the impact of level of rurality on suicide rates: an

61
Haffey 1 1 THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN ANALYSIS OF COMBINED EFFECTS OF KNOWN RISK-FACTORS AT THE COUNTY LEVEL By Kieran Thompson Haffey University of Colorado Boulder Defended April 7th, 2020 Thesis Advisor: Dr. Tim Wadsworth, Department of Sociology Thesis Committee: Dr. Tim Wadsworth, Department of Sociology Dr. Lori Hunter, Department of Sociology Dr. Karl Hill, Depart of Psychology

Upload: others

Post on 27-Feb-2022

4 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

1

1

THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES:

AN ANALYSIS OF COMBINED EFFECTS OF KNOWN RISK-FACTORS AT THE

COUNTY LEVEL

By

Kieran Thompson Haffey

University of Colorado – Boulder

Defended April 7th, 2020

Thesis Advisor:

Dr. Tim Wadsworth, Department of Sociology

Thesis Committee:

Dr. Tim Wadsworth, Department of Sociology

Dr. Lori Hunter, Department of Sociology

Dr. Karl Hill, Depart of Psychology

Page 2: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

2

2

Abstract

Suicide remains a leading cause of death in the United States. Suicide risk is shaped by a

number of sociodemographic characteristics in addition to where people live. In this project,

county suicide rates were examined alongside indicators of known risk factors for suicide

including level of urbanization, geographic division, economic disadvantage, religious

adherence, marriage and divorce, gender and racial categories, and education. Regression models

indicate that rurality influences the relationship between sociodemographic variables and rates of

completed suicide. Central findings include the combined effect of rurality and education,

rurality and region, and rurality divorce on suicide rates. One implication of this is research is a

greater need to consider how context shapes suicide risk when designing programs and policies

to prevent suicide.

Page 3: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

3

3

Acknowledgements

I would like to thank Dr. Wadsworth for his immeasurable contributions to this work and

to my academic experience in my fourth year. I am grateful for everything he has taught me and

for the generous time he committed to teaching me.

I would like to thank Dr. Hunter for her continued advice and shaping of this work, and

for helping me overcome difficulties in each step of the process. Her contributions to structuring

this process were foundational to this work.

I would like to thank Dr. Peterson-Gallegos for her contributions during the inception of

this process and indulging every direction I would have liked to go, but could not have. Her class

and teaching inspired this topic and it would not be what it is now without that.

I would like to thank Dr. Karl Hill for taking time to provide me with relevant literature

and the taking time to participating in this process.

Finally, I would like to thank my parents for being so supportive during my college

career. I would surely not have been able to do it without everything they’ve done for me.

Page 4: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

4

4

Table of Contents

I. Introduction ............................................................................................................................. 5

II. Lit Review ............................................................................................................................... 6

a. Suicide in Sociology ............................................................................................................... 6

b. Predictors of Suicide ............................................................................................................... 8

i. Level of Urbanization ............................................................................................................. 8

ii. Regional Differences in Suicide ............................................................................................. 9

iii. Economic Disadvantage ....................................................................................................... 10

iv. Marriage and Divorce ........................................................................................................... 12

vi. Race and Ethnicity ................................................................................................................ 15

vii. Education .............................................................................................................................. 17

c. Level of Urbanization as a Moderating Variable .................................................................. 18

III. Data and Methods ................................................................................................................. 19

a. Data ....................................................................................................................................... 19

i. CDC Data .............................................................................................................................. 20

ii. USDA Data ........................................................................................................................... 21

iii. NHGIS Data ......................................................................................................................... 22

iv. ARDA Data ........................................................................................................................... 23

v. BSGU Data ........................................................................................................................... 23

b. Method .................................................................................................................................. 24

IV. Analysis ................................................................................................................................ 26

a. Descriptive Statistics ............................................................................................................. 26

b. Multivariate Models .............................................................................................................. 28

c. Multivariate Models with Interaction Terms ........................................................................ 36

d. Does Level of Rurality Matter? ............................................................................................ 45

V. Conclusion ............................................................................................................................ 46

a. Main Findings ....................................................................................................................... 46

b. Limitations ............................................................................................................................ 46

c. Suggestions for Further Research ......................................................................................... 48

Works Cited .................................................................................................................................. 51

Page 5: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

5

5

I. Introduction

Since 1975, suicide has remained among the 12 leading causes of death in the United

States and ranked 10th in 2017 (Stone et al. 2017). From 2005-2015, suicide rates have risen by

more than 10% in 99% of counties. These increases have been highest in rural areas (Rossen et

al. 2018).

Suicide rates do not only vary over time, but between individuals, households, and places

as well. Demographic characteristics such as age, gender, race, and ethnicity are associated with

suicide risk, but their influence is also shaped by additional factors, such as religion, economic

shock, and education (Barranco 2016; Carriere et al. 2019; Kubrin and Wadsworth 2009; Shah

and Bhandarkar 2009). Mental illness and substance abuse are also relevant to suicide risk

(Heron 2019; Stone et al. 2017). Risk factors at the household and national levels include social

support, financial distress, access to mental health resources, and means to commit suicide

(Stone et al. 2017).

Known risk-factors for suicide have been studied extensively, some of the most

prominent in the sociological canon being economic deprivation, marriage and divorce, religion,

race, and education, many of which were studied by Durkheim in his seminal study, Le Suicide

(Allen and Goldman-Mellor 2018; Barranco 2016; Bjorkenstam et al. 2011; Durkheim and

Simpson 1999; Kubrin, Wadsworth, and DiPietro 2006; Stack and Kposowa 2016). One gap in

the understanding of suicide is how intersecting risk-factors impact suicide rates and how these

intersections vary. For example, one frequently studied correlate of suicide rates, level of

urbanization, was shown to influence the effect of economic shock (Carriere et al. 2016).

Beynon, Crawley, and Munday (2016) show that level of urbanization is inherently

difficult to quantify. Level of urbanization cannot be exclusively measured by population size or

Page 6: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

6

6

density, rather, urbanization encompasses a number of diverse characteristics including

institutions and socioeconomic indicators (Beynon, Crawley, and Munday 2016). Moreover,

level of urbanization may impact factors such as economic deprivation, educational attainment,

or religious adherence. In the interest of public health, it is prudent that the combined effect and

level of urbanization and sociodemographic correlates of suicide rates be investigated. The

pertinent research questions are: How does race, gender, education, religion, economic

deprivation, and marriage and divorce relate to a given area’s suicide rate? And do these

relationships vary between rural and urban areas?

II. Lit Review

a. Suicide in Sociology

The sociological study of suicide stems from Emile Durkheim’s foundational 1897 work,

Le Suicide (Durkheim and Simpson 1999). His study influenced both the method and

conceptualization of the topic, illustrated by the expansion of statistical analyses, and studying

suicide alongside social, cultural, and economic components of social order. His empirical

methods marked a deviation from traditional moral and theological discussion of suicide and

instead examined it as a product of the structures and patterns of society. One of his greatest

contributions was his examination of structural factors, such as religion, economics, culture,

rurality, and social interaction.

One of the first incantations of a structural explanation of suicide comes from Peuchet’s

1838 work in Memoirs from the Police Archives of Paris which noted the interplay between

individuals and social contexts such as the family unit, and poverty (Tierney 2010). Peuchet’s

proto-statistical approach studied suicide mortalities in Paris and the connection to the expansion

of liberalism in the social environment at the time. In doing so, Peuchet expanded the focus on

Page 7: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

7

7

suicide to be of societal concern, and not just that of the individual (Tierney 2010).

Notwithstanding some of his psychological analysis, his work signifies an intermediate step

between Hobbesian perspectives of insanity and illness, and Durkheim’s refutation of such

individualistic causes of suicide (Durkheim and Simpson 1999). Between Peuchet and

Durkheim, suicide’s categorization as a social issue illustrated its institutionalization and paved

the way for a multi-dimensional approach to suicide, beginning with Durkheim’s theory of social

integration and regulation (Tierney 2010).

In Le Suicide, Durkheim proposed social integration and social regulation as the chief

causes of suicide (Durkheim and Simpson 1999). Too much or too little of either resulted in

excessive or insufficient influence of social norms and values. He posited that the largest

plurality of suicides were what he called ‘egoistic’, which stemmed from too little integrative

forces (Durkheim and Simpson 1999). Social integration has since been reconceptualized and

defined in terms of, but not limited to: economic opportunity and its relationship to social class,

access to institutional resources and liberties, symbolic power stemming from social capital,

subsequent opportunities and resources provided by social, economic, and cultural linkages

between individuals and groups (Leončikas 2004; Portes 2000).

The dynamic definition of social integration has dominated the sociological discussion of

suicide for some time. It can be argued that social integration has accommodated societal

context. That is, social integration has been defined differently over time, and has accompanied

discourse of suicide during the transition from traditional to modern societies alongside the

proliferation of post-materialist values (Kamali 1999). As such, indicators of social integration

that have guided the sociological study of suicide merit attention. Some of the most prominent

Page 8: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

8

8

include economic disadvantage, family dynamic, religion, race, education, and level of

urbanization.

b. Predictors of Suicide

i. Level of Urbanization

Research on how level of urbanization effects suicide rates has long produced similar and

significant results (Carriere et. al 2018; Durkheim 1897; Fontanella et. al 2015; Gopal and

Siahpush 2002; Helbich 2017; Levin and Leyland 2005). Rural areas are consistently associated

with higher rates of suicide, even when considering more than a dozen models of classifying

urban and rural areas (Helbich 2017). Additionally, there appears to be a gradient in suicide rates

dependent on how urban or rural a given area is. Levin and Leyland (2005) find that rural areas

with greater accessibility to urban areas experience lower suicide rates compared to rural areas

with little urban access. While this finding is corroborated by other research, some metrics used

to measure level of urbanization are similar to those that measure economic disadvantage, which

is itself related to suicide (Law, Snider, and De Leo 2014).

Research has shown the widening gap in suicide rates between urban and rural areas.

From 1970-2010, the gap between rural and urban suicide has grown, most significantly for men

(Fontanella et. al 2015; Rossen et. al 2018). Interestingly, there is evidence to support that

suicidal ideation does not occur in rural populations at higher rates. Higher levels of suicidal

ideation occur in more urban populations, although it is possible certain norms of religiosity,

individualism, and stigmatization impacts how the ideation is handled (Hirsch and Cukrowicz

2014). This is partly indicated by the finding of lower use of mental health resources in rural

areas after controlling for availability issues. Further, the finding that non-fatal suicidal actions

are more prevalent in urban populations illustrates that differences in mental health and access to

Page 9: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

9

9

mental health resources paint an incomplete picture of the differences between urban and rural

suicide (Allen and Goldman-Mellor 2017).

Means-restriction has received attention in recent literature following findings that

differences in the approach to suicide exist between rural and urban areas (Barber 2014; Hirsch

and Cukrowicz 2014; Levin and Leyland 2005; Widger 2018). Some examples include

disproportionate use of pesticide self-poisoning and firearms in rural versus urban areas (Hirsch

and Cukrowicz 2014). Suicide by firearm is more prevalent in rural areas than in urban ones.

Pesticide self-poisoning is a pattern in rural areas on the global scale, where it is the most

common means to commit suicide. Up to 2/3 of pesticide self-poisonings in rural areas are

intentional and suicidal (Hirsch and Cukrowicz 2014). There is considerable research showing

there are disproportionately high suicide rates in rural areas, it is less clear how rurality shapes

the influence of mechanisms and covariates of suicide. Investigating this is one of the main goals

of this research.

ii. Regional Differences in Suicide

In the United States, region has consistently been associated with suicide rate (Haws et al.

2009; Hemenway 2002; Miller, Azrael, and Hemenway 2002; Pepper 2017). The highest rates of

suicide are in the west and midwest, and lowest in the northeast. The four census regions (west,

midwest, south, and northeast) encompass nine smaller geographic divisions: pacific west,

mountain west, west north central, east north central, west south central, east south central, new

England, middle Atlantic, and south Atlantic. The relationship between census division and

suicide rates is clearer than using region and is more frequently used as a unit of analysis

(Hemenway 2002; Miller, Azrael, and Hemenway 2002; Pepper 2017). The higher suicide rate in

the western U.S is centralized in the mountain west, followed by the pacific west (Hemenway

Page 10: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

10

10

2002). Pepper (2017) has found that suicide rate in the mountain west is relatively higher than

most other divisions across demographic groups, indicating there may not be an

overrepresentation of an at-risk group, rather, there is “support for a possible culture-of-suicide

script that pervades the region, crossing demographic distinctions” (Pepper 2017: 348). Level of

urbanization may contribute to this relationship given the different proportions of regions made

up of urban and rural counties.

iii. Economic Disadvantage

Economic disadvantage has been a thoroughly researched predictor of suicide rates since

Durkheim first included it in Le Suicide (Durkheim and Simpson 1999). Economic disadvantage

(more specifically, poverty) was initially designated as a function of excessive social regulation.

More recent literature has since revised the concept to include access to economic resources, and

its relationship to status formation and symbolic resources (Leončikas 2004). There is mixed

evidence supporting the relationship between socioeconomic status and suicide rate (Daly,

Wilson, and Johnson 2007; Kim et al. 2016; Purselle et al. 2009). It can be discerned from the

literature that the direction of the relationship is highly conditional to other sociodemographic

factors.

Economic recession has been associated with increased suicide rates, with the largest

effect on female suicide rates. This finding is strongest in rural communities where children

comprise more of the population (Carriere, Marshall, and Binkley 2019). Carriere et al. (2019)

suggest that economic hardship may encourage a higher female labor force participation rate.

The acute stressors that accompany the dual role many women play in domestic labor and the

formal labor market contribute to exacerbated effects of economic disadvantage. Interestingly,

female labor-force participation rates have been found to reduce suicide rates in both men and

Page 11: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

11

11

women (Chen et al. 2017). The discrepancy between these two findings is suggested to reflect

risk factors associated with suicide, emphasizing whether female labor-force participation stems

from more egalitarian norms versus economic necessity (Chen et. al 2017).

Some studies suggest that economic disadvantage is not restricted to unemployment rates

and income, but the type of employment as well. Kubrin et al. (2006) have found that Wilson’s

deindustrialization thesis is highly relevant to racial disparities in social ills stemming from the

economic transition away from manufacturing-oriented economies. The effects on unstable

employment and more segregated communities have a two-fold effect. Kubrin et al. (2006)

suggest that the concentration of low-level service-oriented jobs for African Americans in urban

communities hinders social networks used to find employment. The increased suicide-risk in

these communities was thought the be a result of the economic racial inequality, when in fact, it

was found to be based on absolute disadvantage (Kubrin et al. 2006). This kind of economic

marginalization is a risk-factor for suicide in both white and black adolescents, although when

gun availability is included in the analysis, suicide risk increases for black adolescents only

(Kubrin and Wadsworth 2009).

In contrast to Kubrin et al. (2006), Allen and Goldman-Mellor (2017) have found that

suicide rates are not significantly different between economically disadvantaged and non-

disadvantaged communities. Suicide rates were instead associated with the perception of

economic advantage or disadvantage between communities. These findings obfuscate the

prevailing literature which links lower socioeconomic status to increased suicide risk. Thus, the

importance of subjectivity and perception in understanding one’s socioeconomic status and

economic deprivation calls for a more nuanced approach to studying the interplay of economic

disadvantage and other community characteristics and encourages us to look at both absolute and

Page 12: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

12

12

relative disadvantage (Allen and Goldman-Mellor 2017). Level of urbanization may be relevant

to the relationship between economic disadvantage and suicide because of varying distribution of

wealth or the type of industry that supports the economy (agrarianism, for example).

iv. Marriage and Divorce

Marriage and divorce have been found to impact suicide rates since Durkheim’s work in

Le Suicide. He suggested there is a protective effect of marriage on both men and women

(Durkheim and Simpson 1999). The mechanism by which marriage and divorce are associated

with suicide rates remains debated. There is mixed evidence whether it is due to economic stress

caused by divorce, psychological effects, reduced social integration, or composition of the family

unit (Carriere et al. 2019; Durkheim and Simpson 1999; Kposowa 2000). The association

between divorce and suicide has been commonly studied through status integration theory and

frameworks of social capital that stress bonding effects in social support. Status integration

theory adopts a neo-Durkheimian approach that postulates suicide rates are inversely related to

the stability of social relationships, predicated upon the compatibility between an individual’s

role and its respective social expectations. This theory is supported by gender disaggregated

divorce and suicide data but fails to address the full suite of differences in suicide between men

and women when consider other covariates of suicide (Stack and Kposowa 1990).

Kposowa’s (2000) findings support status integration theory by showing that single

people do not experience such an advanced risk of suicide when compared to their divorced

counterparts. This may be due to the consistency in norm expectation and fulfillment for single

people and lack thereof for divorcees. That is, the adjusted norms divorcees and widows

experience are incongruous with their established roles as spouses, whereas singles’ roles are in

accordance with the norms expected of them. Stack and Kposowa’s (1990) findings also suggest

Page 13: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

13

13

that the effect of divorce may be more psychological among men, and economic among women.

For men, they enjoy greater economic independence based on conventional gender roles of

breadwinner vs. wife/mother. Meanwhile women may enjoy greater social support, but not the

economic benefits. Correll, Benard, and Paik (2018) find that many women face a reduction in

wages after becoming mothers, which further entrenches the gendered differences between men

and women when it comes to participation in formal and informal labor. This gendered division

of labor can thereby contextualize the respective economic and social impacts of divorce

between men and women (Carriere et. al 2018; Stack and Kposowa 1990). Level of urbanization

may be important to the relationship of marriage, divorce, and suicide because of the role of

economic dependence, or the importance of social capital in marriage and divorce.

v. Religion and Religious Adherence

Religion plays a significant role in establishing meaning and interpretations of suicide.

Major Judeo-Christian belief systems condemn suicide and deem it an act of murder, or rejection

of purpose given by god (Kastenbaum 2007). The study of religion as a protective force against

suicide was pioneered by Durkheim, who theorized that the structure of the belief system and the

subsequent bonds between the individuals, church, and clergy all contribute to differential rates.

This was particularly salient in his discussion of Catholics and Protestants, citing the relatively

lower social regulation of the Protestant Church (Durkheim and Simpson 1999). This framework

that suggests the structure of church is the protective mechanism is contrasted by Barranco’s

(2016) research on religion, nationality, and suicide. Barranco (2016) found that U.S-born

Latinos that remain in the U.S benefit from religious adherence regardless of the denomination,

whereas Latinx immigrants strictly benefitted from adherence to a Catholic church. A theoretical

approach to this relationship is the community norms thesis, which posits localities which share

Page 14: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

14

14

values and norms foster stronger and more cohesive adherence to such norms. The community

support thesis proposes this adherence is the result of stronger social bonds and networks. In the

case of religion, both of these theses support the protective effect against suicide. However, the

reduction in social bonds and networks faced by some immigrants provides greater support for

the community norms thesis (Barranco 2016). Teasing apart shared norms and the social capital

provided through religious networks may not be feasible, given the relationship between them.

Religion’s impact on suicide rates remains unclear, both in effect, and in mechanism.

Some research that suggests protective effect of religion are rooted in the belief system and

moral commitment. Osafo et al. (2013) finds that hope provided through prayer is a common

route many take to address underlying mental health issues. However, this leads to the

understanding that religion only manages suicidal ideation, though does not prevent it. In

contrast to this nuance, Lawrence, Oquendo, and Stanley (2016) find that those with a moral or

religious objection to suicide report less suicidal ideation than those who do not. Regardless,

preventing suicide is valuable whether it is palliative or preventative.

Some research has found that not all religious commitment provides protective effects.

Pargament et. al (1998) finds that while the majority of coping mechanisms provided through

religion have positive psychological effects, a minority of respondents experience a greater sense

of abandonment by God and cite the hopelessness and powerlessness at the hand of a higher

power. These coping mechanisms were associated with higher levels of stress and depression,

which are associated with suicide risk (Placidi et al. 2000). The single strongest body of evidence

for the role of religion is the impact of religious attendance, which reduces suicide risk (Burr et

al. 1994; Garroutte et al. 2003; Lawrence, Oquendo, and Stanley 2016; Moore 2015). Moreover,

Page 15: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

15

15

it is suggested that the role of religion be understood and included in approaches to preventing

suicide (Gearing and Lizardi 2009).

The relationship between religion and education complicates previous findings of the

relationship between religion and suicide. Research shows that increasing educational attainment

is associated with lower rates of suicide, although higher levels of education attainment are also

associated with decreasing adherence to religion, or secularization (Phillips and Hempstead

2017; Stack and Laubepin 2019). In this way, religion is both a protective factor and a risk factor

for suicide. It follows that the effect of religion is, in part, dependent on the effect of education.

The effect of education capitulating that of religion has been speculated to be the result of

materialistic values absorbing roles that were previously held by religious values. This

connection between secularization and suicide is in reference to economic strain theory and

deviance (Stack and Laubepin 2019). Level of urbanization may play a role in the relationship

between religion and suicide rate when considering the cohesion of social norms and networks.

The size of a county could indicate the number of religious organizations that may influence

such norms and networks.

vi. Race and Ethnicity

In the United States, there is significant variation in suicide across racial and ethnic

groups. Native Americans commit suicide at the highest rate, though predictions vary based on

whether undetermined intent is included (Stone et. al 2017). That is, different racial and ethnic

groups have substantially different mortality rates with undetermined intent, meaning the actual

suicide rates between groups may be different than is reported. Non-Hispanic whites have the

second highest suicide rate, distantly followed by Asian and Pacific Islanders, Hispanic and

Latinx, and African American populations (Suicide Prevention Resource Center 2019). In 2017,

Page 16: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

16

16

the suicide rate for Native American populations was 22.15/100,000. In descending order of

white, Hispanic, African American, and Asian/Pacific Islander, the rates for 2017 were 17.83,

6.89, 6.85, and 6.75, respectively (Suicide Prevention Resource Center 2019). Some scholarship

suggests the relatively high rate among Native Americans is the result of higher rates of

substance abuse, specifically alcoholism (Kastenbaum 2007). An additional factor that varies

between racial groups is lethality of attempt, or method.

Durkheim argued the difference in suicide rates between racial and ethnic groups can be

explained by the different exposure to certain social, political, cultural, and economic conditions

(Durkheim and Simpson 1999). In the late 20th century, this basic principle was applied in status

integration theory. It posited that the narrowing of gaps between White and Black populations in

social and political liberties stemming from more equitable economic opportunities could

contribute to a rise in the suicide rate among African Americans (Joe et al. 2007; South 1984).

South (1984) argued the assimilation effects of the civil rights period led to the mainstreaming of

social liberties for African Americans, potentially reducing constraining social norms. In a

Durkheimian framework, “adequate” social regulation was protective against suicide (Durkheim

and Simpson 1999). It would follow that underlying sociopolitical or socioeconomic norms that

could influence the suicide rate in white populations could impact black populations in the same

manner. Suicide rates over time have not shown this (Suicide Prevention Resource Center 2019).

The varying proportions of race and ethnicity in the population across the U.S. may indicate a

significant role of level of urbanization in the relationship between race and suicide as various

racial and ethnic groups are unequally distributed across rural and urban areas.

Page 17: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

17

17

vii. Education

Educational attainment is associated with suicide risk and has received much attention in

the sociological canon (Abel and Kruger 2005; Dubow, Boxer, and Huesmann 2009; Durkheim

and Simpson 1999; Geulayov, Metcalfe, and Gunnell 2016; Shah and Bhandarkar 2009). Some

findings have suggested that yearly increases in education reduce suicide rates by 2% in some

populations (Abel and Kruger 2005). Individuals with high levels of education illustrate a

significantly lower suicide risk compared to those with low levels, though the effect is less

pronounced in women (Lorant et al. 2005). It is suggested that the relationship between

educational attainment and suicide is conditional to the degree that education impacts

intelligence and its relationship to economic prosperity1 (Durkheim and Simpson 1999; Shah and

Bhandarkar 2009). Higher levels of education can also predispose individuals to suicide if the

expected economic gains from education are not reached (Shah and Bhandarkar 2009).

Therefore, the relationship could potentially be curvilinear (peaking both at low and high

educational attainment) dependent on what factors correlate with higher levels of educational

attainment (Pompili et. al 2012; Shah and Bhandarkar 2009). Lorant et al. (2005) corroborates

that there is a relationship between education and suicide risk, finding that lower class men and

women experience divergent effects of education on suicide. The relationship was positive for

women, and negative for men (Lorant et al. 2005).

In addition to the literature produced on the subject of educational attainment, increased

performance in school has been shown to be negatively associated with suicide risk. Even after

controlling for certain aspects of home living and family dynamics, the existing associations of

1 Intelligence in this cross-national study (Shah and Bhandarkar 2009) was measured by the UN Education Index.

Intelligence as both an independent and dependent variable illustrates a methodological difficulty in measurement.

Measures of intelligence span a number of approaches, and it is rare that individuals do well on each (Neisser et al.

1996).

Page 18: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

18

18

school performance and suicide risk persisted (Bjorkenstam et. al 2010). However, the

mechanisms that contribute to this are unclear. They could span the relationship between poor

school performance and antisociality or conduct disorders, the subsequent effects on income and

economic status, or the reduced ability to respond to mental health education (Bjorkorkenstam et.

al 2010). Research shows that there are regional patterns to educational attainment, and as such,

there is reason to believe that level of rurality may shape the relationship between education and

suicide rates (Wheeler and Pappas 2019).

c. Level of Urbanization as a Moderating Variable

Level of urbanization has been shown to moderate the relationship between economic

disadvantage and suicide, showing a stronger effect (of higher rates) in rural areas (Carriere et al.

2018). There is a lack of literature regarding level of urbanization as moderating the relationship

between a number of other known risk factors and suicide. This illustrates the need for more

research into the interaction effect of level of urbanization on known risk factors for suicide. The

evolving categorization of counties as urban or rural illustrates the variability of institutional and

societal characteristics relative to other counties. The significant differences in suicide rates

between level of urbanization calls for known indicators of suicide to be analyzed while

considering the level of urbanization. Based on the discussions above, my hypotheses are as

follows:

H1: Increasing level of rurality will be associated with higher suicide rates.

H2: Economic disadvantage, measured by poverty, unemployment rate, wealth

distribution, income per capita, and percent of workers in a professional occupation will be

associated with higher suicide rates.

H3: Higher educational attainment will be negatively associated with suicide rates.

Page 19: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

19

19

H4: Religious adherence will be negatively associated with suicide rates.

H5: Increasing racial homogeneity and percent of the population that is white will be

positively associated with suicide rates.

H6: Geographic divisions in the west will be associated with higher suicide rates.

H7: Marriage rates will be negatively associated with suicide rates and Divorce rates will

be negatively associated with suicide rates.

H8: The combined effect of rurality with each independent variable will be significantly

associated with suicide rates.

III. Data and Methods

a. Data

Suicide rates were obtained from the Underlying Cause of Death Dataset from the

Centers for Disease Control spanning 2013-2017, aggregated at the county level (Center for

Disease Control and Prevention 2018). Suicide rates are based on completed suicides under

injury intent. Measures of urbanicity and rurality were obtained through the 2013 Rural-Urban

Continuum Codes provided by the USDA Economic Research Service (2019). Data regarding

occupation, economic characteristics, education, and racial demography were obtained from the

2017 American Community Survey: 5-Year Data [2013-2017, Tracts & Larger Areas], provided

by IPUMS National Historical Geographic Information System (Manson et. al 2019). Religious

adherence data was obtained from the U.S. Religion Census: Religious Congregations and

Membership Study, 2010 (County File), provided by the Association of Religious Data Archives

(Grammich et. al 2010). Marriage and divorce data were obtained from the County-Level

Marriage & Divorce Data, 2010, provided by the National Center for Family & Marriage

Research (BGSU Data Compass 2016).

Page 20: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

20

20

i. CDC Data

Data are compiled and published by U.S Center for Disease Control and Prevention, the

Department of Health and Human Services, the National Center for Health Statistics, and the

Office of Analysis and Epidemiology (Centers for Disease Control and Prevention 2018). The 57

districts of the Vital Statistics Cooperative provide mortality data based on death certificates.

Data accessed through the CDC provided suicide rates per 100,000 people by geographic area.

Only deaths designated as suicides were used in this research. This excludes mortalities with

‘undetermined intent’, which are sometimes used in suicide research. Undetermined intent

mortality data was excluded due to the potential for overrepresenting or underrepresenting some

segments of the population, evident in the substantially different rates between gender and racial

groups, particularly in the case of self-poisonings in women and Native Americans (Suicide

Prevention Resource Center 2019a; Bjorkorkenstam et. al 2010). Some research suggests suicide

rates overall are underestimated by nearly 10% and that these suicides are instead classified as

being of undetermined intent; the majority of these are from self-poisoning (Bjorkorkenstam et.

al 2010; Donaldson et al. 2006). A potential issue with suggesting that ‘undetermined intent’ was

in fact ‘suicide’ is shown in methodology of retroactively determining intent. For example,

Donaldson et al. (2006) classified more than half of self-poisonings (overdoses) as suicides if the

given individual displayed suicidal behavior. This appears a priori and would be thereby remiss

to neglect research finding that suicidal ideation and attempts relatively common and do not

necessarily reflect suicide rates (Turecki and Brent 2016). Moreover, some literature contends

that while some suicides are misclassified as ‘undetermined intent’, the rates of misclassification

are contingent on age, sex, and race (Mohler and Earls 2001). Correcting the suicide rates of

Page 21: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

21

21

some groups by incorporating ‘undetermined intent’ would merely misrepresent the values of

other groups instead.

ii. USDA Data

The Economic Research Service of the United States Department of Agriculture

developed the Rural-Urban Continuum Codes scheme (RUCC) to distinguish counties by their

population size and proximity to other metropolitan centers. Given counties and county-

equivalents (boroughs, census tracts, or parishes), “counties” hereafter, are assigned a value of 1

through 9. Codes 1, 2, and 3 indicate a metro county, codes 4-7 indicate a non-metro county

with an urban population, and 8-9 indicate complete rurality. The definitions of each are

provided below.

1: Counties in metro areas of 1 million population or more.

2: Counties in metro areas of 250,000 to 1 million population.

3: Counties in metro areas of fewer than 250,000 population.

4: Urban population of 20,000 or more, adjacent to a metro area.

5: Urban population of 20,000 or more, not adjacent to a metro area.

6: Urban population of 2,500 to 19,999, adjacent to a metro area.

7: Urban population of 2,500 to 19,999, not adjacent to a metro area.

8: Completely rural or less than 2,500 urban population, adjacent to a metro area.

9: Completely rural or less than 2,500 urban population, not adjacent to a metro area.

(USDA Economic Research Service 2019).

RUCC codes with the same population range were collapsed to determine whether

proximity to a metro area influenced model fit and direction of the included variables. Model fit

was weaker, and direction of associations were unchanged after collapsing RUCC codes from

nine categories to five. The original scheme was used so that level of urbanization reflects more

than population size. The Office of Management and Budget (OMB) defined the urban and rural

indices by the population size and worker commuting data gathered in the 2010 U.S Census

Page 22: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

22

22

(USDA Economic Research Service 2019). Nine dummy variables were created for geographic

division: pacific west, mountain west, west north central, east north central, east south central,

south Atlantic, middle Atlantic, and new England. Each county was assigned a binary value for

membership based on U.S Census methodology (United States, and Bureau of the Census 1993).

iii. NHGIS Data

The IPUMS National Historical Geographic Information System (NHGIS) is part of the

Institute for Social Research and Data Innovation at the University of Minnesota. The NHGIS is

supported by the National Science Foundation and compiles various geographically based data

covering population, agriculture, and economic data. The NHGIS provided all data pertaining to

economic indicators, and racial and populational composition. For economic deprivation,

relevant variables are income, wealth inequality, and division of labor between professional and

labor/service-oriented jobs (Carriere et. al 2018). Professional and labor/service distinctions were

made on the basis of occupation. Those in fields of finance and insurance, real estate,

professional, scientific, and management and administrative services, educational services,

health care, and social assistance were coded as ‘professional’. Labor/service were all other

classes of occupations listed and not included in ‘professional’. For racial data, racial

composition as a percentage of a given area’s population was used. Racial groups were: non-

Hispanic or Latinx white, non-Hispanic or Latinx African American, non-Hispanic or Latinx

Native American, non-Hispanic or Latinx Asian, non-Hispanic or Latinx Pacific Islander, non-

Hispanic or Latinx other single race or more than one race, and Hispanic or Latinx. With

exception to the final category, each will be referred to without the ‘non-Hispanic or Latinx’

qualifier hereafter e.g. ‘White’ or ‘Asian’. For education, level of educational attainment was

measured as a binary variable, the two categories being those who graduated high school (or got

Page 23: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

23

23

a GED) or less, and those who have some college education or more (also includes technical

degrees).

iv. ARDA Data

The Association of Religion Data Archives hosts data recorded by a number of

researching organizations. The U.S Religion Census: Religious Congregations and Membership

Study, carried out by the Association of Statisticians of American Religious Bodies, documented

the rates of membership of 236 religious groups across the United States. Adherence is measured

by those who attend religious services in addition to those who are full members of a religious

group; membership connotes a religious confirmation ceremony such as communion (Grammich

et al. 2018). Given the variety of mechanisms religion has in regard to suicide, determining the

variable of interest is not easy. For that reason, and for the variability of characteristics in

different religious bodies (such as subjective adherence, religious influence in government or

school, or structure of religious community), this measure of adherence rates is a feasible

approach to macro-level study (Rasic et al. 2011).

v. BSGU Data

The National Center for Family and Marriage Research and the Center for Family

Demographic Research, with support from Bowling Green State University and the National

Institutes of Health produced all data used for marriage and divorce counts. The number of

marriages and divorces were based on estimates from county court records and the American

Community Survey. Marriage are divorce rates, measured per 100,000 are available at the county

level and thereby be readily included in macro-level study (BGSU 2016).

Page 24: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

24

24

b. Method

All data was merged and cleaned using Microsoft Excel version 16.33 and Stata MP

version 13.0 with two additional data analysis and data presentation packages: estout, and univar

(Microsoft Corporation 2019; StataCorp 2013; Jann 2019; Gleason 1997). Data was merged on

one-to-one terms of the corresponding FIPS code. Areas with unavailable data from the NHGIS

were dropped, reducing the number of the observations by 9 (n=3138). Each county was

assigned a binary value for geographic division according to their categorization by U.S Census

division and region (United States and Bureau of the Census 1993). Data was presented

geographically using Tableau version 9.2.2 (Tableau Software 2020).

In cases of data being presented in raw counts, such as marriage and divorce, rates per

100,000 were made by using population estimates of the same year. Suicide rates were obtained

from counties with more than 20 suicides. Values are suppressed when there were fewer than 10

and considered not reliable when there are between 10 and 20. As a result, smaller, and typically

more rural counties had suppressed or unreliable data when compared to their urban

counterparts. Suppressed and unreliable counties were not used in this research, leaving counties

with 20 or more suicide mortalities (n=1723).

A variable to measure racial diversity of a county was generated using the Blau index.

The index measures racial heterogeneity using the equations below:

𝐵𝑙𝑎𝑢 = 1 − (∑ 𝑛 12 , 𝑛 2

2 , … )

where 𝑛 is the proportion of the population belonging to a given racial group. The maximum

value (indicating full racial heterogeneity) is .8571 and the minimum value (indicating full racial

homogeneity) is 0. In Models including the percent of a population comprised by a given racial

Page 25: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

25

25

group, the Asian and Pacific Islander category was the reference group because it had the lowest

suicide rate, inferred from the literature (Suicide Prevention Resource Center 2019).

Religious adherence was adjusted in 29 counties when the reported rate per 100,000

exceeded 100,000. This discrepancy was likely due geographic factors where congregations may

be reporting members that live in a different county, or more likely church overcount (Grammich

et. al 2010). It nonetheless represents a county with a high level of religious adherence and was

assigned 100,000 to represent this, but to not skew distributions.

Descriptive analyses and correlation matrices were run to investigate relationships

between variables and any confounding effects. An initial multivariate model of economic

indicators, demographic of race, educational attainment, marriage and divorce rates, and

religious adherence was run. Sequential models included racial groups and region to discern if

the associations of the original model were capturing the effect of race or region instead.

Interaction terms were created for each variable found to be statistically significant (p<.05) in

Model 4, and relevant to the concepts presented in the hypotheses. Significant variables were

then multiplied by the county’s respective RUCC value to create and interaction term and were

then run again in the same model. To avoid collinearity, each interaction term was run

separately. The effects of interaction terms were calculated by setting variables in the regression

to their means, except for when manipulating variables of interest, to investigate the change in

suicide rate.

Residual plots were tested for heteroscedasticity in Stata MP 13 using the Breusch-Pagan

test, the White test, and Bartlett’s test for equal variance to determine whether variance

supported ordinary least squares values.

Page 26: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

26

26

IV. Analysis

a. Descriptive Statistics

Descriptive statistics of mean, minimum and maximum values, and standard deviations

for significant variables are presented in Table 1. Most variables show relatively normal

distributions, except for suicide rate, which is right-skewed. The relationship between the urban-

rural index and suicide rate is shown in Figure 1. There is a positive relationship between the

level of rurality and suicide rates, with a steeper trend beginning when RUCC=4. A geographic

representation of suicide rates at the state level is presented in Figure 2. This shows the

particularly high rates of suicide in the mountain west, pacific west, and part of New England.

Table 1: Descriptive Statistics of Major Variables, N= 1723.

Variable Mean Std. Dev. Min Max

Suicide Rate 17.69832 6.257146 5.1 68.8

Urban-Rural Index 3.59083 2.157629 1 9

% Below Poverty 15.4866 5.559747 3.034668 51.95788

% With Some College or More 54.37548 10.57966 20.76609 85.93239

Income Per Capita (USD) 27192.19 6505.33 11665 69529

Marriage Rate 714.326 555.0968 38.33112 13792.57

Divorce Rate 411.2285 179.2536 0 2895.315

Gini Index 0.4460615 0.031889 0.3522 0.5976

Blau Index 0.3326374 0.1801079 0.0161961 0.7779214

% Employed 93.4121 2.196636 79.86759 98.65183

% Professional 48.67573 6.683808 23.55725 73.82757

White 76.19336 18.53638 0.6354463 98.36216

Black 8.838693 12.0978 0.037843 75.60945

Native American 1.277615 5.471585 0 82.02372

Hispanic or Latinx 9.478079 12.5479 0.0198807 99.1848

Other Race(s) 2.235411 1.630664 0.037843 23.52782

% Male 49.65357 1.642345 45.64196 66.05021

Page 27: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

27

27

Figure 1: Suicide Rate by Urban-Rural Index (RUCC).

Figure 2: Suicide Rates Across the United States

Page 28: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

28

28

b. Multivariate Models

All associations presented represent sociodemographic characteristics of counties to the

suicide rates at the county level, not how any one variable is associated with suicide on an

individual level. That is, there is no way of asserting that someone who committed suicide is

represented in the associations found here. Models 1-4 show the cumulative associations of level

of rurality, socioeconomic variables, geographic division, and race with suicide, respectively.

Models 2-4 incorporate socioeconomic variables, geographic division, and race incrementally to

investigate what variables were capturing the effects of others. Model 4 includes each

independent variable stated in the hypotheses. Results show that poverty, employment rate,

divorce rate, percent of population with some college education or more, percent male, each

geographic division with exception to east north central, and all racial groups were significantly

associated with suicide rates.

Level of rurality was significant and positively associated with suicide rate across Models

1-4 (p<.001). This corroborates previous findings and literature on the subject (Durkheim 1897;

Gopal and Siahpush 2002; Levin and Leyland 2005; Fontanella et. al 2015; Carriere et. al 2018;

Helbich 2017). The coefficient decreases incrementally as more variables were added to the

model, indicating that level of rurality was capturing their effect. This finding supports H1.

The percent of the population living below the poverty line was negatively associated

with suicide rates (p<.05), suggesting that counties with more poverty experience less suicide, all

else held equal. Interestingly, economic inequality measured by the gini index and income per

capita were not significant predictors of suicide rates. This is notable because it indicates that

economic disadvantage measured in terms of the standard of living in given county may be more

significant than measures of wealth distribution. Another notable finding is that the percent of

Page 29: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

29

29

population employed was very significant (p<.001) and negatively associated with suicide rate

across each model. The regression coefficient of percent employed was larger than that of

poverty, however, standardized coefficients show that poverty has a larger effect, due to the

poverty being more prevalent than unemployment in most counties (See Appendix A). Percent

employed and poverty both being negatively associated with suicide rate is interesting because it

indicates that there is less suicide in areas with higher poverty as well as are with higher

employment rates, but absolute measures of wealth and wealth inequality do not shape this. This

suggests that the protective effect of employment may not be based in wealth or income, and that

the focus on income or its distribution is not as pertinent, which corroborates previous research

(Sawada, Ueda, and Matsubayashi 2017). It could be that employment’s protective effect may

pertain to social capital or other integrative forces that are characteristic of some forms repeated

social interaction (Wanberg 2012). These findings offer mixed support for H2.

Percent of the population with some college education or more was significant (p<.001)

and negatively associated with suicide rates. The results cannot indicate whether there are

marginal increases in the protective effect of education, but it does indicate that there is some

protective effect of the percentage of college educated individuals in a county. However,

education level is associated with a number of other socioeconomic variables, such as income

per capita, poverty, and employment, calling into question where the protective effect is coming

from. The results provided here cannot determine where that may be. One notable finding is the

interplay between percent with some college education or more, geographic division, and race.

The sign change in Model 3 indicates that percent with some college or more was capturing the

effect of geographic division to the to the extent that “true” direction of the effect of percent of

some college or more was being commanded by geographic division. The strong connection

Page 30: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

30

30

between educational attainment and geographic region is supported by research (Wheeler and

Pappas 2019). This finding supports research that suggests there is a protective effect from

increases in educational attainment and therefore, H3 is supported by these findings (Abel and

Kruger 2005; Lorant et al. 2005).

Religious adherence was not significantly associated with suicide rates when geographic

region and race were included. This contrasts previous research that suggests religion important

in shaping suicide (Gearing and Lizardi 2009). However, this could be due to the unit of analysis

used in this study and the aggregation of all religious groups into one adherence rate. Individual

denominations may be significant given the differences between religious groups, although this

research shows that the adherence rate in a county is not related to suicide rate, effectively

rejecting H4.

Figure 3: Relative Effect of Independent Variables on Suicide Rate

Note: Standardized beta coefficients measure the change of the standard deviation of suicide rates

-(6.25/100,000) associated with a one standard deviation change in each variable. All variables here were -

-statistically significant in Model 4

Page 31: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

31

31

Table 2: Multiple OLS Linear Regressions Predicting Suicide Rates

Model 1 Model 2 Model 3 Model 4

Urban Rural Index 1.312*** 0.945*** 0.759*** 0.545***

(0.0623) (0.0751) (0.0666) (0.0627)

Religious Adherence -0.0000181 -0.0000204* -0.0000111

(0.00000991) (0.00000921) (0.00000858)

Racial Heterogeneity -5.840*** -11.67*** -3.092*

(0.879) (0.874) (1.428)

Poverty -0.0741 -0.149** -0.143** (0.0592) (0.0524) (0.0493)

Gini Index -3.975 -1.857 8.962

(6.655) (5.859) (5.510)

Income -0.160*** -0.0446 -0.0383

(0.0462) (0.0407) (0.0392)

% Male 0.489*** 0.204** 0.157*

(0.0857) (0.0775) (0.0724)

% Employed -0.476*** -0.497*** -0.327***

(0.0808) (0.0735) (0.0709)

% Professional -0.0205 0.0495 0.0332 (0.0308) (0.0293) (0.0279)

Marriage Rate 0.000481* 0.000122 0.000163

(0.000234) (0.000205) (0.000191)

Divorce Rate 0.00387*** 0.00180** 0.00156*

(0.000737) (0.000657) (0.000617)

Some College or More 0.0801*** -0.0640** -0.0982***

(0.0223) (0.0228) (0.0215)

Pacific West 7.009*** 7.233***

(0.659) (0.652)

Mountain West 11.95*** 11.82*** (0.644) (0.627)

West North Central 3.674*** 3.119***

(0.590) (0.549)

West South Central 6.380*** 5.601***

(0.561) (0.532)

East North Central 0.983 0.816

(0.518) (0.480)

East South Central 3.872*** 3.503***

(0.567) (0.526)

South Atlantic 4.061*** 4.017***

(0.494) (0.464)

New England 1.398 1.430* (0.729) (0.676)

% White 0.228***

(0.0439)

% Black 0.121**

(0.0442)

% Hispanic or Latino 0.104*

(0.0453)

% Native American 0.405***

(0.0449)

% Other Race(s) 0.414*** (0.109)

Constant 12.99*** 39.34*** 56.53*** 17.47

(0.261) (8.726) (7.909) (9.074)

N 1723 1723 1723 1723 R2 0.205 0.291 0.467 0.549

Standard errors in parentheses

* p < 0.05, ** p < 0.01, *** p < 0.001

Page 32: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

32

32

All racial groups included in the model were significantly and positively associated with

suicide rates. All associations were positive because the reference group (Asian or Pacific

Islander) had the lowest suicide rate, which was inferred from the existing race disaggregated

suicide statistics (Suicide Prevention Resource Center 2019). While the largest coefficients were

shown in percent native American and percent other race(s), Appendix A illustrates that the

relative effect of percent white is sizably larger than both native American and other race(s) due

to the larger percentage of whites in most counties. Model 4 shows that after including racial

groups, the significance and magnitude of the association between racial heterogeneity and

suicide decreased. This is unsurprising, and it can be inferred that lower suicide rates associated

with higher levels of racial heterogeneity is a primarily function of the proportion of the

population that is white. These findings support H5.

Geographic division was significant in seven of the eight divisions included in the model.

Each was positively associated with suicide rates because the division with the lowest suicide

rate, middle Atlantic, was the reference group. The mountain west, pacific west, and west south

central had the had the highest rates. Illustrated by Figure 3, geographic divisions are three of the

leading five variables that have the largest relative effect on suicide rate. This illustrates that the

regional variation between counties has a substantial effect on suicide rates when compared to

many sociodemographic variables. Geographic division and sociodemographic variables are by

no means mutually exclusive. Rather, geographic division is indicative of varying rates of

poverty, income, education level, divorce, religiosity, and racial demography. However, as

Pepper (2017) notes, there are some counties where consistently higher rates of suicide transcend

demographics, indicating another factor not captured by sociodemographic analysis, and is

Page 33: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

33

33

instead related to cultural norms that differ between geographic regions. Nonetheless, this result

supports H6.

After including race and geographic division in the model, divorce rates were statistically

significant and positively associated with suicide, while marriage was not. This could be due to

the regional and racial disparities in marriage and divorce rates, with divorce rates being higher

in states and regions that experience higher suicide rates with higher suicide rates, as well as

other at-risk populations (Raley, Sweeney, and Wondra 2015; United States Census Bureau

2020). It could be that higher divorce rates are related to another risk factor for suicide, one that

is not related to marriage rates, given their low correlation to one another (r=.02). However,

without investigating the influence of non-marriage as well, this mechanism or associated factor

may be difficult to uncover. No significant variables in Model 4 have notable associations to

divorce rate and marriage rate, indicating that no single risk factor or protective factor in this

analysis could explain this discrepancy between marriage and divorce. The results do not support

H7, since marriage rates were not significant.

In sum, hypotheses 1, 3, 5, and 6 were supported by the findings. H2 received mixed

support given the non-consensus between indicators of economic disadvantage. H4 was flatly

rejected. H7 received mixed support, as marriage was not associated as predicted, while divorce

was. These findings indicate that after including all variables in the model, rurality, education,

race, and geographic division shaped suicide as predicted and as outlined in much literature.

Economic disadvantage, religious adherence, and marriage did not influence suicide as predicted.

Page 34: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

34

34

Table 3: Multiple OLS Linear Models Predicting Suicide Rates, with Interaction Terms

Model 5 Model 6 Model 7 Model 8 Model 9 Model 10 Model 11 Model 12

U-R Index 0.471** 3.077 0.796*** -0.277 -2.691 0.607* 0.443*** 0.440***

(0.143) (2.218) (0.134) (0.263) (1.615) (0.253) (0.0606) (0.0917)

Poverty -0.0853 -0.0687* -0.0718* -0.0726* -0.0680* -0.0699* -0.0664* -0.0705*

(0.0441) (0.0293) (0.0293) (0.0292) (0.0292) (0.0294) (0.0288) (0.0293)

Racial Heterogeneity -3.457* -3.380* -3.315* -3.100* -3.394* -3.472* -1.915 -3.323*

(1.399) (1.396) (1.395) (1.396) (1.395) (1.409) (1.386) (1.397)

% Employed -0.346*** -0.227 -0.356*** -0.366*** -0.348*** -0.351*** -0.338*** -0.353***

(0.0697) (0.127) (0.0695) (0.0695) (0.0694) (0.0698) (0.0683) (0.0695)

Divorce Rate 0.00158** 0.00159** 0.00394** 0.00161** 0.00160** 0.00158** 0.00160** 0.00161**

(0.000612) (0.000611) (0.00124) (0.000610) (0.000611) (0.000612) (0.000601) (0.000612)

Some College or More -0.0794*** -0.0808*** -0.0799*** -0.133*** -0.0817*** -0.0788*** -0.0896*** -0.0796***

(0.0147) (0.0148) (0.0147) (0.0224) (0.0148) (0.0147) (0.0145) (0.0147)

Pacific West 7.070*** 7.066*** 6.986*** 6.891*** 7.140*** 7.045*** 7.197*** 7.124***

(0.606) (0.605) (0.605) (0.606) (0.606) (0.606) (0.595) (0.607)

Mountain West 11.63*** 11.63*** 11.62*** 11.45*** 11.71*** 11.61*** 12.29*** 11.69***

(0.599) (0.598) (0.597) (0.599) (0.599) (0.599) (0.594) (0.600)

West North Central 2.850*** 2.814*** 2.831*** 2.858*** 2.964*** 2.852*** 2.917*** 2.886*** (0.521) (0.522) (0.521) (0.520) (0.523) (0.521) (0.512) (0.521)

West South Central 5.487*** 5.418*** 5.498*** 5.549*** 5.589*** 5.501*** 5.732*** 5.524***

(0.516) (0.520) (0.515) (0.515) (0.517) (0.516) (0.508) (0.516)

East North Central 0.598 0.575 0.553 0.608 0.646 0.594 0.643 0.620

(0.448) (0.448) (0.448) (0.447) (0.448) (0.448) (0.441) (0.448)

East South Central 3.300*** 3.245*** 3.292*** 3.458*** 3.398*** 3.321*** 3.346*** 3.349***

(0.507) (0.510) (0.505) (0.505) (0.506) (0.505) (0.496) (0.505)

South Atlantic 3.981*** 3.943*** 3.942*** 4.087*** 4.061*** 3.999*** 4.034*** 4.038*** (0.456) (0.457) (0.455) (0.454) (0.455) (0.455) (0.447) (0.455)

New England 1.555* 1.574* 1.558* 1.523* 1.562* 1.552* 1.658* 1.599*

(0.672) (0.672) (0.671) (0.670) (0.671) (0.672) (0.660) (0.672)

%White 0.223*** 0.227*** 0.215*** 0.187*** 0.221*** 0.220*** 0.255*** 0.208*** (0.0417) (0.0417) (0.0413) (0.0425) (0.0413) (0.0413) (0.0409) (0.0423)

% Black 0.117** 0.122** 0.106* 0.0751 0.108** 0.110** 0.130** 0.0997*

(0.0427) (0.0427) (0.0419) (0.0434) (0.0419) (0.0427) (0.0412) (0.0429)

% Hispanic or Latino 0.100* 0.105* 0.0914* 0.0607 0.0966* 0.0941* 0.119** 0.0827

(0.0435) (0.0435) (0.0429) (0.0443) (0.0429) (0.0438) (0.0423) (0.0441)

% Native American 0.394*** 0.393*** 0.382*** 0.364*** 0.391*** 0.389*** -0.137 0.376***

(0.0434) (0.0434) (0.0437) (0.0443) (0.0434) (0.0469) (0.0809) (0.0454)

% Other Race(s) 0.428*** 0.441*** 0.408*** 0.349** 0.424*** 0.417*** 0.497*** 0.206

(0.106) (0.106) (0.105) (0.107) (0.105) (0.106) (0.104) (0.189)

% Male 0.175* 0.172* 0.163* 0.176* -0.131 0.174* 0.145* 0.176*

(0.0696) (0.0696) (0.0697) (0.0694) (0.168) (0.0698) (0.0685) (0.0695)

Interaction Term Poverty % Employed Divorce Some

College or More

% Male % White % Native

American

% Other

Races(s)

0.00416 -0.0272 -0.000622* 0.0155** 0.0648* -0.000886 0.0883*** 0.0455

(0.00854) (0.0237) (0.000283) (0.00492) (0.0324) (0.00299) (0.0114) (0.0333)

Constant 21.30* 9.729 22.44** 29.43** 36.72** 21.91* 19.19* 23.59**

(8.632) (13.44) (8.613) (8.944) (11.48) (8.697) (8.474) (8.739)

N 1723 1723 1723 1723 1723 1723 1723 1723

R2 0.547 0.547 0.548 0.550 0.548 0.547 0.562 0.547

Standard errors in parentheses

* p < 0.05, ** p < 0.01, *** p < 0.001

Page 35: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

35

35

Table 4: Multiple OLS Linear Models Predicting Suicide Rates, with Interaction Terms (Cont’d)

Model 13 Model 14 Model 15 Model 16 Model 17 Model 18 Model 19

U-R Index 0.474*** 0.529*** 0.561*** 0.554*** 0.450*** 0.595*** 0.510*** (0.0623) (0.0625) (0.0639) (0.0632) (0.0608) (0.0640) (0.0614)

Poverty -0.0657* -0.0694* -0.0696* -0.0681* -0.0608* -0.0598* -0.0695*

(0.0292) (0.0293) (0.0293) (0.0293) (0.0289) (0.0294) (0.0292)

Racial Heterogeneity -3.415* -3.423* -3.300* -3.359* -3.194* -3.130* -3.173*

(1.391) (1.397) (1.399) (1.397) (1.378) (1.397) (1.399)

% Employed -0.357*** -0.349*** -0.350*** -0.354*** -0.334*** -0.344*** -0.359***

(0.0693) (0.0695) (0.0695) (0.0696) (0.0686) (0.0694) (0.0696)

Divorce Rate 0.00153* 0.00157* 0.00160** 0.00157* 0.00166** 0.00151* 0.00151* (0.000609) (0.000611) (0.000611) (0.000611) (0.000603) (0.000610) (0.000611)

Some College or More -0.0804*** -0.0791*** -0.0790*** -0.0795*** -0.0894*** -0.0822*** -0.0801***

(0.0147) (0.0147) (0.0147) (0.0147) (0.0146) (0.0147) (0.0147)

Pacific West 7.097*** 7.061*** 7.019*** 7.028*** 2.338** 7.068*** 7.048***

(0.603) (0.606) (0.605) (0.605) (0.904) (0.604) (0.604)

Mountain West 8.722*** 11.63*** 11.58*** 11.60*** 11.96*** 11.57*** 11.67***

(0.977) (0.599) (0.599) (0.598) (0.592) (0.597) (0.598)

West North Central 2.925*** 2.640** 2.829*** 2.856*** 3.077*** 2.827*** 2.905***

(0.519) (0.825) (0.521) (0.521) (0.515) (0.520) (0.521)

West South Central 5.567*** 5.506*** 6.120*** 5.493*** 5.647*** 5.457*** 5.518***

(0.514) (0.516) (0.709) (0.516) (0.509) (0.515) (0.515)

East North Central 0.644 0.598 0.568 0.584 0.763 0.549 0.635

(0.447) (0.448) (0.449) (0.448) (0.443) (0.447) (0.448)

East South Central 3.374*** 3.331*** 3.289*** 3.963*** 3.461*** 3.220*** 3.362***

(0.503) (0.505) (0.506) (0.767) (0.498) (0.505) (0.505)

South Atlantic 4.034*** 4.004*** 3.984*** 4.004*** 4.110*** 5.071*** 4.015***

(0.453) (0.455) (0.454) (0.454) (0.448) (0.588) (0.454)

New England 1.612* 1.557* 1.529* 1.546* 1.728** 1.556* -0.190

(0.669) (0.672) (0.672) (0.672) (0.663) (0.670) (1.065)

%White 0.221*** 0.220*** 0.222*** 0.216*** 0.163*** 0.216*** 0.221***

(0.0411) (0.0413) (0.0413) (0.0414) (0.0416) (0.0412) (0.0413)

% Black 0.110** 0.113** 0.113** 0.107* 0.0493 0.104* 0.110** (0.0417) (0.0419) (0.0418) (0.0421) (0.0423) (0.0419) (0.0418)

% Hispanic or Latino 0.0952* 0.0966* 0.0969* 0.0925* 0.0397 0.0893* 0.0959*

(0.0428) (0.0429) (0.0429) (0.0431) (0.0431) (0.0429) (0.0429)

% Native American 0.390*** 0.394*** 0.394*** 0.388*** 0.315*** 0.384*** 0.393***

(0.0433) (0.0434) (0.0434) (0.0437) (0.0443) (0.0434) (0.0434)

% Other Race(s) 0.428*** 0.421*** 0.430*** 0.414*** 0.332** 0.395*** 0.420***

(0.105) (0.105) (0.105) (0.105) (0.104) (0.105) (0.105)

% Male 0.173* 0.176* 0.175* 0.170* 0.145* 0.173* 0.174* (0.0693) (0.0696) (0.0695) (0.0697) (0.0687) (0.0694) (0.0695)

Interaction Term Mountain

West

West North

Central

West South

Central

East South

Central

Pacific West South Atlantic New England

0.695*** 0.0588 -0.180 -0.164 1.389*** -0.345** 0.489*

(0.186) (0.173) (0.141) (0.148) (0.200) (0.121) (0.232)

Constant 22.66** 21.56* 21.48* 22.65** 27.99** 21.50* 22.61** (8.586) (8.616) (8.612) (8.669) (8.547) (8.595) (8.619)

N 1723 1723 1723 1723 1723 1723 1723

R2 0.551 0.547 0.547 0.547 0.559 0.549 0.548

Standard errors in parentheses

* p < 0.05, ** p < 0.01, *** p < 0.001

Page 36: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

36

36

c. Multivariate Models with Interaction Terms

Models run with interaction effects show promising support for level of rurality shaping

the influence of certain independent variables on suicide rates. Eight of the twelve variables that

were significant in Model 4 had significant interactions with the level of rurality. Four of these

interactions were from geographic division: mountain west, pacific west, south Atlantic, and new

England. The other three were divorce rate, percent of population with some college education or

more, percent male, and percent Native American (See Table 3 and Table 4). Again, these

interaction effects must be understood as an association between the counties with particular

characteristics to suicide, not the characteristics of individuals themselves. For example, while

there is an association between divorce rates and suicide rates that varies by level of rurality, we

would be committing the ecological fallacy if we suggested that divorced individuals in urban

counties had a higher risk of suicide than their rural counterparts.

The equation below will be used to show the combined effects of two interactions,

education and level of rurality, and divorce and level of rurality. These two interactions will be

focused on in particular because they have a greater potential for policy application. Since no

individual suicide outcomes are presented here, the combined effects of percent male and

rurality, and percent native American and rurality would not indicate an advanced risk for native

Americans or males, rather, their presence. These findings would be without utility. Table 5

shows the combined effect of divorce and rurality. Table 6 shows the combined effect of the

percentage of population with some college or more and rurality. Table 7 shows the absolute

change in suicide rate, comparing the joint effects of these two interactions. Table 8 shows the

combined effect of geographic division and rurality compared to the mean suicide rate in those

geographic divisions. Each geographic division’s level of rurality is set to its respective mean.

Page 37: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

37

37

For the purposes of these examples, we first estimate the suicide rate when all of the other

covariates are set to the mean. We then see what happens when we increase the percentage of the

population with a college education by 10% and the divorce rate by 40%. In accordance with

Table 1, these manipulations represent a shift of one standard deviation from the mean values,

that is, a statistically reasonable difference when comparing counties. These data manipulations

are not to indicate what a given change within one county could do, rather, the potential

difference between two-nearly identical counties, with the manipulated variable being the only

difference.

𝑦 = 𝑏0 + 𝑏1𝑥1 + 𝑏2𝑥2 + 𝑏3𝑥1𝑥2

y= change in suicide rate

b1,2= coefficient of variable in interaction model

b3= coefficient of the interaction term

x1,2=manipulated value for independent variable

Table 5 shows that the combined effect of divorce and rurality is strong in both urban and

rural areas, and that the direction of the effect of increasing divorce changes in moderately rural

areas (RUCC=6). These results show that in a very urban county, low divorce rates would

contribute to a lower suicide rate relative to the average divorce rate, whereas in very rural areas,

lower divorce rates would contribute to a higher suicide rate. Inversely, higher divorce rates in

urban areas contribute to higher suicide rates while in rural areas, high divorce rates contribute to

lower suicide rates. The inversion of the effect of divorce rate when RUCC=6 indicates that the

combined effect of divorce and rurality may be curvilinear. This means that the direction of the

effect of divorce rate changes when a county has an urban population of 2,500 to 19,999 and is

adjacent to a metro area (RUCC=6). This exact point cannot be determined with certainty

because it may be describing a non-linear relationship using linear modeling.

Page 38: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

38

38

Table 5: Interaction Effects of Divorce and Rurality on Suicide Rate

It is unclear why the effect of higher divorce rates changes direction based on rurality,

though one possibility is the relationship between norms regarding deviance, in terms of both

divorce and suicide. Some research has shown that cognitive deviance, the belief in some value

that is not commonly accepted, increases the acceptability of that value depending on the level of

social bonding within a community (Pals and Engin 2019). Applied to this case, it could be

argued that high divorce rates are supported by the norms in rural areas, and that this relaxation

of social constraints would also undercut any norms that dissuade suicide or suicide attempts.

This interpretation is reliant on the assumption of relatively higher social acceptance of divorce

in rural areas, though it is more likely that this dynamic is typified by norms that accept a

number of behaviors, resulting from low social bonding in rural areas. This assumption is

Page 39: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

39

39

partially supported by research that shows mortality rates are associated with level of rurality,

and once social capital and social stability is controlled for, the influence of rurality decreases

(Yang, Jensen, and Haran 2011). This is to say that the effects of social bonding and its

relationship to acceptance of norms is highly contingent of level of rurality, and would thereby

support the findings presented here, in terms of the combined effect of divorce and rurality. This

mechanism is inherently broad on account of these associations being social addresses rather

than individual circumstances.

Table 6: Interaction Effects of Education and Rurality on Suicide Rate

Table 6 shows that combined effect of education and rurality is stronger in urban areas

than in rural ones. It also shows the combined effect of increasing education rates and rurality is

protective in most all areas, except for the most rural (RUCC=9). Following suit with the

Page 40: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

40

40

analysis of the interaction between divorce and rurality, the decreasing marginal benefit of

percent of the population with college education or more could be the result of social bonding.

The effect that social bonding has on reducing the spatial dependence of rural areas in predicting

social phenomena, shown by Yang et al. (2011), would support the interpretation that education

has an increased effect in rural areas. This is not shown by the results here, indicating that the

higher education may be related to other social phenomena that is protective in urban areas, and

less so in rural areas. As noted in the multivariate analysis above, the results provided here do

not show associations to education that can explain its decreasing protective effect.

This dynamic could also be result of a mechanism related to educational attainment and

religiosity. Data in this research, in addition to outside literature shows there are higher rates of

religious adherents in rural areas, as well as an inverse relationship between religious adherence

and suicide rate (Chalfant and Heller 1991). This trade-off between education and religious

adherence is significant. It may indicate that in the urban areas, secularization may be

representing a protective effect that religiosity would otherwise hold. This secularization could

manifest in something related to infrastructure, such as access to mental health resources, or

reduce social stigmas surrounding mental health issues.

The combined effect of education and rurality is distinctly different from that of divorce

and rurality. Table 7 illustrates this comparing the absolute change in suicide rate based on the

joint effects studied above, by level of rurality. It should be noted that since the change is

measured in absolute terms, the curvature seen in divorce when RUCC=6 and in education when

RUCC=8 marks the inversion in the combined effect of rurality and the variable of interest. That

is, when high divorce rate changes from being a risk factor to a protective factor, and when

higher education rate changes from being a protective factor to a risk factor, respectively.

Page 41: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

41

41

Table 7 shows that the combined effect of divorce and rurality is strongest in urban areas

and weakest when RUCC=6, an area that is not fully rural, nor urban. When RUCC=4 and when

RUCC=9, the combined effect of divorce and rurality is similar in magnitude, however, the

direction of the effect is opposite. For education, a 10% change in the percent of the population

with some college education or more has a much stronger effect in urban areas than in rural ones.

The same type of inversion occurs in the effect divorce occurs in the effect of education between

RUCC codes 8 and 9. Table 7 also shows that the combined effect of education and rurality is

stronger than that of divorce and rurality in RUCC codes 1-7. A proper interpretation of the

differences in Table 7 would require a more in-depth analysis of both rurality and divorce, and

rurality and divorce. Here, the most appropriate interpretation may be related to a structural

explanation, where education rates may be more intimately related to other risk factors, such as

race, and region (de Brey et al. 2019; Wheeler and Pappas 2019).

Page 42: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

42

42

Table 7: Absolute Change in Suicide Rate by Interaction Effect

Table 7 shows that the combined effect of divorce and rurality is strongest in urban areas

and weakest when RUCC=6, an area that is not fully rural, nor urban. When RUCC=4 and when

RUCC=9, the combined effect of divorce and rurality is similar in magnitude, however, the

direction of the effect is opposite. For education, a 10% change in the percent of the population

with some college education or more has a much stronger effect in urban areas than in rural ones.

The same type of inversion occurs in the effect divorce occurs in the effect of education between

RUCC codes 8 and 9. Table 7 also shows that the combined effect of education and rurality is

stronger than that of divorce and rurality in RUCC codes 1-7. A proper interpretation of the

differences in Table 7 would require a more in-depth analysis of both rurality and divorce, and

rurality and divorce. Here, the most appropriate interpretation may be related to a structural

Page 43: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

43

43

explanation, where education rates may be more intimately related to other risk factors, such as

race, and region (de Brey et al. 2019; Wheeler and Pappas 2019). The significant takeaway here

is the relative importance of divorce and education in different levels of rurality (based on the

combined effect). This is to say that the nature of risk-factors and protective factors for suicide is

highly conditional on level of rurality. This conclusion can be applied the other significant

interaction effects in as well, those being percent male, percent Native American, and the

following geographic divisions: mountain west, pacific west, south Atlantic, and New England.

This illustrates that divorce rates are more effective in shaping suicide rates more in

urban areas. One of the biggest takeaways between these two interactions is the degree to which

rurality moderates the relationship between sociodemographic variables and suicide rates. The

interaction effects above are not to indicate that some variables are more important in explaining

suicide than others, as much as it illustrates that significant predictors of suicide also have joint

effects, and that level of rurality may contribute to many of them. Additionally, many predictors

of suicide covary, meaning the likelihood of having cases like those above (nearly identical

characteristics except for one) is scarce, and as such, the predictive ability of single independent

variables may not be generalizable across different social environments. Further, the

generalizability these findings is conditional to geographic division as well. The combined effect

of geographic division and rurality is shown in Table 8.

Page 44: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

44

44

Table 8: Average Suicide Rate Versus Combined Effect of Geographic and Rurality

Table 8 shows that the combined effect of geographic division and rurality, and that each

joint effect leads to higher suicide rates than the average of the division. The largest effect is

shown in the pacific west and the smallest effect in New England. Beyond these existing

differences in suicide rate between geographic divisions, this figure illustrates that the theoretical

variation between urban and rural areas is more substantial in the pacific west and mountain

west. The combined effect indicates some correlates of rurality may contribute to suicide in some

geographic divisions and that these effects are not constant between geographic division. It

follows that predictors of suicide must be studied in reference to the relative effect based on both

geographic division and rurality. Speculating what differences between geographic division may

Page 45: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

45

45

be frivolous, as Pepper (2017) has shown that regional differences in suicide transcend

demographic distinctions, thereby hindering any explanatory power of this research.

d. Does Level of Rurality Matter?

In Models 1-4, the association between level of rurality and suicide rate is clear. In

Models 5-18, the interaction effect was significant in more than half. This provides promising

support for H8, as the joint effects of rurality and sociodemographic were significantly associated

with suicide rates. That is, level or rurality moderated the effect of predictors of suicide. This is

shown further by the very fact that that regressing suicide rate by level of urbanization yields

r2=.204. This research shows that level of urbanization does, in fact, matter when considering

county level suicide rates.

More importantly, this research shows that suicide rates vary based on a combination of

factors, and these are relevant to the how urban or rural an area is. Since divorce rates strong

effects in both urban and rural areas while higher education has a stronger effect in only urban

areas, it goes to say that the pathologies of suicide may be based in different realms of the social

sphere depending on where one lives. Additionally, the combined effect of geographic division

and rurality shows that there are geographically distinct trends that are explained by predictors

included in this research. It follows that level of urbanization does not just matter by itself, but it

may serve as a condition or context that moderates the influence of other variables on suicide.

This finding is true for divorce rates, percent of the population with some college education or

more, percent male, percent Native American, and four geographic divisions (pacific west,

mountain west, south Atlantic, and New England). H8 partly supported, considering only some

independent variables in H1-H7 has significant interactions.

Page 46: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

46

46

V. Conclusion

a. Main Findings

The main objective of this research was to evaluate the role of level of rurality in suicide.

The main findings indicate the role of rurality is significant both by itself and alongside other

predictors in a multivariate model. Level of rurality was found to have a combined effect with

divorce rate, percent of population with some college education or more, gender, percent Native

American, and four geographic divisions. Many predictors of suicide found by past research are

corroborated here as well. Religious adherence stands out as being less important than predicted

and theorized in the literature. Nuances in the findings related to economic disadvantage call for

more in-depth analyses of mechanisms that provide the protective effect against completed

suicide rates. The joint effects of some predictors of suicide maintain support the conclusion that

there is elevated suicide risk in more rural areas, but also that these predictors have different

effects based on rurality. In regard to geographic division, the findings presented here illustrate

multiple aspects of spatial analysis provide more nuanced sociodemographic relationships to

suicide at the county level. The non-linear standard errors of this research shows that there is still

need for further evaluation of what predicts suicide rates at the county-level. It echoes maxims

that suicide is difficult to explain at the macro-level, nor can it be reduced to such a small

handful of predictive variables. This advances the understanding of suicide because it addresses

the intersection of a number of sociological concepts that cannot, by themselves, explain such a

complex phenomenon.

b. Limitations

Heteroscedasticity tests were run in Model 4 and rejected the null hypothesis of

homoscedasticity in both the Breusch-Pagan and White tests (p<.001). This indicates that the

Page 47: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

47

47

residuals are not constant, thereby cautioning the reliance on r2 values. That said, the number of

factor variables used to predict outcomes of a continuous variable with a large range makes these

models prone to heteroscedasticity. Numerous scatter plots illustrate this (See Appendix B). The

primary variable of interest, level of rurality, shares little evidence of equal covariance with

suicide rate using Bartlett’s test for equal variances (chi-square= 165.27, p<.001, when df=8).

Some valuable predictors (at the individual level) such as sexual orientation, non-binary

gender identification, whether or not someone has experienced sexual assault or rape, served as a

veteran, or have a history of attempted suicide could not be analyzed here. Not all of these data

are readily available at the county level. Additionally, the existing number of variables in the

models provide a limitation to adding more and potentially overcomplicating the models. There

are additional variables of interest that could not be added and accounted for, but for the sake of

the concepts discussed at length above, not all could be effectively studied.

It must be recognized that at this unit of analysis, this research does not offer

individualistic explanations of suicide outcomes, nor that suicide rates can be interpreted as the

product of varying proportions of the studied groups. That is, higher percentages of populations

with some college education or more reducing suicide rates does not mean that those with a

college education are less prone to suicide, but that there might be other spurious factors that

reduce overall suicide in a population. Moreover, this analysis not applicable to mechanisms that

motivate suicidal ideation or attempts. The largest barrier to this study was the lack of full data.

For counties where the suicide mortality was less than 10 individuals, the data was suppressed.

For counties where the suicide mortality was between 10 and 20, the data is considered

unreliable based on the statistical influence of each suicide. Assuming a relatively normal

distribution of suicide at the county level, this method excludes counties with lower populations,

Page 48: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

48

48

as their rates may unduly influence the analysis where each county is weighted equally. The

drawback of this is the systematic elimination of rural counties where there is a small

population. Moreover, the necessary rate for these counties to be unsuppressed and reliable is

thereby larger than more urban counties that easily surpass these thresholds.

c. Suggestions for Further Research

An underlying theme of the interplay of explanatory variables illustrates the need to

research the interaction between known correlate of suicide, such as geographic region, race, and

economic disadvantage (specifically, employment). The combined effects of known risk factors

should be evaluated to determine previously unknown at-risk populations that may not be

evident in simple bivariate or multivariate analyses. In the interest of public health and reducing

suicide rates, current policy approaches should be evaluated against findings that relate

seriocomic, populational, and geographic variables to suicide. Such findings may illuminate risk-

factors not previously known or understood and could direct allocation of resources to prevent

suicide.

Page 49: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

49

49

Appendix

Appendix A: Multiple OLS Linear Regressions Predicting Suicide Rates (Standardized Beta Coefficients)

Model 1 Model 2 Model 3 Model 4

Urban Rural Index 0.452*** 0.326*** 0.262*** 0.188***

(0.0623) (0.0751) (0.0666) (0.0627)

Religious Adherence -0.041 -0.046* -0.025

(0.00000991) (0.00000921) (0.00000858)

Racial Heterogeneity -0.168*** -0.336*** -0.089* (0.879) (0.874) (1.428)

Poverty -0.066 -0.132** -0.127**

(0.0592) (0.0524) (0.0493)

Gini Index -0.020 -0.009 0.046

(6.655) (5.859) (5.510)

Income -0.166*** -0.046 -0.040

(0.0462) (0.0407) (0.0392)

% Male 0.128*** 0.054** 0.041*

(0.0857) (0.0775) (0.0724)

% Employed -0.167*** -0.174*** -0.115*** (0.0808) (0.0735) (0.0709)

% Professional -0.022 0.053 0.035

(0.0308) (0.0293) (0.0279)

Marriage Rate 0.043* 0.011 0.014

(0.000234) (0.000205) (0.000191)

Divorce Rate 0.111*** 0.052** 0.045* (0.000737) (0.000657) (0.000617)

Some College or More 0.135*** -0.108** -0.166***

(0.0223) (0.0228) (0.0215)

Pacific West 0.286*** 0.295***

(0.659) (0.652)

Mountain West 0.515*** 0.509***

(0.644) (0.627)

West North Central 0.170*** 0.145***

(0.590) (0.549)

West South Central 0.344*** 0.302*** (0.561) (0.532)

East North Central 0.060 0.050

(0.518) (0.480)

East South Central 0.196*** 0.177***

(0.567) (0.526)

South Atlantic 0.268*** 0.265***

(0.494) (0.464)

New England 0.041 0.042*

(0.729) (0.676)

% White 0.675*** (0.0439)

% Black 0.233**

(0.0442)

% Hispanic or Latino 0.209*

(0.0453)

% Native American 0.354***

(0.0449)

% Other Race(s) 0.108***

(0.109)

N 1723 1723 1723 1723

R2 0.205 0.291 0.467 0.549

Standardized beta coefficients; Standard errors in parentheses

* p < 0.05, ** p < 0.01, *** p < 0.001

Page 50: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

50

50

Appendix B: Scatter Plots of Suicide Rate by Predictor Variable

Page 51: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

51

51

Works Cited

Abel, Ernest L., and Michael L. Kruger. 2005. “Educational Attainment and Suicide Rates in the

United States.” Psychological Reports 97(1):25–28.

Allen, Kristina, and Sidra Goldman-Mellor. 2018. “Neighborhood Characteristics and

Adolescent Suicidal Behavior: Evidence from a Population-Based Study.” Suicide and

Life-Threatening Behavior 48(6):677–89.

Anderson, Laura M., Lynda S. Lowry, and Karl L. Wuensch. 2015. “Racial Differences in

Adolescents’ Answering Questions About Suicide.” Death Studies 39(10):600–604.

Anderson, Robert N., and Betty L. Smith. 2005. “National Vital Statistics Reports: Deaths--

Leading Causes for 2002: (509562006-001).”

Barber, Catherine W., and Matthew J. Miller. 2014. “Reducing a Suicidal Person’s Access to

Lethal Means of Suicide.” American Journal of Preventive Medicine 47(3):S264–72.

Barranco, Raymond E. 2016. “Suicide, Religion, and Latinos: A Macrolevel Study of U.S.

Latino Suicide Rates.” The Sociological Quarterly 57(2):256–81.

Beynon, Malcolm J., Andrew Crawley, and Max Munday. 2016. “Measuring and Understanding

the Differences between Urban and Rural Areas.” Environment and Planning B:

Planning and Design 43(6):1136–54.

BGSU Data Compass, County-level Database. (2016). County-level marriage and divorce

counts, 2010 [Data File]. Bowling Green State University. Retrieved from

(http://www.bgsu.edu/ncfmr/resources/data/original-data/county-level-marriage-divorce-

data-2010.html)

Page 52: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

52

52

BGSU. 2016. “County-Level Marriage & Divorce Data, 2010.” Bowling Green State University.

Retrieved February 4, 2020 (https://www.bgsu.edu/ncfmr/resources/data/original-

data/county-level-marriage-divorce-data-2010.html).

de Brey, C., Musu, L., McFarland, J., Wilkinson-Flicker, S., Diliberti, M., Zhang, A.,

Branstetter, C., and Wang, X. (2019). Status and Trends in the Education of Racial and

Ethnic Groups 2018 (NCES 2019-038). U.S. Department of Education. Washington, DC:

National Center for Education Statistics. Retrieved [date] from

https://nces.ed.gov/pubsearch/

Burr, Jeffrey, Patricia McCall, and Eve Powell-Griner. 1994. “Catholic Religion and Suicide:

The Mediating Effect of Divorce.” Social Sciences Quarterly 75(2):300–308.

Bjorkenstam, C., G. R. Weitoft, A. Hjern, P. Nordstrom, J. Hallqvist, and R. Ljung. 2011.

“School Grades, Parental Education and Suicide--a National Register-Based Cohort

Study.” Journal of Epidemiology & Community Health 65(11):993–98.

Carriere, Danielle E., Maria I. Marshall, and James K. Binkley. 2019. “Response to Economic

Shock: The Impact of Recession on Rural–Urban Suicides in the United States.” The

Journal of Rural Health 35(2):253–61.

Centers for Disease Control and Prevention, National Center for Health Statistics. Underlying

Cause of Death. 1999-2017 on CDC WONDER Online Database, released December

2018. Data are from the Multiple Cause of Death Files, 1999-2017, as compiled from

data provided by the 57 vital statistics jurisdictions through the Vital Statistics

Cooperative Program. Accessed at http://wonder.cdc.gov/ucd-icd10.html on Sep 27, 2019

11:59:51 AM,

Page 53: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

53

53

Centers for Disease Control and Prevention. n.d. “Data Summary Descriptions.” Centers for

Disease Control and Prevention. Retrieved February 4, 2020

(https://wonder.cdc.gov/DataSets.html).

Chalfant, H. Paul, and Peter L. Heller. 1991. “Rural/Urban versus Regional Differences in

Religiosity.” Review of Religious Research 33(1):76.

Chen, Ying-Yeh, Mengni Chen, Carrie S. M. Lui, and Paul S. F. Yip. 2017. “Female Labour

Force Participation and Suicide Rates in the World.” Social Science & Medicine 195:61–

67.

Coleman, James S. 1993. “The Rational Reconstruction of Society: 1992 Presidential Address.”

American Sociological Review 58(1):1.

Correll, Shelley J., Stephen Benard, and In Paik. 2018. “Getting a Job: Is There a Motherhood

Penalty?” Pp. 391–99 in Inequality in the 21st Century, edited by D. B. Grusky and J.

Hill. Routledge.

Curtin SC, Warner M, Hedegaard H. Suicide rates for females and males by race and ethnicity:

United States, 1999 and 2014. NCHS Health E-Stat. National Center for Health Statistics.

April 2016.

Daly, Mary C., Daniel J. Wilson, and Norman J. Johnson. “Relative Status and Well-Being:

Evidence from U.S. Suicide Deaths.” Federal Reserve Bank of San Francisco Working

Paper #2007- 12, 2007.

Dubow, Eric F., Paul Boxer, and L. Rowell Huesmann. 2009. “Long-Term Effects of Parents’

Education on Children’s Educational and Occupational Success: Mediation by Family

Interactions, Child Aggression, and Teenage Aspirations.” Merrill-Palmer Quarterly

55(3):224–49.

Page 54: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

54

54

Durkheim, Emile, and W. D. Halls. 1984. The Division of Labor in Society. New York: Free

Press.

Durkheim, Émile, and George Simpson. 1999. Suicide: A Study in Sociology. New York: Free

Press.

Fontanella, Cynthia A., Danielle L. Hiance-Steelesmith, Gary S. Phillips, Jeffrey A. Bridge,

Natalie Lester, Helen Anne Sweeney, and John V. Campo. 2015. “Widening Rural-Urban

Disparities in Youth Suicides, United States, 1996-2010.” JAMA Pediatrics 169(5):466.

Foucault, Michel. 1990. The History of Sexuality. Vintage Books ed. New York: Vintage Books.

Garroutte, Eva Marie, Jack Goldberg, Janette Beals, Richard Herrell, and Spero M. Manson.

2003. “Spirituality and Attempted Suicide among American Indians.” Social Science &

Medicine 56(7):1571–79.

Gearing, Robin E., and Dana Lizardi. 2009. “Religion and Suicide.” Journal of Religion and

Health 48(3):332–41.

Geulayov, G., C. Metcalfe, and D. Gunnell. 2016. “Parental Suicide Attempt and Offspring

Educational Attainment during Adolescence in the Avon Longitudinal Study of Parents

and Children (ALSPAC) Birth Cohort.” Psychological Medicine 46(10):2097–2107.

Gleason, John R. 1997. “'UNIVAR': module to generate univariate summary with box-and-

whiskers plot”. Statistical Software Component, Boston College Archives.

Goldman-Mellor, Sidra, Kristina Allen, and Mark S. Kaplan. 2018. “Rural/Urban Disparities in

Adolescent Nonfatal Suicidal Ideation and Suicide Attempt: A Population-Based Study.”

Suicide and Life-Threatening Behavior 48(6):709–19.

Page 55: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

55

55

Grammich, Clifford, Kirk Hadaway, Richard Houseal, Dale Jones, Alexai Krindatch, Richie

Stanley, and Richard Taylor. 2018. “U.S. Religion Census Religious Congregations and

Membership Study, 2010 (County File).”

Haws, Charlotte A., Douglas D. Gray, Deborah A. Yurgelun-Todd, Michelle Moskos, Laurence

J. Meyer, and Perry F. Renshaw. 2009. “The Possible Effect of Altitude on Regional Variation in

Suicide Rates.” Medical Hypotheses 73(4):587–90.

Helbich, M., V. Blüml, T. de Jong, P. L. Plener, M. P. Kwan, and N. D. Kapusta. 2017. “Urban–

Rural Inequalities in Suicide Mortality: A Comparison of Urbanicity Indicators.”

International Journal of Health Geographics 16(1):39.

Hemenway, D. 2002. “Association of Rates of Household Handgun Ownership, Lifetime Major

Depression, and Serious Suicidal Thoughts with Rates of Suicide across US Census

Regions.” Injury Prevention 8(4):313–16.

Hirsch, Jameson K., and Kelly C. Cukrowicz. 2014. “Suicide in Rural Areas: An Updated

Review of the Literature.” Journal of Rural Mental Health 38(2):65–78.

Jann, Ben. 2019. “'ESTOUT': module to make regression tables. Statistical Software Component,

Boston College Archives.

Joe, Sean, Steven C. Marcus, and Mark S. Kaplan. 2007. “Racial Differences in the

Characteristics of Firearm Suicide Decedents in the United States.” American Journal of

Orthopsychiatry 77(1):124–30.

Kamali, Masoud. 1999. “Distorted Integration: Problems of Monolithic Order.” Innovation: The

European Journal of Social Science Research 12(1):81–97.

Kastenbaum, Robert. 2012. Death, Society, and Human Experience. 11th ed. Boston: Pearson.

Page 56: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

56

56

Keage, Hannah A. D., Graciela Muniz, Lisa Kurylowicz, Miranda Van Hooff, Levina Clark,

Amelia K. Searle, Michael G. Sawyer, Peter Baghurst, and Alexander McFarlane. 2016.

“Age 7 Intelligence and Paternal Education Appear Best Predictors of Educational

Attainment: The Port Pirie Cohort Study: Predictors of Educational Attainment.”

Australian Journal of Psychology 68(1):61–69.

Kim, Jeong Lim, Ji Man Kim, Young Choi, Tae-Hoon Lee, and Eun-Cheol Park. 2016. “Effect

of Socioeconomic Status on the Linkage Between Suicidal Ideation and Suicide

Attempts.” Suicide and Life-Threatening Behavior 46(5):588–97.

Kposowa, A. J. 2000. “Marital Status and Suicide in the National Longitudinal Mortality Study.”

Journal of Epidemiology & Community Health 54(4):254–61.

Kubrin, C. E., T. Wadsworth, and S. DiPietro. 2006. “Deindustrialization, Disadvantage and

Suicide among Young Black Males.” Social Forces 84(3):1559–79.

Kubrin, Charis E. and Tim Wadsworth. 2009. “Explaining Suicide Among Blacks and Whites:

How Socioeconomic Factors and Gun Availability Affect Race-Specific Suicide Rates.”

Law, Chi-kin, Anne-Marie Snider, and Diego De Leo. 2014. “The Influence of Deprivation on

Suicide Mortality in Urban and Rural Queensland: An Ecological Analysis.” Social

Psychiatry and Psychiatric Epidemiology 49(12):1919–28.

Lawrence, Ryan E., Maria A. Oquendo, and Barbara Stanley. 2016. “Religion and Suicide Risk:

A Systematic Review.” Archives of Suicide Research 20(1):1–21.

Leončikas, Tadas. 2004. Social Adaptation of Ethnic Minorities. Kaunas: Vytautas Magnus

University.

Levin, Kate A., and Alastair H. Leyland. 2005. “Urban/Rural Inequalities in Suicide in Scotland,

1981–1999.” Social Science & Medicine 60(12):2877–90.

Page 57: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

57

57

Lorant, Vincent, Anton E. Kunst, Martijn Huisman, Giuseppe Costa, and Johan Mackenbach.

2005. “Socio-Economic Inequalities in Suicide: A European Comparative Study.” British

Journal of Psychiatry 187(1):49–54.

Manson, Steven. et al. 2019. IPUMS National Historical Geographic Information System:

Version 14.0 [Database]. Minneapolis, MN: IPUMS. 2019.

http://doi.org/10.18128/D050.V14.0

Microsoft Corporation. 2019. Microsoft Excel. Retrieved from

https://office.microsoft.com/excel.

Miller, Matthew, Deborah Azrael, and David Hemenway. 2002. “Household Firearm Ownership

and Suicide Rates in the United States.” Epidemiology 13(5):517–24.

Mohler, Beat, and Felton Earls. 2001. “Trends in Adolescent Suicide: Misclassification Bias?”

American Journal of Public Health 91(1):150–53.

Moore, Matthew D. 2015. “Religious Heterogeneity and Suicide: A Cross-National Analysis.”

Social Compass 62(4):649–63.

Neisser, Ulric, Gwyneth Boodoo, Thomas J. Bouchard Jr., A. Wade Boykin, Nathan Brody,

Stephen J. Ceci, Diane F. Halpern, John C. Loehlin, Robert Perloff, Robert J. Sternberg,

and Susana Urbina. 1996. “Intelligence: Knowns and Unknowns.” American

Psychologist 51(2):77–101.

Nestadt, Paul S., Patrick Triplett, David R. Fowler, and Ramin Mojtabai. 2017. “Urban–Rural

Differences in Suicide in the State of Maryland: The Role of Firearms.” American

Journal of Public Health 107(10):1548–53.

Page 58: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

58

58

Osafo, J., B. L. Knizek, C. S. Akotia, and H. Hjelmeland. 2013. “Influence of Religious Factors

on Attitudes Towards Suicidal Behaviour in Ghana.” Journal of Religion and Health

52(2):488–504.

Pals, Heili, and Ceylan Engin. 2019. “Attachment to Society and Cognitive Deviance: The Case

of Turkey.” Deviant Behavior 40(7):799–815.

Pargament, Kenneth I., Bruce W. Smith, Harold G. Koenig, and Lisa Perez. 1998. “Patterns of

Positive and Negative Religious Coping with Major Life Stressors.” Journal for the

Scientific Study of Religion 37(4):710.

Pepper, Carolyn M. 2017. “Suicide in the Mountain West Region of the United States.” Crisis

38(5):344–50.

Phillips, Julie A., and Katherine Hempstead. 2017. “Differences in U.S. Suicide Rates by

Educational Attainment, 2000–2014.” American Journal of Preventive Medicine

53(4):e123–30.

Placidi, Giovanni P. A., Maria A. Oquendo, Kevin M. Malone, Beth Brodsky, Steven P. Ellis,

and J. John Mann. 2000. “Anxiety in Major Depression: Relationship to Suicide

Attempts.” American Journal of Psychiatry 157(10):1614–18.

Pompili, Maurizio, Monica Vichi, Ping Qin, Marco Innamorati, Diego De Leo, and Paolo

Girardi. 2013b. “Does the Level of Education Influence Completed Suicide? A

Nationwide Register Study.” Journal of Affective Disorders 147(1–3):437–40.

Portes, A. 2000. “Social Capital: Its Origins and Applications in Modern Sociology.” Pp. 43–67

in Knowledge and Social Capital. Elsevier.

Page 59: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

59

59

Purselle, David C., Michael Heninger, Randy Hanzlick, and Steven J. Garlow. 2009.

“Differential Association of Socioeconomic Status in Ethnic and Age-Defined Suicides.”

Psychiatry Research 167(3):258–65.

Rasic, Daniel, Jennifer A. Robinson, James Bolton, O. Joseph Bienvenu, and Jitender Sareen.

2011. “Longitudinal Relationships of Religious Worship Attendance and Spirituality with

Major Depression, Anxiety Disorders, and Suicidal Ideation and Attempts: Findings from

the Baltimore Epidemiologic Catchment Area Study.” Journal of Psychiatric Research

45(6):848–54.

Raley, R. Kelly, Megan M. Sweeney, and Danielle Wondra. 2015. “The Growing Racial and

Ethnic Divide in U.S. Marriage Patterns.” The Future of Children 25(2):89–109.

Rossen, Lauren M., Holly Hedegaard, Diba Khan, and Margaret Warner. 2018. “County-Level

Trends in Suicide Rates in the U.S., 2005–2015.” American Journal of Preventive

Medicine 55(1):72–79.

Sawada, Yasuyuki, Michiko Ueda, and Tetsuya Matsubayashi. 2017. Economic Analysis of

Suicide Prevention: Towards Evidence-Based Policy-Making. Singapore: Springer.

Senior, Martyn, Huw Williams, and Gary Higgs. 2000. “Urban–Rural Mortality Differentials:

Controlling for Material Deprivation.” Social Science & Medicine 51(2):289–305.

Shah, Ajit, and Ritesh Bhandarkar. 2009. “The Relationship Between General Population

Suicide Rates and Educational Attainment: A Cross-National Study.” Suicide and Life-

Threatening Behavior 39(5):463–70.

Singh, Gopal K., and Mohammad Siahpush. 2002. “Increasing Rural–Urban Gradients in US

Suicide Mortality, 1970–1997.” American Journal of Public Health 92(7):1161–67.

Page 60: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

60

60

South, Scott J. 1984. “Racial Differences in Suicide: The Effect of Economic Convergence.”

Social Science Quarterly 65(1):172–80.

Stack, Steven. 1990. “New Micro-Level Data on the Impact of Divorce on Suicide, 1959-1980:

A Test of Two Theories.” Journal of Marriage and the Family 52(1):119.

Stack, Steven, and Augustine J. Kposowa. 2016. “Culture and Suicide Acceptability: A Cross-

National, Multilevel Analysis.” The Sociological Quarterly 57(2):282–303.

Stack, Steven, and Frederique Laubepin. 2019. “Religiousness as a Predictor of Suicide: An

Analysis of 162 European Regions.” Suicide and Life-Threatening Behavior 49(2):371–

81.

StataCorp. 2013. Stata Statistical Software: Release 13. College Station, TX: StataCorp LP.

Stone, D.M. et al. (2017). Preventing Suicide: A Technical Package of Policies, Programs, and

Practices. Atlanta, GA: National Center for Injury Prevention and Control, Centers for

Disease Control and Prevention. 2010.

von Stumm, Sophie, Sally Macintyre, David G. Batty, Heather Clark, and Ian J. Deary. 2010.

“Intelligence, Social Class of Origin, Childhood Behavior Disturbance and Education as

Predictors of Status Attainment in Midlife in Men: The Aberdeen Children of the 1950s

Study.” Intelligence 38(1):202–11.

Suicide Prevention Resource Center. n.d. “Racial and Ethnic Disparities | Suicide Prevention

Resource Center.” Retrieved March 1, 2020b (https://www.sprc.org/racial-ethnic-

disparities).

Tableau 9.2.2. 2020. Seattle, WA: Tableau Software Inc.

Tierney, Thomas F. 2010. “The Governmentality of Suicide: Peuchet, Marx, Durkheim, and

Foucault.” Journal of Classical Sociology 10(4):357–89.

Page 61: THE IMPACT OF LEVEL OF RURALITY ON SUICIDE RATES: AN

Haffey

61

61

Turecki, Gustavo, and David A. Brent. 2016. “Suicide and Suicidal Behaviour.” The Lancet

387(10024):1227–39.

United States Census Bureau. 1993. Statistical Abstract of the United States, 1993.

Washington, D.C.: U.S. Dept. of Commerce, Bureau of the Census : For sale by the Supt.

of Docs., U.S. G.P.O

United States Census Bureau. 2020. “U.S. Marriage and Divorce Rates by State.” The United

States Census Bureau. Retrieved March 16, 2020 (https://www.census.gov/library

/visualizations /interactive/marriage-divorce-rates-by-state.html).

USDA Economic Research Service. n.d. “USDA ERS - Documentation.” Retrieved March 1,

2020c (https://www.ers.usda.gov/data-products/rural-urban-continuum-

codes/documentation/#Methodology).

Wanberg, Connie R. 2012. “The Individual Experience of Unemployment.” Annual Review of

Psychology 63(1):369–96.

Wheeler, Lauren B., and Eric C. Pappas. 2019. “Determining the Development Status of United

States Counties Based on Comparative and Spatial Analyses of Multivariate Criteria

Using Geographic Information Systems.” International Journal of Higher Education

8(1):92.

Widger, Tom. 2018. “Suicides, Poisons and the Materially Possible: The Positive Ambivalence

of Means Restriction and Critical–Critical Global Health.” Journal of Material Culture

23(4):396–412.

Yang, Tse-Chuan, Leif Jensen, and Murali Haran. 2011. “Social Capital and Human Mortality:

Explaining the Rural Paradox with County-Level Mortality Data: Social Capital and

Human Mortality.” Rural Sociology 76(3):347–74.