Content uploaded by Bryan J Pesta
Author content
All content in this area was uploaded by Bryan J Pesta on Oct 17, 2018
Content may be subject to copyright.
Toward an index of well-being for the fifty U.S. states
Bryan J. Pesta
a,
⁎, Michael A. McDaniel
b
, Sharon Bertsch
c
a
Cleveland State University, Department of Management, 2121 Euclid Avenue, Cleveland OH 44115, United States
b
Virginia Commonwealth University, United States
c
University of Pittsburgh at Johnstown, United States
article info abstract
Article history:
Received 27 July 2009
Received in revised form 1 September 2009
Accepted 16 September 2009
Available online xxxx
Well-being is a construct spanning multiple disciplines including psychology, economics,
health, and public policy. In many ways, well-being is a nexus of inter-correlated variables,
much like the gnexus. Here, we created an index of well-being for the geographical and
political subdivisions of the United States (i.e., states). The measure resulted from hierarchical
principal components analyses of state-level data on various hypothesized sub-domains of
well-being, including general mental ability, education, economics, religiosity, health, and
crime. A single, general component of well-being emerged, explaining between 52 and 85% of
the variance in the sub-domains. General mental ability loaded substantially on global state
well-being (.83). The relationship between global well-being and other important state-level
outcomes was examined next. We conclude by offering parallels between the gnexus and the
well-being nexus.
© 2009 Elsevier Inc. All rights reserved.
Keywords:
Intelligence
Well-being
gnexus
U.S. states
1. Introduction
The present study is an attempt to integrate multiple
research streams into one. In particular, we review the multi-
disciplinary literature on well-being, and then derive an index
of the construct for the fifty U.S. states.We show that state well-
being is substantially related to general mental ability and its
covariates. We then test whether well-being can predict other
important outcomes atthe state level. Finally, we offer parallels
between the gnexus and a well-being nexus.
We selected the fifty U.S states as the unit of analysis
because aggregate-level data are becoming increasingly im-
portant in psychology (see, e.g., Diener, 2000; Lynn, Harvey, &
Nyborg,2008; McDaniel,2006; Reeve, 2009;seealsoDavenport
& Remmers, 1950). For example, both Diener (2009) and
Gottfredson (2004a)have argued that researchersshould move
beyond the individual, and test the impact of psychological
constructs (well-being nationally, and gin epidemiology,
respectively) on aggregate-level variables.
Examining the well-being of U.S. states, specifically, is
important for at least five reasons. First, geographic differences
in well-being and their correlates represent important ques-
tions in applied psychology and public policy. State differences
in well-being are potential drivers of regional differences in
social, political, economic and psychological criteria that impact
millions of people. Second, considerable objective data exist at
the state level which can inform the development of a well-
being measure and permit examination of nomonological
relationships. Third, much of these data are collected period-
ically as part of federal and private programs. Thus, one could
track trends in such data over time. Trends in the data can be
used to evaluate the effectiveness of changes in public policy
within states. Fourth, substantial between-state variability
exists in the variables we examine. Such variability permits
cross-sectional analyses of correlates (e.g., government effec-
tiveness) that can help determine the causes of inter-state
differences in well-being. Fifth, as shownbelow, well-beingand
gare intimately linked at the state level. No complete
understanding of well-being (or g) can come without appeal
to its affects and covariates at all levels of analysis. In particular,
it appears that well-being represents a network of inter-
correlated variables, much like the gnexus, described next.
Intelligence xxx (2009) xxx–xxx
⁎Corresponding author. Tel.: +1 (216) 687 4749.
E-mail address: b.pesta@csuohio.edu (B.J. Pesta).
INTELL-00546; No of Pages 9
0160-2896/$ –see front matter © 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.intell.2009.09.006
Contents lists available at ScienceDirect
Intelligence
ARTICLE IN PRESS
Please cite this article as: Pesta, B.J., et al., Toward an index of well-being for the fifty U.S. states, Intelligence (2009),
doi:10.1016/j.intell.2009.09.006
1.1. The g nexus
General mental ability (g) is a ubiquitous predictor of
success in life.
Across dozens of studies, gcorrelates consistently with
important, real-world outcomes, including educational
achievement (Gottfredson, 2004b, 2005); income and wealth
(Lynn & Vanhanen, 2006); health (Batty, Deary, & Gottfred-
son, 2007; Reeve, 2009), longevity (Deary, 2008), job
performance (Hunter, 1993; Schmidt & Hunter, 1998), and
law-abidingness (Herrnstein & Murray, 1994).
The consistent finding that gis essential to predicting a
variety of life outcomes has led researchers to propose the
existence of a gnexus (Jensen, 1998; Nyborg, 2003). As
identified by Jensen (1998), the gnexus is a network of inter-
correlated variables with general mental ability at the center.
It has both horizontal and vertical components. The horizon-
tal component includes real-world variables which co-vary
and interact with general mental ability. Examples include
many of the variables cited above. The vertical component
includes presumed causes of individual differences in g, with
a special focus on biological and neuropsychological variables
(i.e., individual differences in properties of the human brain).
Much recent work has focused on mapping the gnexus.
McDaniel (2006), for example, has shown that state IQ
correlates strongly with a global measure of state health and a
separate measure of violent crime rates across states.
Similarly, Reeve (2009) presented data on state IQ as a
predictor of a variety of health measures, including infant
mortality and fertility rates. Nyborg (2009) reported that
religiosity is inversely related to g, using a large sample of
white, adolescent Americans (see also see also, Bertsch &
Pesta, 2009; Reeve, 2009). Finally, Lynn et al. (2008; see also
Lynn, in press) documented the relationship between IQ and
national poverty rates, income levels, and rates of atheism.
Many of the variables that correlate with gwould likely
also be indicators of well-being. In fact, well-being could be a
central node in the nexus of inter-correlated variables that
includes g. The overlap between the gnexus and a “well-
being nexus”is implied by how psychologists have oper-
ationalized subjective well-being as a scientific construct.
1.2. The well-being nexus
Well-being has been a topic of interest to humanity since
at least the sixth century B.C. (Steel, Schmidt, & Shultz, 2008).
Within psychology, well-being generally lacks a fixed
definition, although much research has focused on the
construct of subjective well-being (e.g., Diener & Lucas, 1999;
Steel et al., 2008). Subjective well-being is thought to be
comprised of four dimensions: life satisfaction, happiness,
affect, and quality of life (Steel et al., 2008).
Life satisfaction is typically based on an overall evaluation
of one's life (Pavot, Diener, Colvin & Sandvik, 1991; Steel et al.,
2008). Happiness refers to an optimistic outlook or mood that
endures over time (Averill & More, 2000; Steel et al., 2008).
When defined as affect, well-being shares close construct
similarity with personality. When well-being is defined with
respect to negative affect, it appears quite similar to the
personality trait of neuroticism and when defined with respect
to positive affect it appears quite similar to the personality trait
of extraversion (Steel et al., 2008; Yik & Russell, 2001). Quality
of life is usually a global measure that is a composite of well-
being across multiple life areas (e.g., financial, health, social;
Steel et al., 2008).
A second way to conceptualize well-being comes from the
industrial/organizational psychology literature, and appeals
to both physical and psychological health (see Warr, 1987,
2007 for a comprehensive discussion). Physical health refers
to factors like the absence of disease, one's level of physical
fitness, or whether one's basic needs (e.g., food, shelter) are
met. Psychological health comprises five facets: affective
well-being (similar to affect as a component of subjective
well-being above), competence (having the intellectual
ability needed to deal with life demands), autonomy (the
ability to resist environmental influences), aspiration (moti-
vation to establish and work toward meeting goals), and
integrated functioning (the ability to balance different life
demands).
In either case, the list of variables comprising measures of
well-being seems to overlap considerably with the list of
variables correlated with g. With this in mind, we attempted
to identify state-level variables that would capture–at least in
a preliminary measure–global, state well-being. In particular,
we identify and review literature on various “sub-domains”of
well-being, including intelligence, religiosity, crime, educa-
tion, health and income. State-level data on these variables
are readily available from the U.S. census and other private
sources. We do not, however, claim that our sub-domains
completely measure well-being, or map one to one with all of
its facets. Instead, we selected the sub-domains because they
are logically consistent with both psychological conceptions
of subjective well-being reviewed above, and because
reliable, state-level data already exist on many variables
that theoretically fit in each sub-domain.
1.3. Sub-domains of well-being at the state level
Cognitive ability is an obvious candidate for inclusion as a
sub-domain of well-being. The variables that gcorrelates
with are featured prominently in most discussions of well-
being (especially with regard to quality of life variables, and
competence—see, e.g., Gottfredson, 2004c). Physical health is
likewise critical to well-being, as it plays a central role in
Warr's (1987) conceptualization of the construct.
Religiosity is often mentioned as a component of subjec-
tive well-being (Aghili & Kumar, 2008; Ellison & Levin, 1998;
Heaven & Ciarrochi, 2007; Joshi, Kumari, & Jain, 2008). This is
especially true in the area of health psychology (Hill &
Pargament, 2008; McCullough, Hoyt, Larson, Koenig, &
Thoresen, 2000; Reeve, 2009). For example, people who
attend church regularly may have longer life expectancies
(Powell, Shahabi, & Thoresen, 2003). Park (2007) argued that
religion improves health by increasing a person's level of
social support and sense of self-meaning, and by offering
prohibitions against certain unhealthy behaviors such as
drinking or smoking. Hence, religion might contribute to
well-being by helping people avoid harmful behaviors, and by
offering a source of support and community to fall back on in
times of need.
Not all studies show positive relationships between religi-
osity and health. Recently, Reeve (2009) reported aggregate
2B.J. Pesta et al. / Intelligence xxx (2009) xxx–xxx
ARTICLE IN PRESS
Please cite this article as: Pesta, B.J., et al., Toward an index of well-being for the fifty U.S. states, Intelligence (2009),
doi:10.1016/j.intell.2009.09.006
data from nearly 200 countries around the world. He found that
religiosity (operationalized as thepercentage of people within a
country who report belief in a god) predictedgreater infant and
maternal mortality rates, deaths from HIV/AIDS, and lower life
expectancies.
Further complicating the issue is that high levels of ghave
been linked to lower levels of religiosity, both at the national
(Lynn et al., 2008; Reeve, 2009) and individual levels (Bertsch
& Pesta, 2009). One possibility is that religiosity is inversely
related to competence (i.e., one of the five facets of
psychological health). Competence includes holding beliefs
and world views that accurately reflect reality (Warr, 1987).
Adopting irrational or mystical belief systems could partly
result from lacking the information-processing capacity
needed to think critically (see, e.g., Bertsch & Pesta, 2009).
1
At any rate, inclusion of religiosity as a well-being sub-
domain, either because it creates happiness and social
support systems for many people, or because it predicts
deficits in competence, justifies its inclusion in an index of
well-being.
Subjective well-being in terms of happiness, quality of life,
and perhaps physical health would be affected by crime rates
in the area where one lives. The relationship between crime
and gis also well-established (see, e.g., Herrnstein & Murray,
1994). Likewise competence, life satisfaction and quality of
life would all be impacted by one's level of education. We
therefore included measures of crime and education as sub-
domains of global, state well-being.
The final sub-domain we identified is economic well-
being. At a macro level, economists often consider well-being
as synonymous with personal or household income (U.S.
Census Bureau, 2008a). Positive correlations between nation-
al wealth and subjective well-being have been reported
(Diener & Biswas-Diener, 2002), and the link between gand
income is well documented (see Lynn, in press, for a recent
example).
1.4. Measuring and analyzing state well-being
Our general strategy was to compile multiple indicators
for each of our six sub-domains of well-being (i.e., g, health,
religion, crime, education, and income levels). We coded data
primarily from the Statistical Abstract of the U.S. Census
(2008b). Other private sources also were included in the data
coding process (e.g., The Pew Foundation; The United Health
Fund). We subjected the variables to principal components
analyses (PCAs). Of key interest was whether the sub-
domains represented independent “facets”of well-being, or
whether they contributed to a single, global component
(much like specific cognitive abilities factor into a single,
global measure of intelligence; e.g., Jensen, 1998). In sum, we
wanted to see whether well-being has a “bottom line”—a
single number that accurately captures a large proportion of
variance across its diverse sub-domains.
After creating the well-being index, we then explored
whether it correlated with other important state-level out-
comes and/or demographic variables. We picked the vari-
ables here not based on any particular theory, but instead to
establish whether the well-being index possessed criterion
validity. Of primary interest were (1) political variables (e.g.,
the percentage of people within a state voting for Barack
Obama in the 2008 election; see e.g., Deary, Batty, & Gale,
2008; Rinderman, 2008, for aggregate-level data on political
variables), (2) religious affiliations (e.g., the percentage of
Catholics within a state), (3) attitudes on gay marriage bans
(e.g., whether a state has amended its constitution to ban gay
marriage), and (4) miscellaneous state-level variables that
did not logically fit with any of our well-being sub-domains
(e.g., traffic fatalities within a state).
2. Method
2.1. Sample and scale construction
The unit of analysis was the U.S. state, yielding a sample
size of 50. Although all states were included in our index, the
small sample size limited the methods available for data
analysis. To build scales, we conducted principal component
analysis (PCA) separately for the variables in each postulated
sub-domain of well-being (except for state IQ, which was a
single measure established by McDaniel, 2006). Based on the
above review, our well-being composites fell in the following
sub-domains: intelligence, religious belief, crime rates,
educational achievement, state health, and state income. A
hierarchical PCA was then conducted on the well-being sub-
domains (including state IQ) to determine whether state
well-being could best be described by a single general
component or whether well-being is more suitably explained
by a set of multiple components.
2.2. Measures
2.2.1. State IQ
State IQ was drawn from McDaniel (2006).
2.2.2. Religiosity
To form a state religiosity composite, we obtained data
from the Pew Forum on Religion and Public Life (2008). Seven
items were available to assess religiosity across states: (1) “I
am certain God exists,”(2) “Religion is very important to me,”
(3) “I attend church at least once per week,”(4) “I pray daily,”
(5) “My prayers are answered at least monthly,”(6) “My holy
book is literally true,”and (7) “Mine is the one true faith.”
Responses represented the percentage of survey respondents
in each state who agreed with each statement.
2.2.3. Crime-rate data
To form a state crime-rate composite, data on burglary,
murder and rape rates, as well as the number of inmates per
state, were obtained from the 2008, Statistical Abstract of the
U.S. Census (2008b). These variables were expressed as
counts per 100,000 of the population within a state. We also
obtained the violent crime rate per 1000 residents available
from McDaniel (2006).
2.2.4. Educational achievement
The educational-achievement composite was formed by
obtaining the percentage of each state's residents with a
bachelor's degree, and the percent of the state workforce with
1
We thank a reviewer for this idea.
3B.J. Pesta et al. / Intelligence xxx (2009) xxx–xxx
ARTICLE IN PRESS
Please cite this article as: Pesta, B.J., et al., Toward an index of well-being for the fifty U.S. states, Intelligence (2009),
doi:10.1016/j.intell.2009.09.006
science and engineering jobs. These data also came from the
Statistical Abstract of the U.S. Census (2008b).
2.2.5. State health
The health composite was comprised of the following
variables: the percentage of births to unwed mothers, the
percentage of births to teenage mothers, and infant mortality
rates (per 1000 births). These three variables came from the
U.S. Census Bureau (2008b). The health composite also
included a global measure of state health taken from the
United Health Foundation (2008). This measure is an index
comprising a battery of health-related variables. Examples
include smoking, obesity, and health insurance coverage.
2.2.6. Income
To form a state income composite, we obtained data from
the U.S. Census Bureau (2008b). Variables included: income
per capita, disposable income per capita, percent of families in
poverty, and percent of individuals in poverty.
2.2.7. Additional measures
The six sub-domains described above were entered into a
hierarchical PCA to create a global index of state well-being.
Thereafter, we obtained data on several other variables to
explore the presence or absence of a nomonological network
of well-being. These additional variables were either demo-
graphic in nature (e.g., the percentage of Catholics, Protestant
or Godless people [i.e., report no belief in God] in each state),
or themselves represented important outcomes (e.g., teacher
salaries; traffic fatalities). The latter variables did not logically
fit with our sub-domains (e.g., traffic fatalities) or did not load
on them (e.g., teacher salaries failed to load on the education
factor). The additional variables, however, served as tests of
whether the global well-being index could itself predict other
important state-level outcomes.
Five of these variables came from the U.S. Census,
including: (1) the number of physicians per capita, (2) the
number of driving fatalities per capita, (3) gross state product
per capita, (4) median teacher salaries (kindergarten through
twelfth grade), and (5) minimum wage. The percentage of
residents in each state voting for Barack Obama in the 2008
presidential election was obtained from the Federal Register
(2008). In addition, we coded the percentage of residents in
each state who are Catholic, Protestant, or Godless. These data
came from the Pew Foundation (2008) survey, and no other
religious group had large enough representations within
states to be included in the analyses.
We also coded the following two variables from the
political polling website Fivethirtyeight (2008): the percent-
age of gun owners in each state, and the ratio of Starbuck
stores to Walmart stores within a state. These two variables
were regularly used by pollsters in the 2008 presidential
election, as they are thought to roughly index a dimension of
liberalism/conservatism for residents within a state.
Two final variables were included in these analyses. As of
the end of the 2008 elections, 30 states have now amended
their constitutions to ban gay marriage whereas 20 have not
(National Conference of State Legislatures, 2009). Whether
gay marriage should be legal is an obviously contentious
issue. Arguments for and against, however, are often framed
around the construct of well-being, at least indirectly.
Denying people who are gay the right to marry would have
obvious effects on components of their subjective well-being,
including happiness, satisfaction and quality of life. Con-
versely, opponents of gay marriage often appeal to the
putative negative effects that gay marriage may have on
both family and societal well-being (see, e.g., Family Research
Council, 2001, 2002). Hence, we explored whether constitu-
tional bans of gay marriage correlate with our index of well-
being at the state level. As an additional measure in this
domain, we included the percentage of same-sex households
within a state from Fivethirtyeight (2008).
3. Results
Table 1 shows the results of five PCAs for the diverse sub-
domains of state well-being; namely, religion, crime, educa-
tional achievement, health, and income. The sixth sub-
domain was state IQ (a single measure). For each sub-
domain, only one principal component emerged with an
Eigen value greater than one. The percentage of variance
accounted for by these principal components ranged from
58% to 86%. The state IQ variable was drawn from McDaniel
(2006) who did not report a PCA for the data used to create
the composite. However, McDaniel did report an alpha
reliability of .99 which indicated the substantial stability in
state IQ data across years. For the five other sub-domains in
Table 1, the lowest alpha reliability was .64 for educational
Table 1
Principal components analysis for the domains of state well-being, with variance accounted for in first principal component, alpha reliabilities and correlation
matrix.
Well-being domain % variance in 1st
principal component
Correlation matrix
1234567
1. Religiosity 86 (.97)
2. Crime 58 .51 (.72)
3. Education 73 −.62 −.26 (.64)
4. Health 79 −.68 −.82 .61 (.93)
5. Income 85 −.72 −.42 .66 .63 (.94)
6. State IQ
a
–−.55 −.76 .41 .75 .57 (.99)
7. Global well-being
b
67 −.83 −.78 .72 .92 .81 .83 (.90)
a
State IQ was a single measure and so was not subjected to PCA. McDaniel (2006), however, reports an alpha reliability of .99 for this variable.
b
This is the global measure of state well-being resulting from a hierarchical PCA on the six sub-domains that precede it. Note that religiosity and crime have
negative loadings on global well-being.
4B.J. Pesta et al. / Intelligence xxx (2009) xxx–xxx
ARTICLE IN PRESS
Please cite this article as: Pesta, B.J., et al., Toward an index of well-being for the fifty U.S. states, Intelligence (2009),
doi:10.1016/j.intell.2009.09.006
achievement, and the alpha of the crime composite (.72) was
the second lowest. The alpha reliabilities for the remaining
three composites were in the .90s.
The bottomrow of Table 1 shows theresults of a hierarchical
PCA conducted on the six sub-domains of well-being. A strong
general component emerged from this analysis. It explained
67% of the variance in the six sub-domains, and no other
principal component emerged with an Eigen value of greater
than one. The alpha reliability of the global component was .90.
As multiple components of well-being did not emerge, the
global well-being index seems to be the most appropriate way
of expressing state well-being based on our data.
Religious belief has a large magnitude negative loading on
global state well-being, indicating that less religiously
inclined individuals are residents of high well-being states.
Though consistent with other aggregate-level data (see, e.g.,
Reeve, 2009), this pattern is inconsistent with results
typically seen at the individual level. There, religiosity is
often associated with subjective well-being and health-
related well-being (see, e.g., Powell et al., 2003). As expected,
crime was associated with negative state well-being but
educational achievement, health, income and IQ were
positively related to state well-being. State IQ, itself, loaded
.83 on the global well-being composite.
Table 2 shows rankings and standard scores for the 50 U.S.
states by global well-being, and by the six sub-domains. All
measureswere standardizedwith a mean of 100 and a standard
deviation of 15. We also reverse-scored religiosity and crime
such that higher well-being scores corresponded to less
religiosity and lower crime rates within a state (i.e., we reverse
scored these variables so that for all sub-domains in the table,
higher scores corresponded to higher well-being). Colorado, for
example, has a global well-being score of 113.0. Its religiosity
well-being score is 114.5, indicating that Colorado is one of the
least religious states in the country. The crime well-being score
for Colorado (98.5) places it slightly above average in terms of
crime rates relative to the other states.
The well-being scores in Table 2 produced roughly normal
distributions. All skew and kurtosis values were less than two
times their standard errors (in 10 of 14 cases, the skew or
kurtosis values were less than one standard error for the
variable in question). States with consistently high values in
all domains include Massachusetts, New Hampshire, and
Connecticut. These states therefore have relatively high state
IQs, income, health, and education levels, and relatively low
rates of crime and religiosity. Conversely, states scoring
consistently low on the well-being indices include Missis-
sippi, Louisiana, and Arkansas. As a general pattern, states in
the South scored lowest on well-being, while states on the
East Coast scored highest (other coastal states and the
Midwest fell somewhere in between).
Table 3 informs discussion of the nomonological network
of global state well-being by considering its correlation with
other political, social, health, and economic variables. For all
correlations in the table, a value of .236 is needed for
statistical significance (pb.05) for a directional test (e.g., one-
tailed) and a value of .279 is needed for a non-directional test
(e.g., two-tailed).
Regarding political variables, global well-being was posi-
tively correlated with the percent of votes cast in the state for
Barack Obama in the 2008 presidential election (r=.47).
Although concerns about many issues shaped the 2008
election, this finding may reflect a greater liberalism in states
with greater well-being. Consistent with our liberalism
inference, states high in well-being also had higher minimum
wages (r=.35), fewer residents owning guns; (r=−.34), and
nominally higher Starbucks to Walmart ratios (r=.23).
Likewise, states high in well-being were less likely to have
amended their constitutions to ban gay marriage (r=−.43),
and had a higher percentage of same-sex households (r=.42).
Religious affiliations by state also correlated moderately-
to-strongly with nearly every other variable in the table. For
example, states with higher percentages of Catholics (and
Godless people) fared better in global state well-being and all
of the economic variables in the table (e.g., state GPD, teacher
salaries, and minimum wages—see Table 3). Conversely,
states with many Protestants scored in the opposite direction
on these variables. Surprisingly, religious affiliations corre-
lated strongly with both active physicians per capita, and the
number of traffic fatalities within a state. These patterns may
be due to the heavy concentrations of Protestants in the South
(and Catholics elsewhere), where well-being scores seem
lowest. In sum, the correlations in Table 3 show a complex but
interconnected nexus of variables; all of which co-vary with
global state well-being.
4. Discussion
We created a multi-dimensional measure of well-being
for the 50 U.S. states. The analytic strategy was to build scales
by identifying variables that logically seemed to represent
hypothesized sub-domains of well-being. These sub-domains
included: IQ, religion, crime, education, health and income.
Hierarchical analysis of the sub-domains showed that a
general factor of well-being could be extracted from the
data. The global index of well-being, created from a PCA on
the sub-domains, predicted other important variables.
4.1. The g/well-being Nexus
At the level of the U.S. state, a nexus of inter-correlated
variables exist that together seem to offer a reliable indicator
of well-being. The well-being nexus also seems to overlap
considerably with the gnexus. State IQ itself predicted most
all of the variables that well-being did, and vice versa. Recall
that the gnexus has both horizontal (variables that correlate
and interact with g) and vertical (presumed causes of
individual differences in g) components (Jensen, 1998).
With regard to a well-being nexus, the horizontal and vertical
components would be similar to those seen with GMA.
Postulated causes of individual differences in well-being
could be identified in the vertical dimension, while the
consequences that follow from differences in well-being
could be identified in the horizontal dimension.
For example, health rate differences across states, mea-
sures of government effectiveness, or levels of pre-natal care
(together with other variables) might represent causes of
group differences in well-being. Social, economic and educa-
tional/cognitive outcomes would likely then co-vary (hori-
zontally) with state differences in the vertical direction. It is
clear, however, that gis a central node in the well-being
nexus.
5B.J. Pesta et al. / Intelligence xxx (2009) xxx–xxx
ARTICLE IN PRESS
Please cite this article as: Pesta, B.J., et al., Toward an index of well-being for the fifty U.S. states, Intelligence (2009),
doi:10.1016/j.intell.2009.09.006
4.2. Religiosity and well-being
Perhaps the most surprising finding in the present study
was the consistent and large negative correlations found
between religiosity and the other well-being sub-domains
(except crime, where the correlation was positive). Specifi-
cally, religiosity had the highest loading on the global well-
being component. Between 26% (crime) and 52% (income) of
the variance in the other diverse sub-domains was explained
just by knowing a state's level of religiosity. In Table 3, the
effects of religiosity seemed linked to the Protestant (versus
Catholic or Godless) faith. This finding is consistent with
Table 2, where well-being seemed lowest in the South
(containing more Protestants) and higher in the East Coast
(with more Catholics and Godless people).
Whatever the cause of the inverse relationship between
religiosity and state well-being, we see the consequences as
being significant. This is perhaps most clearly illustrated by
appeal to the data on constitutional bans of gay marriage.
The argument that same-sex marriage would lead to a
breakdown in morality or of societal/family values (see, e.g.,
Family Research Council, 2001, 2002) is not supported by
Table 2
State ranks and standard scores for the global measure of well-being and its sub-domains.
State Well-being
Rank/score
Religiosity
a
Rank/score
Crime
a
Rank/score
Education
Rank/score
Health
Rank/score
Income
Rank/score
IQ
Rank/score
Alabama 47/76.9 49/72.5 38/86.0 40/88.3 44/83.2 44/82.9 45.5/95.7
Alaska 25/104.4 4/123.3 47/79.8 6/116.4 25/101.0 11.5/109.2 36/99.0
Arizona 36/90.5 17/106.4 41/84.7 26/98.0 37/90.3 35/90.4 43/97.4
Arkansas 48/75.1 44/79.8 43/82.8 48/75.9 47/77.6 47/77.1 42/97.5
California 30/98.6 12/110.4 31/91.6 14/110.4 21.5/103.3 20/104.7 48/95.5
Colorado 10/113.0 8/114.5 28/98.5 2/128.5 19/106.0 10/111.7 20/101.6
Connecticut 3/122.7 6.5/118.8 18/110.0 5/119.6 10/114.3 1/134.2 9/103.1
Delaware 33/94.9 24/102.5 40/84.8 41/86.7 41/87.2 9/112.8 28/100.4
Florida 35/92.2 29/100.3 39/85.9 30/95.5 39/89.6 26/101.7 38.5/98.4
Georgia 42/85.2 40/85.0 33/90.2 34/92.3 43/84.2 36/89.9 40/98.0
Hawaii 28/100.4 33/97.0 22/106.6 18/106.4 16/109.4 11.5/109.2 47/95.6
Idaho 21/105.6 37/92.5 16/111.0 12/112.0 7/116.4 38/89.7 22/101.4
Illinois 27/100.6 19/105.1 30/96.0 23/101.0 28/96.1 15/107.4 31/99.9
Indiana 32/96.5 34/95.7 27/102.0 44/85.4 35/93.4 28/98.0 19/101.7
Iowa 12/109.2 18/105.5 7/115.4 42/86.6 9/115.2 24/103.4 8/103.2
Kansas 23/104.9 36/94.0 25/102.9 16/107.6 21.5/103.3 23/103.5 12/102.8
Kentucky 39.5/86.7 42/84.6 24/103.1 47/77.9 36/93.1 45/79.8 34/99.4
Louisiana 49/69.1 46/77.6 50/72.2 43/85.5 49/66.9 48/75.8 49/95.3
Maine 7/115.3 3/124.0 2/126.5 31/95.0 11/113.7 30/97.1 6.5/103.4
Maryland 22/105.5 26/101.7 36/86.2 3/126.2 33/93.7 3/126.3 32/99.7
Massachusetts 1/127.2 5/122.3 15/111.1 1/135.2 5/120.5 4/124.0 1/104.3
Michigan 29/100.2 20/103.7 34/90.0 11/112.5 27/97.7 29/97.7 27/100.5
Minnesota 5/119.3 14/107.7 10/113.5 9/114.3 3/123.5 7/116.1 5/103.7
Mississippi 50/61.2 50/63.9 37/86.1 50/65.3 50/64.3 50/64.8 50/94.2
Missouri 34/93.7 35/94.8 32/91.2 37.5/89.5 34/93.6 33/95.3 25/101.0
Montana 16/108.1 22/103.2 6/115.5 19/104.9 17/109.3 37/89.8 6.5/103.4
Nebraska 19/107.1 28/100.4 11/113.3 33/93.1 15/111.4 21/104.3 15/102.3
Nevada 37/89.9 16/106.7 46/80.0 45/81.2 30/94.7 16.5/106.9 44/96.5
New Hampshire 2/126.3 1.5/126.8 3/124.8 17/107.2 2/125.4 5/121.6 2/104.2
New Jersey 6/117.6 13/110.0 17/110.8 13/111.9 14/111.7 2/128.7 12/102.8
New Mexico 44/84.7 27/101.4 48/79.1 10/113.5 45/82.1 46/77.3 45.5/95.7
New York 17/107.9 9/114.3 19/108.4 22/103.8 24/102.6 14/107.7 26/100.7
North Carolina 39.5/86.6 43/80.3 35/86.4 32/93.9 42/86.0 39/89.3 29/100.2
North Dakota 9/113.4 30.5/100.2 1/128.4 26/98.0 6/116.7 25/102.1 3.5/103.8
Ohio 31/98.0 25/102.3 29/98.3 39/89.0 29/95.0 31/96.6 18/101.8
Oklahoma 43/84.9 41/84.7 42/83.5 37.5/89.5 40/87.9 42/86.0 35/99.3
Oregon 15/108.3 11/111.2 20/107.7 15/107.9 12/113.3 34/95.2 23/101.2
Pennsylvania 26/103.9 23/103.1 23/105.2 26/98.0 26/100.2 18/105.9 21/101.5
Rhode Island 13/109.1 6.5/118.8 13/112.4 21/104.0 18/108.7 19/105.1 33/99.5
South Carolina 46/77.3 48/76.4 49/75.6 35/90.2 48/75.3 43/83.2 38.5/98.4
South Dakota 24/104.5 30.5/100.2 8/114.6 36/89.9 20/104.0 27/98.6 12/102.8
Tennessee 45/78.5 47/77.3 44/81.8 46/80.9 46/80.1 40/88.9 41/97.7
Texas 38/89.0 39/86.0 45/80.3 24/100.9 31.5/94.0 41/86.9 30/100.0
Utah 20/105.7 45/78.5 5/115.6 20/104.6 1/127.2 32/95.9 24/101.1
Vermont 4/122.5 1.5/126.8 4/122.8 7/116.3 4/121.2 22/103.7 3.5/103.8
Virginia 14/108.9 32/97.3 21/107.0 8/115.5 23/103.2 8/113.7 16.5/101.9
Washington 8/113.5 15/107.0 26/102.5 4/124.9 8/116.2 13/107.9 16.5/101.9
West Virginia 41/86.4 38/88.7 14/111.9 49/75.4 38/90.2 49/75.2 37/98.7
Wisconsin 11/111.8 10/112.1 9/113.9 28/97.0 13/112.0 16.5/106.9 10/102.9
Wyoming 18/107.5 21/103.2 12/112.7 29/96.6 31.5/94.0 6/119.9 14/102.4
Notes. The well-being scores for all variables result from the following conversion: Well-being =100 +Z(15). States with identical well-being scores in a column
may not be listed as tied in the rankings due to rounding error.
a
Religiosity and crime were reverse coded such that higher well-being scores equal lower religiosity and lower crime within a state.
6B.J. Pesta et al. / Intelligence xxx (2009) xxx–xxx
ARTICLE IN PRESS
Please cite this article as: Pesta, B.J., et al., Toward an index of well-being for the fifty U.S. states, Intelligence (2009),
doi:10.1016/j.intell.2009.09.006
considering the data h ere. Correlations bet ween banning gay
marriage and the sub-domains were: religiosity (.45), crime
(.34), education (−.26), health, (−.26), income (−.36), and
IQ (−.36). These values suggest that well-being does not
follow from a state's decision to constitutionally ban gay
marriage. Likewise, the argument that morality stems only
from a higher power (for arguments for and against, see
Garcia & King, 2009) fares even worse, considering the
correlations between the percentage of Godless people in a
state and th e sub-domains: religiosity (−.79), crime (−.32),
education (.52), health (.51), income (.43), and IQ (.29). To
the extent that morality leads to well-being (e.g., lower
crime rates; lower teenage pregnancy rates), these correla-
tions show that well-being can be achieved in the absence of
religiously-derived morality (i.e., the correlations do not
support the hypothesis that societal well-being depends on
the religiosity of its citizens).
4.3. Implications
The global index of well-being could be used to inform
decision makers with regard to important policies, programs
and issues at the state and national levels. It could serve as a
report card, a guide for what needs to be done next and
where, and as a useful “flow chart”illustrating how outcome
variables are not independent of other markers of well-being,
but instead are intimately linked. An empirically created
index of well-being, refined with better measures and more
variables over time, could serve invaluable as an assessment
tool for the well-being of the 50 U.S. states.
Decision makers could also track how changes in one
variable might lead to changes in other variables and sub-
domains.Totheextentthatwhatgetsmeasuredgets
attention, highlighting differences in important societal out-
comes across states might help prioritize the order in which
different issues are addressed. Further, state rankings already
exist on a number of economic and demographic variables via
the U.S. Census and other sources. The data here show that
these variables are not independent, but map logically into
coherent domains, which themselves are statistically
explained by a higher-order domain. Reporting outcome
variables in a series of tables without any consideration of
how variables across tables relate likely results in missing
important patterns and relationships. In sum, well-being is a
hierarchy/nexus of interconnected domains and variables,
and so should be considered as such when reporting data on
specific outcomes at the state level.
4.4. Limitations and directions for future research
Although we can describe the profile of high and low well-
being states, the present data do not permit causal conclu-
sions. For example, states high in well-being are found mainly
in the East Coast; their citizens are more liberal, educated,
wealthy, and intelligent on average, but less religious. All or
none of these variables could be the causal link that binds
them together, and binds them with other important out-
comes (e.g., crime and health rates). Specific hypotheses
about causality await further study.
Our global and domain specific indices should be
replicated, extended to other variables, and recalculated
with passing time. Other reliable indicators of the sub-
domains could be added to the index. More and better
variables might lead to identifying other important sub-
domains, and a clearer picture of the well-being nexus. In
particular, our educational-achievement sub-domain com-
prised only two correlated variables. Future research could
perhaps identify other state-level outcome variables that
increase the validity of the construct.
Also missing from our index are measures of personality
variables at the state level. We are aware of state-level data
on the NEO model of personality (Rentfrow, Gosling, &
Potter, 2008). We did not report analyses related to
personality and state well-being because they produced
inconsistent and un-interpretable results. Specifically, con-
sidering a matrix of the five NEO traits (i.e., neuroticism,
extraversion, openness, agreeableness and conscientious-
ness) and the seven well-being composites used here (i.e.,
the global index and the six sub-domains), only 10 of 35
correlations (29%) were statistically significant. Half of these
involved correlations between conscientiousness and the
sub-domains, but the correlations were in unexpected
Table 3
A nomonological network of global state well-being and its covariates.
Variable 1234567891011121314
1. State well-being –
2. % Obama .47 –
3. Active doctors .52 .71 –
4. Traffic fatalities −.71 −.58 −.67 –
5. State GDP .44 .38 .43 −.50 –
6. Teacher salary .39 .62 .64 −.67 .59 –
7. Minimum wage .35 .65 .49 −.41 .31 .59 –
8. % Catholic .61 .59 .57 −.48 .52 .53 .40 –
9. % Protestant −.68 −.47 −.43 .52 −.44 −.48 −.47 −.72 –
10. % Godless .58 .40 .25 −.38 .32 .31 .50 .28 −.63 –
11. % gun owners −.34 −.77 −.69 .61 −.46 −.70 −.50 −.64 .54 −.22 –
12.S/W ratio .23 .44 .20 −.25 .29 .42 .51 .14 −.39 .43 −.43 –
13. Gay marriage ban −.43 −.49 −.50 .48 −.37 −.39 −.36 −.44 .32 −.27 .37 .02 –
14. Same-sex households .42 .62 .53 −.51 .24 .39 .47 .34 −.55 .55 −.53 .37 −.40 –
Notes. A correlation of r= .236 is statistically significant (pb.05) for a directional test, and a correlation of r= .279 is statistically significant for a non-directional
test. S/W ratio = Starbucks to Walmart ratio.
7B.J. Pesta et al. / Intelligence xxx (2009) xxx–xxx
ARTICLE IN PRESS
Please cite this article as: Pesta, B.J., et al., Toward an index of well-being for the fifty U.S. states, Intelligence (2009),
doi:10.1016/j.intell.2009.09.006
directions (e.g., conscientiousness correlated −.31 with
health; −.30 with education and .30 with crime).
2
Whether
the odd relationships here reflect a problem with the well-
being scales, the state-level personality scores, or both, is
unknown.
A future research direction is to consider refining the levelof
analysesto major U.S. cities,or MetropolitanStatistical Areas,as
well-being may differ across regions and cities within a state.
One problem with creating indices like these, however, is the
relative lack of data on many of the outcome variables included
in this research. In general, the smaller the geographic unit of
analysis, the less available data on that unit.
In sum, well-being at the state level appears to be comprised
of separate but related sub-domains. Though the domains we
selected to measure well-being seemed diverse (i.e., IQ,
religiosity, health, crime, education and income), the general
index of well-being explained 67% of their variance. The
emergence of a strong general component suggests the
existence of a well-being nexus, much like the gnexus identified
in research on human intelligence (Jensen, 1998). Our data
show, however, substantial overlap between the gnexus and
the well-being nexus. It is hoped that f uture research refines the
index to strengthen its psychometric properties, and then uses it
to test hypotheses about possible reasons for why states differ–
often markedly–in our measure of global well-being.
References
Aghili, M., & Kumar, G. V. (2008). Relationship between religious attitude and
happiness among professional employees. Journal of the Indian Academy
of Applied Psychology,34,66−69.
Averill, J., & More, T. (2000). Happiness. In L. Michael & J. M. Haviland (Eds.),
Handbook of emotions (pp. 617−629). New York: Guilford Press.
Batty, G., Deary, I., & Gottfredson, L. (2007). Premorbid (early life) IQ and
later mortality risk: A systematic review. Annals of Epidemiology,17,
278−288.
Bertsch, S., & Pesta, B. (2009). The Wonderlic Personnel Test and elementary
cognitive tasks as predictors of religious sectarianism, scriptural
acceptance and religious questioning. Intelligence,37, 231−237.
Davenport, K., & Remmers, H. (1950). Factors in state characteristics related
to average A-12 V-12 test scores. Journal of Educational Psychology,41,
110−115.
Deary, I. (2008). Why do intelligent people live longer? Nature,456,
175−176.
Deary, I., Batty, G., & Gale, C. (2008). Childhood intelligence predicts voter
turnout, voting preferences, and political involvement in adulthood: The
British Cohort Study. Intelligence,36, 548−555.
Diener, E. (2000). Subjective well-being: The science of happiness, and a
proposal for a national index. American Psychologist,55,34−43.
Diener, E. (2009). National policy of well-being. Observer,22,19−20.
Diener, E., & Biswas-Diener, R. (2002). Will money increase subjective well-
being? Social Indicators Research,57, 119−169.
Diener, E., & Lucas, R. (1999). Personality and subjective well-being. In E.
Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The foundations of
hedonic psychology (pp. 213−229). New York: Russell Sage Foundation.
Ellison, C., & Levin, J. (1998). The religion-health connection: Evidence,
theory and future directions. Health Education & Behavior,25, 700−720.
Family Research Council. (2001). The negative health effects of homosexu-
ality. Retrieved November 14, 2008, from the World Wide Web: http://
www.frc.org/get.cfm?i=IS01B1
Family Research Council. (2002). Homosexuality and child sexual abuse.
Retrieved November 14, 2008, from the World Wide Web: http://www.
frc.org/get.cfm?i=IS02E3
Federal Register. (2008). Popular vote totals, 2008 presidential election.
Retrieved February 1, 2009, from the World Wide Web: http://www.
archives.gov/federal-register/electoral-college/2008/popular-vote.html
Fivethirtyeight. (2008). Today's polls and final election projection. Retrieved
September 1, 2008 through November 3, 2008, from the World Wide
Web: http://www.fivethirtyeight.com
Garcia, R., & King, N. (2009). Is goodness without God good enough? A debate
on faith, secularism, and ethics. Lanham, MD: Rowman & Littlefield
Publishers.
Gottfredson, L. (2004a). Intelligence: Is it the epidemiologists' elusive
“fundamental cause”of social class inequalities in health? Journal of
Personality and Social Psychology,86, 174−199.
Gottfredson, L. (2004c). Schools and the gfactor. The Wilson Quarterly,35−45.
Gottfredson, L. (2004b). Life, death and intelligence. Journal of Cognitive
Education and Psychology,4,23−46.
Gottfredson, L. (2005). Implications of cognitive differences for schooling
within diverse societies. In C. L. Frisby & C. R. Reynolds (Eds.), Compre-
hensive handbook of multicultural school psychology (pp. 515−554). New
York, NY: Wiley.
Heaven, P., & Ciarrochi, J. (2007). Personality and religious values among
adolescents: A three-wave longitu dinal analysis. British Journal of
Psychology,98, 681−694.
Herrnstein, R., & Murray, C. (1994). The bell curve: Intelligence and class
structure in American life. New York: Free Press.
Hill, P., & Pargament, K. (2008). Advances in conceptualization and
measurement of religion and spirituality: Implications for physical
and mental health research. Psychology of Religion and Spirituality,S,
3−17.
Hunter, J. (1993). A causal model of cognitive ability, job knowledge, job
performance, and supervisor ratings. In F. J. Landy, S. Zedeck, & J.
Cleveland (Eds.), Performance measurement and theory (pp. 257−266).
Hillsdale, NJ: Lawrence Erlbaum Associates.
Jensen, A. (1998). The g Factor: The science of mental ability. Westport CT:
Praeger.
Joshi, S., Kumari, S., & Jain, M. (2008). Religious belief and its relation to
psychological well-being. Journal of the Indian Academy of Applied
Psychology,34, 345−354.
Lynn, R. in press. In Italy, north–south differences in IQ predict differences in
income, education, infant mortality, stature and literacy. Intelligence
doi:10.1016/j.intell.2009.07.004.
Lynn, R., & Vanhanen, T. (2006). IQ and global inequality. Athens, GA:
Washington Summit Publishers.
Lynn, R., Harvey, J., & Nyborg, H. (2008). Average intelligence predicts
atheism rates across 137 nations. Intelligence,37,11−15.
McCullough, M., Hoyt, W., Larson, D., Koenig, H., & Thoresen, C. (2000).
Religious involvement and mortality: A meta-analytic review. Health
Psychology,19, 211−222.
McDaniel, M. (2006). Estimating state IQ: Measurement challenges and
preliminary correlates. Intelligence,34, 607−619.
National Conference of State Legislatures. (2009). Same-sex marriage, civil
unions and domestic partnerships. Retrieved May 1, 2009, from the
World Wide Web: http://www.ncsl.org/programs/cyf/samesex.htm
Nyborg, H. (2003). The scientific study of general intelligence: A tribute to
Arthur Jensen. Oxford, U.K.: Pergamon Press.
Nyborg, H. (2009). The intelligence–religiosity nexus: A representative study
of white adolescent Americans. Intelligence,37,81−93.
Park, C. (2007). Religiousness/spirituality and health: A meaning systems
perspective. Journal of Behavioral Medicine,30, 319−328.
Pavot, W., Diener, E., Colvin, C., & Sandvik, E. (1991). Further validation of
the Satisfaction with Life Scale: Evidence for the cross method
convergence of well-being measures. Journal of Personality Assessment,
57, 149−161.
Pew Forum on Religion and Public Life. (2008). U.S. religious landscape
survey: Religious affiliation: Diverse and dynamic. Retrieved November
14, 2008, from the World Wide Web: http://religions.pewforum.org/
pdf/report-religious-landscape-study-full.pdf
Powell, L., Shahabi, L., & Thoresen, C. (2003). Religion and spirituality:
Linkages to physical health. American Psychologist,58,36−52.
Reeve, C. (2009). Expanding the gnexus: Further evidence regarding the
relationship among national IQ, religiosity, and national health out-
comes. Intelligence,37, 495−505.
Rentfrow, P., Gosling, S., & Potter, J. (2008). A theory of the emergence,
persistence, and expression of geographic variation in psychological
characteristics. Perspectives on Psychological Science,3, 339−369.
Rinderman, H. (2008). Relevance of education and intelligence for the
political development of nations: Democracy, rule of law and political
liberty. Intelligence,36, 306−322.
2
We mentioned in the introduction that when subjective well-being is
framed as positive aspect, it seems similar to the pers onality trait,
Extraversion. When framed as negative aspect, subjective well-being seems
similar to the trait, Neuroticism. Neither trait, however, correlated with our
global index of well-being (r's = −.09 and −.18, respectively). Neuroticism
did correlate −.30 with the health sub-domain, and −.30 with education.
Likewise, extraversion correlated −.33 with the education sub-domain,
although we have no theoretical explanation for why the relationship was
negative.
8B.J. Pesta et al. / Intelligence xxx (2009) xxx–xxx
ARTICLE IN PRESS
Please cite this article as: Pesta, B.J., et al., Toward an index of well-being for the fifty U.S. states, Intelligence (2009),
doi:10.1016/j.intell.2009.09.006
Schmidt, F., & Hunter, J. (1998). The validity and utility of selection methods
in personnel psycholo gy: Practical and th eoretical implications of
85 years of research findings. Psychological Bulletin,124, 262−275.
Steel, P., Schmidt, J., & Shultz, J. (2008). Refining the relationship between
personality and subjective well-being. Psychological Bulletin,134,
138−161.
U.S. Census Bureau. (2008a). Dynamics of economic well-being. Retrieved
November 14, 2008, from the World Wide Web: http://www.census.
gov/population/www/socdemo/wellbeing.html
U.S. Census Bu reau. (2008b). The 2008 statistical a bstract. Retrieved
November 1, 2008, from the World Wide Web:: http://www.census.
gov/compendia/statab/
United Health Foundation. (2008). America's health rankings: A call to action
for people and their communities. Retrieved November 1, 2008, from the
World Wide Web: http://www.unitedhealthfoundation.org/
media2007/survey.asp
Warr, P. (1987). Work, unemployment and mental health. Oxford: Oxford
Press.
Warr, P. (2007). Work, happiness, and unhappiness. Erlbaum: Mahwah, NJ.
Yik, M., & Russell, J. (2001). Predicting the Big Two of affect from the Big Five
of personality. Journal of Research in Personality,35, 247−277.
9B.J. Pesta et al. / Intelligence xxx (2009) xxx–xxx
ARTICLE IN PRESS
Please cite this article as: Pesta, B.J., et al., Toward an index of well-being for the fifty U.S. states, Intelligence (2009),
doi:10.1016/j.intell.2009.09.006
