Factors that Influence – ‘Life Expectancy’:

Multivariate Regression Analysis of US counties

Written by Joo Young Park

PUBP 704-002 : Statistical Method in Policy Analysis

Dec. 8th, 2011

1. Introduction

This paper explores statistically significant factors that influence ‘life

expectancy’. With the help of a multivariate regression analysis of all U.S. counties, this

study shows the factors that influence life expectancy. From the results we can discern

priority of health risk factors for individual health management as well as for health

policy.

Background

‘Life expectancy’ is the expected, in the statistical sense, number of years of life

remaining at a given age, on average, by a particular cohort (Sheffrin, 2003).1 It most

commonly refers to life expectancy at birth, the median number of years that a

population born in a particular year could expect to live. For instance, based on recently

released final data, life expectancy at birth in 2003 was 77.5 years. This tells us that, for

those born in calendar year 2003 in the United States, 50% will die before that age; the

other half will live longer (Shrestha,2006)2

Life expectancy from birth is a frequently utilized and analyzed component of

demographic data for the countries of the world. It represents the average life span of a

newborn and is an indicator of the overall health of a country. In this sense, life

expectancy is one of the factors in measuring the Human Development of each nation,

and also used in describing the physical quality of life of an area.

According to several empirical studies, demographic factors such as race,

gender, and income, direct mortality causes such as cancer, stroke or heart disease,

environmental heath, risk factors for premature death, and access to care service have

been proven as influential factors on Life expectancy. However, the specific level of

1 Steven M. Sheffrin (2003). Economics: Principles in action. Upper Saddle River, New Jersey: Pearson

Prentice Hall. p. 473. 2 Laura B. Shrestha (2006). CRS Report for Congress :Life Expectancy in the United States, Congressional

Research Service, The Library of Congress

2

significances for health-related daily behavioral risk factors and general public supports

on Life expectancy has not been clearly verified, excluding direct mortality causes.

In this sense, to provide specific health risk factors for individual health

management and policy implication, this study will focus on statistical impacts of

general daily behavioral attributes and availability of public health service.

Research Question

In terms of individual behavioral risk factors and availability of public health

supports, what are the influential factors/attributes of ‘Average Life expectancy’ of US

citizens?

2. Literature Review

I & II articles assured that Life Expectancy is a valid indicator of representing

Population health statue. From III to IX, offered considerable independent variables for

the analysis, and X shows recent trend with ‘wellbeing’ trend.

I. Jean Marie Robine, Karen Ritchie, 1991, “Health life expectancy : evaluation of

global indicator of change in population health” BMJ vol.302 :457-460

This text outlines ‘Healthy life expectancy’ is a valuable index for the

appreciation of changes in both the physical and the mental health states of the

general population, for allocating resources, and for measuring the success of political

programs. It states that future calculations should also take into account the

probability of recovery and thus extend the applicability of the indicator to

populations in poor health rather than focusing on the well population.

3

II. Gabriel Gulis, 2000, “Life expectancy as an indicator of environmental health”

European Journal of Epidemiology vol. 16, no.2 :161-165

This article questioned that whether or not life expectancy at birth is related to

the quality of life as expressed by global economic, environmental and nutritional

measures. To get an answer, two models set of independent variables and multivariate

analysis was performed. An attempt to estimate the role of studied variables in overall

life expectancy was done, too. Access to safe drinking water per capita gross domestic

product, literacy, calories available as percentage of needs and per capita public health

expenditures were taken as exposure, and compared with life expectancy at birth. A

linear regression model was used to estimate the role of different exposures on life

expectancy at birth. In the result, the correlation coefficient for the linear model was

0.8823 (R2=0.7784)

III. Anna Peeters, 2003, “Obesity in Adulthood and its consequences for life expectancy :

Life table analysis” Annals of Internal Medicine vol.138 number1 :24-33

Main conclusion of this study is that obesity and overweight in adulthood are

associated with large decreases in life expectancy and increases in early mortality. These

decreases are similar to those seen with smoking. Obesity in adulthood is a powerful

predictor of death at older ages. Thus, the paper contends more efficient prevention

from the prevalence of obesity and its treatment should become high priorities in public

health.

IV. Catherine E. Ross, John Mirowsky, 2002, “Family Relationships, Social Support and

Subjective Life Expectancy” Journal of Health and Social Behavior vol.43 :469-489

This work finds that having adult children and surviving parent’s increases the

length of life on expects, but young children in the home does not, and marriage only

contributes years of life expected for older men.

4

According to the interpretation of this article, better current health is associated with

higher subjective life expectancy, but it does not explain the impact of supportive

relationships. Most of the impact of supportive relationships appears to be a direct

result of projected security about the future.

V. Henrik Bronnum-Hansen, Knud Juel, 2001 “Abstention from smoking extends life

and compresses morbidity: a population based study of health expectancy among

smokers and never smokers in Denmark” Tobacco Control, vol.10, no.3 :273-278

This study analyzes health expectancy never smokers, ex-smokers and smokers

in Denmark with comparing smoking attributable mortality rate. The results confirmed

that Smoking reduces the expected lifetime in good health and increase the expected

lifetime in poor health

VI. Mira M.Hidajat, Mark D.Hayward, Yasuhiko Saito, 2007 “Indonesia’s social capacity

for population health: the educational gap in active life expectancy” Population

Research and Policy Review, vol.26, no.2 :219-234

This study develops a model within the analytic framework of a Markov-based

multistate life table model to calculate an important indicator of the burden of disease,

the expected years of active life of elderly Indonesians. The result shows that having

some education increases life expectancy but it also expands the expected years with a

major functional problem.

VII. Eileen M. Crimmins, Yasuhiko Saito, 2001 “Trends in healthy life expectancy in the

United States, 1970-1990: gender, racial, and educational differences” Social Science

& Medicine 52 :1629-1641

5

This paper examines healthy life expectancy by gender and education for Whites

and African Americans in the United States at three dates: 1970, 1980 and 1990. There

are large racial and educational differences in healthy life expectancy to each date and

differences by education in healthy life expectancy are even larger than differences in

total life expectancy. Large racial differences exist in healthy life expectancy at lower

levels of education.

Educational differences of healthy life expectancy have been increasing over time

because of widening differentials in both mortality and morbidity. In the last decade, a

compression of morbidity has begun among those of higher educational statues; those

of status are still experiencing expansion of morbidity

VIII. R G Wilkinson, 1992 “Income distribution and life expectancy” BMJ vol. 304 :165-168

This paper suggests that the association between health and income distribution

is a result of factors to do with relative rather than absolute income. Increasingly social

scientists have emphasized the importance of relative poverty.

These results should caution against using the lack of a close relation between national

mortality and gross national product per head to infer that health inequalities within

societies cannot be a reflection of income differentials. Indeed, if health differences

within the developed countries are principally a function of income inequality itself, this

would explain why social class differences in health have not narrowed despite growing

affluence and the fall of absolute poverty.

IX. Robert A. Hahn, Steven Eberhardt, 1995 “Life Expectancy in Four U.S. Racial/Ethnic

Populations:1990” Epidemiology, vol.6, no.4 :350-355

This report used information on population undercounts by race/ethnicity in the

census and on misclassification of race/ethnicity on death certificates to calculate life

expectancy for black, white, American Indian, and Asian men and women in the United

6

States in 1990. The result shows Asian men had life expectancies of 82 years and Asian

women 85.8 years-the highest life expectancies reported for any population in the

world and beyond the limit predicted by some current theories

X. R.J.M. Perenboom, L.M.Van Herten, H.C. Boshuizen, G.A.M. Van Den Bos, 2004

“Trends in life expectancy in wellbeing” Social Indicators Research, vol.65, no.2 :

227-244

According to this paper, contrary to life expectancy in good perceived health and

to disability free life expectancy-which show a decreasing trend- the overall wellbeing of

the population is increasing. It seems that aspects in human life that contribute to

wellbeing or quality of life other than physical health are gaining importance. This

makes life expectancy in wellbeing a less appropriate instrument to monitor changes in

population health, but a useful instrument to measure population quality of life.

3.1 Data

The data utilized in this study comes from CHSI 2009 (Community Health Status

Indicators3) which is a collection of nationally available indicators for counties. CHSI

2009 reported about 205 health indicators for 3,141 current counties in 50 states and

District of Columbia. The data used to construct the Community Health Status Indicators

(CHSI) were obtained from a variety of federal agencies including the Department of

Health and Human Services, Environmental Protection Agency, Census Bureau, and

3 ‘Data Sources, Definitions, and Notes ; Community Health Status Indicators 2009’ : provide health indicator

definitions, sources, and methods used in the Community Health Status Indicators Reports created by the Community

Health Status Indicators (CHSI) Project Working Group. It is not intended to stand alone but to be used as a reference

for the user of the county health profile provided for every U.S. County and is available at

http://www.communityhealth.hhs.gov/homepage.aspx?j=1

7

Department of Labor. The CHSI data is reported at the county-level and publicly

available data. (http://www.communityhealth.hhs.gov/homepage.aspx?j=1)

The Community Health Status Indicators (CHSI) Project was initially launched in

2000 and archived in 2004 when the data became outdated. Since then the project was

updated with funding from Robert Wood Johnson Foundation and re-launched by an

expanded partnership that included the Centers for Disease Control and Prevention

(including NCHS and ATSDR), the National Institutes of Health/National Library of

Medicine, the Health Resources Services Administration, the Public Health Foundation,

the Association of State and Territorial Health Officials (ASTHO), National Association of

County and City Health Officials (NACCHO), National Association of Local Boards of

Health (NALBOH), and Johns Hopkins University School of Public Health in 20084. Annual

updates are anticipated with 2009 being the latest.

The CHSI 2009 contains data of direct causes of death such as cancer, heart

disease, homicide, stroke, motor vehicle injuries and others, and represents public

health such as access to and utilization of healthcare services, birth and death measures,

average life expectancy, vulnerable populations, risk-factors for premature deaths,

communicable diseases, and environmental health. In addition, It also covers health

related behavioral factors such as tobacco use, diet, physical activity, alcohol and drug

use, sexual behavior and others substantially contribute to deaths.

As mentioned earlier, the goal of this study is finding implications about the

relationship between ‘average life expectancy’ representing the current health indicator

and general behaviors & public health supports in daily life. In this sense, I excluded the

direct cause factors of death and initially selected 16 variables5.

4 Data Sources, Definitions, and Notes, Community Health Status Indicators 2009

5 Population size, population density, Poverty, Race, unemployed, no exercise, few fruits/vegetables, obesity, smoker,

uninsured, elderly Medicare, Medicaid beneficiaries, primary care physicians, dentist rate, community health center and health professional shortage area.

8

Definitions and Data source of Variables

The definitions of dependent variable and 16 candidates of independent

variables are given as below; Table 1 &2

Table 1. Dependent Variable Used in Analysis6

Dependent

Variable Definition Data source Year

Life expectancy

The average number of years that a baby born

in a particular year is expected to live if current

age-specific mortality trends continue to apply

The Community Health Status Indicators Report

by Department of Health and Human Service 2008

Table 2. 16 Candidates of Independent Variables in Analysis7

6 Data Sources, Definitions, and Notes, Community Health Status Indicators 2009 7 Data Sources, Definitions, and Notes, Community Health Status Indicators 2009

Independent

Variable

Expected

sign Definition Data source Year

Population

Size unclear Annual estimates of the resident population US Census Bureau 2008

Population

Density negative Population density (people per square mile) US Census Bureau 2008

Poverty negative individuals living below poverty level % US Census Bureau 2008

Population

Race/Ethnicity unclear

Race and ethnicity-specific population size %

; White/Black/

American Indian/Asian/Hispanic

US Census Bureau 2008

Unemployed negative Unemployed % US Bureau of Labor Statistics 2008

No Exercise negative

% of adults reporting of no participation

in any leisure-time physical activity or

exercises in the past month

Centers for Disease Control

and Prevention 2006

9

3.2 Methods

Data Cleaning

Data sets were visually inspected for missing entries and obvious errors. No

missing value was detected, but minor data errors8 in 15 data were observed and

cleaned as missing data. Verification of each race proportion is conducted by checking

8 Not understandable numbers such as decimal errors or minus quantity etc.

Few fruits/

vegetables negative

% of adults reporting an average

fruits/vegetables

consumption of less than 5 servings per day

Centers for Disease Control

and Prevention 2006

Obesity negative Calculated % of adults of overweight,

based on body mass index (BMI)

Centers for Disease Control

and Prevention 2006

Smoker negative % of adult smoker Centers for Disease Control

and Prevention 2006

Uninsured negative Estimated % of uninsured individuals

under age 65 US Census Bureau 2006

Medicaid

Beneficiaries positive Medicaid beneficiaries

Centers for Medicare and Medicaid

Services 2008

Primary care

physicians positive Primary care physicians per 100,000 pop % HRSA 2008

Elderly

Medicare positive

% of Medicare beneficiaries

for elderly (age 65+)

Centers for Medicare

and Medicaid Services 2008

Dentist Rate positive dentists % per 100,000 pop HRSA 2008

Community

Health Center positive

Indicator for any Community/Migrant Health

Centers located in the county HRSA 2009

HPSA positive Indicator for single county designated

Health Professional Shortage Area HRSA 2009

10

the sum of all races, White, Black, Native American, Indian, Asian and Hispanic, should

be 100%.

Data Recoding

In order to compare relative proportions in each county, data conversions from

the raw number to ratio divided by the population of county is conducted for

Unemployment, Insurance, Elderly Medicare and Medicaid Beneficiaries. Plus, dummy

variable data processing applies for Community health center and Health professional

shortage area, which are composed of nominal data9

Tests for Normality

Normality was evaluated by preparing histograms (See Appendix 2 & 3). Visual

inspection of histograms indicated that each of the distributions were sufficiently

normal to proceed to multivariate analysis.

Identification of Outliers

Outliers were identified by examining histograms (see Appendix 2&3) and box

plots for the dependent variable and each of the independent variables. However, a

decision was made to not eliminate any of the outliers in each variable because there

were so few cases in total10. Moreover, for a number of the outliers, plausible

explanations for the extreme data or unique points of county that relate to the subject

under investigation were apparent; thus, elimination of these outliers could have

inappropriately biased result

Screening for Multicollinearity Prior to Multivariate Regression Analysis

Pearson’s correlation test was performed on the independent variables to

determine whether multicollinearity might be an issue during multivariate regression.

9 Data Sources, Definitions, and Notes, Community Health Status Indicators 2009 10 Less than 10 cases per each independent variable

11

According to the result of Pearson’s correlation coefficients, eliminations were made for

the variables showing more than 0.8 : Population size, Population density, Race portion

of White, Native American and Hispanic, Unemployed, Medicaid beneficiaries, and

Primary care physicians.

In order to double check the potential for multicollinearity , variation inflation

factors (VIFs) were inspected for the initially selected 16 independent variables. (See

Appendix 5) As aligned with Pearson test result, Unemployed, Uninsurance, Elderly

Medicare and Medicaid Beneficiaries appear significant level of correlations.

Multivariate Regression Analysis

Initial model was generated by considering all candidate independent variables

in the order they were entered into SPSS; subsequent several models were constructed

using stepwise deduction of variables, considering regression results as well as VIF

correlation index.

The followings are the specific process and reasons for finalizing the composition

of 11 independent variables for Multivariate regression.

Exclusions of Population size and Population density are decided as their

limited contributions to R2

As the significant level of VIF and Pearson correlation test result

between Unemployment and Uninsurance rate, Uninsurance rate is

chosen to get implications about public health policy

Regarding races, Black and Asian have a clear direction of coefficient in

regression analysis. However, the result of Hispanic is unsecured, and

White & American Indian have no meaningful consequences.

Three variables: Medicaid Beneficiaries, Elderly Medicare and Prime care

physicians show the significant correlation level in VIF. Thus, Elderly

Medicare is selected as its higher coefficient level than others.

Exclusion of Dentist rate is made as its limited coefficient level.

12

4. Results

Descriptive Statistics

Table 3: Descriptive Statistics for Dependent Variable

Variable Mean Median Std.Deviation Minimum Maximum Percentile

25 50 75

Average life

expectancy 76.323 76.500 1.9975 66.6 81.3 75.0 76.5 77.7

Table 4: Descriptive Statistics for Independent Variables

Variable Mean Median Std.Deviation Minimum Maximum Percentile

25 50 75

Poverty 15.24 14.30 6.06 3.10 54.40 10.90 14.30 18.30

Black 9.12 2.30 14.39 0.00 86.00 0.60 2.30 10.60

Asian 1.18 0.50 2.77 0.00 55.60 0.30 0.50 1.00

No Exercise 26.51 26.00 6.70 8.30 52.40 21.90 26.00 30.80

Few fruits/

vegetables 78.92 79.00 5.16 63.10 96.40 75.50 79.00 82.40

Obesity 24.15 24.30 4.90 4.20 42.60 21.10 24.30 27.20

Smoker 23.11 23.00 5.73 3.60 46.20 19.40 23.00 26.70

Uninsured 15.10 14.41 5.08 0.00 41.91 11.33 14.41 18.02

13

Elderly

Medicare 14.76 14.30 4.26 0.00 38.10 11.98 14.30 17.16

Community

Health

Center

0.51 1.00 0.50 0.00 1.00 0.00 1.00 1.00

HPSA 0.75 1.00 0.43 0.00 1.00 1.00 1.00 1.00

Regression Results

Final multivariate regression models were generated (See Table 5 below).

Table 5: Summary of Final Model

Model R R2 Adjusted R2 Std. Error of

the Estimate

Regression 2 0.860* 0.739 0.737 1.0451

* a. Predictors: (Constant), Poverty, Uninsurance%, Black, Asian, Few_Fruit_Veg, Obesity, No_Exercise, Smoker, Elderly_Medicare%, Community_Health_Center_Ind, HPSA_Ind, b. Dependent Variable: ALE

The regression equation for the final model is:

Average life expectancy = 83.640 -0.081(Poverty)-0.052(Black)+0.050(Asian)

-0.057(no exercise)-0.008(Few Fruit and vegetable)

-0.048(Obesity)-0.109(Smoker)-0.019(Uninsurance)

+0.030(Elderly Medicare)+0.138(Community HC)

-0.059(HPSA)

Positive correlation11 : Community health center > Asian > Elderly

Medicare

Negative correlation : Smoker > Poverty > HPSA > No exercise >

Obesity > Black > Uninsurance > Few fruit and vegetable

11 In order of coefficient

14

Table 6: ANOVA of Final Model

Model Sum of

Squares df

Mean

Square F Sig.

Regression

Residual

Total

5506.510

1943.000

.000

3

45

48

.000

.000

16.829

.000*

* Predictors: (Constant), Poverty, Uninsurance%, Black, Asian, Few_Fruit_Veg, Obesity, No_Exercise, Smoker, Elderly_Medicare%, Community_Health_Center_Ind, HPSA_Ind, b. Dependent Variable: ALE

Table 7: Coefficients* and Collinearity Statistics of Final Model

Coefficientsa

Model

Unstandardized

Coefficients

Standardized

Coefficients

t Sig.

Collinearity

Statistics

B Std. Error Beta Tolerance VIF

(Constant) 83.640 .534 156.654 .000

Poverty -.081 .006 -.229 -13.530 .000 .512 1.952

Black -.052 .002 -.345 -23.612 .000 .687 1.455

Asian .050 .009 .076 5.418 .000 .744 1.344

No_Exercise -.057 .006 -.183 -9.658 .000 .407 2.459

Few_Fruit_Veg -.008 .006 -.019 -1.337 .181 .730 1.369

Obesity -.048 .008 -.107 -6.363 .000 .518 1.931

Smoker -.109 .006 -.290 -18.763 .000 .613 1.632

Uninsurance% -.019 .006 -.041 -2.947 .003 .767 1.304

Elderly_Medicare% .030 .007 .057 4.113 .000 .766 1.306

15

5. Public Policy

A key objective of Public Policy at the local, state and national levels is to

promote Public Health. As discussed in the literature, Average Life Expectancy is a

valuable indicator to monitor Public Health of population and its environments. In this

regard, opportunities to promote Average Life Expectancy deserve the attention of a

wide spectrum of policy makers.

This study generated data which demonstrate that Average Life Expectancy at

the US county level is significantly a function of Poverty, Race (Black, Asian), health

related daily behavioral risk factors (no exercise, Few Fruit and vegetable, Obesity,

Smoker) and political health supports (Uninsurance, Elderly Medicare, Community health

center, HPSA). Therefore, results from this study represent data that policymakers may

want to consider in order to develop or refine county-level strategies to promote Public

health.

County-level policymakers who are interested in developing strategies to

stimulate Public health enhancement may want to consider the following possibilities

which are supported by the specific findings of this analysis:

Mobilize and enhance Community Health Centers which are for low

income and uninsurance care

Develop policies and plans that support HPSA

Assures the quality and accessibility of health services, especially in HPSA

Community_Health_Center_I

nd

.138 .054 .034 2.542 .011 .843 1.186

HPSA_Ind -.059 .074 -.010 -.787 .431 .866 1.155

a. Dependent Variable: ALE

16

Link people to needed personal health services and assure the provision

of health care when otherwise unavailable

Inform and educate people about healthy behaviors

Intensify anti-smoking policies

6. Conclusion

The results of this US county level analysis show some similarities to the results

obtained in previous studies. The final model of this multivariate regression analysis

indicates that Poverty, Black, Asian, no exercise, Few Fruit and vegetable, Obesity,

Smoker, Uninsurance, Elderly Medicare, Community health center, HPSA are significant

inputs to Average Life Expectancy at the US county level. In terms of statistical

coefficient, Community health center, Smoking, and Poverty show relatively higher

influences.

7. Technical appendix

Appendix1. Original Source and Reference of Independent Variables

Poverty Level

The percentage of individuals living below the poverty level in 2008 is data obtained

from the “Small Area Income Poverty Estimates (SAIPE),” U.S. Bureau of the Census and

can be obtained at

http://www.census.gov/did/www/saipe/data/statecounty/data/index.html. Poverty

percent for the composite county of Hoonah-Angoon-Skagway is the weighted (total

census 2008 population) poverty percentage (census poverty percent 2008) for the

17

constituent counties (02-105 and 02-230)

Population by Race/Ethnicity

Race- and ethnicity-specific population sizes are from “Annual estimates of the resident

population by age, sex, race, and Hispanic origin for counties: April 1, 2000 to July 1,

2008.” These data are mid-year estimates of the resident population of 2008, and

reflect standard race and ethnicity categories in use by the U.S. Bureau of the Census,

and can be obtained at http://www.census.gov/popest/counties/asrh/CC-EST2008-

alldata.html. Note, the percentages of white, black, Asian American/Pacific Islander, and

American Indian do not total to 100% due to the multiple race category. The percent

Hispanic is non-additive with the race categories. The reader is advised that populations

cross-classified by race and ethnicity (e.g., non-Hispanic white; non-Hispanic black, etc.)

are available at http://wonder.cdc.gov/Bridged-Race-v2008.HTML.

No Exercise

The percentage of adults reporting of no participation in any leisure-time physical

activities or exercises in the past month.

Few Fruits/Vegetables

The percentage of adults reporting an average fruit and vegetable consumption of less

than 5 servings per day.

Obesity

The calculated percentage of adults at risk for health problems related to being

overweight, based on body mass index (BMI). A BMI of 30.0 or greater is considered

obese. To calculate BMI, multiply weight in pounds by 703 and divide the result by

height (in inches) squared.

18

Smoker

The percentage of adults who responded “yes” to the question, “Do you smoke

cigarettes now?”

Uninsured Individuals

The estimated number of uninsured individuals under age 65 in the county in 2006 is

from the U.S. Census Bureau, Small Area Health Insurance Estimates Program (SAHIE).

The SAHIE program models county-level health insurance coverage by combining survey

data with population estimates and administrative records. Data and information on

survey methodology and confidence intervals are found at Area Resource File, Health

Resources and Services Administration, 2008.

Community Health Centers

These centers are a source of care for low-income and uninsured individuals and

families and receive a portion of their funding through grants from HRSA. The data is

current as of September 30, 2009. Source: HRSA. Geospatial Data Warehouse,

http://datawarehouse.hrsa.gov/.

Health Professional Shortage Area

These are counties that have been designated as single-county, primary medical care,

health professional shortage areas, as determined by the Secretary of Health and

Human Services, current as of September 30, 2009. Source: HRSA. Geospatial Data

Warehouse, http://datawarehouse.hrsa.gov/.

19

Appendix 2: Histogram of Dependent Variable

20

Appendix 3: Histogram of Independent Variables

21

22

23

Appendix 4: Pearson Test Result of Finally Utilized Dependent Variables

Correlations

Povert

y Black Asian

No_

Exercise

Few_

Fruit_Veg Obesity Smoker

Uninsurance

%

Elderly

Medicar

e%

Dentist

Rate

Community

Health_

Center

HPSA

_Ind

Poverty 1 .462 -.177 .556 .312 .427 .332 .216 -.058 -.293 -.188 -.226

Black .462 1 .006 .263 .099 .285 .033 .038 -.254 -.041 -.219 -.099

Asian -.177 .006 1 -.250 -.269 -.248 -.202 -.025 -.286 .346 -.158 .135

No_Exercise .556 .263 -.250 1 .404 .572 .526 .176 .139 -.388 .014 -.204

Few_

Fruit_Veg .312 .099 -.269 .404 1 .377 .266 .067 .076 -.386 .179 -.265

Obesity .427 .285 -.248 .572 .377 1 .414 -.058 .047 -.336 -.035 -.126

Smoker .332 .033 -.202 .526 .266 .414 1 -.062 -.082 -.236 -.041 -.044

Uninsurance% .216 .038 -.025 .176 .067 -.058 -.062 1 -.040 -.209 -.022 -.270

Elderly_

Medicare% -.058 -.254 -.286 .139 .076 .047 -.082 -.040 1 -.135 .242 -.076

Community_H

ealth_

Center_Ind

-.188 -.219 -.158 .014 .179 -.035 -.041 -.022 .242 -.120 1 -.038

HPSA_Ind -.226 -.099 .135 -.204 -.265 -.126 -.044 -.270 -.076 .330 -.038 1

24

Appendix 5: VIF Test Result of All Candidate Dependent Variables

Model

Unstandardized Coefficients

Standardized

Coefficients

t Sig.

Collinearity Statistics

B Std. Error Beta Tolerance VIF

1 (Constant) 83.168 .508 163.737 .000

Poverty -.090 .006 -.256 -15.026 .000 .504 1.986

Black -.054 .002 -.354 -24.499 .000 .698 1.433

Asian .030 .010 .045 3.105 .002 .700 s1.428

Hispanic .015 .003 .078 5.568 .000 .738 1.355

Unemployed -2.659E-5 .000 -.171 -2.293 .022 .026 38.181

No_Exercise -.059 .006 -.189 -10.249 .000 .428 2.339

Few_Fruit_Veg -.007 .006 -.017 -1.150 .250 .705 1.419

Obesity -.034 .007 -.077 -4.692 .000 .539 1.856

Smoker -.103 .006 -.275 -17.848 .000 .613 1.631

Uninsured -5.194E-7 .000 -.020 -.526 .599 .098 10.178

Elderly_Medicare 7.981E-6 .000 .162 3.270 .001 .059 16.881

Medicaid_Beneficiaries 6.261E-7 .000 .036 .890 .374 .087 11.443

Prim_Care_Phys_Rate .000 .001 .006 .320 .749 .443 2.258

Dentist_Rate .000 .001 .006 .311 .756 .438 2.282

Community_Health_Center_Ind .252 .056 .061 4.517 .000 .791 1.264

HPSA_Ind -.056 .075 -.010 -.751 .453 .848 1.179

25

a. Dependent Variable: Average Life Expectancy