An Assessment of Affordable Housing in West Virginia by Brian Lego RESEARCH PAPER 9735 1997-98 National Science Foundation Undergraduate Research Fellow Regional Research Institute P.O. Box 6825 West Virginia University Morgantown, WV 26506-6825 Mentor: Associate Professor Scott Loveridge, WVU Extension Service 404 Knapp Hall P.O. Box 6031 Morgantown, WV 26506-6031 Brian Lego, Division of Resource Management P.O. Box 6108 Morgantown, WV 26506-6108 blego@wvu.edu 304/293-4832 ext. 4457 Abstract: This research assesses affordable housing in West Virginia. Regression and χ 2 analysis are used to examine the relationship of demographic characteristics and geographical location to homeowner and home-renter cost burdens. Analysis shows that owner and renter cost burdens have different determinants. An analysis of affordable housing non-profit developers indicates these groups have made significant contributions to the supply of affordable housing, but current productivity is below national norms and residents in large geographic areas with substantial housing issues do not have access to services. Acknowledgments: I would like to acknowledge Lynn Talley of Community Works, Inc. and the housing directors who participated in this study s survey. Also, I would like to thank Scott Loveridge, Dennis Smith and Brian Cushing for their valuable suggestions and guidance. The author reserves responsibility for all errors.
Introduction Without decent housing, neither a family nor an individual can achieve the American dream. With this in mind, the National Housing Act of 1949 declared the general welfare and security of the Nation and the health and living standards of its people require a decent home and suitable living environment for every American family (Blair, 1995). That legislation built the foundation for today s federal subsidy and assistance programs. With real housing costs increasing at a faster rate than real income, the goal of safe and affordable housing remains a pipe dream for many American families, especially those in rural communities and central city neighborhoods. Steve Cocheo (1992) believes the issue of affordable housing has been studied to death, yet the challenge of providing it still remains. More importantly, housing affordability is normally considered an urban problem, whereas housing quality is the primary concern in rural areas. Federal legislation classifies a household as suffering an excess housing cost burden when it spends 30 percent or more of income on housing. In 1983, 19.6 percent of rural households were cost burdened compared with 25.3 percent of urban households. Additionally, 50.4 percent of poor rural households suffered excess cost burden, in contrast to 68.3 percent of poor urban households (Andrews & Jaffe, 1989). McDonald (1990) notes that in recent years, however, this situation has changed. Now almost twothirds of low-income rural homeowners spend 30 percent or more of their incomes on housing, and one-fourth spend 70 percent or more. Rural renters face an even greater affordability problem. Approximately four-fifths of lowincome rural renters spend at least 30 percent on housing and one-third spends 70 percent or more. Housing quality remains a problem in rural areas although the problem s magnitude is smaller than what was observed in previous years. Why should federal legislators concern themselves with rural communities since the majority of people in America live in urban areas? Rural America accounts for approximately one-fourth of the U.S. population, but 80 percent of the total land area (U.S. Bureau of the Census, 1992). To many, rural life is idyllic--thoughts of small family farms and towns where everyone knows your name are envisioned. These images may hold truth but problems associated with rural living set it apart from urban life. Rural people tend to have less educational attainment, higher rates of unemployment and poverty, and lower median household incomes when compared with the urban population. A lack of banking services reduces access to financing, providing opportunities to make housing safer and more affordable (Daniels et al, 1989). Rural areas tend to have a higher proportion of
3 homeowners, single-family detached units, mobile homes and units lacking complete plumbing and/or kitchen facilities (Hession et al, 1987). My study assesses the affordable housing situation in West Virginia. Initially, I examine the state s housing conditions and owner/renter affordability problems. Then I use econometric analysis to consider how selected householder demographic characteristics, geographic location and other factors relate to owner s or renter s housing affordability. Also, I look at differences in various housing quality conditions across geographic regions and residential classifications. Finally, I assess West Virginia s affordable housing non-profit sector based on survey data. West Virginia s Situation 1 West Virginia has some of the poorest economic conditions in the country, but has shown steady improvement over the past decade. According to 1990 Census data, West Virginia s disposable personal income per capita ranks 47 th in the nation and its percentage of population receiving public assistance income ranks in the top five. 2 This state s overall poverty rate of 19.6 percent is among the nation s highest. Figure 1 provides a spatial illustration of county-level median household income levels. Upon inspection, one realizes that a significant proportion of the state s counties have median household incomes below $19,000 per year. Additionally, there is regional clustering. High levels of unemployment, high poverty rates and low levels of income have long plagued two regions, the Southwestern and West-Central. This latter problem is captured in figure 1. Figure 1: Distribution of Median Household Income Levels 3 Dollars 12855-15607 15608-18739 18740-21255 21256-24372 24373-30941 As with other rural 4 areas, in West Virginia a county s economic well being correlates strongly with 1 All statistics used in the following section, unless noted otherwise, are from the 1990 Census of Population of Housing, Summary Tape File 3C, March 1993. 2 1994 Statistical Abstract of the United States, U.S. Bureau of the Census. Includes recipients of AFDC and SSI. 3 Class breaks on GIS maps created with natural class breaks option on ArcView GIS Version 3.0a. 4 West Virginia ranks as the second most rural state in the U.S., with approximately 65 percent of its population classified as rural. Because of its rural dominance, I consider the whole state as rural for my analysis unless a specific situation makes it useful to distinguish between rural and urban residency.
4 Figure 2a: Percentage of county housing stock lacking complete plumbing facilities Figure 2b: Percentage of county housing stock lacking complete kitchen facilities Percent of HousingStock 1-3 4-5 6-8 9-12 13-19 Percent of Housing Stock 1-2 3-4 5-6 7-8 9-15 the quality of its housing stock. According to the 1990 Census, 25,069 housing units (3.2%) lacked complete plumbing facilities while 18,276 (2.3%) lacked complete kitchen facilities. The counties that have the highest proportion of homes lacking complete plumbing and kitchen facilities also tend to have the lowest median household income levels. Inspecting these housing quality maps reveals a regional cluster. The West-Central region has the highest percentage of homes lacking plumbing and kitchen facilities. Webster and Pendleton counties share the distinction of having the highest percentage of their respective housing stocks lacking complete kitchen and plumbing facilities. Another great housing quality concern is the source from which households receive water. In 1990, 218,104 (28%) housing units used wells or other sources as their primary source of water. In many circumstances, water from wells has been found to be unsafe for human consumption due to significant levels of disease-causing pathogens and excessive levels of Iron particulates. Figure 3a: Percentage of county housing stock Figure 3b: Percentage of county housing stock constructed 1939 or earlier constructed 1940-1949 Percent of Housing Stock 10-15 16-21 22-27 28-34 35-46 Percent of Housing Stock 4-6 7-8 9-10 11-13 14-16 West Virginia s housing stock shows signs of aging. In 1990, 371,405 (47.5%) houses were built prior to 1960 and 185,500 (23.7% of the total) were built prior to or during 1939. Persons migrating to West Virginia searching for employment in the formerly booming coal and steel industries during the late 19 th and early 20 th centuries accounts for a large proportion of homes constructed in 1939 or earlier.
5 Household overcrowding used to be a major problem in West Virginia. In 1990, only 13,123 (1.3%) housing units contained more than one person per room, compared to 28,232 (4.1%) in 1980. Extreme overcrowding, defined as more than two persons per room, has also decreased. In 1990, only 0.05 percent of the occupied housing units had cases of extreme overcrowding. In addition to high frequencies of old, dilapidated and substandard houses, West Virginia s householders are experiencing their housing costs rise at faster rates than their incomes. This is also known as the housing costincome squeeze. For example, the median household spends 28.7 percent of its income on owner costs, just 1.3 percent below the federally mandated excess cost burden level. Renters suffer greater affordability problems; 37.2 percent of West Virginia s renters pay 30 percent or more income on rental costs. Figure 4 provides a spatial distribution of median gross rental costs a percentage of household income. Counties with the highest median gross rental rates typically have the lowest median household income levels, except for Monongalia County. This particular county contains a high percentage of college students who earn very little annual income, but must pay a high percentage of income for rent. Figure 4: Median Gross Rental Costs as a Percentage of Household Income Percent of Income 21-23 23-26 26-28 28-31 31-35 How do West Virginia s householders affordability problems compare to those in other states? Although only 15.4 percent of West Virginia s householders are classified as suffering excess cost burden, many low-income householders in West Virginia cannot afford to pay 30 percent of their income on housing costs. Median owner costs are $498 per month and median renter costs are $303 per month, amounts that are among the least expensive in the nation. California has the highest proportion of rural cost-burdened households, 31.9%, while New Jersey has the most expensive rural rental rates and rural median owner costs, $668 per month and $1,333 per month, respectively. Owner and renter costs encompass a wide array of costs in the monthly and yearly rate. Included in these costs are property taxes and home insurance; these costs are relatively high in states such as California and
6 New Jersey but are relatively low in West Virginia. Furthermore, West Virginia is a poor state whereas California and New Jersey rank in the top five of states with the highest per capita personal income. Figure 5a: Percentage of county households earning $10,000 or less and paying 30% or more of household income on owner costs Figure 5b: Percentage of county households earning $10,000-$19,999 and paying 30% or more of household income on renter costs Percent of Income Class 15.7-28.5 28.6-36.9 37-41.1 41.2-45.6 45.7-52.6 Percent of Income Class 8-11 12-14 15-17 18-21 22-32 Although it may appear that West Virginia householders do not have housing affordability problems, these cross-state comparisons do not reveal that a large percentage of West Virginia s counties have extreme cases of housing cost burden. For example, almost half of West Virginia s counties have at least 40 percent of their householders earning less than $10,000 per year and spending at least 30 percent of that income on owner costs. Figure 6a: Percentage of county households earning $10,000 or less and paying 30% or more of household income on renter costs Figure 6b: Percentage of county households earning $10,000-$19,999 and paying 30% or more of household income on renter costs Percent of Incom e Class 46-53 54-59 60-63 64-68 69-81 Percent of Incom e Class 5-11 12-20 21-27 28-36 37-47 Also, 95 percent of the counties have at least 30 percent of householders earning less than $10,000 in income spending 30 percent or more of income on owner costs. In most cases, renters face greater income constraints. Onefourth of West Virginia s counties have at least 68 percent of their renters earning $10,000 or less spending 30 percent or more of household income on renter costs. All counties in West Virginia contain at least 45 percent of renter-occupied households earning $10,000 or less paying 30 percent or more of household income on renter costs. Hession et al. (1987) note that rural areas contain a disproportionate amount of mobile homes in comparison to urbanized areas. Mobile homes account for nearly 15 percent of West Virginia s occupied housing
7 stock; the national average this housing characteristic is five percent. Six counties have 25 percent or more of their total housing stock classified as mobile homes and trailers. Four of these six counties are located in the southwestern portion of the state, an area often referred to as the coal fields. This high proportion of mobile home occupancy is simple to explain. Costs associated with owning a mobile home or trailer are significantly less than those of a comparably equipped single-family detached unit. Many of the families living in mobile homes cannot afford to pay 30 percent of their incomes on housing so they must live in this style of home to minimize ownership costs. Figure 7: Percentage of county housing stock classified as mobile home or trailers Percent of Housing Stock 4-10 11-16 17-20 21-25 26-33 The poor housing conditions found throughout West Virginia, in combination with its poor economic and social conditions, pose a challenge to families searching for affordable housing. There is hope for these families. In recent years West Virginia has witnessed the creation of several community-based organizations that see it as their responsibility to aid families in the process of finding affordable housing. Howell (1993) explains their roles succinctly: These corporations mushroomed in the late 1980s and early 1990s. Today, these elements represent a major resource for addressing the nation s critical affordable housing needs. The role of non-profit affordable housing organizations (NPOs) has become significant in West Virginia and other states. A later section of this paper assesses the activities and roles of NPOs in alleviating West Virginia s affordable housing problem. Methods This study uses two statistical tools to analyze geographic differences in housing characteristics as well as the relationship between selected demographic characteristics and owners and renters affordability problems in different regions. Data for this study come from the 1990 Census of Population and Housing Public Use Microdata
8 Sample, 5 percent sample. 5 Observations were limited to householders; housing and person weights were not applied in the data analysis (n=33,986). SAS Version 6.12 was used to perform the data analysis. Over the years, researchers have developed several methods to measure housing affordability. For this study, housing affordability will be measured using the percentage of income measure. For a detailed explanation of this and other methods strengths and weaknesses see Bogdon and Can (1997). Data for the percentage of income measure are most readily available and its results are relatively simple to understand. Classifications of the percentage of income measure will be based upon Bogdon and Can (1997): Households paying 30-49 percent of their income on housing have an excess cost burden while households paying 50 percent or greater are defined as paying severe cost burdens. Chi-square analysis is the first tool used in this study. It will aid in examining regional differences in housing quality characteristics and relationships between owner/renter costs and householder demographics. The other tool used in this study is an OLS estimation of two regression models. The dependent variables are owner s affordability and renter s affordability, respectively. Since each variable is originally expressed in different units, the parameters will be expressed as standardized beta coefficients. The first model, which tests owner affordability problems, is: Owner Cost = f (Age, Race, Family Type, Place of Birth, Year House is Built, Mobility Status, Work Status in Previous Year, Years of Education, Military Service, # of Bedrooms, # of Automobiles available, # of Elderly in Household, Geographic Region, Household Income, Residential Classification, Travel time to work, # of Related Children in Household, Value of Property, error) The second model, which tests renter affordability problems, uses the same explanatory variables as the first model, although it omits two variables: the year house is built and the property value. Three explanatory variables, family type, geographic region and residential classification, are coded as a series of dummy variables in each model. For family type a series of three dummy variables are used to represent married households, single-male households and single-female households. All family types are expected to have negative signs in the owner cost model while single-female households are hypothesized to have positive signs in the renter cost model. 5 The map displaying the nine individual PUMAs and their relative locations can be found in the 1990 PUMS Technical Documentation.
9 According to Muth and Goodman (1989), younger and older households, as defined by the age of the head, tend to spend larger amounts on housing than do households with middle-age heads. Based on this observation, age was determined to be a non-linear variable when modeling a householder s costs. Thus, this study codes age as a quadratic variable. Previous research on age s effects on housing affordability suggests that age will have negative estimate for the renter and owner cost models. Affordable housing literature suggests that minorities have greater affordability problems, particularly in the case of renter affordability. Race was coded as a dummy variable with 1 for white householders and 0 for nonwhite householders. The expected sign is negative for both models, which shows that minorities spend a greater proportion of income on housing. The author is not aware of previous studies of affordable housing which included the place of birth variable, but researchers who look at the mobility status of householders have indirectly examined its importance. This study examines this variable s effects because it is hypothesized that householders who are born in their state of residence will have fewer cases of cost burden than those who moved here from a different state will. Movers are searching for new employment and/or a better quality-of-life, but they initially incur higher housing costs. The place-of-birth residence is coded 1 for householders born in West Virginia and 0 for those who were not. The expected sign is negative for both models. Muth and Goodman (1989) note that moving is of special interest. The costs associated with moving are so large that moves are made only in rare circumstances, such as changing employment or the nature of the household due to death or divorce. Moreover, if the household plans to remain in the unit for some time, price and income effects assume more importance in the household s mobility calculation. In other words, householders are believed to remain in their current residence if they perceive their long-run housing costs to decrease relative to their long-run income levels. The mobility decision variable s importance in estimating housing affordability has been understood but has been left out of analyses of affordable housing. The mobility variable used in this study examines whether a householder lived in his current residence five years earlier. The variable was coded as 1 for householders living in the same residence and 0 for those who have moved in the past five years. The hypothesized sign for this explanatory variable is negative for both models. Another variable not previously investigated is the year in which a house is built. Including this variable as an explanatory variable can be easily explained. Older homes generally cost less to own when compared to newer
10 ones. This variable is omitted from the renter cost model since the age of a rental unit has little to do with its price, except for newly constructed apartment complexes. The expected sign of the year house is built variable is negative. Employment status and educational attainment were included because their effects on housing affordability are easy to comprehend. If a householder is unemployed, he/she will have difficulty paying housing costs since not having a job usually means not getting a paycheck. Thus, any housing costs borne by the owner or renter absorb a large percentage of his/her income. A person s educational attainment allows him/her the ability to find a better job and pay owner or renter costs with greater ease. Employment status was coded 1 for being employed the previous year (1989) and 0 for being unemployed in the previous year. These explanatory variables are expected to have negative coefficients in both models. Military service has not been examined as a possible cause of a householder s affordability problems. In this study, it is believed that persons join the military because they are poor and military service handicaps their ability to accumulate wealth. Military service was coded as a dummy variable, 1 for service in the military and 0 for no duty. The expected sign of this variable is negative for both models. The number of automobiles available to a household normally serves as a good indicator of that household s income; i.e. having more cars usually means having more money. Households that have only one or two vehicles do so not out of taste, but out of necessity. A poorer household that owns more vehicles would have less money available to pay for housing costs. Given this situation, this variable is expected to have a negative coefficient. Household income is probably one of the most important explanatory variables in a study of affordable housing. Its relationship to housing costs is evident and quite easy to understand. According to Muth (1989), in virtually any set of cross sectional data for individual consumer units, the fraction of income spent on housing units would appear to decline as income increases. For this study the hypothesized sign of the income variable is negative for both models. Residential classification has not been directly addressed in previous literature since most researchers believe affordable housing to be a metropolitan (urban) problem. This study seeks to reveal the extent to which rural households are more cost-burdened than urban households. As previously mentioned, the residential classification variable will be coded as a series of dummies. The first dummy variable, rural residency, will be coded 1 for rural non-farm householders and 0 otherwise. The second dummy variable, urban residency, will be
11 coded 1 for urban householders and 0 otherwise. These variables are expected to be positive in the owner and renter cost models. Travel time to work is rarely mentioned in affordable housing literature. Very little work using this specific variable has been performed because there is a lack of data. The PUMS data include a specific travel time to work variable where a householder s commuting time to work is provided. For this study, low-income householders are assumed to live a significant distance from their jobs in town in order to minimize their housing costs, which would otherwise be high if they lived within city limits (because of property taxes, sewage, etc). Middle-income householders locate close to their workplaces while high-income householders live a significant distance from their workplace because they dislike congestion and they can also afford the extra expense on commuting costs. This variable, just as age, has a non-linear relationship with a householder s cost burden. The travel time to work variable is coded as a quadratic variable. Travel time to work is expected to have a negative sign in both models in both models. The presence of children in a household changes the way in which a person perceives potential housing. A family with children that is searching for a house or apartment must find larger and safer units to accommodate its needs. In some situations this might force a householder to spend too much of his/her income on housing costs merely to meet the needs of his/her family. Researchers recognize this variable s importance but it has not been included in previous empirical research. For this study, the number of related children in a household is expected to have a positive sign for the owner and renter cost model, representing that a household with more children spends a greater proportion of income on housing. The number of bedrooms in a household usually predicts its price. The more rooms in a house or apartment the more it will cost. For a homeowner, if his/her house contains a large amount of bedrooms, they will ultimately spend more than if they purchased a home with fewer bedrooms. Apartments, however, pose a different situation altogether. Apartments allow unrelated individuals to pool their money together and pay an even share of an apartment that would be otherwise very unaffordable. For example, a $400 apartment would account for 40% of a person s income if he/she earned $1000 per month. Splitting the cost with another person would cut the rental rate in half and make the apartment more affordable. In the owner and renter cost models, the hypothesized signs are positive and negative, respectively.
12 Property value increases the amount a householder pays for his/her home; when a house and its surrounding property are assessed, the county assessor estimates the value of the property and the amount of property taxes a householder must pay is based upon this assessed property value. It is evident that the higher a property s value the higher its property taxes. These property taxes are included in the calculation of owner s monthly household costs. This variable has not been included in previous work, but its relevance to increasing the cost burden has been recognized by most researchers. The property value variable will only be included in the owner cost model since renters do not directly pay property taxes. Property value is hypothesized to have a positive sign. Results χ 2 Analysis The first set of results from this study were generated using χ 2 contingency tables of householder demographic variables, geographic location, household characteristics, and residential classifications (i.e. urban vs. rural). Housing characteristics were examined first to determine the extent to which they differ across different regions in West Virginia. Contingency tables of plumbing and kitchen facilities revealed that overall there was an even geographic distribution of homes where the householder reported having incomplete kitchen and plumbing facilities. The Northern Panhandle region, however, had a low percentage of homes in the sample with incomplete kitchen and plumbing facilities. χ 2 statistics were large because of the large sample size; the test of association, Cramer s V, for these household characteristics was low 0.055 for plumbing facilities and 0.041 for kitchen facilities. Additional contingency tables of plumbing and kitchen facilities were constructed to examine their statistical relationships with the age of a house. Results determined that a large percentage of homes lacking plumbing and kitchen facilities were built before 1940. Geographic region was used as a control variable and the resulting proportions remained almost equal to the original tables. Again, χ 2 statistics were too large to interpret but Cramer s V was slightly larger. The subsequent chi-square tests were performed to look at the relationships of owner and renter costs against various householder demographics. An owner cost-income contingency table was constructed to test Muth s (1989) assumptions. It was determined that the inverse relationship exists, even when the geographic region variable was used as a control variable. An overwhelming majority (95%) of excess cost-burdened owners and
13 χ 2 Analysis Results: Table 1. Housing Characteristics CONTINGENCY TABLE χ2 STATISTIC CRAMER S V Geographic Region vs. Plumbing Facilities 144.596*** 0.065 Geographic Region vs. Kitchen Facilities 61.018*** 0.042 Kitchen Facilities vs. Year Built 28.648*** 0.029 Plumbing Facilities vs. Year Built 51.916*** 0.039 ***P-value: 0.001 renters had annual household incomes less than $15,000 while 85 percent of severely cost-burdened householders (owners and renters) had household incomes less than $10,000. Residential classification was examined in a contingency table with the owner and renter cost variables. Results show that the highest percentages of excessive and severe cost-burdened householders are classified as rural non-farm residents. It appears, however, rural renters experience greater cost burdens than do rural homeowners. Gender and race are applicable variables even though a large percentage of household heads in West Virginia are male and white. For gender versus owner cost, female householders are twice as likely as male householders to have a low cost burden (1-10% of income on owner costs) but were twice as likely to have a severe cost burden. Nearly identical results were found when geographic region was introduced as a control variable. As for renter cost burden for males and females, the previous outcome did not occur. Even with geographic region as a control variable, women were always more likely to have affordability problems, particularly severe cost-burden. More specifically, Female renters were found to be 3.5 times more likely to have a severe housing cost burden. Why are there more instances of females having severe and excess cost burdens than males in the same income classes and geographic regions? While the answer might seem relatively straightforward to many, it is not easy to explain because many gender inequalities are at work here. One possible explanation is that many female householders in West Virginia are single mothers and one child s presence in a household decreases the amount of money they can use to spend on housing costs. Many of these single mothers must live in apartments with rents that are too expensive for their relatively low incomes, which accounts for the fact that women appear to have more housing affordability problems.
14 Table 2. Householder Demographics CONTINGENCY TABLE χ2 STATISTIC CRAMER S V Household Income vs. Owner Cost 10389.535*** 0.276 Household Income vs. Renter Cost 7553.628*** 0.236 Residential Classification vs. Owner Cost 594.755*** 0.094 Residential Classification vs. Renter Cost 870.727*** 0.113 Gender vs. Owner Cost 1819.738*** 0.231 Gender vs. Renter Cost 1598.957*** 0.217 Race vs. Owner Cost 241.114*** 0.084 Race vs. Renter Cost 173.354*** 0.071 Household Income vs. Commuting Time 7531.787*** 0.192 ***P-value: 0.001 Race displayed different characteristics than gender when it was compared to owner and renter costs. When race versus owner cost burden was examined, white and non-white householders were found to be equally as likely to have excess and severe cost burdens. In examining renter cost burden, however, non-white renters were found to be twice as likely as white renters to have a severe cost-burden. Both relationships maintained similar properties when geographic region was incorporated as a control variable. Regression Analysis Regression model estimation for this study was done using OLS estimation of two classical linear regression models. If the geographic regions represented in the data were not delineated properly to where they would have unlike population proportions then a WLS estimation would have been necessary to avoid heteroskedasticity, caused by large variations in populations across regions, resulting in larger error variances in smaller regions. Standardized Beta Coefficients and t-statistics for the owner and renter cost models are given in tables later in this section. The regression models in this study contain a lot of unexplained variance. The owner cost model has an adjusted-r 2 value of.1778 and the renter cost model adjusted-r 2 is.2810. The F-tests for both models are significant at the 0.0001 level, as are the F-tests for joint significance of the two dummy variable series, geographic region and family type.
15 Regression Results. Table 3: Householder Cost Models OWNER COST MODEL RENTER COST MODEL VARIABLE COEFFICIENT T-STATISTIC SIGNIFICANCE VARIABLE COEFFICIENT T-STATISTIC SIGNIFICANCE Intercept 14.40385 12.499 0.0001 Intercept 7.17702 39.947 0.0001 Age -0.03260-3.257 0.0011 Age -0.18536-20.073 0.0001 Race -0.02434-4.843 0.0001 Race -0.02597-5.537 0.0001 Married -0.03384-4.512 0.0001 Married -0.10091-14.409 0.0001 Single-Male -0.00831-1.571 0.1163 Single-Male -0.01960-3.963 0.0001 Single Female -0.03369-5.701 0.0001 Single-Female 0.07771 14.066 0.0001 Birthplace -0.01014-1.985 0.0473 Birthplace -0.03007-6.295 0.0001 Year Built -0.04424-8.198 0.0001 Mobility -0.23020-44.317 0.0001 Mobility 0.06518 11.464 0.0001 Work Status -0.07270-10.964 0.0001 Work Status -0.03728-5.257 0.0001 Years School -0.02760-4.934 0.0001 Years School -0.06355-10.403 0.0001 Military 0.02136 4.254 0.0001 Military -0.03539-6.592 0.0001 # Bedrooms -0.15139-28.789 0.0001 # Bedrooms 0.08838 14.784 0.0001 # Autos -0.11237-18.760 0.0001 # Autos -0.02900-4.492 0.0001 # of Elderly 0.00541 0.717 0.4736 # of Elderly -0.09248-11.434 0.0001 Income -0.11477-19.905 0.0001 Income -0.33713-53.032 0.0001 Rural -0.03533-1.726 0.0843 Rural 0.07906 3.604 0.0003 Urban 0.07438 3.602 0.0003 Urban 0.04121 1.864 0.0623 Travel Time -0.01950-4.115 0.0001 Travel Time -0.00293-0.578 0.5631 # of Children 0.06444 11.415 0.0001 # of Children 0.05422 8.959 0.0001 Property Value 0.37130 56.355 0.0001 All variables exhibit the hypothesized coefficient values in the owner cost model. Race, year house is built, work status in previous year, years of schooling, military service, number of automobiles, number of elderly in household, household income, age, travel time to work, and householder s place of birth are negative and significant. Positive significant estimates are: rural residency, mobility status, number of related children in household, property value and number of bedrooms. Additionally, a few dummy variables exhibit individual significance. In the geographic region series of dummies, puma2, puma6 and puma9 have positive and individual significance. For the family type dummy series, married households and female households are negative and significant. Insignificant explanatory variables are presence of elderly in household, urban residency, and travel time to work. Each insignificant variable exhibits the hypothesized sign.
16 In the renter cost model, many variables exhibit the hypothesized sign values. Age, race, place of birth, work status in previous year, household income, travel time to work, number of bedrooms, years of schooling, mobility status are negative and significant. Positive and significant explanatory variables are the number of related Table 4: Geographic Region Dummy Series OWNER COST MODEL RENTER COST MODEL VARIABLE COEFFICIENT T-STATISTIC SIGNIFICANCE COEFFICIENT T-STATISTIC SIGNIFICANCE PUMA1 0.00793 1.323 0.1860-0.00265-0.475 0.6344 PUMA2 0.02548 3.913 0.0001 0.02421-3.985 0.0001 PUMA3 0.00089 0.148 0.8825 0.01325 2.342 0.0192 PUMA4 0.00203 0.342 0.7322 0.01030 1.859 0.0630 PUMA5-0.00902-1.442 0.1494 0.00296 0.506 0.6126 PUMA6 0.01993 3.273 0.0011 0.01285 2.259 0.0239 PUMA8-0.00711-1.096 0.2733 0.02716 4.476 0.0001 PUMA9 0.01999 3.055 0.0023 0.00723 1.187 0.2353 children in a household and the single-female householder dummy variable. In the geographic region dummy series, puma2, puma3, puma6, and puma8 exhibit individual significance and are positive. Each family type dummy exhibits individual significance. Married and single-male households are negative and significant while singlefemale households are positive and significant. A few variables contradict the hypothesized results. From the owner cost to renter cost model, rural residency became negative and insignificant whereas military service became positive but retained significance. Many researchers have argued that urban renters are more likely to suffer excess and extreme cost burden when compared to rural renters. Even though the variable did not behave as hypothesized, previous research supports my results. The military service variable s outcome cannot be fully explained at this time. A possible reason military service increases renter costs (or positive coefficient) is because those who serve live in apartments are single and usually end up paying a large amount of income on housing costs. Since military Table 5: Joint Tests of Significance for Dummy Series DUMMY SERIES OWNER COST F-STATISTIC SIGNIFICANCE RENTER COST F-STATISTIC SIGNIFICANCE Geographic Region 7.0449 0.0001 5.0693 0.0001 Family Type 12.2433 0.0001 259.9991 0.0001 service inhibits wealth accumulation, those unmarried householders serving in the military must occupy an apartment with other persons or end up paying a large proportion of income on housing. Number of elderly in a
17 household is the only insignificant ordinal variable in the renter cost model, but its coefficient followed the original hypothesis. As mentioned earlier, householder demographic characteristics account for many differences in the owner and renter cost models. Age and travel time to work were significant in the renter cost model whereas number of elderly persons in the household is significant only in the owner cost model. The age and travel time variables become increasingly significant for different reasons. In the case of apartments, Muth (1989) notes the age variable can be partly explained by life cycle income effects. The money income receipts of younger workers tend to increase more rapidly than do those of middle-aged and older workers. Consequently, since households usually remain in a housing unit for some time since moving costs are so high, average income over the expected period of occupancy is larger given the current income for a younger household. Younger households also typically live in their current residence for a shorter period of time. If the size of a dwelling tends to decline relative to the optimal level determined by current income over the period of occupancy of dwelling, younger households will tend to spend more on housing given their current income. Since income levels of younger households are lower, they must spend a larger percentage of their incomes on housing, regardless of cost. As householders age and receive gradual increases in income, the percentage of earnings spent on housing generally decreases. Older householders encounter a different situation. Householders approaching retirement age have a lower expected period of occupancy of any given dwelling when compared to younger householders. In effect, the price of housing is smaller for older households for this reason, and they consequently spend more on housing at current income levels. Travel (or commuting) costs affect the amount of money a person is able to spend on an apartment. If a poorer renter decides to live a long distance from his/her workplace, then he/she must decide carefully so that the transportation costs do not significantly decrease the amount of income available for rental costs, and child care costs for households with children. Goodman (1989) notes that a householder must know which (if either) worker is considered the primary worker, whether the spouse started working after locating in the new house, and whether the parents anticipate changes in family size. In other words, housing location involves optimizing an intertemporal utility function with constraints that might include more than one income, more than one set of commuting costs, two sets of leisure time and child care costs. The household location decision has a small effect on wealthier households simply because they have more money and wealth to invest in housing and they might maximize their utility functions by paying a large proportion of income on housing.
18 The Non-Profit Affordable Housing Sector Bratt et al. (1994) summarize the non-profit sector s importance in providing affordable housing to those who need it. They write: Having established legitimacy in national legislation, and with significant support from state and local governments, foundations, and the private sector, the spotlight is on the non-profit sector to deliver. Such delivery requires both an ability to sustain the production of affordable housing units and a capacity to manage the stock once built. If non-profits can fulfill constituents and supporters expectations, they can capture the momentum for continued growth and thereby become a permanent force in the provision of the nation s affordable housing. If not, the fledgling non-profit industry may be at risk. The opportunities are great but so are the challenges (pg. 24). This quote epitomizes the role of West Virginia s non-profit affordable housing sector. Although their parent community development corporations are established in age and success, many of the housing developers are new to the game. They face many challenges, but they also have many opportunities. The majority of affordable housing is developed by NPOs (Bratt et al., 1994). Howell (1993) reports that nationally, NPOs produced between 23,000 and 30,000 housing units each year for the past several years. Nearly 50% of this production, however, is being produced by 20 to 25 organizations in large metropolitan areas. He also notes that the majority of NPOs produce 25 to 50 housing units each year and another significant percentage produce less than 25 annually. Howell (1993) places NPO affordable housing development organizations into six different classes, based on their size, age, organizational structure, functions and activities. They are: 1) Grassroots, community-based development corporations: similar to CHDOs (community housing development organizations); distinct neighborhoods and focus; activities other than housing production. Board of directors include large numbers of community representatives; further classified by age a) Mature organizations: at least five years in existence and more than 50 housing units per year. Staff includes five to seven full-time equivalents. Operating budgets between $250,00- $300,000 and ¼ income comes from development fees, ½ from government grants and contracts, ¼ from private sources. b) Emerging CDCs: at least 3-5 years existence; produce between 25-30 units per year. Staff of three full-time equivalents. Operating budgets of $150,000-$175,000/ year. c) Start-up Organizations: less than 3 years old and usually have small staffs (typically less than 3); budgets less than $100,000/year and produce less than ten units per year. 2) Citywide, housing development organizations: main focus is housing production; boards include community members with leadership coming from business sector. Involved in housing management and take on large projects. Large and sophisticated with operating budgets exceeding $1 million per year. Staffs include persons with considerable housing expertise who take business approach to development. 3) Local government spin-offs: creations of local and state governments, receive most support (financial and employed) from government sources. Most boards appointed by municipalities. 4) Corporate nonprofit CDCs: started by large corporations; substantial access to equity and tend to function like businesses. Boards include significant amount of community members.
19 5) Special-needs groups: provide specialized housing for elderly, homeless, etc.; Use HUD Section 202 program. 6) Developer-sponsored nonprofit development corporations: emerged in response to National Affordable Housing Act of 1990. Bratt et al. (1994), however, recognize that some NPOs do not exactly fit these criteria. They note that there are a variety of organizations that are not community-based that serve a sizeable geographic area, which are important producers in some locales. Howell s classification system and Bratt et al. s exception will aid in identifying important characteristics of West Virginia s non-profit affordable housing sector. Survey Discussion and Results In order to gain a better understanding of their successes, failures, opportunities and challenges in providing affordable housing to the citizens of West Virginia, this study reports the results of a telephone survey of NPO housing directors in West Virginia. Many NPOs are an integral part of community development corporations (CDCs), while others are totally independent organizations. The survey was developed under the supervision of Associate Professor Scott Loveridge; additional input was received from Professor Dennis K. Smith, Resource Management, and Ms. Lynn Talley, Executive Director of Community Works, Incorporated. A copy of the survey can be requested from this paper s author. follows: The survey examines different aspects of NPOs and can be broken down into nine different segments, as 1) General information about the survey respondent s affiliation with the NPO as well as the organization s tenure in providing affordable housing 2) Challenges and opportunities in affordable housing 3) General description of geographic region and its housing conditions 4) Production numbers from last fiscal year and estimated production for coming FY; Also investigated volume of financing (i.e. loans, grants, etc.) handled in previous FY 5) Sources of financing federal, state and local level; strengths and weaknesses of each; level of difficulty in receiving financial assistance for clients and cooperation of financing organizations; 6) Organizational structure information; number of employed staff and volunteers; Operational funding sources; Board-of-Director composition 7) Marketing services utilized their overall effectiveness 8) Affordable housing services provided to potential clientele 9) Support received from government agencies, individuals and other sources The survey targeted the housing directors of the thirteen NPOs. If respondents were constrained for time and unable to participate in the survey over the phone, a copy was sent by facsimile to their offices. Two respondents could not be reached by either phone or fax while four other housing directors failed to reply to telephone messages and a
20 facsimile version of the survey. The response rate was 54% (n=7). Statistical analysis will not be applicable to the results due to small sample size, but comparisons can be made between the responding NPOs. Survey results reveal that many of West Virginia s affordable housing nonprofit organizations do not match Howell s classification system. This indicates that Howell s system may contain an urban/metropolitan bias. Many of West Virginia s NPOs not only serve a large geographic region, in some cases three counties, but they also employ very small staffs. The average staff size for each member of the sample is five full-time equivalents. Also, every respondent surveyed replied that his/her organization s board of directors was composed of a significant number of local community members. The other B.O.D members for many of these organizations represented local banks, churches and construction companies. Many of these developers also have longevity; the survey sample averaged twenty years of existence. Many housing directors have served only an average of five years in his/her position. Their relatively short tenures can be partly explained in two ways: 1) As noted by Howell, the non-profit industry is relatively new in catering to those who need access to affordable housing, and 2) Directors and staff members can often become frustrated with the relatively low pay-high workload associated with their jobs. After they achieve success many are attracted to the higher paying private sector, resulting in a high turnover of NPO staffs. In terms of their major opportunities, most housing directors perceived the first-time homebuyer market as the most opportune. Additionally, respondents also believed that rehabilitation was a major opportunity for affordable housing given West Virginia s poor housing conditions. All respondents thought finding sources of production capital pose the greatest challenge, while several others also noted that problems with clientele credit histories create barriers to affordable housing financing. In describing housing conditions for their respective geographic regions, all housing directors characterized houses as being old, substandard and unaffordable to current householders or potential householders. Production numbers indicate that rehabilitation of single-family units is the primary activity for NPOs in West Virginia. On average, West Virginia s NPOs rehabilitate approximately five units per year. This is another case where West Virginia does not match Howell s classification criteria since their production numbers are much lower than national norms. In one case, the NPO was responsible for rehabilitating 16 single-family units in one year. Also, affordable housing developers facilitated the production of new single-family units, and the majority of these units were constructed on-site. No survey respondent reported rehabilitating multi-family units in the FY96.