Determinants of poverty in the US state of Virginia: an examination of the impact of rent (the neglected variable)

ABSTRACT

This study provides new insight into factors that influence the poverty rate by testing the following hypothesis: the percentage of the population in poverty is positively related to rent levels. Rent levels constitute an issue effectively overlooked in the poverty determinants literature. The present study estimates a panel data set inclusive of control variables for the US state of Virginia, which provided all the data needed for the analysis. Panel least squares (PLS) estimations using county fixed effects and period fixed effects for the period 2008–17 find poverty inversely related to median income and the percentage of the population with at least a high-school diploma. Poverty also is found to be positively related to the percentage of the population employed in mining; the percentage of the population classified as obese; and the unemployment rate. Finally, poverty in Virginia is, as hypothesized here, an increasing function of rent. Higher monthly rental levels on one-bedroom apartments increase the percentage of the population in poverty; indeed, a US$100 per month increase in rent would elevate the overall poverty rate in the state by 1.20–1.35%.

KEYWORDS:

• poverty
• rent
• median income
• unemployment
• control variables

JEL:

• I32
• P36
• P46

1. INTRODUCTION

In 2019, the official poverty rate in the United States as a whole was 10.5%, down 1.3 percentage points from 2018. Despite the decline in the poverty rate, 34.0 million people in the United States lived in poverty in 2019 (Semega et al., 2020). Although the state of Virginia ranked 10th out of 50 states in terms of the lowest poverty rate in the United States, 9.9% of the state’s population still lived in poverty in 2019.¹ Clearly, given the percentage of Americans who live in poverty, it is imperative that researchers and policymakers continue to examine the factors that influence poverty in order to help ensure that poverty can be addressed in efficient and practical ways.

While reducing poverty continues to be a topic of interest, defining and measuring poverty can be complex and multidimensional. Although definitions vary, historically the definition of poverty in the United States has mostly concentrated on the income that is available to families (Cauthen & Fass, 2007; Blank, 1993). The US Census Bureau (2021) uses money income thresholds based on family size and composition to determine if an individual lives in poverty. When a family’s threshold nominal income exceeds the family’s actual nominal income, every individual in the family is considered to live in poverty.² The US Census Bureau’s statistics only include family members living together who are related by birth, marriage or adoption; therefore, unrelated individuals are not included in this measurement.

The measurement for the official poverty rate was developed in the 1960s by Orshansky, an economist at the Social Security Administration (Hunt, 1996). Whereas some argue that this measurement is outdated and should be adjusted to reflect more modern times (Cauthen & Fass, 2007; Blank, 1993; Parolin, 2019), the intent of the present study is not to discuss the potential flaws in the measurement of poverty. Rather, its objective is to examine factors that may influence poverty rates by analysing panel data for the counties and independent cities in the state of Virginia over the period 2008–17. The focus on Virginia is because this state provides a substantial database for the variables considered in this study, as attested to by other empirical studies (e.g., Blake-Gonzalez, 2021; Blake, 2021). These variables include median household income, educational attainment, the unemployment rate, the obesity rate and the percentage of the population employed by the mining industry. In this empirical analysis, these variables are treated as de facto control variables. An additional explanatory variable, indeed, the variable of primary interest in this study, is the average monthly rent payment for a one-bedroom apartment. Such a variable has heretofore been, in the literal sense, entirely overlooked in the literature on poverty determinants and is, as a consequence (and for reasons explained more fully in section 3), the principal focus of the present study. This study also examines the interaction terms between median household income and education attainment and between the unemployment rate and the population employed by the mining industry; this is because these multiplicative interaction terms were statistically significant in the county fixed effect, period fixed effect PLS estimation provided here.

The remainder of the paper is structured as follows. Section 2 provides a brief review of the literature on factors influencing poverty in the United States and in other nations. Section 3 provides a framework for the analysis that focuses on average monthly rent payments. Lastly, conclusions and implications are provided in Section 4.

2. LITERATURE REVIEW: FACTORS THAT INFLUENCE THE POVERTY RATE

The literature on poverty in the United States is diverse, but it frequently focuses on characteristics related to economic development. For example, Reeves et al. (2016) identify five dimensions of poverty based on the 2014 American Community Survey Public Use Microdata Sample: low household income, lack of education, lack of health insurance, low-income area and unemployment. The study examines the percentage of respondents who are classified as multidimensional, meaning that the respondent belonged to more than one of the identified dimensions of poverty. Based on the study, multidimensional poverty is more prevalent among blacks and Hispanics, with more than 3 million black and 5 million Hispanic adults belonging to at least three of the dimensions. Blacks and Hispanics are at a disadvantage when compared with whites as 18.8% of blacks and 15.7% of Hispanics lived below the poverty level in 2019 as compared with only 9.1% of whites (Semega et al., 2020). The five dimensions identified by Reeves et al. (2016) are often examined in the literature on poverty.³

The literature suggests that factors that influence health also contribute to an increase in the poverty rate. For example, in 2019, the estimated poverty rate was 25.9% for individuals with disabilities compared with 11.4% for individuals without a disability (Houtenville & Rafal, 2020). Overall, when compared with other population groups, individuals with a disability may experience lower educational attainment levels, lower average income levels and higher health-related costs due to higher out-of-pocket expenses (Batavia & Beaulaurier, 2001). Batavia and Beaulaurier further state that individuals with disabilities may have a reduced earnings capacity and little financial reserves to combat financial hardships, leading some individuals to be at risk of homelessness. The percentage of the population that is obese is also positively associated with poverty rates. The literature suggests that there is a positive relationship between obesity, sedentariness and poverty (Levine, 2011).

Income thresholds are frequently used to determine poverty rates; therefore, the lower the income, the more likely that an individual or family will live in poverty. Income gained from employment contributes to the median household income, hence employment is a determinant of poverty rates. A vast array of literature addresses the relationship between poverty rates and employment rates. While the magnitude varies, the literature suggests that higher unemployment rates have very large and negative effects on the poor, while inflation has few effects at all (Blank & Blinder, 1986, 2021; Blank, 2000; Blank et al., 1993; Romer, 2000; Freeman, 2003). Therefore, poverty rates are expected to be an increasing function of the unemployment rate.

Population density is the proportion of a county’s total population to the county’s total land area in square miles (Ladd, 1992). Generally, rural populations have a lower population density, higher unemployment rates, a higher percentage of uninsured residents, and are often more vulnerable to economic downturns since they often have a concentrated economic specialization (Hart et al., 2005). The literature suggests that rural areas have higher poverty rates, but also suggests that there are higher percentages of low-income households in surrounding cities and counties of metropolitan areas (Mieszkowski & Mills, 1993; Mills & Lubuele, l997; Madden, 2003; Gittell & Tebaldi, 2010). For example, in Tinsley and Bishop (2006), population density categories are used to analyse poverty in the state of Georgia. The study shows that poverty rates are higher in urban areas than in suburban areas in Georgia. While the overall poverty rate for Georgia was 13%, the poverty rate in the Atlanta urban core area was 23.2%, and 19.7% in other metropolitan core areas. The study found that the poverty rate in suburban areas in Georgia was only 7.6%.

The type of industries in a particular area may also impact poverty levels. For example, research suggests that there is a negative relationship between poverty rates and counties with high dependence on the mining sector (Tickamyer & Tickamyer, 1988; Johnson et al., 2019). Much of the research related to employment in the mining industry in the United States concentrates on the Appalachian region, which is one of the three major coal mining regions in the United States. The Appalachian Regional Commission divides the Appalachian region into 420 counties across 13 states, including Virginia. Historically, the Appalachia has been characterized by extensive poverty rates, especially in Central Appalachia given the region’s dependence on coal mining (Deaton & Niman, 2012; Ghosh & Cebula, 2021). While Deaton and Niman suggest that the Appalachian region’s dependency on the mining sector increases poverty rates in the long-term, poverty rates may be lower in the short-term. Other research suggests that past the year 2000, there is an inverse relationship between poverty rates and the coal mining sector in Appalachia (Partridge et al., 2013). Thus, there is no consensus in the literature regarding the impact of coal mining employment on poverty.

Of course, poverty is a global problem that extends far beyond the United States. Wealthy nations that have a serious poverty challenge include not just the United States but also other highly developed nations, such as the UK. The Borgen Project (2019), though not a robust empirical study per se, has assembled pertinent information that emphasizes multiple factors that contribute to the extent of poverty, which is estimated to be approximately 13.5 million people.

A major contributor to the rate of poverty in the UK is a deficiency of educational attainment and technical among the population. Indeed, the biggest issue facing policymakers in the UK it is argued is that as many as 5 million adults in the country do not have even rudimentary literacy or basic numerical skills. Choice of lifestyle also is argued to contribute to the extent of poverty, that is, problems such as substance abuse, underemployment and domestic violence poverty. Indeed, these circumstances are found to lead to an increased likelihood of involvement in criminal activity, on the one hand, to mental illness, on the other. Moreover, these problems are exacerbated over time by inflation, especially the rising prices of food.

Causes of poverty in the UK are spotlighted by the BBC (2020/21), but the list, while similar to that in The Borgen Project (2019), has two rather different components. More specifically, the primary factors are alleged by the BBC to be long-term unemployment, lack of education, homelessness and social class background. The latter most of these is said to be reflected in the form of prejudice against people based upon negative stereotyping; along with homelessness, it constitutes two factors not explicitly observed in The Borgen Project. Moreover, while both the BBC and The Borgen Project identify unemployment issues as a culprit, nonetheless underemployment and long-term unemployment are somewhat different.

Mihai et al. (2015) focus on the impact and importance of education on the extent of poverty across nations, including France and neighbouring nations, emphasizing that the impact has at least two major components. In particular, education successfully completed provides one with knowledge and skills as well as ‘credentials’ to succeed in the labour market, and in turn enhances the likelihood of avoiding poverty status. In addition, to the extent that offspring are raised by a family in poverty, they will very often be less well equipped to succeed in a formal educational setting and hence very possibly also less motivated to succeed. Without the knowledge, skills and credentials of formal education that enable offspring raised in poverty to gain the opportunity to secure adequate employment to avoid poverty, a cycle of poverty is perpetuated. Interestingly, the significant role of deficiencies in educational attainment for its own sake and for its role in promoting the cycle of poverty has been found as an insidious root cause of poverty in many nations on the other side of the globe, for example, India (Aasha & Mehta, 2003; Alok, 2020). In many of these nations, an accompanying issue is an excessive dependence upon agriculture as the family ‘business’. It has become commonplace to find two public policies that are strongly advocated in addressing such circumstances, namely, the development and pursuit of policies that increase student retention accompanied by the development of improved infrastructure, for example, access to safe water and improved low-cost public transportation; the success of these policies to data is currently unclear.

Ferguson et al. (2007) also focus upon the critical importance of education to the avoidance of poverty. In the case of Canada, it is argued that the income gap among Canadian families has widened since the 1990s, with educational outcomes (attainment) being significantly affected by family income. It is argued that children from lower income families are frequently at a disadvantage vis-à-vis their counterparts from more affluent family settings in terms of school readiness. In particular, effectively paralleling the cycle of poverty perspective, the milieu of typically lesser parental education, knowledge and income levels, along with concomitant greater access to learning tools such as computers, restricts access to educational opportunities that result in greater financial success in one’s life. It is further argued that paediatricians and family physicians presumably have a variety of opportunities through which they can improve child preparedness for school and educational success.

Brady et al. (2009) analyse how political context, embodied by the welfare state and Leftist political factors, shapes individual poverty. Using the Luxembourg Income Study, they conduct a multilevel analysis of working-aged adult poverty across 18 affluent Western democracies. Their welfare generosity index is found to exercise a negative and statistically significant effect on poverty net of individual characteristics and structural context. Thus, in conjunction with low levels of education and higher unemployment acting to raise poverty, for each 1 SD (standard deviation) increase in the level of welfare generosity, the likelihood of poverty declines.

In closing this literature review for the United States and other nations, in none of the peer-reviewed literature on poverty determinants/causes has there been an empirical analysis that hypothesizes and formally investigates the idea that higher apartment rents lead to a higher incidence of poverty per se. This hypothesis and the background as to why this is an important issue for inquiry are provided in section 3. This fact contrasts to the roles assumed by unemployment and educational attainment, which are almost universally found to influence poverty.

3. FRAMEWORK FOR THE ANALYSIS: THE MATTER OF RENT

As the brief literature review in section 2 illustrates, the issue of poverty and the potential causes thereof have received attention from a variety of researchers. Interestingly, despite the volume and breadth of this literature, however, it is noteworthy that the potential impact of higher rent levels on the relative degree of poverty has been effectively overlooked in the published related research. This is especially perplexing in view of the fact that rent has been found by a variety of studies to contribute significantly to a social problem/issue that parallels poverty, namely, homelessness. By and large, homeless persons would be considered to be poor. And, perhaps unsurprisingly, a variety of studies have found homelessness to be an increasing function of the rental levels of apartments, especially one-bedroom apartments (Grimes & Chressanthis, 1997; Quigley et al., 2001; Lee et al., 2003; Johnson et al., 2019; Cebula & Alexander, 2020). Nevertheless, the poverty literature has not taken rent as a serious contributor to poverty.

3.1. Principal hypothesis

Accordingly, the main objective of this study is to investigate the impact of higher rent levels on one-bedroom apartments on the relative poverty rate. Consider the process of constrained utility maximization, where the representative individual consumer (or family unit) seeks to maximize utility subject to the budget constraint: (1) $\mathrm{Maximize}:U=\left(X1,X2,X3,\dots ,\mathrm{Xn}\right)$(1) (2) $\mathrm{Subject}\mathrm{to}:\mathrm{Yd}=Y-T=\sum _{j=1}^n\mathrm{Pj}\ast \mathrm{Xj}$(2) where Xj is the consumption/purchase of commodity j, j = 1, … , n; where Xn refers to the units consumed of rental housing in the form of a one-bedroom apartment; Pn is the monthly rent paid on Xn; Y is pre-tax income; T refers to tax liabilities; and Yd is disposable income.

Clearly, equation (2) can be rewritten as: (3) $\mathrm{Yd}\mathrm{\#}=Y-T\text{–}\mathrm{Pn}\ast \mathrm{Xn}=\sum _{j=1}^{n-1}\mathrm{Pj}\ast \mathrm{Xj}$(3) where the new disposable income level, Yd#, is actually an after-tax and after-rent level of income available for the consumption/purchase of commodities X1 through to Xn – 1. Accordingly, the higher the rent level on one-bedroom apartments, the less the representative consumer (or family unit) has available for the purchase of everything other than housing. Thus, increasingly expensive rent levels act to decrease the consumer’s purchasing power and thereby to elevate the probability of or degree of poverty for the affected party(ies). Accordingly, it is hypothesized that poverty is an increasing function of the rent level, ceteris paribus. This hypothesis, as clear as it seems, has eluded most published studies focused on the causes of poverty.

Hypothesis (Ho): Ceteris paribus, the percentage of the population in poverty (POV_it) in county i in year t is an increasing function of rent levels.

This focus in the study on rent per se is important for several reasons. To begin with, extensively paralleling the pattern of the overall cost of living in the United States, it is noteworthy that there are and have been enormous geographical differentials not only in the cost of living but also in rent levels across the country, differentials that have existed since the nation’s very inception (McMahon & Melton, 1978; Cebula, 1980, 1989; Ostrosky, 1983, 1986; Hogan, 1983; McMahon, 1991). This is true whether comparing cities, metropolitan areas, counties, rural versus urban areas or even states (Cecchetti et al., 2002; Kurre, 2003; McMahon, 1991).

It logically follows that when the Census Bureau adopts simply nominal income thresholds based on family size, family composition and actual nominal family income to determine the extent of and location of poverty while failing to adjust both the threshold and the actual nominal incomes that take geographical rent differentials into account, that it may fail to accurately measure the extent of poverty (Cebula et al., 1992; Kurre, 2003; Campbell & James, 2020). In other words, the Census Bureau is culpable of ‘money illusion’. Accordingly, public economic and non-economic policies that are considered as measures to alleviate poverty may prove to be seriously ineffective if not defective.

To the extent that higher rent levels contribute to the degree of poverty but are overlooked in the compilation of poverty statistics and by public policymakers, potentially useful public economics policies such as rent subsidies may be severely miscalculated. Indeed, in some geographical areas, it may turn out that such subsidies are altogether unnecessary or at least rather minimally useful, whereas in other such areas much higher rent subsidies might be needed. Moreover, if it is found that poverty is an increasing function of the level of rent, then alternatives to rent subsidies may warrant consideration by policymakers and other government officials, including city planners. For example, re-zoning certain geographical areas to allow for more apartment construction may alleviate upwards pressure on rent levels and possibly even create downward pressure on rent. Furthermore, certain tax subsidies may be worthy of consideration insofar as they encourage landlords to allocate greater resources to the construction and availability of apartment facilities, which in turn would tend to ameliorate rent burdens and increases therein over time. In addition, the possibility of imposing rent controls (Grimes & Chressanthis, 1997) may or may not be found to be more appropriate considering the significant impact of rent on impoverishment. In any case, policies such as these would seem warranted only if serious empirical evidence as opposed to, say anecdotal ‘evidence’ from the media, indicating that higher rent elevates poverty has been obtained and made available.

In any event, in order to test this hypothesis, we are using the rent level on one-bedroom apartments because for any given quality of apartment quality will be less expensive in one-bedroom units than in multi-bedroom units. The rent data were obtained from RentData.org. (2008–17) and the Virginia Department of Housing Authority (2020). Interestingly, previous studies have used panel data for the state of Virginia and the Appalachia region (Blake, 2021; Ghosh & Cebula, 2021), but have not included a variable to correspond to rent levels.

3.2. Control variables

Based on the literature review found in section 2, several control variables are included in this empirical test of Ho above, as follows:

MEDHHINC_it: median household income in county i in year t.

UNEMPRATE_it: percentage unemployment rate in county i in year t.

PCTHSORMORE_it: percentage of the population aged 25 years and over with at least a high-school diploma in county i in year t.

PCTEMPLMINING_it: percentage of the labour force employed in mining in county i in year t (Ghosh & Cebula, 2021).

PCTADOBESITY_it: percentage of the adult population categorized as obese in county i in year t.

PCTBLACK_it: percentage of the population in county i in year t that is Afro-American.

INTEREDINC_it: interaction term between PCTHSORMORE_it and MEDHHINC_it.

INTERUNMINING_it: interaction term between UNEMPRATE_it and PCTEMPLMINING_it.

These two interaction terms were reported because they were the only ones found to be statistically significant in the estimation process. The data for these control variables were obtained from the Federal Reserve Bank of St. Louis, FRED (2020a) and FRED (2020b), NORC (2020), US Energy Information Administration (2019) and US Census Bureau (2012, 2020). Unfortunately, data on economic freedom, such as freedom from excessive government spending, tax freedom and labour market freedom (Stansel et al., 2019), are not available for many of the individual geographical areas within Virginia. Meanwhile, the descriptive statistics for the non-interactive term variables are provided in Table 1.

Table 1. Descriptive statistics.

Variable	Mean	SD	Maximum	Minimum
POVit	15.04	6.61	37.5	2.9
RENTit	740.81	259.32	1,513	423
MEDHHINCit	54,014	20,335	136,191	20,335
UNEMPRATEit	6.28	4.03	17.43	2.5
PCTHSORMOREit	84.74	6.12	98.2	66.9
PCTEMPLMININGit	2.07	4.03	25.7	0.0
PCTADOBESITYit	29.32	3.95	41.5	15.81
PCTBLACKit	19.41	16.77	79.53	0.48

3.3. The synthesized model

Based on section 3, the following model will be estimated: (4) $\begin{array}{rl}& \mathrm{Log}(\mathrm{POV}\mathrm{it})=a_0+a_1\mathrm{RENT}\mathrm{it}+a_2\mathrm{MEDHHINCi}t+a_3\mathrm{UNEMPRATE}\mathrm{it}\\ & +a_4\mathrm{PCTHSORMORE}\mathrm{it}+\mathrm{PCTEMPLMINING}\mathrm{it}+a_6\mathrm{PCTADOBESITY}\mathrm{it}\\ & +a_7\mathrm{PCTBLACK}\mathrm{it}+a_8\mathrm{INTEREDINC}\mathrm{it}+a_9\mathrm{INTERUNMINING}\mathrm{it}+\mathrm{\lambda i}+\text{β}t+\mathrm{\mu it}\end{array}$(4) where ʎi reflects county fixed effects and βt reflects period fixed effects, and the term µ_it is the error term, with subscript i representing county i and subscript t representing year t and with t running from 2008 to 2017. This study adopts the PLS estimation technique and also conducts Hausman (1978) testing. The latter indicates that the estimates should reflect county fixed effects (dummy variables) as well as period fixed effects (dummy variables). Given that the study adopts panel data, our econometric approach is a generally accepted one (Cebula & Duquette, 2022; Blake, 2021; Cebula & Alexander, 2020; Grimes & Chressanthis, 1997). To have simply adopted PLS without including cross-section and period fixed effects would have been a de facto misspecification. Arguably, an alternative empirical analysis could have been undertaken using generalized method of moments (GMM).

The PLS estimation of equation (4) is provided in Table 2, where robust t-values are provided. The semi-log form of the specification is presented because it facilitates the interpretation of the coefficients. The signs on all the estimated coefficients exhibit the expected signs. Furthermore, of the nine estimated coefficients, six are statistically significant at the 1% level, one is significant at the 2.5% level and one is significant at the 5% level. Moreover, adding credibility to the model, the F-statistic is significant at the 1% level. The coefficient of determination is 0.98, so the model appears to explain nearly all the variation in the dependent variable.

Table 2. Panel least squares estimation.

Dependent variable: log (POVit)
Variable	Coefficient	t-value	Probability
RENTit	0.00012	2.73	0.0066
MEDHHINCit	−0.00003	−6.20	0.0000
UNEMPRATEit	0.10810	4.70	0.0000
PCTHSORMOREit	−0.01370	−5.57	0.0000
PCTEMPLMININGit	0.01100	2.04	0.0420
PCTADOBESITYit	0.00123	2.82	0.0049
PCTBLACKit	0.00196	0.39	0.6994
INTEREDINCit	0.0000002	4.86	0.0000
INTERUNMININIGit	−0.00066	−2.24	0.0249
Constant	3.82
F-statistic	345.90
Prob (F-statistic)	0.00000
R^2	0.98
Adjusted R^2	0.97
County fixed effects (dummy variables)
Period fixed effects (dummy variables)

Based on this estimation, the poverty rate is a decreasing function of median family income, and the percentage of the population with at least a high-school diploma. In addition, the percentage of the population in poverty is an increasing function of: the percentage of the population employed in mining (a finding that makes the case that coal mining employment is associated with greater poverty, unlike many previous studies); the percentage of the population classified as obese; and increases in the unemployment rate. The statistical insignificance of the variable PCTBLACK_it can arguably be attributed at least in part to multi-collinearity with the interactive terms. These results are largely compatible with the existing published literature on poverty.

Finally, there is the matter of the poverty-impact of the variable RENT_it. As shown in Table 2, this coefficient is positive and statistically significant at the 1% level. Thus, poverty is, as hypothesized, an increasing function of rent. The higher the monthly rental on one-bedroom apartments, the greater the percentage of the population in poverty, that is, the higher the poverty rate. In fact, as shown, a US$100 per month increase in rent would elevate the percentage of the population in poverty by 1.2%. Given the improvements in the poverty rate in the United States over recent years, this figure may not be regarded as encouraging, especially during the inflationary times prevailing in 2021 and 2022 as well as the foreseeable future.

3.4. A simple robustness test

This subsection provides a simple robustness test of the basic model. In particular, the following modified version of the model is to be estimated: (5) $\begin{array}{rl}\mathrm{Log}\left(\mathrm{POV}\mathrm{it}\right)=& a_0+a_1\mathrm{RENT}\mathrm{it}+a_2\mathrm{MEDHHINC}\mathrm{it}+a_3\mathrm{UNEMPRATE}\mathrm{it}\\ & +a_4\mathrm{PCTHSORMORE}\mathrm{it}+a_5\mathrm{PCTEMPLMINING}\mathrm{it}\\ & +a_6\mathrm{PCTADOBESITY}\mathrm{it}+a_7\mathrm{PCTBLACK}\mathrm{it}+\mathrm{\lambda i}+\text{β}t+\mathrm{\mu it}\end{array}$(5) As shown, the model is now expressed in a form to help ensure that the interactive terms cannot influence outcomes for the other explanatory variables, that is, they are deleted from this re-specification. Otherwise, the interpretation of the estimation results is unchanged from before, with the PLS county fixed effects/period fixed effects estimate of equation (5) being provided in Table 3.

Table 3. Alternative panel least squares estimation.

Dependent variable: log (POVit)
Variable	Coefficient	t-value	Probability
RENTit	0.000135	2.17	0.0305
MEDHHINCit	−0.000008	−5.71	0.0000
UNEMPRATEit	0.0881	2.39	0.0171
PCTHSORMOREit	−0.0051	−2.46	0.0141
PCTEMPLMININGit	0.0085	2.36	0.0184
PCTADOBESITYit	0.0013	1.99	0.0473
PCTBLACKit	0.0007	1.80	0.0716
Constant	3.11
F-statistic	346.22
Prob (F-statistic)	0.00000
R^2	0.98
Adjusted R^2	0.97
County fixed effects (dummy variables)
Period fixed effects (dummy variables)

As shown in Table 3, the poverty rate is a decreasing function of median family income and the percentage of the population with at least a high-school diploma. In addition, the percentage of the population in poverty is an increasing function of: the percentage of the population that is employed in mining (once again, a finding that makes the case that coal mining employment is associated with greater poverty); the percentage of the population that is classified as obese; increases in the unemployment rate; and the percentage of the population that is black/Afro-American.

Of greater relevance for the purposes of this study, however, is the poverty impact of the variable RENT_it. As shown in Table 3, this coefficient is positive and statistically significant at beyond the 5% level. Thus, poverty is, as hypothesized, an increasing function of rent. Hence, the higher the monthly rental on one-bedroom apartments, the greater the percentage of the population in poverty. In fact, as shown in Table 3, a US$100 per month increase in rent would in this estimation appear to elevate the percentage of the population in poverty by 1.35%.

5. CONCLUSIONS

Given the finding that higher rent levels are both a statistically significant and economically significant factor contributing to poverty rate in the United States, the question arises: Is there an efficient and practical public policy or policy mix that can alleviate the problem of high and rising rent levels? Arguably, various plausible solutions can be considered among the potential mix of policy options. This list of possibilities would arguably include the provision of public housing per se or rent subsidies.⁴ However, alternatives to rent subsidies may warrant consideration by policymakers and other government officials, including city planners. For example, re-zoning certain geographical areas to allow for more apartment construction may alleviate upwards pressure on rent levels and possibly even create downward pressure on rent. Furthermore, certain tax subsidies/concessions may be worthy of consideration insofar as they encourage landlords to allocate greater resources to the construction and availability of apartment facilities, which in turn would tend to ameliorate rent burdens and increases therein over time. In addition, the possibility of imposing rent controls (Grimes & Chressanthis, 1997) may or may not be found to be appropriate in view of the significant impact of rent on impoverishment. And, of course, potentially a variety of income supplement policies could be formulated. Moreover, it is stressed that each geographical area within a state may consider its own unique demographic and other characteristics in the process of developing an appropriate and hopefully efficient mix of such policies.

However, yet another perspective approach to the issue could involve exploring two paths. First, investigating the factors that determine rent levels in order to identify potential practical and usable tools to alleviate rent pressure effects on those with limited financial resources. Second, policymakers can peruse the factors aside from rent that influence poverty. For instance, what useful steps could be taken to elevate educational attainment and pertinent training so as to raise peoples’ purchasing power? Additionally, what sorts of policies can be adopted to improve diets and help reduce obesity challenges? Recognizing that obesity is likely in some cases the result of compulsive over-eating, we should be sensitive to the fact that over-eating is sometimes a coping mechanism for economic or non-economic hardships and problems. Furthermore, are there practical ways in which to equip and motivate those who would otherwise seek employment in a dying/contracting industry such as coal mining to seek alternative and more promising employment options. In closing this study, it is observed that, given the study period considered here (2008–17), it was not possible to assess the poverty impact that COVID-19 might have had (did have in 2020–21) on unemployment and hence on poverty. This might be a useful added issue to consider when extending the scope of the present study.⁵ Finally, in terms of study shortcomings, it is observed that, as suggested above, data on economic freedom, such as freedom from excessive government spending, tax freedom and labour market freedom (Stansel et al., 2019), are not available for a large portion of the individual geographical areas within the state of Virginia. Based on the finding that labour freedom reduced the rate of homelessness in the United States (Cebula & Alexander, 2020), there may be reason to believe that this type of variable may have added an interesting dimension to the poverty discussion. Clearly, once such economic freedom data are available, their integration into the estimation process would seem very worthwhile.

The study also has other limitations. Ideally, a longer study period would be highly desirable so as to increase confidence in the conclusions of this paper. Moreover, other econometric techniques, such as GMM or panel vector autoregressive models, could be adopted to further test the main hypothesis of this paper.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1 The state of Virginia’s is based on American Community Survey 2019 estimates (https://data.census.gov/cedsci/table?q=virginia%20poverty%20rateandtid=ACSST1Y2019.S1701andhidePreview=true).

2 For a complete description of the poverty rate calculation, see https://www.census.gov/topics/income-poverty/poverty/guidance/poverty-measures.html.

3 For the related analysis, see Gotham (2007).

4 The analysis by Grainger (2021) in terms of homeless policy may be of relevance to policymakers and of interest if not relevance.

5 Related to rebuilding the economy following the COVID-19 crisis, see Martin (2021). See Morisson and Doussineau (2019) and Mulligan et al. (2019) for insights into the complexities and nuances associated with the concept of policy mix. One facet of this perspective involves factor interaction, that is, the fact that policies such as public housing, rent subsidies and rent control do not act in isolation but interact and do so in ways that affect people per se, the economy and society.