RURAL AND URBAN DIFFERENCES IN HOUSEHOLD WEALTH ACCUMULATION: WHAT ROLE DO DEMOGRAPHICS, WAGES AND PROPERTY VALUES PLAY?

Transcription

1 RURAL AND URBAN DIFFERENCES IN HOUSEHOLD WEALTH ACCUMULATION: WHAT ROLE DO DEMOGRAPHICS, WAGES AND PROPERTY VALUES PLAY? Alexander W. Marré U.S. Department of Agriculture Economic Research Service 1400 Independence Ave., SW Mail Stop 1800 Washington, DC Selected Paper to be presented at the 2014 Agricultural and Applied Economics Association Annual Meeting in Minneapolis, Minnesota, July The views expressed here are the author s and should not be attributed to the Economic Research Service or the U.S. Department of Agriculture.

2 1. Introduction This paper explores the dynamics of household wealth for residents in metropolitan (metro) and nonmetropolitan (nonmetro) areas in the United States during the 2000s. 1 An empirical model is used to estimate the drivers of differences in nonmetro and metro household wealth and wealth accumulation, with a focus on demographic and wage differences between nonmetro and metro areas. Such research has only recently garnered attention in the U.S. rural development literature and since people and households draw on wealth for economic well-being, particularly during hard times, the subject is critically related to economic prosperity and resiliency. Given the impacts of the last recession on home values, jobs and wages, a better understanding of the characteristics of nonmetro household wealth and what leads to its accumulation or depletion is greatly needed. Much of the existing nonmetro development research has focused on traditional topics of interest to economists -- such as income, employment and poverty -- that are flows. In contrast to these measures of economic activity, wealth is a stock of assets, net of liabilities, that generates flows that are of value to people or society as a whole (Pender et al., 2012). Households can use assets to support consumption during economic downturns and use their wealth to make investments or take on other risks that allow for increased asset accumulation over time when economic conditions are better, such as opening a new business, or paying for a college education or training. Household net worth also provides additional information about the economic status of a household not captured in income. While income is related to wealth, the correlation between the two can be quite low. Lastly, household net worth may be related to political capital and prestige in a community and as such can affect future economic opportunities. In the final

3 analysis, if nonmetro households have less net worth than their metro counterparts, then they have fewer opportunities to leverage their assets into new and successful enterprises. In this way, the wealth of nonmetro households is influenced by the economic prosperity of the places they live in. 2. Conceptual Framework In terms of the literature on household wealth, recent research on household net worth has focused on the wealth effects of the national recession beginning in Many homeowners saw the value of their homes decline while a decline in the stock market affected retirement savings and other investments. Using data from the Survey of Consumer Finances, the Federal Reserve estimated that median and mean household net worth declined by 38.8 percent and 14.7 percent, respectively, between 2007 and 2010 and that the median net worth for all families decreased from $126,400 to $77,300 (Bricker et al., 2012). Previous research has shown the important role that living in a rural area plays in conditioning and determining economic flows, such as income, employment and poverty. In terms of research focusing stocks rather than flows, Fisher and Weber (2004) used the PSID to analyze the effect of nonmetro residence on asset poverty and found that living in nonmetro counties increases the likelihood of asset poverty, even after controlling for demographic characteristics like age, race and education. Similarly, Marré and Pender (2013) showed that differences in net worth and assets are at least in part affected by living in nonmetropolitan areas. This echoes other literature that has shown a nonmetro effect on a wide variety of other outcomes (Weber et al., 2005).

4 Only recently has place and people-based wealth received much attention in the rural economic development literature, although the idea of using wealth as a framework for rural development research is not new (Pender, Weber, Johnson and Fannin, 2014; Pender, Marré and Reeder, 2012; Castle, 1998). In terms of recent research on rural household net worth, Marré and Pender (2013) used the Panel Study of Income Dynamics (PSID) to examine trends in household net worth between 2001 and 2009 for USDA-ERS defined resource regions and metropolitan status. A key finding from this study was that the geographic context of households can help explain changes in net worth. Marré (2014) also uses data on household net worth from the PSID to look for geographic differences in household net worth accumulation by controlling for residence in a nonmetro county. Living in a nonmetro household resulted in a less financial capital and home equity in Nonmetro differences in demographic factors and wages with metro areas explained all of the difference in financial capital and part of the difference in home equity. This study extends Marré (2009) by examining the role of adjacency to a metro area on differences in household net worth and by controlling for regional variation by Census regions. Marré and Pender (2013) hypothesize that adjacency to metropolitan areas could influence net worth in three ways. First, home prices in more remote rural areas were not as affected by the recent recession (Housing Assistance Council, 2011). Second, the farm economy fared well in the most recent recession relative to other industries and farms are more likely to be found in more remote rural areas (Henderson and Akers, 2009). Third, economic growth associated with shale oil and gas development is more likely to have occurred in more remote rural areas.

5 There are a number of factors that could potentially drive differences in nonmetro and metro net wealth: a demographic effect, a wage effect, and a property value effect. Demographic characteristics have been found to influence net worth. The life-cycle hypothesis of wealth accumulation suggests that people accumulate an increasing amount of wealth as they age until their retirement, whereupon they draw on savings for consumption during retirement. Educational attainment is also likely to increase net worth by increasing wages, improving financial literacy and increasing access to economic opportunities. Racial and ethnic differences in wealth accumulation have received attention in the literature, too (Oliver and Shapiro, 1995; Hurst, Ching Luoh and Stafford, 1998). In particular, African-American households overall have lower net worth than Caucasian households, even when controlling for other characteristics such as educational attainment. Gittleman and Wolff (2004) found that at least some of the difference can be attributed to differences in inheritances. While the relationship between income and wealth can sometimes be weak, the amount of income earned by a household can clearly affect its ability to make new investments (Keister and Moller, 2000). New growth theory suggests that the lower wages offered in nonmetro labor markets could be due to a lack of knowledge spillovers that metro areas have as a product of population density. Glaeser and Maré (2001) found that when nonmetro residents move to cities, they are able earn higher wages and wage growth. Therefore the difference in the overall wage structure between nonmetro and metro areas may lead to lower net worth for nonmetro households.

6 Property values are another potential source of differences in household net worth between nonmetro and metro areas. Home equity and the net value of any other form of real estate and farmland are probably the forms of net worth most tied to geography. For many households, home equity is the largest component of overall net worth. Home prices are generally lower in nonmetro areas relative to metro areas. During the most recent recession, nonmetro households were spared the worst of the burst of the housing bubble (Figure 1). On the other hand, as Marré and Pender (2013) note, farmland values, especially in the Corn Belt region grew substantially in the latter half of the 2000s. For nonmetro households that own farmland, high farmland values during the 2000s may have boosted net worth relative to metro households. Living in a nonmetro county adjacent to a metro county may also increase home and land prices. 3. Data and Empirical Strategy We use household-level data from the Panel Study of Income Dynamics (PSID) from the University of Michigan, one of three national household wealth surveys and the only one with complete, publicly-available metropolitan/nonmetropolitan identifiers. The PSID is a nationally representative, longitudinal data set begun in 1969 (Brown, Duncan and Stafford, 1996). As discussed in Marré (2014) and Marré and Pender (2013), the PSID s major strengths are that it provides the Beale Nonmetro-Metro Continuum Code for some survey years, which is particularly helpful for nonmetro researchers. Although the survey is relatively small compared to the other two national surveys of household net worth (the Survey of Consumer Finances and the Survey of Income and Program Participation), it does provide relatively good estimates of net worth and its major components (Curtin, Juster and Morgan, 1989).

7 Household heads are the unit of analysis in the PSID with the most data available. Therefore, household heads who responded to the survey in 1999 and 2009, two survey years for which Beale codes are available, were selected for the analysis. A comparison of the sample with data from the 2000 Census is shown in Table 1. The sample is a bit older and oversamples female, black and nonmetro household heads. The educational attainment of the household heads roughly matches the population as a whole. To explore differences in net worth by nonmetro and non-adjacent residency, a series of regressions is estimated for the change in four broad categories of household net worth between 1999 and 2009: financial capital, home equity, farm and business assets, and other property. Home equity and the net value of farm and business assets are provided in aggregate form in the PSID data set. The financial capital and other property categories were constructed for this analysis as the sum of various assets and liabilities. 1. Financial Capital: the sum of the value of checking and savings accounts, certificates of deposit, savings bonds, value of private annuities or individual retirement accounts, stocks, mutual funds, investment trusts, rights in estates, less credit card, student loan, or other forms of debt; 2. Home Equity: net value of primary home; 3. Farm and Business: net value of farm and/or business (including farmland); 4. Other Property: net value of any second homes, land, rental real estate plus the net value of vehicles, recreational vehicles, boats, etc. The strategy pursued in the following section is to introduce demographic variables, regional fixed effects and household wages to determine what role these factors play in explaining any

8 differences by nonmetro and non-adjacent residency. Indicator variables for two specific geographic variables of interest were included as regressors: nonmetropolitan residence and nonadjacency to a metropolitan area. As in Marré (2014), the nonmetropolitan variable indicates PSID household heads that lived in a nonmetropolitan county in each survey year between 1999 and The non-adjacency indicator variable for nonmetropolitan household heads indicates household heads that live in a county non-adjacent to a metropolitan county in Therefore, the non-adjacency variable flags households that are more remote from metro areas. In this way, the nonmetro indicator variable tests for differences between metro and nonmetro areas. When the non-adjacency variable is included with the nonmetro variable in the regressions, it tests for differences in the components of net worth between nonmetro/non-adjacent households and metro households. 4. Empirical Results As a starting point for this analysis, a series of simple regressions were estimated to look for nonmetro and metro differences in household net worth and its subcomponents (net financial capital, home equity, net value of farm and business assets, and other property). Results from these regressions are shown in Table 2. When total net worth is considered, the results suggest that there is no statistically significant difference between nonmetro households in general and metro households, and between metro households and the more remote non-adjacent nonmetro households. However, there were statistically significant differences across these categories for three of the four subcomponents of net worth. Similar to Marré (2014), these results show that nonmetro households have less financial capital, all else equal, but there was no statistically significant effect found for non-adjacency to a metropolitan area. The results also show less home equity for nonmetropolitan households and an additional penalty for remoteness that was

9 not captured in Marré s model. Finally, the results show that the net value of farm and business assets has more to do with non-adjacency and remoteness than nonmetropolitan residence. These results guide the following sets of regressions examining the potential determinants of these differences in household net worth in financial capital, home equity, and net farm and business assets. 4.1 Financial Capital Results Table 3 shows results from a series of three regression models. The first column, Model I, essentially replicates the findings from Table 2, except with the non-adjacency indicator variable removed. In Model II, a series of demographic variables are added as regressors as well as the net value of financial capital in 1999, called initial financial capital. Adding these variables shows that demographic factors play a significant role in explaining nonmetro and metro differences in household net worth. Life cycle effects, reflected by the age and age squared variables, take the expected signs. Similarly, there are significant differences in household financial capital between whites, blacks and Hispanics, as expected. Financial capital was estimated to be higher for household heads with college and/or postgraduate education, all else equal, while the only statistically significant effect from marital status was for household heads that became divorced between 1999 and In Model III, household wages are included as an explanatory variable. With the inclusion of the wage effect, the difference in nonmetro and metro household financial capital was eliminated, suggesting that any nonmetro effect on financial capital is through demographic and wage differences. Note, too, that the effects of age and education become statistically insignificant in Model III.

10 4.2 Home Equity Results Results from three models of home equity in 2009 are presented in Table 4. As with the case of financial capital, the first model shows again the negative effect of residence in a nonmetropolitan area and residence in a non-adjacent county on home equity. In Model II, both the nonmetro and remote residence effects become much smaller and statistically insignificant with the inclusion of demographic factors. Life-cycle, race and ethnicity variables have the expected signs, with home equity increasing substantially with educational attainment beyond high school, all else equal. The regional fixed effects variables show that relative to the Northeast, households in the South and North Central regions have lower home equity. In the West, where many respondents live in California, it is perhaps not surprising to see no statistically significant difference with the Northeast, given that home prices can be relatively high in both these regions. Adding household wages in Model III somewhat reduces the size of these regional fixed effects, suggesting that lower wages in the North Central and Southern regions relative to the North East and west explains, in part, lower home equity in these regions too. 4.3 Net Farm and Business Assets Results In Table 5, results from three models of the net value of farm and business assets in 2009 are reported. The positive and statistically significant effect of remoteness on the net value of farm and businesses persists across all three models, even with the inclusion of demographic and wage variables. In contrast to financial capital and home equity, the demographic effects such as age, education and marital status seem to play a less important role. In Model III, statistically significant and positive effects were found for residences in the North Central, South and West

11 regions relative to the Northeast. The coefficient associated with non-adjacency actually increases in Model III compared with Models I and II and bears more examination in future research. In sum, it appears that the non-adjacency effect on this type of net worth is better explained by factors outside these models. Perhaps regional factors, such as soil quality and access to markets, for example, play a larger role than demographic factors. 5. Conclusion This study has used publicly-available data from the PSID to analyze differences in household net worth and its components between nonmetro and metro households and non-adjacent households. Similar to Marré (2014), this analysis finds no statistically significant difference in overall household net worth between nonmetro and metro households overall and no statistically significant difference between nonmetro non-adjacent households and metro households. However, there does appear to be differences for nonmetro overall and non-adjacent nonmetro households, in particular in some of the subcomponents of household net worth, namely financial capital, home equity and the net value of farm and business assets. Additional models of financial capital, home equity and the net value of farm and business assets were estimated, successively adding in demographic variables and regional fixed effects, followed by household wages. The models show that the nonmetro difference in household financial capital is almost entirely explained by differences in wages between nonmetro and metro areas and other select demographic factors. The differences due to a residence in a nonmetro area or a nonmetro nonadjacent area in home equity was entirely explained by demographic and wages too. Finally, the non-adjacency indicator variable was found to be

12 statistically significant in the net value of farm and business assets model, while the nonmetro residence indicator was not. This seems to indicate that much of the assets captured in this category are for farms located in more remote areas. Successively adding in demographic regressors, regional fixed effects and wages increased the explanatory power of the model, but did not explain away the effect of non-adjacency, suggesting that geographic characteristics matter more for this type of net worth. The results show that demographic, geographic and structural differences in the labor market explain much, if not all, of the differences between nonmetro and metro household net worth. As such, it adds to the large body of research on nonmetro and metro differences in poverty rates, wages and employment by identifying the key factors driving these differences. From a policy standpoint, these results suggest that increasing educational attainment, attracting highly educated in-migrants to rural areas and finding strategies to improve nonmetro wages could decrease the difference in nonmetro and household net worth. References Bricker, J., A.B. Kennickell, K.B. Moore and J. Sabelhaus Changes in U.S. Family Finances from 2007 to 2010: Evidence from the Survey of Consumer Finances. Federal Reserve Bulletin, Vol. 98, No. 2, Board of Governors of the Federal Reserve System. Brown, C., G.J. Duncan and F.P. Stafford Data Watch: The Panel Study of Income Dynamics. Journal of Economic Perspectives 10(2): Castle, E.N A Conceptual Framework for the Study of Rural Places. American Journal of Agricultural Economics 80:

13 Curtin, R.T., F.T. Juster and J.N. Morgan Survey Estimates of Wealth: An Assessment of Quality. In R. Lipsey and H. Stone, eds., The Measurement of Saving, Investment, and Wealth. Chicago: University of Chicago Press. Economic Research Service Nonmetro-Metro Continuum Codes. Available online at: July. Fisher, M. and B.A. Weber Does Economic Vulnerability Depend on Place of Residence? Asset Poverty across Metropolitan and Nonmetropolitan Areas. Review of Regional Studies 34(2): Gittleman, M. and E.N. Wolff Racial Differences in Patterns of Wealth Accumulation. Journal of Human Resources 39(1): Glaeser, E.L. and D.C. Maré Cities and Skills. Journal of Labor Economics 19(2): Henderson, J. and M. Akers Recession Catches Rural America. Economic Review First Quarter: Federal Reserve Bank of Kansas City. Housing Assistance Council Foreclosure in Rural America: An Update. Rural Housing Research Note, March. Available at: Accessed November 15, Hurst, E., M. Ching Luoh and F.P. Stafford The Wealth Dynamics of American Families, Brookings Papers on Economic Activity 1: Irwin, E.G., A.M. Isserman, M. Kilkenny and M.D. Partridge A Century of Research on Rural Development and Regional Issues. American Journal of Agricultural Economics 92(2):

14 Isserman, A.M In the National Interest: Defining Nonmetro and Metro Correctly in Research and Public Policy. International Regional Science Review 28(4): Jenkins, S.P. and M. Jäntti Methods for Summarizing and Comparing Wealth Distributions. Institute for Social and Economic Research Working Paper No , University of Essex. Keister, L.A. and S. Moller Wealth Inequality in the United States. Annual Review of Sociology 26: Lerman, D.L. and J.J. Mikesell Rural and Urban Poverty: An Income/Net Worth Approach. Policy Studies Review 7: Lovenheim, M Housing Wealth and Higher Education: Building a Foundation for Economic Mobility. The Pew Charitable Trusts, December. Marré, A The Net Worth of Households: Is There a Rural Difference? In Rural Wealth Creation, J. Pender, T. Johnson, B. Weber and M. Fannin, eds., Routledge Press. Marré, A. and J. Pender The Distribution of Household Net Worth Within and Across Rural Areas: Are There Links to the Natural Resource Base? American Journal of Agricultural Economics 95(2): Oliver, M.L. and T.M. Shapiro Black Wealth / White Wealth. New York: Routledge. Panel Study of Income Dynamics, public use dataset Produced and distributed by the Survey Research Center, Institute for Social Research, University of Michigan, Ann Arbor, MI. Pender, J., B. Weber, T. Johnson and J.M. Fannin Rural Wealth Creation. Routledge Press.

15 Pender, J., A. Marré, and R. Reeder Rural Wealth Creation: Concepts, Strategies and Measures. ERR-131, U.S. Department of Agriculture, Economic Research Service. March. Weber, B., L. Jensen, K. Miller, J. Mosley and M. Fisher A Critical Review of Rural Poverty Literature: Is There Truly a Rural Effect? International Regional Science Review 28(4): Wolff, E.N Recent Trends in Household Wealth in the United States: Rising Debt and the Middle-Class Squeeze an Update to Working Paper No. 589, Levy Economics Institute of Bard College, March. 1 For a discussion of the various ways commonly used to delineate rural and urban geography, see Isserman (2005). 2 Somewhat surprisingly, the results are more robust with the non-adjacency indicator variable defined in this way, rather than indicating households that live in non-adjacent, nonmetropolitan counties in each survey year between 1999 and Future research should examine potential explanations for this phenomenon.

16 Figure 1: Housing Prices in Metropolitan and Nonmetropolitan Areas, Percent Change from Same Quarter in the Previous Year, Source: ERS calculations using Federal Housing Finance Agency s metropolitan area housing price indices and state-level nonmetropolitan price indices. These are combined into aggregate metropolitan and nonmetropolitan averages by weighting each area by its share of total owner-occupied housing units, taken from the Census Bureau s American Community Survey, using the five-year averages by county.

17 Table 1: Sample Household Head Characteristics PSID Sample (1999) Census (2000) Householders Total Population Median Age Female 24.4% 12.2% 50.9% Black 11.1% 7.8% 12.9% Hispanic 1 8.4% 7.0% 12.5% Education 2 Less than a High School Diploma High School Diploma or Equivalent Some College Bachelor s Degree Postgraduate 18.6% 30.5% 22.7% 16.3% 11.9% % 28.6% 27.4% 15.5% 8.9% Married % 51.7% 54.4% Nonmetropolitan Residence % 17.8% 17.4% Source: Author s estimates, using data from the Panel Study of Income Dynamics, University of Michigan, Institute for Social Research and 2000 Census, U.S. Census Bureau. The PSID sample size is N = 4,482 household heads. Notes: 1 The Hispanic variable comes from the 2005 PSID wave. 2 Age and educational attainment data in the 2000 Census is not available for householders. The figures in the total population column are for the population age 25 and older. 3 Census figures reported here for marital status are for the population age 15 and older. 4 The Census figures on nonmetropolitan residence use the 2003 metropolitan area definitions from the Office of Management and Budget.

18 Table 2: Differences in Net Worth and its Subcategories, 2009 dollars Net Worth Est. Coeff. (S.E.) Financial Capital Est. Coeff. (S.E.) Home Equity Est. Coeff. (S.E.) Farm and Business Est. Coeff. (S.E.) Other Property Est. Coeff. (S.E.) Nonmetro Household 64, (199,751.80) -69,178.48*** (15,034.10) -20,862.47*** (6,319.94) (16,328.93) 154, (196,731.10) Nonadjacent Household -139, (179,825.90) 6, (16,685.24) -15,536.04** (6,676.24) 53,682.96** (26,955.75) -184, (174,485.40) Constant 392,293.40*** (37,100.99) 130,060.30*** (10,795.34) 104,121.30*** (3,629.32) 57,796.31*** (10,289.38) 100,315.50*** (30,967.25) Notes: Asterisks indicate statistical significance: ***, ** and * correspond to 0.01, 0.05 and 0.1 levels of statistical significance, respectively. Standard errors are Huber-White heteroskedasticity robust. Source: Author s estimates, using data from the Panel Study of Income Dynamics, Institute for Social Research, University of Michigan

19 Table 3: Models of Household Financial Capital in 2009, 2009 dollars Model I Model II Model III Est. Coeff. (S.E.) Est. Coeff. (S.E.) Est. Coeff. (S.E.) Nonmetro 1-69,178.48*** (15,034.10) -19, (16,655.71) -9, (16,392.77) Nonadjacent 6, (16,685.24) 3, (19,586.02) 5, (19,493.29) Age 3,820.12** (1,890.10) 2, (1,857.92) Age Squared ** (21.27) (21.24) Female -7, (16,446.49) -8, (16,390.64) Black -34,691.88*** (12,365.67) -24,886.78** (12,058.25) Hispanic -52,437.69*** (18,691.95) -40,560.95** (18,777.16) High School Diploma 3, (8,576.58) -1, (8,579.36) Some College 13, (13,589.94) (13,735.76) College Degree 62,231.86** (27,501.74) 23, (27,181.07) Postgraduate 90,708.98** (44,502.79) 52, (43,101.79) Married 2 3, (18,099.27) -29, (18,538.48) Widowed 2 66, (78,286.46) 54, (79,183.77) Became Married 3-14, (19,256.09) -13, (19,052.80) Became Divorced 3-62,474.38* (32,082.12) -67,690.70** (32,002.66) Initial Financial Capital 0.92*** (0.18) 0.92*** (0.19) Household Wages *** (0.17) North Central 2-42,612.61** (20,250.25) -23, (19,031.15) South 2-23, (19,928.38) -6, (19,002.63) West 2-10, (24,287.99) 4, (23,069.99) Constant 130,060.30*** (10,795.34) -18, (48,091.01) -53, (48,961.43) Notes: Asterisks indicate statistical significance: ***, ** and * correspond to 0.01, 0.05 and 0.1 levels of statistical significance, respectively. Standard errors are Huber-White heteroskedasticity robust. The R 2 for Models II and III is 0.59, and 0.61, respectively. 1 Household resided in a nonmetropolitan county for all survey years between 1999 and Variable measured in Variable measured between 1999 and Wages are for household head and wife/partner in Source: Author s estimates, using data from the Panel Study of Income Dynamics, Institute for Social Research, University of Michigan

20 Table 4: Models of Household Home Equity in 2009, 2009 dollars Model I Model II Model III Est. Coeff. (S.E.) Est. Coeff. (S.E.) Est. Coeff. (S.E.) Nonmetro 1-20,862.47*** (6,319.94) -6, (4,509.76) -3, (4,636.47) Nonadjacent 2-15,536.04** (6,676.24) -3, (4,912.90) -3, (4,908.65) Age 1,917.47*** (698.07) 1,706.66** (700.97) Age Squared *** (7.31) ** (7.47) Female 1, (8,270.85) (8,319.40) Black -12,128.14** (5,089.44) -9,398.61* (5,216.19) Hispanic -35,916.68*** (11,149.94) -32,325.77*** (10,964.74) High School Diploma (4,077.72) -2, (4,142.39) Some College 9,718.93* (5,538.50) 5, (5,801.12) College Degree 35,696.00*** (10,211.91) 23,373.80** (11,662.66) Postgraduate 44,974.27*** (13,398.81) 33,112.06** (14,854.41) Married 3 22,257.85** (10,016.27) 11, (11,020.51) Widowed 3-13, (16,595.02) -16, (16,566.26) Became Married 4-3, (8,497.93) -4, (8,306.32) Became Divorced 4-34,080.21*** (11,146.36) -35,270.57*** (11,039.34) Initial Home Equity 1.08*** (0.16) 1.05*** (0.15) Household Wages ** (0.10) North Central 2-45,491.29*** (7,409.95) -39,296.07*** (7,562.67) South 2-26,009.37*** (7,592.57) -20,439.98*** (7,677.68) West 2 2, (11,241.82) 7, (11,246.17) Constant 104,121.30*** (3,629.32) 7, (20,186.00) -7, (20,226.49) Notes: Asterisks indicate statistical significance: ***, ** and * correspond to 0.01, 0.05 and 0.1 levels of statistical significance, respectively. Standard errors are Huber-White heteroskedasticity robust. The R 2 for Models II and III is 0.49 and 0.50, respectively. 1 Household resided in a nonmetropolitan county for all survey years between 1999 and 2009 or household. 2 Household resided in a nonadjacent county in Variable measured in Variable measured between 1999 and Wages are for household head and wife/partner in Source: Author s estimates, using data from the Panel Study of Income Dynamics, Institute for Social Research, University of Michigan.

21 Table 5: Models of Household Net Value of Farm and Business Assets in 2009, 2009 dollars Model I Model II Model III Est. Coeff. (S.E.) Est. Coeff. (S.E.) Est. Coeff. (S.E.) Nonmetro (16,328.93) 6, (15,064.29) 17, (16,131.35) Nonadjacent 1 53,682.96** (26,955.75) 46, (29,180.35) 48,504.53* (28,978.88) Age 5,126.13* (3,046.72) 4, (2,976.95) Age Squared * (27.90) (29.42) Female 26, (33,409.73) 26, (32,093.30) Black -61,224.84*** (13,566.28) -48,779.89*** (14,758.62) Hispanic -58,754.91*** (19,678.45) -43,983.46** (20,195.69) High School Diploma 28,118.75** (13,577.20) 21, (14,355.25) Some College 21, (15,512.49) 4, (18,919.59) College Degree 56,943.29* (30,275.81) 7, (38,047.01) Postgraduate 29, (34,386.73) -19, (50,000.37) Married 2 40,828.41* (24,344.91) (31,081.07) Widowed 2-88, (72,240.08) -98, (71,455.62) Became Married 3 33, (48,423.91) 35, (46,999.65) Became Divorced 3 30, (35,462.95) 23, (34,312.64) Initial Farm & Business Assets 0.28* (0.17) 0.27* (0.17) Household Wages (0.60) North Central 2 29, (25,893.99) 53,385.26** (24,141.63) South 2 27, (24,643.35) 49,519.10** (22,541.54) West 2 62,790.62* (37,121.81) 81,996.81** (36,326.13) Constant 57,796.31*** (10,289.38) -136, (90,340.19) -183,316.10* (95,114.42) Notes: Asterisks indicate statistical significance: ***, ** and * correspond to 0.01, 0.05 and 0.1 levels of statistical significance, respectively. Standard errors are Huber-White heteroskedasticity robust. The R 2 for Models II and III is 0.13, and 0.15, respectively. 1 Household resided in a nonmetropolitan county for all survey years between 1999 and Variable measured in Variable measured between 1999 and Wages are for household head and wife/partner in Source: Author s estimates, using data from the Panel Study of Income Dynamics, Institute for Social Research, University of Michigan