Appendix
STATISTICAL ANALYSES AND ORAL HISTORY SOURCES
This study relies on evidence from a variety of sources. Most are cited in the book’s endnotes, but it is helpful to discuss some of them further. This brief appendix provides additional details about the statistical analyses described in chapter 2, along with information about oral history sources used throughout the book.
Statistical Analysis of Uneven Educational Advancement
As the maps and data discussed in chapter 2 indicate, educated adults were distributed quite unevenly across the metropolitan Kansas City landscape in both 1960 and 1980. There can be little doubt, however, that these patterns were highly correlated with wealth, race, and a number of other factors. One way to establish this is to explore spatial dissimilarities with the use of ordinary least square (OLS) regression.
To identify the spatial organization of adult educational attainment in each of these years, I created fixed dummy (binary categorical, 0–1) variables with the assistance of Sanae Akaba to represent the principal geographic units discussed in chapter 2. We ran separate cross-sectional regressions for each year, with the percentage of college-educated adults (those with any college in 1960 and at least four years in 1980) as dependent variables. We then added a number of additional factors to control for the effects of race and the age structure of the adult population, and home ownership and wealth (property values) across 120 census tracts in each year. The results are presented in tables A.1 and A.2.
In conducting this analysis, we ran three models in stepwise fashion for each point in time, displayed from left to right in tables A.1 and A.2. The first features just the five geographic areas discussed in chapter 2, which roughly correspond to four school districts and the Country Club district (a residential and commercial area located within the boundaries of KCMPS) established by J. C. Nichols in southwest Kansas City. Suburban Jackson County, excluding Raytown, serves as the comparison category. All results are expressed in the form of standardized regression coefficients (Beta) with tolerance coefficients (Tol) to indicate inflation of standard errors.
TABLE A.1 Regression analysis, adult education levels, 1960 (census tract data)
Dependent variable: Percent of adults over 25 with any college attendance
* Significant at .05 level
** Significant at .005 level
As can be seen in both years, this analysis reflects patterns readily evident in maps throughout the book. In 1960, adult attainment levels were lowest in KCMPS, even before desegregation controversies had occurred there and well before it had become majority African American. Adult attainment was highest, on the other hand, in the Country Club district and SMSD, areas developed by J. C. Nichols and associated with highly regarded secondary schools. The coefficients are positive, modestly robust, and highly significant. This was the “higher education sector” identified on map 2.8, and it is clearly evident in the first models in both tables A.1 and A.2. The two other suburban districts in these models, on the other hand, are not appreciably different from the comparison Missouri suburban tracts, failing to achieve significance in either table. Altogether, these fixed geo-spatial dummy variables account for about half the variation in adult attainment in both years, though slightly less in 1980, and all tolerance coefficients are satisfactory.
TABLE A.2 Regression analysis, adult education levels, 1980 (census tract data)
Dependent variable: Percent of adults over 25 with four years or more of college
* Significant at .05 level
** Significant at .005 level
The picture changes in models 2 and 3 in both tables when additional variables are included. Controlling for race (proportion black in each tract) and adult age structure (adults ages 25 to 35) increases the explained variance somewhat, but only race is significant in both years. Not surprisingly, the sign is negative, and the effect size is about twice as large in 1980, when the African American population was greater and resided in more tracts. Including race in the analysis also reduced the negative coefficient for KCMPS in both tables, more appreciably in 1980, when it became insignificant. This reflects changes that had occurred in the region’s urban core, such as white flight, declining property values, and growing perceptions of school problems, all associated with race. Controlling for race made adult attainment levels in KCMPS statistically similar to suburban Jackson County in 1980; in other words, it was the African American population, much of it from the South, that accounted for much of its negative association with attainment levels in model 1. This is clear evidence of the impact that segregated southern educational systems, white flight, growing numbers of poor minority students, and other issues associated with social inequality had on this dimension of educational resources in the urban core.
Tellingly, the other geo-spatial dummy variables in these analyses were affected much less dramatically by the inclusion of race in model 2, particularly those for the Country Club district and SMSD. They represented areas, of course, with much smaller African American populations. Coefficients for the latter areas were reduced substantially, on the other hand, by the inclusion of a wealth variable (home values) in model 3 for both years. This too is hardly a surprise, as these areas had relatively high property values, especially the prestigious Country Club district. Controlling for home values reduced the coefficient for this area more substantially in 1960 than in 1980, and the reverse was true for SMSD. As noted earlier, there had been turnover in certain census tracts in northern Johnson County, which was still growing faster than the older neighborhoods of southwest Kansas City. This may have accounted for somewhat less change in the college-educated adult population on the Kansas side of the border. It is revealing on this count that the raw (uncontrolled) variable for SMSD is slightly less robust than the coefficient for the Country Club district in 1980, whereas the reverse had been true in 1960. This suggests that there was a higher tendency for college-educated adults to settle on the Missouri side of the border by 1980, at least in this part of the metro area.
It is noteworthy, of course, that the coefficients for both these wealthy areas, one inside the city limits and the other an adjoining suburb, remained positive and significant after controlling for race and property values, along with other factors. This was true for both years, although the magnitude of this independent effect varied, as noted above. Both areas—together comprising one contiguous sector divided by the state line—exhibited high levels of adult educational attainment, even controlling for economic status and key demographic factors. This part of the metropolitan area was wealthy, to be sure, but its residents were also highly educated, beyond the level generally associated with their “economic capital.” In short, they were better educated than their property values alone suggested.
A bit farther to the east, the rapidly growing community of Raytown offers a telling point of comparison. As noted earlier in table 2.3, property values were relatively high there in 1980, but adult education levels were considerably lower than in SMSD, and median income was less too. Raytown’s predominantly white schools were known as good academically but not as exemplary as the SMSD institutions, which received national recognition. Raytown was a largely blue-collar town, solidly middle class but not nearly as fashionable or sophisticated as some of the neighborhoods in SMSD. Its inhabitants may have possessed considerable economic capital, reflected in the value of their homes, but they lacked the cultural capital of SMSD and Country Club residents to the west. In this respect, the legacy of J. C. Nichols and his success as a developer was clear; his communities continued to be associated with a degree of refinement and status that other places lacked.
Other factors that contributed to these differences are not altogether clear.Perhaps it was the allure of high-quality schools in the vicinity of the wealthier neighborhoods, particularly two high schools rated among the best in the country in the later 1950s, one on either side of the border. On the other hand, perhaps it was a more subtle form of influence, with individuals and families deciding to settle in these areas because people with similar tastes and backgrounds already lived there. Whatever the reason, the result is plain to see in model 3 in both tables A.1 and A.2. In addition to considerable economic wealth, this quadrant of greater Kansas City clearly contained its most educated segment of the population.
As a final point, it is noteworthy that the coefficient for the “young adult males” factor in model 3 for the 1980 analysis is positive and statistically significant. This suggests that areas with larger numbers of these individuals and their young families exhibited higher levels of adult education. The areas with the highest numbers of residents in this category were SMSD and the Country Club district, with levels about 25 percent higher than the rest of the metro area. This group represented the leading edge of the baby boom generation, born between 1945 and 1955, and they were starting families during the 1970s. The fact that they were settling in larger numbers in Johnson County and the adjacent Country Club district is evidence of the attraction this “higher education sector” held for families in search of educational opportunity. This group of younger adults also would be most likely to exhibit higher four-year-college levels of attainment than older adults. This was evidence that the next generation of collegiate adults was choosing the very best-educated neighborhoods to settle in and raise their children.
The sign on the KCMPS variable in model 3 is positive, although it failed to achieve significance. This may be due to a portion of the urban core that exhibited higher levels of educational attainment than normally associated with the relatively low stocks of economic capital (property wealth) that existed there. This was certainly evident in the higher levels of adult education exhibited to the west of the notorious “Troost Wall” in Westport and other neighborhoods in its immediate vicinity. Much of this area was undergoing the early stages of gentrification in the latter 1970s, and this may have accounted for some of this effect. In short, once race and property values were held constant, there were parts of Kansas City, Missouri, outside of the Country Club district that exhibited levels of adult education similar to that seen to the south and west in the metro area.
Statistical Analysis of Educational Success
The book’s analysis of the likelihood of graduation considers whether seventeen-year-old youth are enrolled at the junior year or higher, including having graduated, or not. It thus indicates whether they have been successful in school, having kept more or less on track to graduate and not dropped out. At the national level, this variable is correlated with nineteen-year-old graduation rates at .9, accounting for more than 80 percent of the variation across states.1
Because the dependent variable is dichotomous, binary logistic regression was used to assess the likelihood of high school success for individuals in the sample. This technique produces odds ratios to express the chances of such outcomes, net of other factors in the model. These are expressed as coefficients by exponentiation to produce values, with positive and negative signs for odds greater or less than even. Each coefficient (or odds ratio) is an expression of the likelihood that a given characteristic is associated with the outcome in question, controlling for other specified factors. In this case, the question of interest is odds of school success for seventeen-year-olds in Johnson County, Kansas, and urban and suburban settings in metro Kansas City, Missouri.
The dependent variable is coded as a binary measure, with 1 for enrollment in grade 11 or 12 or graduated, and 0 for enrollment in a lower grade or out of school without a diploma. It is thus a broad measure of success in school, or attainment of a fixed standard of accomplishment. To address this question of who succeeded in school and why, this analysis was conducted in a stepwise fashion with four models. Model 1 includes the two geo-spatial dummy variables described above, and other factors are added in successive stages. The final model includes all variables in the analysis and exhibits odds ratios for each of them. This approach is useful for identifying how various factors influence one another, and particularly their interaction with the geo-spatial variables in model 1. The results are presented in the table below.
To begin, model 1 leaves little question that major differences in school success distinguish youth in these places. Seventeen-year-olds in Johnson County were 70 percent more likely to have reached their junior year in high school or better than their Missouri suburban counterparts (the comparison group for geo-spatial factors), and the same was true of their success compared to youth in Kansas City, despite the fact that the latter group included many students attending suburban schools.
Subsequent models in the analysis introduce additional factors, all of them characteristics of students or their families. In model 2, race and gender are added, with binary categorical variables for being African American (as opposed to white) or female (as opposed to male). As noted in chapter 2, being black was associated with nearly 25 percent lower odds than whites of school success in 1980, and controlling for this improved the odds of attainment in Kansas City to almost 10 percent greater odds than suburban Missouri residents. Since more than 90 percent of African American youth in the sample lived in Kansas City, their lower odds of success clearly affected overall city attainment levels. The inclusion of these variables, on the other hand, had no meaningful effect on the odds of school success in Johnson County, where few blacks lived at the time.
TABLE A.3 Binary logistic regression: junior year status or higher, 17-year-olds, 1980
N=958
* Significant at .05 level
** Significant at .01 level
*** Significant at .001 level
The picture changed somewhat in models 3 and 4, however, where socioeconomic, educational, and family structure factors were introduced, each in binary categorical (0–1) form. Here too, the results conform closely to findings of other studies of educational attainment in the postwar era and beyond. Living in a home owned by the family more than doubled the odds of school success for youth in the sample, compared to renters, while living in a household below the poverty level lowered the odds of attainment by nearly 50 percent, compared to families above it. These are robust effects, and controlling for them reversed the negative sign on the African American variable and reduced it to insignificance, along with boosting the fixed Kansas City dummy. Controlling for poverty and home ownership, also binary variables, African American youth were just as likely to experience school success as whites, and Kansas City residents were nearly 13 percent more likely to be successful than their Missouri suburban counterparts. The effect of being female was enhanced as well, with girls exhibiting a nearly 17 percent advantage over boys in this respect. Controlling for these factors also reduced the advantage exhibited by Johnson County youth to about 44 percent greater odds of school success, especially the home ownership variable, which accounted for much of this change alone. As noted in chapter 2, part of the Johnson County advantage in school success was attributable to the area’s high level of home ownership.
In the final step of the analysis, model 4, the picture changes once again with the introduction of categorical variables for living in a single-parent, female-headed household (compared to two-parent households) and having at least one parent with a college degree (compared to those without one). These too are very robust factors. As also noted in chapter 2, living in a single-parent family was associated with a 45 percent reduction in the likelihood of school success, and having a college-educated parent increased the odds of attainment by more than a multiple of 3 (300 percent). A college-educated parent exhibited the strongest relationship to student educational attainment, with a wide range of additional variables controlled. Its inclusion dropped the Johnson County fixed dummy substantially, reducing the advantage of its resident students over their Missouri suburban counterparts to less than 14 percent greater likelihood of school success. The family structure variable in model 4 had very little effect on the Johnson County dummy, although its inclusion did make the black categorical variable both positive and significant. The effect of home ownership was reduced as well, along with Kansas City residence.
From the standpoint of this analysis, however, the interaction of parent collegiate education with the Johnson County dummy helps to demonstrate the manner in which parental education affected the high levels of educational success attained by local students. As argued in chapter 2, it also points to the implications of this for geo-spatial differences in educational attainment in the Kansas City metropolitan area. Johnson County, Kansas, was clearly part of a “higher education zone” in the region. Kansas City, Missouri, on the other hand, was marked by the lack of such attributes, as demonstrated by its rise to significance when they were controlled in the analysis above. Much of the observable variation in attainment appears to have been linked to these family background factors, a finding consistent with decades of research on such questions.
Oral History Sources
In the course of the study I conducted interviews with thirty-one individuals who lived in greater Kansas City between 1950 and 1980 and were connected in one way or another with the region’s public schools. They included former students, teachers, administrators, board members, and community members (including family members of people who worked in the schools). About a third were African American, and the remainder were white. They had lived in all major parts of the metropolitan area and were identified by school district personnel, local historical societies, and by other interview participants. Most of the interviews were conducted with a single subject and lasted between forty minutes and an hour. Others included two or more subjects and took somewhat longer. All were recorded digitally and professionally transcribed. I performed checks of the transcriptions, along with the several transcribers employed on the project.
Per University of Kansas IRB guidelines, the names of oral history subjects were changed to pseudonyms for use in the book. I believed that it was important to guarantee participants anonymity in order to increase the likelihood of open and frank discussion of race, inequality, and related issues, even though several stated that using their identities would be fine. These are socially sensitive topics, and a large body of research suggests that people are often reluctant to discuss them in public forums.2 While anonymizing the interviews surely did not remove such inhibitions entirely, it did reassure some subjects that there would not be repercussions for their remarks.
Validity checks were performed by comparing oral history interviews across subjects, and with other sources of historical perspective, including newspapers and other documentary materials. While just a subset of the interviews are quoted in the book, all offered background information about the period and its educational controversies that was very helpful. As suggested in the narrative, the interviews were especially useful for identifying distinctive black and white perspectives on education and the changing geo-spatial organization of the metropolitan area.
In addition to oral history interviews, depositions of school district leaders, other school employees, and expert witnesses of various sorts conducted by plaintiffs’ lawyers in the Jenkins v. Missouri desegregation case also proved to be very helpful. Since the case began as a suit against the suburban districts by the Kansas City, Missouri, Public Schools, much of the questioning in these depositions focused on issues quite pertinent to this study. In particular, the Jenkins lawyers were especially interested in any behavior of suburban districts or their personnel that could be construed as discriminatory or racially biased, and thus contributing to extant patterns of racial segregation in the schools and the region.
As often occurs in such depositions, many of the participants assumed a largely defensive posture, answering very briefly, but in some cases the responses were quite revealing. Again, much of this provided background for the book, even if such sources are quoted just a few times. Altogether, nine of these depositions were reviewed closely for purposes of this study. This put the total number of oral history sources at forty, representing a substantial contribution to the project.