How Do We Get There? The Impact of Transportation Costs on Affordable Housing. J. Eric Moore. Spring A paper submitted to the faculty of

Transcription

1 How Do We Get There? The Impact of Transportation Costs on Affordable Housing By J. Eric Moore Spring 2010 A paper submitted to the faculty of The University of North Carolina at Chapel Hill in partial fulfillment of the requirements for the degree Master of Public Administration This paper represents work done by a UNC-Chapel Hill Master of Public Administration student. It is not a formal report of the School of Government, nor is it the work of School of Government faculty. Executive Summary Recent research shows affordable housing coincides with increased transportation costs in major metropolitan areas. This study applies that research to explore housing and transportation affordability rates in rural and urban areas across North Carolina. Findings show the median household in North Carolina to be unaffordable before and after transportation costs are included. Statistics show population density has no impact on cost burdens. Geographic comparison shows a disparity between urban and rural areas in housing burden, a disparity that diminishes when transportation is added.

2 Background The United States defines affordable housing as housing consuming 30 percent or less of annual household income. 1 Theoretically, if housing costs rise above this threshold, residents cannot purchase enough other necessities like food, education, healthcare, etc. With that understanding, the US Department of Housing and Urban Development (HUD) administers multiple programs designed to address the need for affordable housing. Research suggests more attention should be given particularly to transportation costs when determining the affordability of housing. A Center for Housing Policy (CHP) study indicates a link between the cost of housing and that of transportation. Families whose housing consumes more than half of their income spend an additional 7.5 percent of household income on transportation. On the other end of the spectrum, families spending under 30 percent on housing dedicate nearly 24 percent of their income on transportation. Families between the 30 and 50 percent markers spend 12.3 percent on transportation. 2 These numbers strongly suggest a cost tradeoff exists in consumer choice between housing and transportation the less families spend on housing, the more they spend on transportation. A few policy institutes have attempted to take transportation costs into account when examining housing affordability. A Brookings Institute study suggests adding a threshold of 18 percent for transportation atop HUD s 30 percent for housing in determining affordability. 3 The Center for Transit-Oriented Development (CTOD) has established an affordability index, which combines housing and transportation costs into one value without setting an affordability threshold. CTOD found that in most cases, transit-rich environments have a positive effect on household disposable income. 4 CTOD s literature focuses on neighborhoods within major metropolitan areas and does not explore the realities of rural America, where per-capita income tends to be lower and transit options are often scarce. According to the U.S. Department of Agriculture, about 30 percent of North Carolina s population resides in rural areas, where workers earn 26 percent less than their urban counterparts. 5 This research will extend the investigation across North Carolina to determine if any tradeoffs between housing and transportation costs exist in this state, inside and outside rural areas. Research Question A report that combines current economic literature with real-world data, including transportation costs, could provide North Carolina policymakers with valuable insight into their housing policy decisions. To that end, this study seeks to address the question: Does a macro analysis reveal a disparity between rural and urban affordable housing rates in North Carolina when transportation costs are taken into account? Methodology This study relies on two types of analysis. First, a statistical analysis of relevant census-block data for North Carolina s 5,278 census block groups 6 evaluates the relationship between demographic factors, particularly population density as indicator of rurality, and the percent of household income spent on housing and transportation. The second uses QuantumGIS software to merge NC OneMap layers with expenditure data 7 from Geolytics, Inc. to show North Carolina s housing and transportation burden. Three maps result: (1) housing burden, (2) housing plus transportation burden, and (3) ratio of the two burdens. Limitations This research has three limitations. First, this study excludes a variety of factors that impact transportation and housing costs from a planning perspective, e.g. distance to job centers, walkability, transit access. The 1

3 studies from CTOD and CHP did make use of these variables. Gathering statewide data to generate these variables is beyond the scope of this project. Second, this study relies upon information compiled by a private company and cannot attest to the reliability of the Geolytics Inc. software s methodology in combining census and GIS data. Third, only 2008 data were available at the time this study began. Newer information may alter the results. Results The median household in North Carolina is unaffordable under the current HUD definition. In fact, 79 percent of the census block groups in the state have median cost burdens of above the 30 percent threshold. Moreover, the median NC home spends more than half of its income (52 percent) on housing and transportation. This figure exceeds the Brookings Institute threshold of 48 percent (30 percent housing, 18 percent transportation). Table 1: North Carolina Housing and Transportation Burden Housing Burden (% household income) Housing + Transportation Burden (% household income) Mean 35% 54% Median 34% 52% The high combined burden results from high housing costs rather than transportation costs, which consume 18 to 19 percent of household income. That is, high housing costs overwhelm any savings in transportation. In fact, by including transportation costs and using the 48 percent threshold, North Carolina becomes slightly more affordable. Where 79 percent of the block groups are unaffordable under the housing alone definition, the figure drops to 74 percent under the combined definition. See Appendices A and B for burden frequency distributions. Table 2: North Carolina Affordability under Different Definitions Affordable 21% Housing Burden (>30% Income) Housing Plus Transportation Burden (>48% Income) Affordable 26% Burdened 79% Burdened 74% Similarly, the maps depict the dire state of North Carolina s housing affordability (Appendix C) and its improvement under the combined definition (Appendix D). The change map (Appendix E) highlights the impact of including transportation costs. The affordability maps show areas clustered around urban 2

4 centers (particularly in the Piedmont) to be the least burdened areas in North Carolina. Moreover, these areas experience the greatest burden increase when transportation costs are included. The regression models (Table 3) evaluate the impact of demographic variables on housing and transportation burdens. The first model investigates impact on housing costs; the second model focuses on combined housing and transportation costs; the third model evaluates the ratio between the combined burden and the housing burden. Coefficients indicate tenth-percent changes in burden, e.g. 1 means for every one unit change in the independent variable, there is a corresponding increase of 0.1 percent of household income consumed. Positives and negatives indicate raised and lowered burdens, respectively. Zeros indicate no material impact. More asterisks (smaller p-value, higher statistical significance), indicate less likelihood the results are due to statistical randomness. Many variables exhibit high statistical significance and no material significance, e.g. 0*** ; the large sample size (N) causes the statistical software to detect minute patterns that have no real-world bearing on cost burdens. The high adjusted R 2 values show the models predict a substantial 75 percent of the variability observed in the three dependent variables. See Appendices F through I for full regression models and results. Table 3: Regression Analysis of Variables Contributing to Housing and Transportation Burden 8 Dependent Variable Model Independent Variable 9 Housing Burden F = 743*** R 2 =.73 Adj. R 2 =.73 Housing + Transport Burden F = 746*** R 2 =.73 Adj. R 2 =.73 Housing + Transport Housing F = 946*** R 2 =.77 Adj. R 2 =.77 Metropolitan Area? (Y/N) 2 3 2*** Population Change (%) Total Estimated Population 0 0** 0*** Population Density 0*** 0*** 0*** Median Age -1*** -1*** 1*** Median HH Income 0*** 0*** 0*** Proportion of Buildings with More than 20 Units 0*** 0*** Proportion of Population that is White 0** 0** 0*** Proportion of Population that is Black 0** 0** 0*** Proportion of Population that is Hispanic 0*** 0*** 0*** Proportion of Population with Above Average Education Proportion of Population with Below Average Education 0*** 0*** 0*** 0*** Proportion of Public Transportation Users 0*** 0*** 0*** Proportion of Population not in Labor Force 1*** 1*** 0*** 0*** 0*** 0*** 3

5 Proportion of Retired 0*** 0*** 0*** Mountains? (Y/N) *** Sandhills? (Y/N) 4 5-3*** Inner Coastal Plain? (Y/N) 4* 4-4*** Tidewater? (Y/N) 6** 6* -8*** Piedmont? (Y/N) excluded to avoid perfect correlation Notes: N = 5,278; Significance: * p.10, ** p.05, *** p.01 Results in Short: Urban-rural cost burden disparity exists in maps but not stats Geography has a slight impact on burden Higher unemployment coincides with higher burdens Older median age correlates with lower burdens Demographic factors have no material impact on cost burdens According to the regressions, rurality (population density) has no impact on cost burdens in any of the models. This suggests housing and transportation in rural areas consume about the same portion of household income as they do in urban areas. Alternatively, the regressions also defined rurality as being outside a Metropolitan Statistical Area (the Metropolitan area? variable). According to the housing and combined models, that distinction is not a significant predictor of cost burden. 10 This result conflicts with the GIS findings. Perhaps enough low-burden, low-density areas exist near urban centers to offset highburden, low-density areas elsewhere in the state. The regressions also examine the cost burden geography by regions. Each of the five primary regions 11 of North Carolina is coded by county and inputted into the model. 12 Tidewater was the only region to have statistical significance in both the housing and combined models, indicating there to be a slight increase in cost burden (0.6 percent) for the average people in that area. The results in the change model indicate a slight shift in burden from transportation to housing in all four regions on the periphery of the state (ranging from 0.3 percent in the Sandhills to 0.9 percent in the Mountains). In other words, people in these areas spend as a percentage slightly more of their income on housing and slightly less on transportation than those in the Piedmont. This result corroborates the findings in the two affordability maps. Very few of the variables in the model were both statistically and materially significant. Unsurprisingly, the higher the proportion of population not in the labor force, the higher the housing and transportation burden. For each ten percent rise above the national average for unemployment, there is a one percent rise in cost burden. In terms of age, for every ten years older the population grows, there is a corresponding 10 percent drop in cost burden. Material insignificance pervades many of the other demographic factors included in the regression. While housing density, race, education, transit use, and retirement 13 do have statistical significance, their zero coefficients indicate they do not matter in any tangible way. 4

6 Conclusion and Recommendations In referring back to the original question, Does a macro analysis reveal a disparity between rural and urban affordable housing rates in North Carolina when transportation costs are taken into account?, the answer seems to be yes. The descriptive statistics reveal the majority of median households across the state are already unaffordable under the current definition. While the situation improves slightly under the combined definition, a majority of households remain cost burdened. The regressions show population density to have no impact on the burden whether or not transportation is included. However, the maps do show the highest burdened areas to be in rural parts of the state. As a whole, the high housing burden across the state gives rise to a high combined burden. Transportation costs slightly equalize the urban and rural areas and make the state slightly more affordable. In short, the disparity exists in spite of transportation costs. The analyses show conflicting evidence of the trade-off reported by CHP. In the statistical models, rural households are just as likely as urban ones to be cost burdened. If there are different housing and transportation demands in these areas, their costs offset. The disparity between this study and CHP s may result from two different units of measure. CHP compared housing expenditures against transportation expenditures whereas this study compares both against population density and demographics. As a result, no trade-off pattern arose in the statistics. However, the affordability maps indicate a general correlation between rising burdens and increasing distances from urban centers. Perhaps population density divided into census block groups does not adequately capture the rural-urban spectrum in North Carolina. Should transportation costs be ignored in future housing policy decisions? That answer seems to be no. The macroscopic focus of this study potentially masks sub-regional expenditure patterns since it uses median values for each census block group that may offset one another over a large area. Future research should focus on smaller geographies and use a wider range of expenditure figures. Research should also incorporate planning-specific variables (like distance to job centers, urban centers) since common demographics had little material impact in this study. For now, local officials should investigate expenditure patterns within their jurisdictions when making appropriate policy decisions. Perhaps the solution to affordable housing and transportation lies in small-scale progress more so than statewide efforts. There may be lessons in those areas that saw the most and least change when transportation costs were included (Appendix C). The fact the average North Carolinian spends 18 percent of household income on transportation alone confirms the need for thoughtful decision-making as to what constitutes affordability. 5

7 Endnote 1 United States, "Affordable Housing," Department of Housing and Urban Development, Current as of 25 August 2009, accessed 10 October Lipman, Barbara J., "Something's Gotta Give," New Century Housing Vol 5, Iss 2, Pg. 17, Center for Housing Policy, accessed 10 October United States, "Calculating the True Cost of Housing," ResearchWorks, Vol. 5, No. 7, July/August 2008, Office of Policy Development and Research, Department of Housing and Urban Development, accessed 5 January This level is based on 2004 mean expenditure figures from the Department of Labor. 4 Center for Transit-Oriented Development, Center for Neighborhood Technology (Chicago, Ill.), & Brookings Institution, The affordability index: A new tool for measuring the true affordability of a housing choice, Market innovation brief, Jan Washington, D.C.: Brookings Institution, Metropolitan Policy Program, Urban Markets Initiative, accessed on 4 January United States, North Carolina Factsheet, Economic Research Service, Department of Agriculture, Last updated 9 December 2009, accessed 16 January Due to some missing data, one census block was not included in the descriptive statistics. 7 The dataset includes figures for median household income and median expenditures on transportation and housing for each NC census block group. 8 The unstandardized coefficients in this table have been adjusted by a factor of 1000 for clearer presentation. The original coefficients can be found in Appendices G-I. 9 None of the variables in the models suffer from multicollinearity. 10 This is not altogether surprising given that Metropolitan Statistical Areas are defined on a county basis. Thus there are likely to be portions of counties inside MSA s with sparser population densities than many groups in otherwise non-metropolitan areas. 11 Robinson, Peter J., North Carolina's Weather and Climate: Regional and Seasonal Summaries. Geography 111, Chapel Hill, NC: University of North Carolina at Chapel Hill, Spring 2008, Accessed on 10 December 2009 at 12 In order for the regression to function appropriately, one region Piedmont was excluded. Otherwise, the five regions would be perfectly correlated. 13 These are proportional variables, with a value of 100 equivalent to the national average. 6

8 Bibliography Carson, J., Johnson, D., & Steindel, C. (2006). Housing Costs in the CPI: What Are We Measuring? Business Economics. 41 (1), Center for Transit-Oriented Development, Center for Neighborhood Technology (Chicago, Ill.), & Brookings Institution. (2006). The affordability index: A new tool for measuring the true affordability of a housing choice. Market innovation brief, Jan Washington, D.C.: Brookings Institution, Metropolitan Policy Program, Urban Markets Initiative. Downs, A., & Godschalk, D. R. (1992). Growth management: Satan or savior? Journal of the American Planning Association. 58 (4). Estimates Professional 2008/2013. (2009). East Brunswick, NJ: Geolytics, Inc. CD-ROM. Lipman, Barbara. J. (2006). A heavy load: The combined housing and transportation burdens of working families. Washington, D.C.: Center for Housing Policy. Lipman, Barbara J. (2005). "Something's Gotta Give," New Century Housing Vol 5, Iss 2. Center for Housing Policy. Litman, T. (1996). Transportation cost analysis for sustainability. Victoria, BC: Victoria Transport Policy Institute. Malpezzi, S. (1998). Private rental housing markets in the United States. Netherland Journal of Housing and the Built Environment. 13 (3), Muellbauer, J. (2008). Housing, credit and consumer expenditure. Discussion papers, no London: Centre for Economic Policy Research. Nelson, A. C., Pendall, R., Dawkins, C. J., & Knaap, G. J. (2002). The link between growth management and housing affordability: The academic evidence. Washington, D.C.: Brookings Institution Center on Urban and Metropolitan Policy. NC Geographic Information Coordinating Council. (2009). Download/FTP. NC OneMap. Raleigh, NC: North Carolina Center for Geographic Information and Analysis. Robinson, Peter J. (2008). North Carolina's Weather and Climate: Regional and Seasonal Summaries. Geography 111. Chapel Hill, NC: University of North Carolina at Chapel Hill. United States. (1991). "Not in my back yard": Removing barriers to affordable housing : report to President Bush and Secretary Kemp. Washington: U.S. Dept. of Housing and Urban Development. United States. (2007). Housing wealth and consumer spending. Washington: Congress of the U.S., Congressional Budget Office. United States. (2008). "Calculating the True Cost of Housing." ResearchWorks, Vol. 5, No. 7. July/August Washington: Office of Policy Development and Research, U.S. Department of

9 Housing and Urban Development. United States. (2009). Affordable housing. Washington: U.S. Dept. of Housing and Urban Development. United States. (2009). North Carolina Factsheet. Washington: Economic Research Service, U.S. Department of Agriculture. 8

10 Appendix A Frequency Distribution of Housing Cost Burdens This chart shows the distribution of housing cost burdens across North Carolina. The y-axis indicates the number of block groups experiencing a given burden, while the x-axis shows the percent of household income consumed by housing (e.g equals 25 percent). All bars falling to the right of 0.30 (30 percent) are cost-burdened under HUD s definition. 9

11 Appendix B Frequency Distributions of Housing Plus Transportation Cost Burdens This chart shows the distribution of housing plus transportation cost burdens across North Carolina. The y-axis indicates the number of block groups experiencing a given burden, while the x-axis shows the percent of household income consumed by housing (e.g. 0.5 equals 50 percent). All bars falling to the right of 0.48 (48 percent) are cost-burdened the Brookings Institute s combined definition. 10

12 Appendix C Housing Cost Burden This map uses HUD's threshold for housing cost burden with darker areas being more cost burdened. White areas spend 30 percent or less of their household income on housing. (Expenditures that rounded up to 31 percent were classed as unburdened.) 11

13 Appendix D Housing Plus Transportation Cost Burden This map depicts the income burden of households when transportation costs are added to housing costs. HUD lists the threshold for housing costs to be burdensome at 30%. The Brookings Institute suggests an 18 percent threshold for transportation given mean U.S. expenditures. Therefore, white areas on this map spend 48% or less of household income on housing and transportation. Darker areas are more burdened. (Expenditures that round up to 49 percent were still classed as unburdened.) 12

14 Appendix E Change in Burden (Housing Plus Transportation / Housing) This map captures the degree of change in burden experienced when adding transportation costs to the housing costs of each census block group. The lightest areas saw the least change (minimum of 35 percent increase), and the darkest areas faced the largest change (maximum of 65 percent increase). 13

15 Appendix F Multivariate Regression Models C 1 = 1A + 2B + 3D + 4E + 5F + 6G + 7H + 8I + 9J + 10K + 11L + 12N + 13O + 14P + 15Q + 16R + 17S + 18T + C 2 = 1A + 2B + 3D + 4E + 5F + 6G + 7H + 8I + 9J + 10K + 11L + 12N + 13O + 14P + 15Q + 16R + 17S + 18T + C 3 = 1A + 2B + 3D + 4E + 5F + 6G + 7H + 8I + 9J + 10K + 11L + 12N + 13O + 14P + 15Q + 16R + 17S + 18T + Where: C 1 = Housing Cost Burden C 2 = Housing Plus Transportation Cost Burden C 3 = Change in Cost Burden A = Metropolitan Area? (defined as Metropolitan Statistical Area) B = Total Estimate Population D = Population Change (percent change, ) E = Population Density (person/square mile) F = Median Age G = Median Household Income H = Proportion of Buildings with More than 20 Units I = Proportion of Population that is White J = Proportion of Population that is Black K = Proportion of Population that is Hispanic L = Proportion of Population with Above Average Education M = Proportion of Population with Below Average Education N = Proportion of Public Transportation Users O = Proportion of Population not in Labor Force P = Proportion of Retired Q = Mountains? R = Sandhills? S = Inner Coastal Plains? T = Tidewater? 14

16 Appendix G Multivariate Regression 1: Housing Cost Burden Model Summary R R Square Adjusted R Square a Std. Error of the Estimate Model ANOVA Sum of Squares df Mean Square F Sig. Regression a 1 Residual Total Coefficients Unstandardized Standardized Coefficients Coefficients t Sig. B Std. Error Beta (Constant) Metropolitan Area? Total Estimated Population Pop Change (%) Population Density Median Age Median HH Income Proportion of Buildings with More than 20 Units Proportion of Population that is White Proportion of Population that is Black Proportion of Population that is Hispanic Proportion of Population with Above Average Education Proportion of Population with Below Average Education Proportion of Public Transportation Users Proportion of Population not in Labor Force Proportion of Retired Mountains? Sandhills? Inner Coastal Plain? Tidewater?

17 Appendix H Multivariate Regression 2: Housing Plus Transportation Cost Burden Model Summary R R Square Adjusted R Square a Std. Error of the Estimate Model ANOVA Sum of Squares df Mean Square F Sig. Regression a 2 Residual Total Coefficients Unstandardized Standardized Coefficients Coefficients t Sig. B Std. Error Beta (Constant) Metropolitan Area? Total Estimated Population Pop Change (%) Population Density Median Age Median HH Income Proportion of Buildings with More than 20 Units Proportion of Population that is White Proportion of Population that is Black Proportion of Population that is Hispanic Proportion of Population with Above Average Education Proportion of Population with Below Average Education Proportion of Public Transportation Users Proportion of Population not in Labor Force Proportion of Retired Mountains? Sandhills? Inner Coastal Plain? Tidewater?

18 Appendix I Multivariate Regression 3: Change in Burden (Housing Plus Transportation / Housing) Model Summary R R Square Adjusted R Square a Std. Error of the Estimate Model ANOVA Sum of Squares df Mean Square F Sig. Regression a 3 Residual Total Coefficients Unstandardized Standardized Coefficients Coefficients t Sig. B Std. Error Beta (Constant) Metropolitan Area? Total Estimated Population Pop Change (%) Population Density Median Age Median HH Income Proportion of Buildings with More than 20 Units Proportion of Population that is White Proportion of Population that is Black Proportion of Population that is Hispanic Proportion of Population with Above Average Education Proportion of Population with Below Average Education Proportion of Public Transportation Users Proportion of Population not in Labor Force Proportion of Retired Mountains? Sandhills? Inner Coastal Plain? Tidewater?

19 Acknowledgements I want to thank each members of my Capstone Committee: Dr. Karl Smith (chair), Dr. Maureen Berner, and Tyler Mulligan. Their guidance and expertise in the subject area and in statistical methodology was invaluable. I also want to thank members of the North Carolina Department of Commerce for their assistance in narrowing my field of inquiry, and I want to thank the data and GIS librarians at UNC- Chapel Hill s Davis Library, particularly Michele Hayslett for her help in data collection and Amanda Henley for her aid in getting started with GIS. 18