Introduction
This narrative provides some insights into the quality of life in MoDOT's St. Louis Area District by considering the relationships between a number of 2000 census variables. The census variables considered were selected based upon two criteria. First, their relevance to Environmental Justice and Title VI (of the Civil Rights Act of 1964) reporting requirements and second, their ability to both describe generally understood characteristics of quality of life and to be statistically testable. It should be noted that the statistical method used, structural equation modeling, provides preliminary analysis for neighborhoods and communities within a MoDOT Planning District. This analysis cannot be generalized to other districts or the state as a whole. It is important to keep in mind that the unit of analysis refers to the conditions within a census block group, and not to any single protected population. Thus, what is being measured by considering the interaction between variables is the social and economic environment of the communities and neighborhoods that comprise the planning district.
A Quality of Life (QOL) model was selected for two important reasons. First, the purpose of transportation planning is to ensure that all members of a community benefit from planning efforts and none experience disproportionate burden. Second, there is an established use in transportation planning of considering QOL. Forkenbrock (1999) advocated considering the impact of planning on low-income and minority communities to address environmental justice issues including federally funded transportation-related programs, policies, and activities having the potential to adversely affect human health or the environment. Purvis (2001) extended the environmental justice variables to include elderly and disabled populations based on proposed metropolitan and statewide planning regulations released in May of 2000. Purvis suggested the use of a 'discrimination assessment' to include a geographic and demographic profile that addressed these four populations in terms of the positive and negative impacts of transportation services available and planned.
While there is no definitive list of social and economic variables that best measure the quality of life for a geographic area, the Decennial Census of Population and Housing is an exceptional data source to explore this issue. Census 2000 variables used to construct the QOL models include both the populations of importance to MoDOT - low-income, disabled, minorities, and elderly and the variables educational attainment, income, housing, transportation and employment to measure quality of life.
Findings Summary
Preliminary findings reveal that for all but the elderly, the St. Louis Area protected populations were more likely than the general population or other special populations to live in neighborhoods and communities with characteristics indicative of a lesser quality of life than the District in general. The analysis allows planners and other community decision-makers to understand the specific barriers to quality of life and, thus, to address them, as possible, within the context of the planning process.
Model 1. QOL Structural Equation Model for the St. Louis Area District
Model 2. QOL Structural Equation Model for the St. Louis Area District with Percent Poor Removed
Model 3. QOL Structural Equation Model for the St. Louis Area District with Percent Minority Removed
Understanding the St. Louis Area District
The overall fit of the structural equation model for the St. Louis Area District was statistically significant and indicated that, as a whole, the protected populations and the quality of life variables are related to each other. Further, though the model did not indicate a degree of multicollinearity (i.e., populations or quality of life indicators relating to each other in a manner that detracts from the ability to measure the relationship between a single population and a quality of life variable) significant enough to nullify the overall fit of the model, not all results were as strong as anticipated or related in a manner that supports prior research and the common understanding of these relationships. To test the validity of the results of the model, two additional models were constructed; one that omitted the percent poor variable (Model 2) and one that omitted the percent minority variable (Model 3). Based on the three models, the following paragraphs first describe the relationships between each protected population and the quality of life variables and then describe important relationships between the quality of life indicators themselves.
Minority Population
For the model including all protected populations (Model 1), the quality of life indicators showing relationships to the minority population (Map 1) were no vehicles available, unemployment, average age of housing, and median house value. These were all low strength relationships indicating that minorities are more likely to live in communities and neighborhoods with slightly higher unemployment, slightly older houses with lower than the median housing values. They were also more likely to live in neighborhoods and communities with more households that had no vehicle available than the District overall. Conversely, the relationships to median gross rent, median household income, and no high school diploma are negligible. These findings indicate that minority neighborhoods are statistically no more or less likely than all neighborhoods and communities in the St. Louis Area District to have a population that pays more or less in rent, has more or less household income, and has not finished high school.
Because the results for some of the quality of life indicators were not what were anticipated, a model was tested that did not include the variable percent poor (Model 2). This was done to ensure that the quality of life indicators were accurately being measured in regard to the minority population and that the results were not overly influenced by the relationship between minority and poverty. In this model, the strength of the relationships between the percent minority and the variables for unemployment, households without vehicles, and average age of housing units were increased to moderate-strength relationships. Indicating that minorities are more likely to live in neighborhoods characterized by a higher percent of the population unemployed, more households without vehicles, and older housing. However, this model indicated no relationship between the minority population and the median household value, and like the first model, no relationship to median gross rent. The model also described the minority neighborhoods as slightly more likely than the overall district to be characterized by lower household incomes and a greater likelihood of not having a high school diploma.
Map 1
click map for larger version
|
|
Map 2
click map for larger version
|
|
Poor Population
As anticipated, the model that includes all protected populations (Model 1) tends to live in neighborhoods characterized by households with no vehicle available, high unemployment, and significantly more persons without a high school diploma than the District overall. Additionally, a low-strength relationship exists between the percent poor and median household income, the age of housing, the median house value and gross rent. Thus, the poor population is more likely to live in neighborhoods comprised of older housing units, lower household income, and lower gross rent costs.
The initial model of the St. Louis Area District shows a negligible, but positive, relationship between the percent of persons poor and median house value. Because this result seemed to contradict previous research as well as community perception, a second model was tested that excluded the minority population (Model 3). The results of this model confirmed and showed a strengthened and substantial relationship between the percent poor variable and the likelihood of increased unemployment and an increased probability of households with no vehicles available. The moderate-strength relationship between this population and the variable no high school diploma stayed the same. The relationship between poverty and median household income is stronger than in the first model, indicating that the poor are more likely to live in neighborhoods characterized by low household incomes. In the model excluding the minority population, there was no relationship found between the poverty population and the quality of life indicators measuring housing costs, median household value and median gross rent.
Disabled Population
There was little variation in results of the three models for the disabled population. The models indicate a moderately strong relationship between the percent of persons disabled (Map 3) and the percent of persons without a high school education. Relationships also exists, but are not as strong, between the percent of the population that is disabled and the median household income and median house value. These findings suggest that the disabled population is more likely to live in neighborhoods and communities with a high percentage of population that does not have a high school diploma and that is poorer and has a housing stock that is overall of lesser value than the median. A similar strength of relationship exists between the percent of the population that is disabled and quality of life indicators, no vehicles and median gross rent. The relationship suggests that the disabled population lives in neighborhoods characterized by a greater percentage of households without vehicles and living in lower cost rental units. There is a negligible positive relationship between the indicator average age of housing units and the percent disabled, indicating that the disabled are nor more likely to live in neighborhoods and communities with housing units aged any different than the district overall. The relationship between the percent of persons disabled and the percent unemployed was also negligible, indicating that disabled persons are not very different than the overall population in regard to employment.
Map 3
click map for larger version
|
|
Map 4
click map for larger version
|
|
65 Years Old and Over
There was little variation in results of the three models for the elderly population. The models reveal that in the St. Louis Area District the elderly population (Map 4) is more likely to live in neighborhoods with higher median house values. A weak relationship exists between the elderly population and the percent of housing units without an available vehicle. Something can be learned from examining the lack of relationship between protected populations and quality of life variables. For the elderly population of the St. Louis Area District the lack of findings of significant strength indicates that they typically live in neighborhoods and communities that are no different from those of the general population.
Relationship Between Dependent Variables
The models also offer the means with which to look at the relationship that exists between the percent of persons without a high school education (Map 5) and the quality of life indicators. For the model that measures all protected populations (Model 1) and the model that measures the poor, disabled, and elderly populations (Model 3), neighborhoods that are occupied by a greater percentage of persons without a high school education are characterized by lower house values, lower rent costs, and lower incomes, though a very weak relationship did emerge indicating a greater likelihood to live in newer than average housing. In the model that included minority, disabled, and elderly populations (Model 2), the strength of all relationships but that of average age of housing were enhanced, finding a substantial relationship between the percent of persons with no high school diploma and lower house values. A moderate relationship exists between percent of persons with no high school diploma and lower rent costs and lower income. A weak relationship exists between the percent of persons without a high school diploma and the likelihood of households with no vehicle available and increased unemployment. The relationship between the percent of population without a high school diploma and the average age of housing units decreased from a weak to negligible relationship, indicating that the age of housing stock is not a reliable predictor of where this population is located.
Map 5
click map for larger version
|
|
Map 6
click map for larger version
|
|
Indicator Selection Criteria
To measure the impact of education on quality of life, the variable of not having a high school education was used. Studies (Rumberger, 1987; Digest of Educational Statistics, 1998) have indicated that persons not completing a high school education are at an increased risk of not finding steady employment, living in less than adequate housing, and earning less when they do work.
Median household income (Map 6) and unemployment status (Map 7) were chosen as indicators of economic well-being. Typically, the less income available to a household, the more difficult it is to acquire the goods and services indicative of a high quality of life. Unemployment status is a useful measure of economic opportunity as well as a predictor of concentrations of poverty within MoDOT districts.
Map 7
click map for larger version
|
|
Map 8
click map for larger version
|
|
Median gross rent (Map 8), median house value (Map 9) and the average age of the housing unit (Map 10) were used as measures of housing quality. Both median gross rent and median house value were included to capture the impact of quality of housing for both households that own and rent. Additionally, there is an established relationship between the market value of housing and the cost. Thus, it is a reasonable assumption that the higher these values the greater the quality of housing units. Because the populations of interest in this model are more likely to live in neighborhoods that are both older and poorer than the general population, the average age of the housing unit was used to complement the variables rent and housing value.
Map 9
click map for larger version
|
|
Map 10
click map for larger version
|
|
To measure the impact of access to transportation on quality of life, the variable of not having a vehicle available (Map 11) was included. The availability of a vehicle is an important indicator of mobility affecting access to employment opportunities as well as the goods and services necessary to maintain an adequate quality of life. Additionally, districts that indicate a significant number of neighborhoods without access to a vehicle will require an increased need for public modes of transportation.
Map 11
click map for larger version
|
Appendix: Definition of Variables
Independent Variables
Percent Minority - The percent minority variable is a measure of the percent of all of the single race categories, other than white, that respondents could have chosen from the census questionnaire. These include: African American, American Indian, Asian and Pacific Islander, and Other Race as well as if they selected Hispanic. Overall, 468,987 people comprised the minority population, representing 24.2 percent of the total population.
Percent Disabled - The percent of individuals that were classified as having a disability if any of the following three conditions were true: (1) they were 5 years old and over and had a response of ''yes'' to a sensory, physical, mental or self-care disability; (2) they were 16 years old and over and had a response of ''yes'' to going outside the home disability; or (3) they were 16 to 64 years old and had a response of ''yes'' to employment disability. Overall, 306,973 people comprised the disabled population, or 17.2 percent of the total population for whom disability status could be determined.
Percent Poor - The percent poor variable is a measure of the percent of persons for whom poverty status was determined. The Census Bureau uses the federal government's official poverty definition. Assigning poverty status takes into account both the family size and total family income. Poverty status was determined for all people except institutionalized people, people in military group quarters, people in college dormitories, and unrelated individuals under 15 years old. The following link is the poverty threshold table for 1999: http://www.census.gov/hhes/poverty/threshld/thresh99.html. Overall, there are 182,864 people, 9.6 percent of the district's population, considered poor by federal guidelines.
Percent 65 and Over - The percent 65 and over variable is comprised of the percent of person's aged 65 years old and over. In total, 12.7 percent of the district's population is elderly (245,541 people).
Dependent Variables
Percent No High School Diploma - The percent no high school diploma variable is a measure of the persons aged 25 years or older who did not graduate high school and have not received a GED. Overall, there were 204,151 people (16.2 percent of people aged 25 years or older) who had not received their high school diploma.
Median Household Income - The median household income variable is a measure of the median household income in 1999 dollars. Household income includes the income of the householder and all other individuals 15 years old and over in the household, whether they are related to the householder or not. The median household income for the St. Louis Area District is $46,184.
Average Age of Housing Units - The average age of housing units variable is a measure of when the unit was built. The average housing unit age for the St. Louis Area District is 37 years.
Median Gross Rent - The median gross rent variable is measured in 1999 dollars. Gross rent is the contract rent plus the estimated average monthly cost of utilities and heating fuels if not included in the contract rent. The median gross rent for the St. Louis Area District is $537.
Median House Value - The median house value variable is a measure of the median value of housing units in 1999 dollars. Housing units are defined as house and lot, mobile home and lot, or condominium unit. Housing value data were determined by asking a sample of respondents to estimate the value of their owner-occupied housing unit, any housing units that they were buying, or housing units they owned that were vacant and for sale. Value is the respondent's estimate of how much the property would sell for if it were for sale. If the house or mobile home was owned or being bought, but the land on which it sits was not, the respondent was asked to estimate the combined value of the housing unit and property. The median value of housing units in the St. Louis Area District is $109,508.
Percent No Vehicles - The percent no vehicles variable is a measure of the percent of occupied housing units whose residents reported having no vehicle present. These data show the number of passenger cars, vans, and pickup or panel trucks of 1-ton capacity or less kept at home and available for the use of household members. Vehicles rented or leased for 1 month or more, company vehicles, and police and government vehicles are included if kept at home and used for nonbusiness purposes. Dismantled or immobile vehicles are excluded. Vehicles kept at home but used only for business purposes also are excluded. Overall, there were 71,602 occupied housing units without a vehicle, or 9.4 percent of all occupied housing units.
Percent Unemployed - The percent unemployed variable measures the percent of persons eligible for work but who were not employed at the time they completed the census. All civilians 16 years old and over were classified as unemployed if they reported that they were neither ''at work'' nor ''with a job but not at work'' during the reference week. Also included as unemployed were those who reported that: they were looking for work during the last 4 weeks and were available to start a job, did not work at all during the reference week, were on temporary layoff from a job, had been informed that they would be recalled to work within the next 6 months or had been given a date to return to work, and were available to return to work during the reference week, except for temporary illness. Overall, there were 55,012 persons classified as unemployed, equaling 5.5 percent of the total population eligible to work.
Interpreting Structural Equation Modeling
A statistical method, structural equation modeling, was used to analyze the relationships between the census variables described above. The value of this statistical method is that it allows consideration of whether or not these variables have an effect on each other, and if they do, the strength of that effect. The responses to the variables were aggregated to the level of the census block group. In total, there are 1,430 block groups in the St. Louis Area District, 1,412 of which were used in the analyses of all three models. If data were missing for any of the eleven variables to be considered in the statistical model, that block group was excluded from the analysis. (In order to better understand the mechanics of the SEM and the terminology associated with the analysis click on the following link: Interpreting the Structural Equation Model.
This statistical method allows interpretation of the relationship between variables in two different ways. First, it measures whether or not the variables included in the analysis, when considered as a group, show a statistically significant relationship to each other. This is called the overall 'goodness of fit'. It is important to keep in mind when interpreting this method (and all other statistical methods that test the relationship between multiple variables) that there is a baseline standard measure that must be met for the overall relationship between variables to be considered significant. Typically this standard is either 90 or 95% agreement between variables. Once that baseline standard has been met, then the strength of the overall relationship of variables can be considered (for example, a .99 score shows a better fit than a .95 score).
If the overall model is determined to be significant, then the relationships of individual variables to one another are significant. What then becomes of importance is the strength of the relationship between variables. Negligible strength relationships between variables in a model that has passed tests of model fit are still not any good regardless if the model has a strong goodness of fit. Additionally, the model measures whether or not the variables are positively or negatively related to each other. For example, there is a strong positive relationship between higher levels of educational attainment and having a higher income. Conversely, there is a negative relationship between having a disability and being employed. However, it is important to remember that what is being measured is the strength of the relationship between the populations of interest and the measures of quality of life of the communities that they live in. So, also measured by the model is the impact of the relationship between populations on the relationship between any single population and a quality of life variable (multicollinearity). If the scores that measure the relationship between populations are too high (above .80), then the score that measures the relationship between individual populations and quality of life variables cannot be considered reliable. Fortunately, multicollinearity was not an issue for the populations of interest in the St. Louis Area District.
Bibliography
Forkenbrock, David J. and Lisa A. Schweitzer (1999). Environmental Justice in Transportation Planning. Journal of the American Planning Association. Vol. 65, No. 1
Purvis, Charles L. (2001). Data and Analysis Methods for Metropolitan-Level Environmental Justice Assessment. Transportation Research Record 1756, Paper No. 01-2907 Sustainability and Environmental Concerns in Transportation 2001
Rumberger, R. W. (1987). High School Dropouts: A Review of Issues and Evidence. Review of Educational Research 57:101-121.
Digest of Education Statistics (1998). Washington, D.C.: U.S. Dept. of Health, Education, and Welfare, Education Division, National Center for Education Statistics.
