Change in U.S. Small Town Community Capitals, 1980–2010

The U.S. is overwhelmingly urban: 94.3 percent of the 2016 population lived in what the Census Bureau defines as metropolitan (85.7 percent) or micropolitan (8.6 percent) areas. We know a great deal about these places, in part because there is a substantial amount of publicly available data for places at least as large as a micropolitan area (a labor market centered on an urban area with 10,000–50,000 residents). We know much less about smaller places often centered on villages, towns and cities with populations as small as a few hundred. This paper is dedicated to the study of these “tiny towns” and their residents.

Our focus is on particularly small communities – those with fewer than 2,500 residents that are located in counties classified by the USDA’s Economic Research Service as “completely rural”. Despite comprising a demographic minority, these areas are likely the most distinct from urban America and, therefore, most in need of rural-centric research and policy (USDA 2018a).

Providing the first nationwide, descriptive place-based analysis of these areas over time, this paper presents basic trends in sociodemographic and economic characteristics between 1980 and 2010. The analyses are structured by three research questions: 1) How has small town America changed socio-demographically in the past several decades?; 2) Are trends in small town change distinct according to proximity to metropolitan areas?; and 3) Are trends distinct as related to longer-term patterns of population growth and decline? To respond, we contrast trends across communities with fewer than 2,500 residents in counties adjacent to metropolitan areas to those in counties not adjacent to metropolitan areas. Such a contrast is useful since metro adjacency expands the labor market, improves market access, and increases proximity to a wide array of services. We further explore patterns of sociodemographic change as related to place-based population growth or decline. We do so through the lens of the “Community Capitals Framework” which considers place-based assets that shape opportunity and potential. This comparison across growing and declining places allows for a more nuanced understanding of shifts in community capitals that have occurred simultaneously to shifts in aggregate population size. In all, this investigation provides an important foundation for future work into the particular challenges faced by tiny towns in the U.S. and the potential demographic, social, economic, and health futures they might face.

Background

It has been over thirty years since the publication of the last broad-based study of remote small towns in the U.S.. Johansen and Fuguitt (1984) undertook an analysis of “nonmetropolitan villages” of 2,500 population or less, from 1950 to 1980, using a 5 percent random sample of 572 small places. They found that most places had fewer retail and service establishments in 1970 than in 1950, likely due to increased competition from retailers in larger cities. Yet this economic decline was somewhat slower in places that experienced at least some population growth and, as such, their results suggest that metro proximity and population trends both shape small town change – a finding that influenced the analytical perspective used here as we contrast trends in places in counties adjacent to metro areas to those that in counties that are not.

More recently, Porter and Howell (2016) offer insight into sub-county demographic trends from 1980 to 2010, by decomposing population change spatially along the rural-to-urban continuum. As in our study, they examined nonmetro counties not adjacent to metro counties and observed within-county concentration between 1980 and 1990, meaning that Census-recognized places grew more than areas outside those places in the same county. That pattern reversed from 1990 to 2010. The processes of concentration and deconcentration tell us about overall demographic trends; what remains unexamined are their socio-economic characteristics, correlates, and implications. These are gaps that the current project begins to fill.

In contrast to the work of Johansen and Fuguitt (1984) and Porter and Howell (2016), almost all research on broad scale rural population change has been carried out at the county level. The chronology is well known: From the 1920s to the 1960s, natural increase in rural counties barely outpaced outmigration. By contrast, the 1970s brought a brief “rural renaissance,” with 80 percent of rural counties gaining population and nonmetro growth proportionally exceeding urban growth, before slowing in the 1980s and 1990s. Another period of rural decline began in 2010, reversing only in 2016–2017. These aggregate trends have an important shortcoming in that they mask variation at the local level and ignore changes in the delineation of analytical units over time, which is especially important for the research presented here.

Of course, place-based factors other than metro proximity also shape population change. For instance, demographic research suggests that migrants also typically prefer rural communities that are relatively dense and/or have natural amenities. Accordingly, counties with persistently declining populations are clustered in low-density, remote areas in the Great Plains, Appalachia, and in higher poverty areas of the Southeast and Southwest (Cromartie and Vilorio 2019). Such tiny towns are encompassed in the research presented here.

To structure the descriptive analyses, we make use of the theoretical perspective commonly called the Community Capitals Framework, which highlights seven types of capital needed by communities to thrive. Where feasible, we use place-level measures that reliably proxy for capitals and while much current work using the CCF examines individual communities or undertakes small sets of comparisons, we employ a data-oriented approach to examine communities across the U.S. As emphasized above, we examine tiny town trends according to their county’s adjacency to a metropolitan area and to their overall pattern of population gain/loss.

Rural Case Studies and the Community Capitals Framework

The Community Capitals Framework (CCF) offers an approach for analyzing the effects of community and economic development on a variety of outcomes such as institutions, infrastructure, social cohesion, and environmental conditions (Emery et al. 2007). The capitals represent seven types of resources on which communities can draw to create change (Pigg et al. 2013). Human capital includes residents’ skills and abilities that are potentially useful for enhancing their community and accessing outside resource (Emery and Flora 2006; Rasmussen et al. 2011), while financial capital represents the means for investment in the community such as in new business or tourism development (Pigg et al. 2013). Built or physical capital is the infrastructure that supports, and is supported by, such investment (Jacobs 2007), and social capital reflects the connections between people and organizations as well as residents’ sense of belonging (Emery and Flora 2006; Rasmussen et al. 2011). Cultural capital includes language, tradition, and the way people “know the world” (Emery and Flora 2006), while political capital is the access to power of individuals, groups, and institutions as well as the distribution of this power (McCrea et al. 2014). Finally, natural capital includes assets linked to place, such as weather, natural amenities, and resources (Stofferahn 2012).¹

Research applying the CCF typically examines the effects of different capitals on economic or community development efforts in specific places. For example, in a cluster of communities in northeastern Iowa, investments in human capital through leadership training yielded increases in financial and built capital by helping community members apply for grants for infrastructure improvements (Emery et al. 2007). Another study in rural British Columbia used a modification of the CCF to show that adapting community services, such as in health and education, to changing rural conditions helped to build social capital (Sullivan et al. 2015). Both examples illustrate Emery and Flora’s (2006) argument that investing in one form of community capital can lead to gains in other forms and ultimately have important impact on community viability.

Community Capital Challenges Facing Rural America

Community capitals are often reciprocally related to local challenges that include, for example, rural America’s high concentrations of working poor, low levels of service provision, limited cell and internet coverage, lack of opportunities for rural youth, and the social inequalities that can emerge with development. Lack of community capital plays a role in generating these challenges, while efforts to reduce these challenges may be hindered by community capital shortfall. Illustrative examples follow.

A lack of diverse and well-paid employment opportunities represents a fundamental rural challenge and plays a role in low levels of human capital engagement. Indeed, rural employment growth has been slow in recent years; in many rural areas, employment has been slow to return to pre-recession levels (Thiede et al. 2016). This lack of opportunity affects not only adults; youths who perceive poor job prospects are typically more likely to consider leaving their rural communities after high school (Kirkpatrick Johnson, Elder and Stern, 2005). A relatively higher proportion of rural employment requires only low-level academic, technical and reasoning skills, which yield low wages, also challenge the development of community capital (Gibbs, Kusmin and Cromartie 2005). Another human capital distinction characterizing rural areas is the relatively higher proportions of retired individuals, which manifests in lower levels of individuals in the labor force. Combined, these issues regarding employment and income levels compromise the availability of financial capital at the community scale.

Technological limitations – reflective of built or physical capital -- also pose a challenge to economic growth in some rural regions. As the internet economy has flourished, access to high-speed internet service is becoming increasingly essential for both businesses and households. Yet while about 55 percent of U.S. adults had broadband access at home in 2008, this number reduces to 41 percent in rural households (Stenberg et al. 2009). Broadband installation in low-density areas offers less return on investment, and also results in higher utilization costs at the household level. As evidence of the economic impact of this capital constraint, rural counties with broadband access by 2000 experienced higher employment growth and greater earnings compared to counties with little or no such access in 2000 (Stenberg et al. 2009).

Another form of community capital -- social capital -- can be compromised as a result of demographic change, particularly when communities attract in-migrants from urban origins. The resulting “social asymmetry” can yield widening inequalities, particularly in rural counties experiencing gentrification (Golding 2016). Indeed, even during the 1970s rural renaissance, long-time residents suffered as a result of asymmetrical in-migration that led to cost-of-living increases without concomitant increases in higher wage employment (Hunter, Boardman and Saint Onge 2005; Saint Onge, Hunter and Boardman 2007). Outside capital can also displace low-income residents, for example when petroleum-driven local economies brought high-wage transient workers to the West (Brasier et al. 2011). These demographic changes can negatively impact local social safety nets as neighbors are increasingly likely to be strangers (Golding 2016).

Bringing Together Demographic Trends and Community Capitals in Tiny Towns

In all, prior nationwide research has documented broad rural community change while case studies have examined the influence of various forms of capital on economic and community development. We bring these literatures together to inform our descriptive analyses. We present decadal trends, 1980–2010, in several community capitals at the national scale for small rural places. As Porter and Howell (2016:180) lament “… owing largely to the technical labor involved … [most studies of the smallest size population settlements and villages] include only a sample of all such settlements in the U.S.” We contrast trends across tiny towns in metro adjacent and nonadjacent counties. Such a contrast is useful since adjacency expands the labor market, improves market access, and increases proximity to a wide array of services. As such, this study’s broad lens on small places reflects a key contribution of this research, as does the focus on remote small places and the unique challenges they may face.

As noted above, the analyses are structured by three research questions designed to shed light on overall sociodemographic trajectories as well as distinctions by metro proximity and overall population growth or decline: 1) How has small town America changed socio-demographically in the past several decades?; 2) Are trends in small town change distinct according to proximity to metropolitan areas?; and 3) Are trends distinct as related to longer-term patterns of population growth and decline?

Data & Methods

What is a “tiny town”? The tiny towns included in this study meet two criteria. First, they had fewer than 2,500 residents at the beginning of our study window. This definition is consistent with the US Census Bureau’s designation of “rural” as all population, housing, and territory not included in an urban area (pop. 50,000+) or urban cluster (pop. 2,500 – 49,999) (US Census Bureau 2018).

Second, our tiny towns are within “completely rural” counties as classified by the USDA Economic Research Service’s (ERS) “Rural-Urban Continuum”. The continuum includes nine categories with three “metro” county classifications and six “nonmetro” classifications. The continuum codes were first developed in 1974 and have been updated four times since; we use the 1983 codes, because they are based on demographic characteristics from the 1980 decennial census which is closest to the start of our study window. Our study communities are in the two most rural classificatiosn within the Rural-Urban Continuum Codes (see Table 1).² The included places were also present in all four data points and are only those in the continental U.S. Using these criteria, we identify 823 communities in “metro-adjacent” counties and 1,873 in “nonadjacent” counties ³

Table 1:

USDA Rural-Urban Continuum Codes, 1983

It is important to note that the analyses presented here are not intended to be generalizable to small places more generally as we have chosen to focus on those places in the most rural counties per the ERS continuum (Table 1). Also, a key contribution of this work is presentation of place-based trends, as opposed to rural demography’s more commonly used county scale. The importance of examining variation at the more local level, as opposed to county, presumes that places within counties are sociodemographically different than the counties in which they are embedded. To explore this presumption, the last two columns of Table 2 present bivariate associations between place and county values on all included community capitals. The correlations range from 0.43 (income inequality) to 0.89 (percentage Hispanic residents). The R-squared indicates that, on average, 43 percent of the variation in place-based capitals can be explained by county values, suggesting the utility of an analytical lens focused on the place.

Table 2:

Descriptive Profile of Community Capitals, 1980–2010

	Tiny Towns in Not-adjacent metro counties					Tiny Towns in Metro-adjacent counties					County versus Place Estimates
	Min	Max	Mean	SD	N	Min	Max	Mean	SD	N	Correlation Coefficient	R2
Human capital
College graduates (%)	0.00	83.87	11.89	8.46	1877	0.00	72.07	11.56	7.36	825	0.54	0.30
Population 65 years + (%)	0.00	100.00	21.50	8.10	1879	0.00	66.67	19.06	6.79	825	0.54	0.30
Employment-to-population (%)	0.00	100.00	52.02	11.73	1877	0.00	92.41	52.01	10.12	825	0.57	0.32
Financial capital
Median household income	0	105,625	23,626	12,649	1877	0	105,417	25,229	12,876	825	0.86	0.74
Individuals in poverty (%)	0.00	100.00	17.78	11.45	1877	0.00	84.48	17.17	10.00	825	0.51	0.26
Unemployment (%)	0.00	100.00	6.78	6.69	1873	0.00	75.00	7.61	5.96	825	0.52	0.27
Physical/Built capital
Market Access	1.99	771.08	109.45	87.40	1879	1.56	654.14	123.83	94.80	825	--	--
Broadband Access*	0.00	5.00	3.00	0.94	1879	1.00	5.00	3.00	0.98	825	--	--
Natural capital
Natural Amenities Scale	−5.34	10.60	−0.23	2.49	1879	−5.20	9.72	−0.24	2.50	825	--	--
Social capital
White residents (%)	0.00	100.00	92.32	15.21	1879	0.00	100.00	88.50	18.33	825	0.79	0.63
Hispanic residents (%)	0.00	94.92	3.15	9.25	1879	0.00	90.71	3.45	9.00	825	0.88	0.77
Income inequality (Gini Index)	0.00	0.58	0.38	0.06	1877	0.00	0.54	0.39	0.06	825	0.43	0.19
Cultural capital
National Historic Landmarks*	0	102	1	3.07	1879	0	45	1	2.95	825	--	--

^*Median presented in place of mean

Data Sources and Variables

We use a variety of data representing “places,” which are named population concentrations locally recognized and independent of other places. “Places” can be either incorporated places – defined by criteria within their respective states – or census-designated places (CDPs), which are population concentrations that are not incorporated and lack a municipal government. CDP boundaries have no legal status, but the 2010 Census required that a CDP name be recognized and used in daily communication by residents.

This project includes core demographic variables such as age, race, and Hispanic origin based on the full count census collected nationwide through the short form, thus providing 100 percent count data”. More detailed information such as education attainment, income, poverty status, civilian unemployment rates are derived from the long form circulated to a sub-sample. For these years, small areas were more heavily sampled to maximize accuracy for such places. In order to represent the U.S. population, Census weighting areas were closely aligned with census tabulation areas within a county, or full counties for regions with small sample counts. Within weighting areas, ratio-estimation procedures estimated characteristics of persons and housing units. For the 1980–2000 censuses, estimates from the sample were mostly consistent with the 100-percent figures for population and housing unit groups.

For 2010, basic demographic information is available from the 100 percent count 2010 census, but there was no long form for 2010 because implementation of the American Community Survey (ACS) in 2005 created a new mechanism for assessing detailed information about the U.S. population. The ACS is a questionnaire sent to approximately 250,000 addresses every month. Given the sampling framework, for smaller communities – like our tiny towns -- five years of data are required to generate statistically reliable estimates of population and housing characteristics. For 2010, the Census Bureau combines ACS data for 2008 through 2012 to estimate educational attainment, civilian employment, and median household income, for example. We use these data but recognize the lack of methodological consistency with the long forms for 1980, 1990, and 2000. All of the sociodemographic variables used in this study were extracted from the National Historical Geographic Information System (NHGIS), which provides both the 1980, 1990, and 2000 decennial census estimates, and the 2010 estimates based on a combination of the decennial census and the 5-year ACS data (Manson et al. 2017).

To represent community capitals, we first measure human capital by indicators of age and educational composition (see Table 2 for descriptions of community capital measures). Education is well understood to be an important aspect of human capital and we include measures of place-based educational levels (percentage of college graduates). Age composition is also important as it can reflect outmigration of young adults, which shapes prospects for community economic development. Importantly, some aging, such as elderly amenity-related migration, can provide opportunities by fueling real estate and commercial sectors (Thiede et al. 2017). Age composition, with a focus on the elderly population, is reflected here as a measure of proportion of the population age 65 and over. Age also factors in to the calculation of employment-to-population ratio which represents employed adults (age 16 and over) as a percentage of all adults. Here, the ratio represents the relative size of the potential pool of local workers.

Community financial capital is initially reflected by census-derived measures of median household income, poverty levels, and civilian unemployment data. These measures reflect the financial resources of a community. Prior research suggests that income gains due to growth accrue primarily to those in low-wage service sector employment (Saint Onge et al. 2007). The economic characteristics of a community may also impact its growth and development.

Physical/Built capital connects rural areas to the larger economy, both physically and through communication. One example is roads, which lower the cost of moving goods and people in and out of the community and can encourage economic opportunity (Chandra and Thompson 2000). In rural America, deteriorating roads increase the cost of transportation for farmers and ranchers and unsafe conditions cause higher mortality due to accidents, as compared to urban areas (ASCE 2017). We use information on the U.S. highway network to construct a measure of connectedness based on travel time via highways graded for different speeds and proximity to economic activity. We calculate this “market access” measure, which is drawn from Redding and Venables (2004) and recently used by Donaldson and Hornbeck (2016) and Jaworski and Kitchens (2019) to analyze the impact of transportation improvements on economic development in the United States. More specifically, the market access variable combines information on travel time via the highway network between all county pairs in the contiguous U.S. with information fuel costs and hourly truck driver wages, to approximate the costs of domestic trade within the United States. We use digitized maps of the highway network for each decade from 1980 to 2010 to calculate the minimum travel time route between all county pairs in each year in order to reflect the change in the density and quality of roads over time.

Broadband access information is from the Federal Communications Commission (FCC) and reflects connectedness to telecommunications infrastructure. Data from 2010 includes residential fixed internet access per 1,000 households, at the census tract level, for service over 200 kbps in at least one direction. The information is categorized from 0 (zero households) to 5 (greater than 800 households per 1,000 households) (FCC 2010).⁴

Natural capital is, in part, the environmental amenities that may shape a place’s demographic and socioeconomic development. Environmental amenities represent another important aspect of place’s broader context and the demographic patterns experienced within high-amenity places are quite distinct in their draw of migrants that tend to be older and of higher socio-economic status. We generate a place-based amenities measuring using characteristics similar to those underlying the USDA’s ERS county-level amenities index. Specifically, the index reflects climate (temperature, days of sunlight, humidity), topographic variation, and water area -- environmental qualities most people prefer (McGranahan 1999).

Data constraints make the final two community capitals particularly challenging to measure at the place scale. On social capital, we draw on Putnam’s characterization (2000) of “bonding” social capital, which is thought to be particularly strong in rural areas (Sørensen 2016); because some research suggests that bonding social capital is strongest among individuals with similar characteristics, such as social class, income inequality and this form of social capital are inversely related (e.g. Anderson and Paskeviciute 2006; Delhey and Newton 2005).⁵ As such, we use a measure of income inequality – the Gini Coefficient -- to reflect the prospect of place-based class divisions (Abounoori and McCloughan 2003). The coefficient ranges from 0 to 1 where 0 indicates perfect equality (all residents have same income), and 1 indicating maximum inequality (one resident earns all the income). In addition, we look at the racial composition of rural places by considering two measures, the percentage of the total population that identifies as white and the percentage that identifies as Hispanic. Previous research in the U.S. has found that racial heterogeneity can be associated with lower social cohesion, which leads to lower levels of bonding social capital (Putnam 2007), although this view remains contested (Hooghe 2007; Van der Meer and Tolsma 2014). Research has also demonstrated that social capital is strongly correlated with active citizenship, such as membership in nonprofessional organizations, involvement in local politics, and church attendance, but these measures are difficult to obtain at the place scale. For instance, religious data, as well as other indicators of social capital, are typically reported at the county-scale (e.g. Rupasingha, Goetz, and Freshwater 2006), while political engagement and social networks, such as collected in the General Social Survey, tend to be gathered in survey form with insufficient data to characterize especially small places.

Cultural capital represents community assets that create bonds between residents and can become part of place identity. This form of community capital is also difficult to represent quantitatively. We incorporate measures of local historic properties since they can play a role in community identity as well as foster the branding and, therefore, economic development potential. Specifically, the National Register of Historic Places includes over 90,000 properties that tell stories “that are important to a local community, the residents of a specific state, or to all Americans.” (National Historic Landmarks 2018). We include information from a publicly available database, overseen by the National Park Service, that includes date of listing with locations that are then spatially linked to places (USNPS 2018).⁶

Analytical Approach

As noted above, our analyses are structured by three research questions related to general trends in sociodemographic trajectories and distinctions by metro proximity which is indicated by the USDA’s county-scale categorization (continuum codes 8 and 9 per Table X). We also examine sociodemographic trajectories and distinctions according to tiny towns’ overall population growth and decline. To do so, three categories of tiny towns are created for counties both metro proximate and not. The categories reflect those with stable population 1980–2010 (≤5 percent change) and those with increases or decreases greater than five percent. In addition to describing trends, we execute a formal statistical test of these differences. Specifically, we use a Least Squared Dummy Variables (LSDV) approach with time interactions to test for trend differences in our ten community capital indicators.⁷ These models are estimated from an OLS equation of the following form:

$$Y_{i,t}=Non-adjacent{}_i+Yea{r}_i+Non-adjacent{}_i\times Year{}_i$$

where our dependent variable (Y) represents one of our variables of interest for place i in time t. Our first two right-hand-side variables are a dummy variable representing the difference between metro-adjacent and nonadjacent, and three time categorical variables (1990, 2000, 2010) contrasted against our base period (1980). To test whether our variables of interest are shifting differentially by county metro status, we estimate a set of interaction terms between our nonadjacent and year dummy variables. Significant differences in the interaction terms imply that, relative to tiny towns in metro-adjacent c-adjacent places, our variables of interest are changing at different rates in those in nonadjacent counties.

Results

We first examine overall population change and, interestingly, regardless of county metro proximity, the included tiny towns experienced nearly identical trajectories of relative change between 1980 and 2010 (Figure 1, panel 1a). For instance, all the included places lost population between 1980 and 1990 although those in more isolated, nonadjacent counties lost relatively more. During the subsequent decade, 1990–2000, small places in both categories similarly grew, although those in more isolated counties grew less. This 1990’s shift toward rural growth is more generally reflective of the decade’s “rural rebound” during which over 70 percent of rural counties across the nation gained population predominantly due to shifts in migration (Johnson 2006). Specifically, the historically consistent outflow of young adults (age 20–29) slowed, although still yielding loss for this age group. But at the same time, in-migration of older adults (age 50–59) was substantially higher than in prior decades.

Figure 1:

Population Trajectories

Divergence in population change by metro proximity occurred during the 2000–2010 time period, where tiny towns in metro-adjacent counties continued to grow, albeit at a slower pace than the decade prior, while those more isolated were again characterized by population decline. All told, during the three-decade study period, tiny towns in metro-adjacent counties gained, on average 11 percent population, while those not-adjacent lost approximately eight percent (Figure 1, panel 1b).

Change in population size is, of course, a very basic measure although as both a consequence and cause of other forms of socioeconomic change, it can reflect underlying complexity. For instance, population change is fundamentally the consequence of sociodemographic processes such as fertility, mortality, migration – as well as a potential cause of change related to, for example, economic development. Natural growth has trended downward in rural areas since 1980 due to lower fertility rates, and aging population and rising earlier life mortality due to the opioid crisis (Cromartie and Vilorio 2019). This suggests that an increasing proportion of population change is due to migration.

Forty percent of places in metro-adjacent counties gained population from 1980 to 2010, as contrasted with only 22 percent in nonadjacent counties. Indeed, fully 69 percent of tiny towns in nonadjacent counties experienced decline (versus 46 percent adjacent). Also intriguing is the scale of change. More isolated places with declining populations lost relatively more residents (average 22 percent) than declining places proximate to metro areas (average 19 percent). Basically, the scale of loss was greater in more isolated places. And while fewer tiny towns in nonadjacent counties grew between 1980–2010, among those that did grow, the population gain was comparable to growing places in metro-adjacent counties (48 percent and 45 percent, respectively).

Tiny town stability is also intriguing. A similar proportion of tiny towns remained demographically stable in both nonadjacent and metro-adjacent counties, with 9 and 14 percent respectively experiencing 5 percent or less in population size over the three-decade period from 1980 to 2010.

As put forward in our guiding research questions, for each cluster of community capitals, we present trends for tiny towns categorized by tertiles of loss and gain for both nonadjacent and metro-adjacent counties. This approach allows a focus on the correlates of types of change. Specifically, the tertiles allow for contrast of places experiencing similar levels of overall population change, a foundational factor in all community capitals. In all, Figure 1, panel 1b reveals more commonality than distinction in relative levels of change across tiny towns categorized by county metro proximity. Specifically, those places that experienced large relative population gains increased substantially regardless of their metro-adjacency. Likewise for those that experienced large relative losses.

Given differences in regional socioeconomic and environmental conditions, spatial variation would logically be anticipated in these patterns of change, and Figure 2 reveals such distinction. For the most part, the maps reveal that gain, loss and stability occurred in all census regions, across county categories by metro proximity. Particularly noticeable, however, is the substantial spatial clustering of declining remote tiny towns in the upper Great Plains (panel 2a). Yet although there are fewer tiny towns in metro-adjacent counties in this region, for the most part these too experienced loss (panel 2b). As to growth, as revealed by other research, the isolated small towns experiencing population increases are most noticeably in the amenity regions of the mountain west, upper Michigan peninsula, and portions of the Appalachians. Similar to the comparable spatial distribution of loss, however, such gains also occur in places within metro-adjacent counties in the same regions. Additional pockets of metro-adjacent growth are apparent in the amenity-laden regions of the coastal southeast.

Figure 2:

Spatial Distribution of Population Decline/Growth

Of course, population size and change represent a narrow view of shifts that shape place-based well-being; Population composition and other community assets also play critical roles. Using the CCF to develop a baseline understanding of other dimensions of change, the series of figures below present trends in community capitals to respond to our three research questions.

Throughout the remainder of the presentation of community capitals, each place-scale characteristic is presented with common line attributes and color themes to ease interpretation. Tiny towns (in counties not adjacent to metro areas) are consistently represented by solid lines while dashed lines indicate trends in those in metro-adjacent counties. Also, places in the highest tertile of gain are green, while those in the greatest tertile of loss are red.

Human Capital

Like population size more generally, the availability of community-level human capital represents both a cause and consequence of other forms of sociodemographic and economic change. For example, the pursuit of higher education can be a reason for outmigration and substantial research has documented the “brain drain” experienced by rural America (e.g. Carr and Kefalas 2009). That said, highly educated individuals may be drawn to particularly types of places and the ability of communities to retain or attract relatively more educated individuals reflects an important aspect of local human capital.

Overall, there has been an increasing proportional presence of college graduates in tiny towns regardless of county adjacency to metro areas. In fact, the trend is nearly identical and is in line with that experienced by the nation as a whole, as the proportion college graduates increased from around 15 to nearly 40 percent between 1980 and 2010. The trend also reflects recent increases in college aspirations among rural youth (Meece et al. 2014).

Yet Figure 3 demonstrates some place-based distinction. Although there has been an overall increase in tiny town educational levels, places losing population had lower levels of this particular measure of human capital in both metro-adjacent and nonadjacent counties. These comparatively low levels may be the product of relatively lower in-migration or relatively higher out-migration of college-educated residents. Research does suggest that higher wages in urban areas attract and keep workers, while rural students moving to urban areas to attend college may subsequently stay for work (Carr and Kerfalas 2009). In general, the pattern suggests that population loss has human capital dimensions in that such loss is associated with lower levels of human capital. Future research with restricted migration data will facilitate improved understanding of these dynamics.

Figure 3:

Human Capital by Population Trajectory and County Metro Proximity

Place-based age composition also reflects human capital through its relationship with the strength of the labor force. Tiny towns in both nonadjacent and metro-adjacent counties have proportionally more residents aged 65 and older as compared to the national population (~12 percent over study period). Yet, while more tiny towns in nonadjacent counties have older population profiles, the pattern of demographic change has been equivalent across tiny towns regardless of metro proximity. What is distinct is the substantially higher proportion of older residents characteristic of nonadjacent tiny towns that have experienced population loss over the period 1980–2010. In fact, the largest distinction is between remote tiny towns characterized by overall population loss or gain; Declining, nonadjacent tiny towns have the highest proportion elderly, while growing ones have the lowest. This likely reflects youth outmigration from declining rural spaces, and retiree inmigration as part of growing rural spaces, each process with important implications for place-based human capital. The flow of retirees is also represented in the employment-to-population ratio which reflects the relative size of the potential pool of local workers (panel 3c). The clear decline in this ratio for growing tiny towns, 2000–2010, has occurred for places in both nonadjacent and metro-adjacent counties suggesting a decline in this particular form of human capital.

We can see these trends more specifically in the observed population pyramids (Figure 4). Baseline population (1980) by age and sex are presented in hatched lines, while the 2010 population is presented in solid red. A quick glance reveals highly similar patterns of gain/loss across tiny towns in not-adjacent-metro counties and those adjacent. For instance, focusing on places in the highest tertile of population gain, it is clear that a substantial decline between 1980–2010 was experienced in proportional population below age 35 years regardless of metro proximity. These places also proportionally gained population in the ages between 35–64 years, both men and women. Of course, some of the increases at older ages is likely aging-in-place although the dearth in younger years suggests a new generation is not replacing those aging. Some of the loss in younger years is also likely due to the process of rural brain drain documented elsewhere. Shifting our focus to those places losing population between 1980–2010, again, regardless of metro proximity, the pyramids reveal aging populations.

Figure 4:

Age Composition: Population Pyramids, Proportion of Population by Age and Sex, 1980–2010*

Overall, with regard to human capital, national gains in human capital in the form of college graduates have not been matched in small town America. Loss of human capital in the form of those in early working ages characterizes small town America, the recognizable “brain drain”, and patterns of such loss are virtually identical for small town America regardless of metro proximity. As related to our second research question, this suggests that proximity to larger communities does not necessarily help draw or retain young residents. However, the third research question asks about human capital as to overall pouplation gain or loss and we find that the small towns that have been consistently losing population (regardless of metro proximity) suffer most from low educational levels, potentially impacting future economic development potential.

Financial capital

Levels of income, unemployment, and poverty are useful indicators of community financial capital and, although related, they reflect different dimensions of local well-being. To be classified as unemployed, an individual must consider themselves in the civilian labor market and, therefore, unemployment measures do not include children nor retirees. On the other hand, poverty is more broad-based and reflects family income as compared to a minimum threshold of resources, based on family size and composition, deemed necessary for family well-being. If a family’s income is below the poverty threshold, all members are considered to be living in poverty. Per Figure 5 and as the case with human capital, income trends are similar regardless of county metro proximity (panel 5a). What differs is the overall level of individuals living under the poverty threshold, which is consistently higher in small places as compared to the national average.

Figure 5:

Financial Capital by Population Trajectory and Metro Proximity

As the case with median household income, trends in poverty are also similar regardless of county metro proximity. What differs is the overall level. The trends align with the broader rural experience over the past several decades in that poverty rates peaked following the recessions of 1980–82 and 2007–09. Recovery from the latter recession has been modest for rural areas overall and stagnant for most rural groups (Kusmin 2015).

Disaggregating by overall growth/decline reveals diverging trajectories specifically for 2000–2010. In particular, growing tiny towns actually experienced the greatest increases in poverty during 2000–2010, while those with population loss actually experienced lower increases in poverty (panel 5b). Such patterns challenge common sense, but recall from above that remote, declining tiny towns tend to have older age compositions. Rural residents over age 65 are less likely than those younger to officially be in poverty given the social safety net for older adults. If we look at poverty rates by age (panels 5c-d), we see that growing tiny towns experienced higher levels of poverty for working adults (age 15–55) and lower levels of poverty for residents over age 65 compared to declining tiny towns over the same period (2000–2010).

Also important is further investigation into metro-adjacent counties. While rural poverty rates have historically been higher than those in urban areas, suburban poverty has recently been on the rise. Further, the highest concentration of suburban poverty is now in rural, unincorporated places outside of large cities (Kneebone and Berube 2013). This pattern may be represented in the relatively larger increase in poverty between 2000–2010 for tiny towns within metro-adjacent counties, although such a potential must be approached with caution as both county categories examined here are considered “nonmetropolitan.”

Unemployment represents another indicator of community-scale financial capital and this measure also exhibits divergence during the decade 2000–2010 (panel 5e). While tiny towns in metro-adjacent and nonadjacent counties tracked similar unemployment levels between 1980–2000, the subsequent decade saw metro-adjacent tiny towns experience higher unemployment levels – fully 2 percentage points distinction by 2010. Again, this is likely associated with age composition given the proportionally older populations in remote, declining tiny towns due to outmigration of working-age adults. Such individuals may report not being part of the civilian labor force. This is reflected in the substantially lower unemployment rates in remote tiny towns that have experienced population loss over the period 1980–2010.

Overall, with regard to financial capital, national gains in median household income have not been matched in small town America, although poverty levels have predominantly tracked national trends in that it appears improvements in poverty are related to relatively older populations and the social saftey net. Indicators of labor force attachment further indicate the growth of elderly populations in tiny towns with increasing populations. While holding some benefit, such shifts in age composition can negatively impact potential labor pools and, therefore, financial capital.

Physical/Built Capital

Physical and built capital represents infrastructure and can usefully indicate economic development potential. For our purposes, we use two measures of physical and built capital that indicate connections between rural areas and the larger economy, both physically and through communication. Indicators of market access logically suggest that tiny towns in metro-adjacent counties have better access to markets, and this better access characterizes both places that gained and lost population between 1980 and 2010. Even remote tiny towns that gained population exhibited lower levels of market access, suggesting that such connectivity does not necessarily lead to population gains. In all cases, small towns both metro-adjacent and not, had lower levels of market access than the national average (Figure 6, panel 6a).

Figure 6:

Physical/Built and Natural Capital by Population Trajectory and County Metro Proximity

Broadband access also represents connectivity. Panel 6b reveals little distinction by metro proximity for household access to broadband in small towns across the conterminous U.S. Again, the primary distinction is as compared to the national average -- access is more generally lower in small towns regardless of adjacency. Disaggregating by gain/loss does not add substantial understanding, although it is intriguing that the more isolated tiny towns with population gains exhibit higher connectivity, perhaps related to documented migration to amenity regions by those working remotely (Ulrich-Schad and Qin 2018).Natural Capital Prior research using county scale data clearly suggests that natural amenities shape both demographic change and economic development potential (McGranahan 1999). The patterns revealed making use of place scale data concur. Recall that we generate a place-based amenities measure using the characteristics similar to those underlying the USDA’s ERS county-level amenities index. Specifically, the index reflects climate (temperature, days of sunlight, humidity), topographic variation and water area -- environmental qualities most people prefer (McGranahan 1999).

High levels of natural amenities are associated with growth in more remote, typically not metro-adjacent tiny towns, consistent with prior research documenting in-migration to amenity-rich regions (Ulrich-Schad and Qin 2018). On the other hand, low levels of amenities characterize tiny towns in in both nonadjacent and metro-adjacent counties with declining population from 1980 to 2010. Tiny towns with growing populations, proximate to metro areas have relatively lower levels of natural amenities, suggesting that natural capital may not be as salient to growth in communities with other forms of capital (panel 6c).

Social and Cultural Capital

As noted above, racial and ethnic composition can influence social connectedness. While the population of tiny towns in the two categories reflecting metro adjacency is predominantly white, this is decreasingly so (Figure 7). Tiny towns in metro-adjacent counties show higher levels of racial diversity, yet the overall similarity of patterns of change is clear. Although generally less diverse, nonadjacent tiny towns have also experienced increases in racial diversity. Notably, racial diversity is higher – reflected by lower levels of percent-white – in tiny towns with population increases from 1980 to 2010. These demographic shifts are primarily resultant of large-scale movement of Hispanic populations into rural America (Jensen 2006; Lichter 2012). That said, overall proportion Hispanic population remains much lower in tiny towns as contrasted with the national average, although higher in areas experiencing growth (Figure 7, panel 7b) – consistent with the broader literature (Lichter 2012).

Figure 7:

Social and Cultural Capital by Population Trajectory and County Metro Proximity

As noted above, some argue that income inequality can influence social capital due to community-level class divisions. A higher Gini index suggests higher inequality, where a small percentage of the population receives the majority of income. As shown in Figure 7c, tiny towns in metro-adjacent and not-adjacent counties tracked similar levels of inequality and the Gini Indices were lower than the national average. Interestingly, while levels of inequality nationally increased in 2010 – reflected by a higher Gini Index – inequality declined for both metro-adjacent and nonadjacent tiny towns during the same time period. Lower levels of inequality are evident in towns that experienced population declines from 1980 to 2010 compared to towns that experienced population gains during the same time period.

In all, tiny towns fare relatively well on indicators of social capital given that racial homogeneity and low levels of income inequality may reflect higher levels of connection. That said, scholars have noted two types of social capital – bonding and bridging. Bonding social capital is developed within homogenous groups that share a similar identity (Szreter and Woolcock 2004) while bridging social capital refers to those social connections that link individuals across different identities, such as gender or occupation (Hogg and Abrams 1988). In the case of racial homogeneity and income inequality, bonding social capital may be particularly strong, although the indicators used here reveal little in relation to bridging social capital.

Our final community capital, cultural capital, is especially challenging to measure. As a start, we examine the presence of historic sites in tiny towns, and explore any variation for places according to their county’s metro proximity. As shown in Panel 7d, over 50 percent of tiny towns do not have any historic landmarks – disaggregated by metro proximity, 43 percent and 40 percent of counties adjacent to metro and not-adjacent (respectively) have at least one. For towns that experienced population growth between 1980 and 2010, a higher percentage of these towns had at least one historic landmark (48 percent) compared to towns that experienced population declines (38 percent) during the same time period.

Up to this point, our descriptive analyses suggest that while metro-adjacent and nonadjacent places differ from the national average in terms of our variables of interest, they appear to be changing in similar fashion to one another. We test this hypothesis of ‘no difference’ using the regression framework described in equation 1 and report our results in Table 3. For our purposes, we focus on the interaction terms to identify whether the attributes of metro- adjacent places are increasing or decreasing relative to nonadjacent places, more quickly than would be expected based on the differences observed in the 1980 base period.

Table 3:

Test of Statistical Significance of Difference in Trends

	College graduates (%)	Population 65+ (%)	Employment-to-population	Median household income(1)	Individuals in poverty (%)	Unemployment (%)	Market Access	White residents (%)	Hispanic residents (%)	Gini coef.
Not-adjacent (ref = adjacent)	0.222	1.741***	−0.214	−0.066***	0.896	−0.830**	−7.627*	3.517***	−0.006	−0.007**
	(0.333)	(0.321)	(0.464)	(0.012)	(0.459)	(0.269)	(3.383)	(0.673)	(0.380)	(0.003)
Year (ref = 1980)
1990	0.904*	1.108**	2.675***	0.48***	2.492***	−0.348	81.315***	−1.515	0.419	0.018***
	(0.391)	(0.379)	(0.546)	(0.015)	(0.540)	(0.316)	(3.988)	(0.794)	(0.448)	(0.003)
2000	2.876***	−0.879*	5.191***	0.919***	−0.376	−1.488***	111.892***	−3.746***	1.760***	0.002
	(0.391)	(0.379)	(0.546)	(0.015)	(0.540)	(0.316)	(3.988)	(0.794)	(0.448)	(0.003)
2010	4.609***	−1.089**	3.566***	1.151***	2.423***	1.798***	80.535***	−5.134***	3.500***	−0.021***
	(0.391)	(0.379)	(0.546)	(0.015)	(0.540)	(0.316)	(3.988)	(0.794)	(0.448)	(0.003)
Interactions
1990 × Non-adjacent	0.107	0.555	−0.402	−0.018	0.255	0.312	−8.528	0.083	−0.084	−0.001
	(0.470)	(0.454)	(0.656)	(0.018)	(0.648)	(0.380)	(4.784)	(0.952)	(0.537)	(0.004)
2000 × Non-adjacent	0.127	0.995*	−0.156	−0.019	−0.343	0.495	−11.327*	−0.071	−0.391	0.002
	(0.469)	(0.454)	(0.656)	(0.018)	(0.648)	(0.380)	(4.784)	(0.952)	(0.537)	(0.004)
2010 × Non-adjacent	0.199	1.250**	1.424*	0.012	−1.071	−0.825*	−7.140	0.006	−0.690	0.003
	(0.469)	(0.454)	(0.656)	(0.018)	(0.648)	(0.380)	(4.784)	(0.952)	(0.537)	(0.004)
Constant	9.458***	19.273***	49.149***	9.369***	16.033***	7.623***	55.391***	91.400***	2.032***	0.385***
	(0.277)	(0.268)	(0.386)	(0.01)	(0.382)	(0.224)	(2.820)	(0.561)	(0.317)	(0.002)

N = 10743

^*p<0.05

^**p<0.01

^***p<0.001

Notes:

⁽¹⁾Median household income values were log-transformed.

Our models produce significant interactions for four of our variables of interest: the percentage of the population aged over 65, the employee-to-population ratio, percentage of unemployed, and market access. The significant interaction terms in Table 3 show that the percentage of the population aged over 65 has been growing more quickly in nonadjacent than metro-adjacent places. In nonadjacent places in 2000 and 2010, the share aged over 65 is roughly a full percentage point higher relative to metro-adjacent places than would be expected based on the 1980 difference. Employment-to-population ratios also grew more quickly in nonadjacent places in 2010 relative to metro-adjacent places relative to 1980. Conversely, unemployment declined more quickly in nonadjacent places in 2010 relative to 1980. Table 3 shows very large and statistically significant effect size for the difference in market access in 2000, which imply dramatic differences in the degree to which adjacent places have gained their market access relative to nonadjacent places. These estimates imply, therefore, that nonadjacent places are becoming significantly older and more economically isolated than metro-adjacent places through time. But conversely, nonadjacent places experienced lower rates of unemployment gains (and fewer declines in employment-to-population ratios) relative to metro-adjacent places in 2010. With this said, it should be noted that as our remaining six variables of interest show little significant variation, there is not evidence of widespread differences between places in nonadjacent and metro-adjacent counties.

Discussion and Conclusion

This descriptive profile of small town America is unique in its place-based lens – the county scale has thus far dominated rural demographic research. To inform both future research and policy, we contend that it is important to acknowledge the distinctions between counties and the places within them – and data presented here suggest there is often mistmatch that characterizes county vs. place-based data. Another important extension of this project is the nationwide application of the Community Capitals Framework (CCF) which builds on the body of research examining capitals within small numbers of case studies.

The descriptive analyses provided above offer a baseline understanding of sociodemographic conditions and trends in small communities across the contiguous U.S. The analyses have been structured by three research questions, the response to each is noted below.

1). How has small town America changed socio-demographically in the past several decades?

Our study communities – places in completely rural counties and with less than 2,500 population in 1980 – generally lost population between 1980–2010 (1.5 percent on average). That said, nine percent of such places experienced growth of over 41 percent (the top tertile of growth), with clusters characterizing amenity regions such as the mountain west, the coastal southeast, and northern Michigan.

Although population sizes have, on average, declined, tiny town community capitals have trended similar to the nation as a whole – this is true for most indicators of human, financial, physical, social and culture capitals. Only three key distinctions emerged. First, the proportion of residents with college education, an important human capital, lacked the same level of growth in tiny towns relative to the U.S. Also lacking was comparable growth in median household income, an important financial capital. On the other hand, income inequality as measured by the Gini Index has been declining in small town America while increasing at the national scale. As included here, this inequality is reflective of social capital as lower inequality is anticipated to yield more social cohesion.

2). Are trends in small town change distinct according to proximity to metropolitan areas?

Another key finding is that overall population loss/gain varies dramatically by county proximity to metro areas. Between 1980–2010, tiny towns in counties not adjacent to metro areas lost 8 percent population, on average, as contrasted with 11 percent gains by small towns in metro-adjacent counties. Yet, the socio-demographic trajectories of tiny towns in both metro-adjacent and nonadjacent counties are in many ways comparable. Instead of distinct trends, it is overall levels of community capitals that differ by county metro proximity. For instance, tiny towns in nonadjacent counties have older populations, but the trajectory of change parallels tiny places elsewhere. Also, tiny towns in both metro-adjacent and nonadjacent counties demonstrate relatively low levels of financial capital in the form of household income as contrasted with national trends, while also being characterized by relatively low levels of physical/built capital. More specifically, market access is low, especially in nonadjacent places, while relatively low levels of broadband access characterize places regardless of county metro adjacency. Community natural capital is also lacking, except for a noticeable connection between amenities and small places with growing populations, a finding that aligns well with current research on amenity migration.

3). Are trends distinct as related to longer-term patterns of population growth and decline?

There are not substantial differences in trends across time or across places that are losing/gaining population. Instead, similar to the response to Research Question #2, what differs are levels of capitals characteristic of small places as contrasted with national data. For instance, human capital in the form of higher educational levels increased across the nation, 1980–2010, including in our study towns of < 2,500 residents (1980) in completely rural counties. However, small towns that have experienced population decline have, overall, the lowest proportion college graduates. The same pattern is revealed for the percentage of residents over age 65 such that while this is increasing across the nation, including in tiny towns, small towns that have experienced population decline have relatively higher proportions of this older demographic group. Combined, these trends may reflect rural brain drain, and suggest the process may characterize tiny towns regardless of their metro proximity.

Overall, the analyses suggest that change in small towns in “completely rural” counties in 1980 have experienced trends similar to the nation as a whole, although the overall levels of human, financial, and physical capitals tend to be lower. Also important, evidence is provided of the mismatch between place and county-based data indicating the necessity of considering sub-county units for both research and policy.

The work presented here forms an important foundation for ongoing research exploring multidimensional trajectories of small town America. But of course there are data limitations, both with regard to included measures and definitions, and to those measures that remain missing. For example, there are inherent limitations in some measures such as the inability of unemployment data to reflect underemployment nor informal work, both important dimensions of rural economic well-being (Slack, Thiede and Jensen 2018). Social and cultural capital also represent key challenges with regard to operationalization and we continue to explore additional indicators of this form of community connectedness and identity. This work could also be enhanced by moving away from the use of county scale definitions of the urban-rural continuum. The current analysis is shaped by USDA county categories, but even more precise categorization will use place-based characteristics to inform categorization that will better reflect local conditions. This also relates to a limitation regarding generalizability. This project has emphasized particularly small places in “completely rural” counties per USDA categorization. Given this focus, the intent is not to generalize but to draw attention to unique conditions and challenges within such rural spaces. The development of place-based distinctions instead of county-based characterizations will likely result in a more nuanced reflection of variation in local conditions. This remains part of our ongoing research and longer-term contribution. Another aspect of our ongoing research is investigation of population change making use of restricted data reflecting in- and out-migration streams at the place scale. This work will add substantial depth to understanding of socioeconomic and demographic trends and transitions within small town America.

Limitations aside, this study provides the first systematic, national examination of recent demographic and socioeconomic trends in very small communities – and one of the few studies focusing on small places since Johansen and Fuguitt’s analysis nearly four decades ago research with several case studies. While much commonality exists across tiny towns regardless of county metro proximity, it is important to emphasize that those in counties not adjacent to metro areas have, indeed, demonstrated distinct patterns of aggregate population change, 1980–2010, losing 8 percent population as contrasted with 11 percent gains by small towns in metro-adjacent counties. Such trends may portend future sociodemographic challenges and much work remains to be done to inform development of place-based policy. Further, while overall trends in tiny town community capitals may parallel broader, national shifts, the lower levels of specific community capitals likely necessitate distinctive programs and policies.

As noted at the onset, this investigation has provided an important foundation for future work designed to improve understanding of the particular challenges faced by tiny towns in the U.S. and the potential demographic, social, economic, and health futures they might face. This effort has but scratched the surface. Our own agenda will continue to shed light on the unique circumstances of tiny towns and we hope to have inspired others to also explore rural America at the place-scale. As an evidence base emerges for place-focused analyses, the unique challenges of small town America will come into greater focus. Based on this insight, informed decisions can then be made to overcome the urban-centric nature of much U.S. social, economic, and environmental policy.