but do not extend their reach to the rural areas.

Descriptive research on geographic variation in educational opportunity that includes rural areas is sparse but growing. A recent paper used national 2010–2011 data to examine racial-ethnic educational disparities and segregation across rural communities and found that rural elementary schools, on average, have poverty rates that are only modestly lower, and test scores that are only modestly higher, than urban schools in nearby metropolitan areas. These findings complicate the typical narrative comparing predominantly Black and Hispanic students in city schools with predominantly White students in wealthy suburban schools and point to the need to better understand the experience of students in poorer rural schools (Logan and Burdick-Will 2017). Because the paper includes only within-state comparisons, however, it does not provide a comprehensive national portrait of educational opportunities in rural America. Another recent working paper that focused on rural student learning rates found evidence suggesting that rural White students appear to learn more in elementary and middle school than White students elsewhere. Rural Black, Hispanic, and Native American students’ learning rates in the same grades, however, appear slower rates than their groups’ respective national averages (Johnson, Kuhfeld, and Soland 2021). It relies on data from a large but nonrepresentative sample of schools that includes 25 percent of all public schools in the United States in the 2016–2017 school year. Although these papers have made useful contributions to the literature, neither provides a thorough description of educational patterns in rural communities.

We have identified one paper that uses SEDA to explore rural educational outcomes. Douglas Gagnon and Marybeth Mattingly (2018) find evidence that racial-ethnic test score gaps are smaller in rural districts than they are in urban districts—a difference that persists even after accounting for socioeconomic factors, levels of segregation, and other district characteristics. Although the authors were unable to determine whether this smaller gap is attributable to higher performing White students in high-gap districts or to higher performing Black and Hispanic students in low-gap districts, their finding suggests the importance of local context in mediating test score gaps. The authors additionally find that, unlike city and suburban districts, the level of relative affluence in a rural district has little bearing on local test score gaps. These recent findings are a helpful start, but more evidence is required for a comprehensive picture of educational opportunity across America’s wide variety of rural communities.

Our analysis aims to fill this gap in the literature by providing a detailed descriptive analysis of third-grade achievement and third-through eighth-grade learning across rural America. We conduct a set of analyses to address the following questions:

How do patterns of average academic performance and learning rates differ for rural versus nonrural students? Do these patterns vary for different student groups?

Among rural districts, do student achievement and learning rates vary across region of the country, relative geographic isolation, and type of local economy?

If present, are regional differences in student achievement and learning rates simply an artifact of demographic factors such as community socioeconomic status and racial-ethnic composition?

EDUCATIONAL OUTCOMES IN RURAL COMMUNITIES

By definition, rural communities have smaller and more widely dispersed populations than their metropolitan counterparts. Beyond this, they differ in ways that might be linked to educational opportunity: through community socioeconomic and racial-ethnic composition, through relative geographic isolation, and through local economic conditions. These factors might lead to differences in educational opportunities, which we discuss in the following section.

DATA

The primary data source for this article is the Stanford Education Data Archive (version 4.0), which reports nationally standardized measures of student achievement for nearly all schools and districts in the United States, including measures disaggregated by student race-ethnicity, gender, and socioeconomic subgroup. SEDA is based on assessment data from the EDFacts database at the Department of Education, which annually collects third- through eighth-grade standardized assessment results in mathematics and English Language Arts (ELA) from all fifty states and Washington, D.C. Following collection from EDFacts, achievement score estimates are linked to a common scale using the National Assessment of Educational Progress math and reading scales, enabling comparisons of achievement scores across the nation (for details on methods used to construct measures of achievement comparable across the country, see Reardon, Kalogrides, and Ho 2019). Estimates from SEDA v4.0 span academic years 2008–2009 through 2017–2018, including scores from approximately 430 million assessments. Further details on sample construction are available from the SEDA v4.0 technical documentation (Fahle et al. 2021).

The SEDA data include estimates of the average test scores, average learning rates (within cohorts, from grade three through eight), and trend in average test scores (across cohorts); these estimates are constructed by pooling test score data from ten years (2009–2018), six grades (grades three through eight), and two subjects (math and reading) within each school and district (for details, see Fahle et al. 2021). The estimated parameters are available for schools and districts; estimates disaggregated by race-ethnicity, gender, and economic status are available at the district level. In addition, SEDA v.4.0 includes standard errors for these estimates, as described in the technical documentation (Fahle et al. 2021). The public v4.0 data suppresses estimates for small districts and schools and those for whom the estimates are not precise: this has the effect of disproportionately removing rural schools and districts from the public data. We therefore use an unsuppressed version of the data in an effort to maximize the number of rural observations in our analysis.

The outcomes of interest here are average third-grade achievement and learning rates. As stated, we consider students’ third-grade achievement scores to be a reflection of their early childhood opportunities, including the in-school and out-of-school resources often tied to a neighborhood’s socioeconomic status (Reardon 2019). Our second outcome of interest is the linear grade slope on average achievement across grades three to eight (we multiply this rate by five and refer to it as the five-year learning rate), which provides a measure of how much the average student learns from third through eighth grade. Following Sean Reardon (2019), we consider this to be a reflection of middle childhood opportunity. Although these measures can be biased if in- or out-migration of high- or low-achieving students from a district or school is systematic, Reardon, John Papay, and colleagues (2019) find that these measures are, on average, unbiased and are highly correlated with average learning rate measures based on student-level longitudinal data. Estimates for both third-grade achievement and learning rates are standardized and expressed in terms of standard deviations (SD) of the national student-level test score distribution.

To interpret the magnitude of our learning rate measures, we note (following Fahle et al. 2021) that nationally, the average student’s score improves by one-third of a standard deviation per grade. This measure can be used to convert differences in average test scores to rough grade-level equivalents for interpretability. For example, in a district where average test scores are 0.33 of a standard deviation higher than the national average, we can say that the average student is roughly one grade level ahead of the national average for students in her grade. Since the average student test score improves by 0.33 SD per year, a district with an average five-year learning rate of 0.033 SD per grade is one where the average student’s score improves from grade three to grade eight by 0.10 grade levels relative to the national average; since the national average five-year learning rate is 5.0 grade levels per grade (by definition), students in this district are learning at a rate 2 percent faster than the national average. A district with an achievement or learning rate score of zero sits at the average of the national distribution for that outcome.

We use measures of socioeconomic status and racial-ethnic composition for both our school- and district-level analysis. The school-level covariate data is drawn from the Common Core of Data (CCD), which provides the percentage of students eligible for free or reduced-price lunch and the racial-ethnic composition of students in each school. The district-level socioeconomic status measure is estimated from the National Center for Education Statistics (NCES) Education Demographic and Geographic Estimates (EDGE) program data, which tabulates American Community Survey (ACS) data within geographic school district boundaries. The ACS and EDGE data are reported as five-year averages; we use the 2005–2009 through 2014–2018 waves of EDGE data. The socioeconomic status (SES) measure constructed by taking the first principal component of six variables reported in the EDGE data: median family income, proportion of adults with a bachelor’s degree or higher, household poverty rates, proportion of unemployed adults, proportion of households receiving Supplemental Nutrition Assistance Program benefits, and proportion of households with children headed by a single mother. We use an empirical Bayes “shrunken” estimate of the SES composite in our analyses. The district-level racial-ethnic composition measure is derived from the school-level CCD data (Fahle et al. 2021).

Our analysis uses a number of community classification measures to describe a school or district’s context: community locale (rurality-urbanicity), geographic region, and type of local economy. For community locale, we use SEDA’s school- and district-level urban-centric locale codes sourced from the NCES. This geographic indicator categorizes communities into four primary types—rural, town, suburban, and city—that each have three subtypes according to population size or proximity to urban centers. This scheme differs from the Economic Research Service’s Rural-Urban Continuum Codes commonly used by researchers performing county-level analyses in that the units are both more granular and defined by proximity to densely populated areas rather than by specific municipal boundaries. Under this scheme, rural locales are census-defined rural territories that are located outside of densely settled areas known as urbanized areas or urban clusters (for more detail, see NCES 2006). Our analysis uses these locale classifications in two ways. First, we use the four primary types to distinguish between rural and nonrural students: we consider students in rural locales to be rural students, and we consider students in city, suburb, and town locales to be nonrural students. Second, we use the three rural subtypes—rural fringe, rural distant, and rural remote—to discern the influence of relative geographic isolation on rural student achievement (for additional definitions, see table A.1).

We use two additional community classifications. First, we classify districts according to their state’s 2017 Census Bureau division code to understand how educational opportunity varies across large regions of the country (for a geographic distribution of districts by their Census Bureau division code, see table A.2). Second, we classify districts according to the economic subtype of their county using the 2015 U.S. Department of Agriculture (USDA) county typology categorization. This typology includes six mutually exclusive categories of economic dependence, including farming, mining, manufacturing, federal or state government, recreation, and nonspecialized counties. The USDA’s Economic Research Service assigns economic dependence classifications based on the proportion of labor earnings or employment in relation to a set threshold for each industry, averaged from 2010 through 2012. For example, if a county’s farm-based earnings accounted for at least 25 percent of the county’s overall earnings, then that county was eligible for the farming-dependent classification in this period (Pender 2019). Counties are classified as nonspecialized if they did not fall into one of the other five other categories. This measure provides an understanding of the extent to which the concentration of jobs in a certain industry potentially influences local educational opportunity (for a distribution of districts by economic subtype, see table A.3).

ANALYTIC SAMPLE

We focus in this article on rural students. However, the sample of students classified as rural varies across different definitions of rurality and different levels of aggregation. We use the twelve-category urban-centric definition created by NCES and the Census Bureau, which broadly defines rural areas as regions located several miles from urbanized areas and clusters. NCES categorizes schools within this framework according to the community in which the school is physically located, and categorizes districts according to the locale type assigned to the majority of its students (Geverdt 2018). As a result, a small number of students living in rural areas but attending schools in nonrural areas may be classified as nonrural; further, districts and the schools within them sometimes have different classifications (not all schools in rural districts are classified as rural; not all rural schools are in rural districts). For this reason, our sample of rural students differs depending on whether we use schools or districts as the unit of analysis.

Figure 1 provides information on the differences in the size of the rural student population depending on whether school, district, or county classifications are used. Of an average nationwide enrollment of 3.75 million students per grade, approximately 715,000 (19 percent) attend rural schools; 540,000 (14 percent) attend rural districts; and 553,000 (15 percent) are in rural counties (as defined by the NCES and USDA). Of the 814,000 students per grade (22 percent of U.S. students) in either rural schools or rural districts, only 441,000 (54 percent) are included in the sample under both classifications; 99,000 attend nonrural schools within rural districts and 274,000 attend rural schools within nonrural districts.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Rural Student Enrollment per Grade, by Classification, Grades Three Through Eight, 2009–2018.

Source: Authors’ calculations based on the Common Core of Data (https://nces.ed.gov/ccd/).

Note: Figures based on the sum of student enrollment within each classification type.

Because our goal is to understand geographic variation in educational opportunity for rural students, we conduct our initial analyses at both the school and district level and include both sets of results in this article. For the remainder of our analysis, however, we examine educational achievement at the district—rather than school—level. Although analysis at the school level may better capture the set of students in what we think of as locally rural places, a district-level analysis may better capture students whose broader educational context is “rural.” Districts wield significant administrative authority, as Gagnon and Mattingly (2018) note, through actions such as teacher hiring, student assignment and school choice policies, and resource allocation decisions. The strong role of districts in shaping schooling contexts suggests a value in understanding district-level patterns of educational opportunity in rural America. In addition to the conceptual value of studying district-level achievement, reasons for this choice are practical: although SEDA v4.0 incorporates school-level achievement data, achievement and socioeconomic status data broken down by subgroup are not available at the school level.

Our final samples include all U.S. schools and districts for which SEDA achievement estimates and covariate data are available. This amounts to 12,448 districts and 75,080 schools, representing 97 percent of all districts and 93 percent of all schools; 52 percent of these districts and 29 percent of these schools are classified as rural. In total, 99 percent of all students in the United States are included in this sample; 99.9 percent of students attending rural districts are included, along with 94 percent of students attending rural schools (for a breakdown of districts by rurality and demographic subgroup, see table A.4).

METHODS

We first provide estimates of the average achievement and learning rates in rural and nonrural communities for students from different demographic backgrounds. For these estimates, we weight SEDA’s pooled achievement estimates by the per-grade enrollment of each subgroup within a rural or nonrural district (table 1) and then perform the same analysis for students within a rural or nonrural school; we do these parallel analyses in order to compare results from these slightly different samples. These estimates reflect the average third-grade achievement and learning rate of students from a given background who attend school in rural versus nonrural environments.

Next, we restrict our sample to students from rural districts and fit a set of regression models to estimate the associations between test score outcomes (third-grade average achievement and average learning rates) and district characteristics. The models take the form

Where Ŷ_d is the estimated average third-grade achievement (or learning rate) in rural district d, averaged across subjects (mathematics and ELA) and years, and X_d is a vector of district covariates. The model has two error terms; ε_d represents the true residual for district d, net of the characteristics in X_d, and ε_d represents the measurement or estimation error in Ŷ_d. The true residuals are assumed homoscedastic with variance to be estimated; the measurement error is assumed to be normally distributed with district-specific variance equal to the estimated sampling variance of Ŷ_d (that is, equal to the squared standard error Ŷ_d). We fit these models via maximum likelihood, using the – metareg – command in Stata 15 (Harbord and Higgins 2008).

RESULTS

We start by comparing the characteristics of the rural student population when drawn at the school level versus the district level. Defining the rural student population at the school level results in a sample with higher proportions of Asian students, Hispanic students, and Black students, whereas defining the rural student population at the district level results in a sample with higher proportions of Native American and White students (table 1). Although the magnitude of these differences is relatively small, the shift is important to acknowledge. Because our composite measure for socioeconomic status is available only at the district level, we include the percentage of students eligible for free or reduced-priced lunch for both schools and districts to compare any differences; we find that these rates are essentially the same. Overall, though these samples are not directly comparable they are similar enough to suggest the potential importance of careful attention to how we define rural students.

We next compare achievement and learning rates between rural and nonrural students to answer our first research question: Do students achieve and learn at higher or lower levels in rural versus nonrural communities, and do these patterns vary by race-ethnicity, economically disadvantaged status, gender, or region? Table 2 reports the mean values for third grade achievement and learning rates for the average student in rural districts compared to the average student in nonrural districts, whereas table 3 reports a subset of these values at the school level. Because not all schools include third-grade students, table 3 reports average achievement in the middle grade of each school’s tested grades in lieu of third-grade achievement.

The first row of table 2 shows that the average rural third grader performs similarly to the nonrural third grader, with only a 0.01 SD difference at the district level. We see similar results at the school level, though the difference is slightly larger (0.03 SD); in other words, the average student attending a rural school scores approximately one-tenth of a grade level higher than the average student attending a nonrural school. This is a very small difference, much smaller, for example, than the difference in scores resulting from being exposed to teachers of higher or lower effectiveness (Rivkin, Hanushek, and Kain 2005; Rockoff 2004; Kane and Staiger 2008).

Within each racial-ethnic group, rural students have lower third-grade achievement than those in nonrural districts, with the exception of Hispanic students, where rural students perform marginally better than their nonrural peers (0.02 SD). White students and Native American students experience the largest differences in achievement, the typical rural student scoring about a half a grade lower than their nonrural counterparts (0.18 SD and 0.15 SD, respectively). Asian students attending rural districts are scoring about a third of a grade level lower than their nonrural peers (by 0.10 SD). It may seem paradoxical that average achievement is lower among rural students within each racial-ethnic group, but higher among rural students overall. This apparent contradiction (a version of what is known as Simpson’s paradox) is due to the fact that White students, who have above-average test scores in both rural and nonrural places, make up a much larger proportion of the rural student population than the nonrural population (see table 1).

Whereas economically disadvantaged students in rural districts have higher third-grade achievement than their peers in nonrural districts (by 0.09 SD), non-economically disadvantaged students in rural districts actually have lower scores than their nonrural counterparts (by 0.11 SD). In other words, wealthier rural students are about a third of a grade level behind wealthier nonrural students, and poorer rural students are nearly a third of a grade level ahead of their nonrural counterparts. Although not the focus of our analysis, we find that both male and female third graders in rural districts marginally outperform their nonrural counterparts (0.01 SD).

Tables 2 and 3 likewise present rural-nonrural comparisons in learning rates over five years. These differences are statistically significant, given the enormous samples, but are very small: the five-year learning rates differ by just 0.01 SD (the average student’s score improves by 0.33 SD per year, or by 1.67 SD over five years, so a difference of 0.01 SD in five years is less than 1 percent of the average learning rate). The typical student in a rural district learns at a rate just 0.01 SD slower over five years than that of the typical student in a nonrural district, whereas the typical student attending a rural school learns at a rate just 0.03 SD slower over five years than students in nonrural schools. District-level patterns differ slightly by race-ethnicity, but the differences remain small: Black rural students improve 0.01 SD more over five years than their nonrural counterparts, and Hispanic rural students improve 0.05 SD more over five years than theirs. In contrast, rural White students improve 0.02 SD less than their nonrural peers. Asian and Native American students’ scores do not differ between rural and nonrural districts. Rural economically disadvantaged students improve 0.01 SD less than economically disadvantaged students in nonrural districts, and non-economically disadvantaged students in rural districts improve 0.02 SD less than their nonrural counterparts. Finally, male and female students in rural districts learn at rates 0.02 SD and 0.01 SD slower than their nonrural counterparts, respectively.

Most regional differences between rural students and nonrural students are small. Rural students in the district-level sample are performing marginally worse than their nonrural peers in most divisions. This pattern has a few exceptions: rural students in the New England and east-north-central regions of the United States are outperforming their nonrural peers (by 0.07 SD and 0.09 SD, respectively), and rural students in the Pacific and south Atlantic regions are underperforming their nonrural peers (by 0.13 SD and 0.09 SD respectively). Regional differences in learning rates are also small when using the district-level sample: the largest difference occurs in the Pacific region of the United States, where rural students are learning 0.07 SD more over five years than their nonrural peers. For both achievement and learning rates, these differences become more consistent at the school level (table 3).

Because our nonrural student category represents a very heterogenous group of communities, we provide a detailed breakdown of achievement and learning rates in urban, suburban, and town environments. Although the nonrural students in aggregate are not dissimilar from rural students on these measures, table 4 shows considerable variation both within and across locales. At both the school and district level, the average city student scores well below the national average and the average suburban student scores well above it. Students living in towns—communities within an urban cluster but located outside an urbanized area—fall slightly below. Although the stark city-suburban differences in achievement harken to the dominant narrative in education research and policy, the outcomes for rural and town students—who together make up all students living outside urbanized areas—highlight the need to better understand what is taking place in the middle. We see little variation in learning rates between locales.

Before turning to our analysis of variation among rural school districts, we investigate the contrasting rural-nonrural trends for different socioeconomic groups by examining the extent to which community socioeconomic status is differentially associated with rural and nonrural test score patterns. To do this, we plotted the relationship between the average socioeconomic status of families in a district (the composite SES variable described above) and the average third-grade achievement or learning rate in a district (figures 2 and 3), weighting by district enrollment so that the fitted lines reflect the test scores of the average student, conditional on district SES.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Third-Grade Average Achievement and Community Socioeconomic Status

Source: Authors’ calculations based on SEDA v4.0 (Reardon et al. 2021).

Note: Thirty-three districts with achievement greater than 1.25/less than –1.25 or SES greater than 3/less than –3 omitted to improve visual display.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Learning Rates and Community Socioeconomic Status

Source: Authors’ calculations based on SEDA v4.0 (Reardon et al. 2021).

Note: Forty-one districts with a learning rate greater than 0.2/less than –0.2 or SES greater than 3/less than –3 omitted to improve visual display

Figure 2 shows a visibly shallower slope between district socioeconomic status and achievement for rural districts, meaning that community socioeconomic status is less predictive of average third-grade achievement in rural districts than of nonrural districts (F_2,12421 = 7.28, p < .001). This difference is modest but meaningful, showing achievement differences at extreme ends of the socioeconomic spectrum to be less dramatic in rural settings. For example, the difference between the average achievement of districts with socioeconomic values of +/-2 is approximately 10 percent smaller among rural districts than nonrural districts. Figure 2 also shows our previous finding that average third-grade achievement is higher in rural districts than nonrural districts, though this pattern reverses among the highly resourced communities. Figure 3, however, shows that the learning rate per SES gradient for rural and nonrural districts is virtually identical (F_2,12421 = 1.24, p > .288).

Having explored differences in student outcomes between rural and nonrural districts, and the extent to which these differences vary between the district and school level, we next explore variation in educational opportunity among rural districts. Because we are interested in the influence of community factors on student outcomes, our analysis focuses on district- rather than student-level variation (for the corresponding unweighted district descriptive statistics, see table A.5).

Regional Variation and Community Demographics

Our second research question asks whether regional differences in rural district achievement and learning rates are systemic. Figures 4 and 5 present this variation over three community classifications: division, rurality, and type of local economy.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Average Achievement and Community Classifications Among Rural Districts

Source: Authors’ calculations based on SEDA v4.0 (Reardon et al. 2021).

Note: New England is the reference category for divisions, rural fringe for rural urban locales, and farming for economic subtypes.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Average Learning Rate and Community Classifications Among Rural Districts

Source: Authors’ calculations based on SEDA v4.0 (Reardon et al. 2021).

Note: New England is the reference category for divisions, rural fringe for rural urban locales, and farming for economic subtypes.

We first examine differences in average achievement and learning rates among the nine U.S. Census divisions. Average achievement is higher in rural New England school districts (the reference category) than in any of the other divisions, and is lowest in the Pacific and the East and West South Central divisions (which include much of the South, from Kentucky, Tennessee, and Alabama west to Texas and Oklahoma). These divisional differences are large relative to the total between-district differences, accounting for 23 percent of between-district variance. Many of the differences between the regions are associated with demographics—as indicated by the smaller effects after we control for community socioeconomic status and racial-ethnic composition—but important geographic variation remains.

In contrast, regional differences in learning rates are not associated with community demographics. Regional patterns in student learning rates (figure 5) differ markedly from regional patterns in achievement: conditional on community socioeconomic status and racial-ethnic composition, districts in the Pacific and Mountain divisions are growing 9 percent and 4 percent faster than New England districts. This regional finding echoes earlier evidence (Reardon 2019) that districts with high levels of early childhood educational opportunity are not necessarily districts with high levels of middle childhood educational opportunity.

The regional variation among rural school districts’ test score patterns is evident in figures 6 and 7. Overlaying the map with census division boundaries makes the relationship clear. Although some divisions are more heterogenous than others, one can clearly see that some divisions are more advantageous for a rural student. In New England, the average rural student is performing one grade level higher than the national average for all students; the opposite is true for the West Coast, where the average rural student is performing one grade level below the national average. Rural districts in the south and west in particular have low test scores, and some—but not all—of that disparity is accounted for by differences in community demographics.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Educational Achievement Across Rural U.S. Districts

Source: Educational Opportunity Project at Stanford University (https://www.edopportunity.org).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Average Annual Learning Rates Across Rural U.S. Districts

Source: Educational Opportunity Project at Stanford University (https://www.edopportunity.org).

We next examine differences in educational outcomes by rurality. We find that average achievement is lower in the more remote districts. Although the amount of variation this explains is quite small relative to that explained by demographics, we find that these differences in achievement are only partially accounted for by community demographics. In other words, geographic isolation appears to have a unique association with achievement beyond what can be attributed to differences in the population who live there. Looking at learning rates, however, we find only marginal differences that disappear once we account for socioeconomic status, suggesting that slower learning rates in more remote areas are the result of fewer resources in these areas.

Finally, we examine variation in rural achievement and learning rates by economic specialization. We use farming-dependent economies as the reference category, given that it is one of the two most common specialized economies across rural districts. Looking at specific economic subtypes, we see that achievement is modestly lower in nonspecialized, government-dependent, and especially mining-dependent economies than in farming-dependent ones, which is partially explained by differences in community composition. Learning rates are modestly higher in nonspecialized, government-dependent, and recreation-dependent economies, but economic subtype explains virtually zero variance in learning rates. We find that the type of local economy is overshadowed by community demographics.

We conclude our analyses by exploring the extent to which average third-grade achievement and learning rates are related. Figure 8 shows that rural districts (but not nonrural ones) with higher average third-grade achievement tend to have a lower learning rate. We explored the extent to which this is due to district size and socioeconomic context. We find that districts with smaller enrollment—both rural and nonrural—exhibited this negative relationship, but that among districts with larger enrollments, only rural districts exhibited this relationship. This suggests that though size is a factor, something is unique happening in rural districts. Although it is expected that districts with smaller enrollments would have a stronger relationship (because the larger measurement error in third grade achievement estimates in small districts will lead to greater apparent regression to the mean in later grades, which results in a negatively biased estimate of the association between third grade scores and learning rates), this alone does not account for the relationship. When exploring these relationships between rural and nonrural districts across the socioeconomic spectrum, we did not find any systemic differences (see figures A.1 and A.2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Third-Grade Average Achievement and Average Annual Learning Rates

Source: Authors’ calculations based on SEDA v4.0 (Reardon et al. 2021).

Note: Thirty-four districts with a learning rate greater than 0.2/less than –0.2 or achievement greater than 1.25/less than –1.25 omitted to improve visual display.

DISCUSSION

No prior study has examined rural achievement and learning rates with the level of granularity as we have here. Thus our primary goal in this article is to establish a set of stylized facts regarding the trends, variation, and correlates of achievement in rural districts. Four particular patterns—and their implications—are worth noting.

First, average achievement and learning rates differ only trivially between rural and nonrural students, on average, but do differ somewhat for some student groups. The largest differences are for White and Native American students. For these groups, third graders attending school in rural districts are scoring about half a grade level lower than their nonrural counterparts. We also note large differences between rural and nonrural students in the Pacific division (Alaska, California, Hawaii, Oregon, and Washington), where third graders in rural districts are achieving, on average, roughly 40 percent of a grade level behind their nonrural counterparts (by 0.13 SD). Rural students in the South Atlantic, too, fall behind their nonrural peers (by 0.09 SD). In contrast, students in the East North Central division (Illinois, Indiana, Michigan, Ohio, and Wisconsin) perform more than a quarter of a grade level ahead of their nonrural peers.

Rural-nonrural differences within economic subgroups were sizable as well: economically disadvantaged students in rural districts are scoring nearly a third of a grade level higher than their peers in nonrural districts, whereas non-disadvantaged students in rural districts are scoring a third of a grade level lower than their nonrural counterparts. These contrasting trends are worth noting because they suggest that the achievement differences at each end of the socioeconomic spectrum appear less dramatic in rural settings. This may be due to a combination of factors. A ceiling effect may be at play in rural communities, such that the wealthiest community members experience an upper limit on the types of advantages they are able to secure for their children. Similarly, a floor effect is also possible by virtue of the fact that rural communities, in general, do not have the critical mass to segregate the public education system: this puts the communities’ poorest children in the same childcare centers and classrooms as the communities’ wealthiest children, which may benefit children with fewer resources.

Outside these important differences, the overall lack of a major disparity in test scores between rural and nonrural students is noteworthy when considering academic achievement as a measure of educational opportunity and the role it plays in college enrollment and attainment. Rural students attend and graduate college at a lower rate than their nonrural peers, and our findings suggest that these trends cannot be explained by differences in achievement (Byun, Meece, and Irvin 2012). Indeed, research indicates that rural students face a number of hurdles to educational attainment, including financial burden, a lack of nearby jobs that require a degree, geographic isolation from higher education institutions, reduced access to rigorous curricula in high school, and family and societal expectations to not attend college (Byun, Meece, and Irvin 2012; Byun, Irvin, and Meece 2015; Roscigno and Crowley 2001; Roscigno, Tomaskovic-Devey, and Crowley 2006).

Second, socioeconomic status is less predictive of achievement in rural America than in nonrural America. In other words, the average SES for families within a district appears to matter more for student achievement in nonrural districts than it does in rural districts. This echoes our previous observation of contrasting rural-nonrural trends for economically disadvantaged versus non-economically disadvantaged students, where disadvantaged rural students are scoring higher than their nonrural counterparts and advantaged rural students are scoring lower. Despite this finding, we still find that, of all covariates examined, socioeconomic status is most strongly predictive of rural student achievement. This suggests that although poverty levels may be a weaker predictor of the early educational opportunity in rural communities than in nonrural communities, it still plays an important role in rural student outcomes.

The third pattern is that place matters for rural students, both in terms of geographic location and community demographic composition. We find that census divisions account for 23 percent of the variance in early childhood educational opportunity (third-grade test scores) and 7 percent of that in middle childhood educational opportunity (five-year learning rate between third and eighth grade). Although the extent to which census divisions accounted for variation in educational opportunity surprised us, we suspect that this may be due in part to differences in the historical development of school systems across the country. Our other community classifications explained very little of the variation in comparison, though our results suggest that achievement is slightly lower in more remote districts and in communities with mining-dependent economies. Compared to these community classifications, community demographics explain a lot more variation: census division, socioeconomic status, and racial-ethnic composition explain more than half of the variation in rural third-grade test scores and 10 percent of variation in learning rates. That community demographics overshadow these regional measures is perhaps not surprising: decades of research have highlighted differences in educational outcomes by socioeconomic status and race-ethnicity that are often the result of policies and practices that produce structural inequalities within and across schools, districts, communities, and states. These findings point to the need to better understand the intersecting ways that racism, economic deprivation, and regional policies may be fundamentally shaping rural educational opportunity across the country.

Fourth is a somewhat surprising finding: third-grade achievement has a strong negative association with learning rates in rural areas and explains three times the variation in rural learning rates that community demographics and geographic location do. In other words, rural communities that provide more early childhood educational opportunity tend to provide less middle childhood opportunity, and those that provide less early opportunity tend to provide more. This pattern is not evident in nonrural communities or among all school districts (Reardon 2019). This negative association is evident even after controlling for regional, economic, and demographic differences among rural communities, though we find it does flatten slightly among larger districts.

One possible explanation for these trends is that both smaller districts and rural districts have fewer resources with which to support high-achieving students. One such resource is high-quality teachers: smaller schools are less likely to employ teachers with high levels of education and training and tend to require passing scores on standardized teaching assessments at a lower rate than their larger peers (Monk 2007). These trends are true for both rural and small schools, though more prevalent in the latter. Further, existing literature on gifted student programs suggests that rural schools struggle to provide differentiated learning options because of staffing and funding limitations (Weinlein 2019). Taken together, it is reasonable to expect that rural schools, especially smaller ones, are limited in their ability to provide challenging, differentiated instruction to support high-achieving students, and must prioritize their efforts on ensuring that all students meet a minimum level of proficiency. These hypotheses, however, require further study.

These analyses provide a more nuanced understanding of the variation in educational outcomes among rural districts and suggest that further attention should be paid to how differences in local policies, practices, and norms in rural communities may account for the varying educational trends we observe. This article has several limitations. One is that the patterns we describe apply only to grades three through eight, so we cannot speak to the trends in rural achievement in earlier or later grades, which may differ from what we observe. In addition, we cannot explain the mechanisms underlying the different trends we see in terms of the learning rates and achievement of rural versus nonrural students; we are merely describing associations.

We hope that future work will build on the insights from this research because more work is needed to consider policy implications for the disparities we find. Future work should consider other features of local context that might influence student achievement and learning rates, such as additional characteristics of the local educator labor market and more detailed measures of the local health and social ecosystems. More research is also needed to understand the negative association we identify between achievement and learning rates in rural communities. Further, our later analyses focus on school districts characterized as rural, which omits many rural schools located in nonrural districts. Future analyses might benefit from using the population of rural schools rather than rural districts, though fewer contextual data are available for schools than for districts. This important analysis would allow for a more precise understanding of rural student outcomes, given that districts are often large and encompass many different types of schools. As figure 1 shows, a sample selected at the district level encompasses the outcomes of many students who do not, in fact, live in a rural environment. The differences in these samples are not inconsequential: analyzing rural districts instead of rural schools disproportionately drops rural southern schools. Finally, it is worth exploring trends in rural achievement over time, particularly because all of the data in our analysis was gathered from years directly following the 2008 recession, which had a varying impact on rural and nonrural communities nationwide (Hertz et al. 2014). These types of additional analyses will be critical for continuing to shed a light on the state of education as well as the conditions that best support students in rural America.

• Download figure
• Open in new tab
• Download powerpoint

Figure A.1.

Third-Grade Achievement to Learning Rate Gradient by District Enrollment, All Districts

Source: Authors’ calculations based on SEDA v4.0 (Reardon et al. 2021).

^a96 percent of these districts are rural (n = 1,626) and 4 percent are nonrural (n = 71).

^b91 percent of these districts are rural (n = 1,625) and 9 percent are nonrural (n = 170).

^c69 percent of these districts are rural (n = 1,627) and 31 percent are nonrural (n = 740).

^d25 percent of these districts are rural (n = 1,625) and 75 percent are nonrural (n = 4,803).

• Download figure
• Open in new tab
• Download powerpoint

Figure A.2.

Achievement/Learning Rate Gradient by District SES, Rural & Nonrural Districts

Source: Authors’ calculations based on SEDA v4.0 (Reardon et al. 2021)

^a47 percent of very poor districts are rural.

^b59 percent of poor districts are rural.

^c63 percent of middle class districts are rural.

^d40 percent of upper class districts are rural.