Perceived risk and expected benefits impact social class differences

The University of Toledo
The University of Toledo Digital Repository
Theses and Dissertations
2015
Perceived risk and expected benefits impact social
class differences in health risk behavior
Heather M. Haught
University of Toledo
Follow this and additional works at: http://utdr.utoledo.edu/theses-dissertations
Recommended Citation
Haught, Heather M., "Perceived risk and expected benefits impact social class differences in health risk behavior" (2015). Theses and
Dissertations. 1836.
http://utdr.utoledo.edu/theses-dissertations/1836
This Dissertation is brought to you for free and open access by The University of Toledo Digital Repository. It has been accepted for inclusion in Theses
and Dissertations by an authorized administrator of The University of Toledo Digital Repository. For more information, please see the repository's
About page.
A Dissertation
entitled
Perceived Risk and Expected Benefits Impact Social Class Differences in Health Risk
Behavior
by
Heather M. Haught
Submitted to the Graduate Faculty as partial fulfillment of the requirements for the
Doctor of Philosophy Degree in Experimental Psychology
_________________________________________
Jason Rose, Ph.D., Committee Chair
_________________________________________
Andrew Geers, Ph.D., Committee Member
_________________________________________
John Jasper, Ph.D., Committee Member
_________________________________________
John Elhai, Ph.D., Committee Member
_______________________________________
Deborah Boardly, Ph.D., Committee Member
_________________________________________
Patricia R. Komuniecki, PhD, Dean
College of Graduate Studies
The University of Toledo
May 2015
Copyright 2015, Heather M. Haught
This document is copyrighted material. Under copyright law, no parts of this document
may be reproduced without the expressed permission of the author.
An Abstract of
Perceived Risk and Expected Benefits Impact Social Class Differences in Health Risk
Behavior
by
Heather M. Haught
Submitted to the Graduate Faculty as partial fulfillment of the requirements for the
Doctor of Philosophy Degree in
Experimental Psychology
The University of Toledo
May 2015
People in lower social classes die younger and suffer a greater burden of disease
than those in upper social classes. This pattern, referred to as the social gradient, is partly
attributable to the fact that people in lower social classes engage in more health risk
behavior. The current studies are the first to examine whether perceived risk and expected
benefits mediate the relationship between social class and health risk behavior and if the
nature of this relationship differs depending upon the category of health risk behavior –
those that derive risk from action vs. inaction. In Study 1, adults recruited from
Amazon’s Mechanical Turk reported how frequently they engage in a variety of health
risk behaviors, indicated their perceived risk and expected benefits for each health
behavior, and provided socio-demographic information (e.g., social class indicators, sex,
and race). Results revealed that people in lower social classes perceived health risk
behaviors to be more risky and have fewer benefits than people in upper social classes.
Additionally, perceived risk was a stronger predictor of inactive health risk behaviors
whereas expected benefits was a stronger predictor of active health risk behaviors. In
Study 2, we manipulated participants’ perceived social status by shifting the reference
iii
point that participants used to make subjective social status judgments and examined
whether this manipulation affected perceived risk and expected benefits in a manner
similar to that expected in Study 1. Results did not replicate those obtained in Study 1.
These findings advance our understanding of the pathways by which social class affects
health risk behavior and inform strategies aimed at reducing health risk behavior among
lower social classes.
iv
Table of Contents
Abstract
iii
Table of Contents
v
List of Tables
viii
List of Figures
ix
I. Introduction
1
A. Perceived Risk & Expected Benefits as Determinants of Risk Behavior
2
B. Measures of Social Class
5
C. Social Class Differences in Perceived Risk & Expected Benefits
6
D. Overview of Study 1
8
II. Study 1 Method
11
A. Participants & Overview of Procedure
11
B. Sociodemographic Questions
11
C. Health Risk Behavior
13
a. Active Risk
13
b. Inactive Risk
13
D. Perceived Risk & Expected Benefits
III. Study 1 Results
15
16
A. Social Class Differences in Perceived Risk & Expected Benefits
16
B. Mediation & Active vs. Inactive Risk Behaviors
18
C. Model Specification
19
a. Hypothesized Model
19
b. Alternative Model
20
v
D. Parameter Estimates & Fit Indices
21
E. Model Modification
a. Hypothesized Model
22
b. Alternative Model
25
F. Model Performance
26
G. Summary of Findings
31
IV. Overview of Study 2
32
V. Study 2 Method
34
A. Participants & Overview of Procedure
34
B. SSS Manipulation
34
C. Perceived Risk & Expected Benefits
35
D. Socio-demographic Questions
35
VI. Study 2 Results
36
A. Manipulation Check
36
B. Primary Analyses
36
C. Exploratory Analyses
37
VII. Discussion
40
A. Social Class Differences in Perceived Risk & Expected Benefits
40
B. Perceived Risk & Expected Benefits as Mediators
43
C. Perceived Risk & Expected Benefits Predict Active & Inactive
Risk Behaviors
44
D. Sex Differences in Risk Behavior
45
E. Limitations
46
vi
F. Basic & Applied Contributions
48
References
50
Appendices
A. SSS Scale
58
B. Health Behavior Questionnaire
59
C. Perceived Risk & Expected Benefits Questionnaire
61
vii
List of Tables
Table 2-1 Sample Characteristics. ...................................................................................12
Table 2-2 Correlations Among Sociodemographic Variables. .......................................12
Table 2-3 Correlations Among Health Risk Behaviors. ..................................................14
Table 2-4 Descriptive Statistics for Health Risk Behavior, Perceived Risk, & Expected
Benefits. ……………………………………………………………………. 14
Table 2-5 Correlations Between Health Risk Behavior, Perceived Risk, & Expected
Benefits………………………………………………………………………15
Table 3-1 Descriptive Statistics for the Class X Judgment Interaction by Indicator (Study
1)……………………………………………………………………………. 17
Table 3-2 Descriptive Statistics for the Judgment X Behavior Interaction (Study 1)…..18
Table 3-3 Fit indices for Social Class Latent Factor (Study 1)…………………………22
Table 3-4 Correlation Between Perceived Risk, Expected Benefits, & Sociodemographic
Variables……………………………………………………………………..23
Table 3-5 Fit indices for the (Modified) Hypothesized & Alternative Models………..27
Table 3-6 Indirect Effects for the (Modified) Hypothesized Model……………………29
Table 3-7 Descriptive Statistics for Active & Inactive Health Risk Behavior by
Indicator……………………………………………………………………..30
Table 6-1 Descriptive Statistics for the Judgment X Behavior Interaction (Study 2)…37
viii
List of Figures
Figure 1-1 Hypothesized Model. ........................................................................................9
Figure 1-2 Alternative Model. ..........................................................................................10
Figure 3-1 Social Class X Judgment Interaction (Study 1). .............................................17
Figure 3-2 Behavior X Judgment Interaction (Study 1)…………………………………18
Figure 3-3 Modified Hypothesized Model……………………………………………....22
Figure 3-4 Loading Plot for Health Risk Behaviors (Study 1)…………………………..24
Figure 3-5 Modified Alternative Model…………………………………………………26
Figure 3-6 Path Estimates for the Modified Hypothesized Model………………………28
Figure 3-7 Path Estimates for the Modified Alternative Model…………………………28
Figure 6-1 Behavior X Judgment Interaction (Study 2)…………………………………37
Figure 6-2 Class X Condition X Judgment Interaction (Study 2)……………………….39
ix
Chapter One
Introduction
The link between social class and health is well-established. People in
lower social classes die younger and are in poorer health than those in upper social
classes (Adler & Rehkopf, 2008; Isaacs, 2004). This pattern, referred to as the social
gradient, has been documented for a range of health outcomes, but is strongest for
cardiovascular disease, diabetes, and chronic respiratory diseases (Adler & Snibbe,
2003). Evidence indicates that class-based health disparities are widening in the United
States and Europe (Elo, 2009; Krostad, Kunst, & Westin, 2002; Marmot, Kogevinas, &
Elston, 1987; Preston & Elo, 1995) and persist even in countries with universal
healthcare (Cohen, Doyle, & Baum, 2006). Consequently, the reduction of these
disparities has become a primary goal of public health practice and research (WHO,
2005).
Although complex, the link between social class and health is partly explained by
differences in health risk behavior. Compared to people in upper social classes, those in
lower social classes are more likely to smoke and drink heavily and are less likely to
exercise, maintain a healthy diet, and seek out or adhere to medical treatments (Barr,
Somers, Speizer, & Camargo, 2002; Darmon & Drewnowski, 2008; Goldman & Smith,
2002; Haught & Rose, 2014; Lynch, Kaplan, & Salonen, 1997). Research in this area has
predominately focused on understanding how characteristics of lower social class
environments shape health risk behavior. For example, studies have shown that people
who live in neighborhoods with high crime rates, unkempt sidewalks, and few
supermarkets are less likely to exercise regularly or maintain a healthy diet than those
1
who live in neighborhoods with low crime rates, quality sidewalks, and many
supermarkets (Caraher, Dixon, Lang, & Carr-Hill, 1998; Kwarteng, Schulz, Mentz, Zenk,
& Opperman, 2013; Moore, Glick, Romanowski, & Quinley, 1996; Powell, Slater, &
Chaloupka, 2004; Virtanen, Kivimäki, Kouvonen, Elovainio, Linna, et al., 2007).
However, environmental factors cannot fully account for social class differences
in health risk behavior (Adler, Boyce, Chesney, Cohen, Folkman, Kahn, & Syme, 1994).
Consequently, researchers have suggested that psychological factors may also be
important for understanding how social class affects health risk behavior (Adler &
Snibbe, 2003; Stephens, Markus, & Fryberg, 2012). Consistent with this idea, research
has demonstrated that how people perceive risk and associated benefits directly impacts
health risk behavior. For example, Haefner and Kirscht (1970) showed that manipulating
people’s perceived risk of disease (i.e., cancer, heart disease, or tuberculosis) and
anticipated benefits of treatment resulted in more frequent doctor visits over an eight
month period compared to controls. However, we are unaware of any study that has
examined whether perceived risk and expected benefits contribute to the link between
social class and health risk behavior. The current studies seek to address this issue.
Specifically, we examine whether people’s perceptions of risk and expected benefits
mediate the relationship between social class and health risk behavior and if the nature of
this effect differs depending upon characteristics of the behavior (active vs. inactive risk).
Perceived Risk & Expected Benefits as Determinants of Risk Behavior
Risk behavior, in health and other domains, can be conceptualized as a trade-off
between the perceived risk and expected benefits of engaging in a behavior (Weber,
Blais, & Betz, 2002; Weber & Johnson, 2008; Weber & Milliman, 1997). On a night out,
2
for example, a person may need to decide whether or not to engage in unprotected sex. If
he (she) perceives the risk of negative outcomes (e.g., sexually transmitted infection,
unwanted pregnancy) to be relatively low and expects that by doing so he (she) will gain
desired outcomes (e.g., a passionate experience, social connection), he (she) may choose
to engage in unprotected sex. This conceptualization suggests that risk behavior can be
motivated by more than simply a person’s propensity for risk (Gullone & Moore, 2000;
Zuckerman & Kuhlman 2000).
Research on how perceived risk and expected benefits affect risk behavior has
been conducted using the Domain-Specific Risk-Taking (DOSPERT) scale (Weber,
Blais, & Betz, 2002). The DOSPERT measures risk behavior, perceived risk, and
expected benefits in five domains: finance (e.g., betting at a race track), social interaction
(e.g., opposing a friend’s viewpoint), health and safety (e.g., not wearing a seatbelt),
ethics (e.g., cheating on an exam), and recreation (e.g., skydiving). Findings indicate that,
although related, perceived risk and expected benefits are independent predictors of risk
behavior (Foster, Shenesey, & Goff, 2009; Harris, Jenkins, & Glaser, 2006) and that
expected benefits may be a stronger predictor of risk behavior than perceived risk
(Hanoch, Johnson, & Wilke, 2006). Additionally, Weber and colleagues (2002) found
that perceived risk and expected benefits were unique predictors of risk behavior in each
DOSPERT domain and that expected benefits was a stronger predictor in every domain
except finance.
Importantly, however, the DOSPERT fails to distinguish between health risk
behaviors that derive risk from action vs. inaction. Active health risk behaviors, like
substance abuse and reckless driving, involve engaging in an action that threatens a
3
person’s current health status whereas inactive health risk behaviors, like unprotected sex
and physical inactivity, involve avoiding an action that protects a person’s current health
status. This distinction between action and inaction is prominent throughout the
psychological literature and has important implications for risk behavior. For instance,
Reinforcement Sensitivity Theory (Gray, 1970), one of the most widely accepted
biopsychological theories to date, argues that behavior is motivated by two systems, the
Behavioral Inhibition System (BIS) and the Behavioral Activation System (BAS). The
BIS is sensitive to punishment and motivates inaction or avoidance whereas the BAS is
sensitive to reward and motivates action. Consistent with this idea, Verbruggen, Adams,
and Chambers (2012) demonstrated that people who are forced to inhibit a manual
response are subsequently less likely to take financial risks than people who are not
forced to inhibit a manual response. In addition, Keinan & Bereby-Meyer (2012) recently
demonstrated that active and inactive risk behavior are differentially associated with
related constructs like sensation-seeking, impulsivity, and procrastination.
The DOSPERT is primarily a measure of active risk-taking, suggesting that
expected benefits is a stronger predictor of active health risk behaviors than perceived
risk. Consistent with this idea, Teese and Bradley (2008) showed that expected benefits is
a stronger predictor of reckless driving among adolescents and young adults than
perceived risk. Additionally, people who are highly sensitive to reward use drugs and
alcohol more frequently and in greater quantities than those who are less sensitive to
reward (Genovese & Wallace, 2007; Kynazev, 2010; O’Conner & Colder, 2005; Rose &
Mossler, 2013). However, expected benefits does not seem to be a stronger predictor of
inactive health risk behaviors than perceived risk. For instance, people with lower levels
4
of perceived risk are less likely to be vaccinated, wear a seatbelt, exercise or obtain
cancer screenings than people with higher levels of perceived risk (Brewer, Weinstein,
Cuite, & Herrington, 2004; Katapodi, Lee, Facione, & Dodd, 2004; Ouimet, Morton,
Noelcke, Williams, Leaf, Preusser, & Hartos, 2008; Stephan, Boiche, Trouilloud,
Deroche, & Sarrazin, 2011). Furthermore, Macdonald, Fong, Zanna, and Martineau
(2000) demonstrated that manipulating the salience of risks associated with unprotected
sex altered safe sex practices among both sober and intoxicated individuals.
In light of these findings, we argue that perceived risk and expected benefits may
be differentially related to active vs. inactive health risk behaviors such that perceived
risk is a stronger predictor of inactive health risk behaviors whereas expected benefits is a
stronger predictor of active health risk behaviors. To our knowledge, no study has
previously addressed this issue; therefore, clarifying how perceived risk and expected
benefits impact active vs. inactive health risk behaviors is one important goal of the
current studies.
Measures of Social Class
Another goal of the current studies is to examine whether perceived risk and
expected benefits mediate the relationship between social class and health risk behavior.
In the United States, social class is typically measured in terms of income or education,
with higher income and more education denoting higher social class (Elo & Preston,
1996; Herd, Goesling, & House, 2007). These variables, referred to as objective social
status (OSS) indicators, govern people’s access to specific sets of resources (Haught &
Rose, 2014) and are among the most consistent predictors of health among adults and
children (Chen, Matthews, & Boyce, 2002; Minkler, Fuller-Thomson, & Guralnik, 2006).
5
Recently, however, research has demonstrated that a person’s perceived social standing,
referred to as subjective social status (SSS), is only moderately correlated with OSS
indicators (Adler, Epel, Castellazzo, & Ickovics, 2000; Goodman, Adler, Kawachi,
Frazier, Huang, & Colditz, 2001) and exerts an independent effect on health (SinghManoux, Adler, & Marmot, 2003). Consistent with this idea, researchers have suggested
that SSS may serve as a global measure of social class that accounts for various OSS
indicators simultaneously (Singh-Manoux, Marmot, & Adler, 2005; Kraus, Piff,
Mendoza-Denton, Rheinschmidt, & Keltner, 2012). Indeed, people report thinking about
a number of variables when making SSS judgments including their income, education,
occupation, parents’ income during childhood, health, social and ethical responsibility,
and future financial situation (Snibbe, Stewart, & Adler, unpublished; Singh-Manoux et
al., 2005). In light of these findings, we decided to measure social class using both OSS
and SSS indicators.
Social Class Differences in Perceived Risk & Expected Benefits
Previous research has demonstrated that sociocultural differences in risk behavior
are largely accounted for by perceived risk and expected benefits (Weber & Morris,
2010). For instance, Weber and colleagues (2002) found that sex differences in risk
behavior were eliminated after controlling for perceived risk and expected benefits in
four of the five DOSPERT domains (all except the social interaction domain). In
addition, Blais and Weber (2006) showed that English participants engaged in more risk
behavior than French participants across the five DOSPERT domains and that this
difference was significantly reduced after controlling for perceived risk.
6
These prior findings have primarily been explained in terms of environmental
factors (Figner & Weber, 2011; Weber & Morris, 2010). In the financial domain, for
example, studies have shown that differences in perceived risk between American and
Chinese participants are evident in each culture’s proverbs and are mediated by
environmental factors like the number of family and friends that would lend the
participant money during financial hardship (i.e., the cushion hypothesis; Hsee & Weber,
1998; Weber & Hsee, 1998; Weber, Hsee, & Sokolowska, 1999). Importantly, social
class is tied to the social and physical environments that people inhabit in their daily
lives, governing where people live, the jobs they hold, and the kinds of interactions they
have with others (Kawachi & Berkman, 2003; Fiske & Markus, 2012). In particular,
people in lower social classes tend to live in poor neighborhoods with high crime rates,
attend low quality schools, and work in hazardous conditions with low autonomy and
control. People in upper social classes, by contrast, tend to live in rich neighborhoods
with quality schools, have more non-familial social relationships, possess greater
geographic mobility, and hold more autonomy and control at work. Despite marked
differences in the environments inhabited by upper vs. lower social classes, we are
unaware of any study that has examined social class differences in perceived risk and
expected benefits. The current studies are designed to address this issue and determine
whether social class differences in perceived risk and expected benefits give rise to social
class differences in health risk behavior. We anticipate that 1) people in lower social
classes will perceive active and inactive health risk behaviors to be less risky and have
more benefits than people in upper social classes; 2) perceived risk and expected benefits
will mediate the relationship between social class and health risk behavior; and 3)
7
regardless of social class, people will weigh expected benefits more heavily for active
health risk behaviors, but weigh perceived risk more heavily for inactive health risk
behaviors.
Overview of Study 1
The current studies examine whether perceived risk and expected benefits mediate
the relationship between social class and health risk behavior and if the relative
contribution of perceived risk vs. expected benefits differs depending upon the category
of health risk behavior (active vs. inactive). Study 1was conducted with a sample of U.S.
adults in order to obtain a wider range of social class than in traditional undergraduate
samples. Participants reported how frequently they engage in a variety of health risk
behaviors, indicated their perceived risk and expected benefits for each behavior, and
provided socio-demographic information, including standard social class indicators, sex,
and race. We expected that people in lower social classes would perceive health risk
behaviors to be less risky and have more benefits than people in upper social classes and
that perceived risk would be a stronger predictor of inactive health risk behavior whereas
expected benefits would be a stronger predictor of active health risk behaviors—among
both upper and lower social class individuals. These hypotheses were tested
simultaneously using structural equation modeling. The hypothesized model is depicted
in Figure 1-1. Sex and race were included in the model because risk-taking differs
systematically by sex and because women and minorities are more concentrated in lower
social classes (Braveman et al., 2005; Byrne, Miller, & Schafer, 1999).
8
Figure 1-1. Hypothesized Model
Additionally, we compared the hypothesized model to an alternative model, which does
not differentiate between active and inactive health risk behaviors. This model is depicted
in Figure 1-2. In the case that the alternative model is superior, we expect that 1) people
in lower social classes perceive health risk behaviors to be less risky and have more
benefits than people in upper social classes; 2) perceived risk and expected benefits
mediate the relationship between social class and health risk behavior; and 3) regardless
of social class and type of behavior (action or inaction), expected benefits is a stronger
predictor of health risk behavior than perceived risk.
9
Figure 1-2. Alternative Model
10
Chapter Two
Study 1 Method
Participants & Overview of Procedure
Participants were 339 adults (106 males) recruited from Amazon’s Mechanical
Turk (MTurk)—a crowdsourcing online marketplace. The MTurk pool was specified to
include adults who reside in the United States and speak English. Research has shown
that data obtained from MTurk are at least as reliable as traditional samples and
significantly more diverse by age, race, and socioeconomic status than college student
samples (Buhrmester, Kwang, & Gosling, 2011). In the current sample, participants
ranged in age from 18 to 65 (M=33.25, SD=11.38), more than a quarter were minorities,
and had annual household incomes ranging from less than $20,000 to greater than
$75,000. Sample characteristics are summarized in Table 2-1. Participants completed
each of the four questionnaires discussed below (randomly ordered) and earned $0.10 in
exchange for their participation. The study took about 10-15 minutes to complete.
Sociodemographic Questions
Participants reported their sex (0 = male, 1 = female), and race (recoded as 0 =
white, 1 = non-white). Social class was measured using standard indicators including
annual household income (1 = <$20,000; 2 = $20,000-$34,999; 3 = $35,000-$49,999; 4 =
$50,000-$74,999; 5 = ≥$75,000), highest level of education (1 = less than high school, 2
= high school graduate, 3 = some college, 4 = bachelor’s degree, 5 = postbaccalaureate), and subjective social status (SSS; Goodman, Adler, Kawachi, Frazier,
Huang, & Colditz, 2001). SSS was assessed using the MacArthur Scale, which consists
of a ladder with 10 rungs, anchored at the top with “those who are best off in society” and
11
Table 2-1. Sample Characteristics
%
Sex
Male
Female
Race
White
Non-White
Education
Less than High School
High School/GED
Some College
Bachelor’s Degree
Post-Bach Degree
Household Income
Less than $20,000
$20,000-$34,999
$35,000-$49,999
$50,000-$74,999
$75,000 or greater
31.5
68.5
74.6
25.4
1.8
11.5
39.5
35.1
12.1
24.3
23.4
20.1
17.5
14.8
at the bottom with “those who are worst of in society” (Goodman et al., 2001; Appendix
A). Participants indicated their position relative to these anchors by placing an X next to
one of the rungs. In the current sample, the correlations among income, education, and
SSS were somewhat weaker (r’s < .37; see Table 2-2) than those typically reported in the
literature (r ≈ .50; Adler, Epel, Castellazzo, & Ickovics, 2000; Haught & Rose, 2014;
Oakes & Rossi, 2003).
Table 2-2. Correlations Among Sociodemographic Variables
Income
Edu
SSS
Sex
Race
Income 1
Edu
.24***
1
SSS
.37***
.28***
1
Sex
.07
.04
.13*
1
Race
.02
-.03
-.08
-.01
†p<.10 *p<.05 **p<.01 ***p<.001
12
1
Health Risk Behavior
To measure engagement in active vs. inactive health risks, we compiled items
from two sources, the DOSPERT scale (Weber et al., 2002) and the Health Information
National Trends Survey (2012). The items associated with each category of health risk
behavior are discussed below. The full scale is also provided in Appendix B.
Correlations, means and standard deviations among each of the health risk behaviors are
provided in Tables 2-3 and 2-4, respectively.
Active Risk. Participants reported the average number of alcoholic beverages
they have on days when they drink (1=0 drinks, 2=1-2 drinks, 3=3-4drinks, 4=5-6 drinks,
5=7-8 drinks, 6=9-10 drinks, 7=11or more drinks), the number of sexual partners that
they have had in the past month (1=0 partners, 2=1 partner, 3=2 partners, 4=3 partners,
5=4 partners, 6=5 partners, 7=6 partners, 8=7 partners or more), the average number of
cigarettes that they smoke per day (1=0 cigarettes, 2=1-2 cigarettes, 3=3-4 cigarettes,
4=5-6 cigarettes, 5=7-8 cigarettes, 6=9-10 cigarettes, 7=11 or more cigarettes), the
number of times that they used an illegal substance during the past month (1=0 times,
2=1-2 times, 3=3-4 times, 4=5-6 times, 5=7-8 times, 6=9-10 times, 7=11or more times),
how frequently they drive more than 10 miles per hour over the speed limit (1 = never, 7
= always), and the average amount of soda they consume per day (1 = none, 2 = 12
ounces or less, 3 = 13-24 ounces, 4 = 25-36 ounces, 5 = 37-48 ounces, 6 = more than 48
ounces).
Inactive Risk. Participants reported the number of days per week that they
exercise for at least 30 minutes (1 = 0 days per week, 2 = 1 day per week, 3 = 2 days per
week, 4 = 3 days per week, 5 = 4 days per week, 6 = 5 days per week, 7 = 6 days per
13
week, 8 = 7 days per week), the average number of fruits and vegetables they consume
per day (1 = none, 2 = less than ½ cup, 3 = ½ cup to 1 cup, 4 = 1-2 cups, 5 = 2-3cups, 6 =
3-4cups, 7 = more than 4 cups), how frequently they wear their seatbelt (1 = never, 7 =
always), the number of times that they had unprotected sex during the past month (1=0
times, 2=1-2 times, 3=3-4 times, 4=5-6 times, 5=7-8 times, 6=9-10 times, 7=11or more
14
Table 2-3. Correlations Among Health Risk Behaviors
Drinking
1
.19**
.13*
.14**
.10†
-.06
.04
-.07
.02
.13*
.02
†p<.10 *p<.05 **p<.01
Drinking
Partners
Smoking
Drugs
Driving
Soda
Exercise
Fruit/Veg
Seatbelt
Unprotect
Sunscreen
Partners
Smoking
Drugs
Driving
Soda
Exercise
Fruit/Veg
Seatbelt
Unprotect
Sunscreen
1
.02
.09
.16**
.03
.01
-.05
.10†
.41**
-.03
1
.18**
-.09†
.16**
.01
.03
.09†
.12*
.04
1
.07
-.04
.05
-.07
.09
.20**
.003
1
.04
-.03
-.04
.10†
.05
.01
1
.18**
.12*
.19**
.11*
.11*
1
.20**
.07
-.04
.08
1
.03
-.04
.08
1
.07
.07
1
.07
1
Table 2-4. Descriptive Statistics for Health Risk Behavior, Perceived Risk, & Expected Benefits
Behavior
Perceived Risk
Expected Benefits
M
SD
M
SD
M
SD
Heavy Drinking
2.13
1.14
5.31
1.67
1.96
1.48
Number of Sexual Partners
1.73
.64
5.56
1.62
2.31
1.64
Smoking
2.11
2.04
6.17
1.28
1.64
1.30
Illegal Substance Use
1.76
1.80
5.97
1.44
2.28
1.79
Reckless Driving
2.73
1.83
4.72
1.56
2.68
1.62
Soda Consumption
2.07
1.20
4.03
1.66
1.85
1.21
Not Exercising
5.33
1.96
4.12
1.61
2.19
1.74
Not Eating Fruits/Veggies
4.34
1.36
3.72
1.48
2.06
1.45
Not Wearing Seatbelt
1.61
1.47
5.99
1.42
1.58
1.29
Unprotected Sex
2.21
1.93
5.84
1.59
2.25
1.66
Not Wearing Sunscreen
5.06
2.11
4.59
1.64
2.09
1.46
14
times), and how frequently they use sunscreen when going outside for more than 1 hour
(1 = never, 7 = always). Frequency of exercise, fruit and vegetable consumption, seatbelt
and sunscreen use were reverse scored in all analyses to aid interpretation, such that
higher numbers represent greater risk-taking.
Perceived Risk & Expected Benefits
Perceived risk and expected benefits were measured in the same way as the
DOSPERT scale. Specifically, participants rated the perceived risk and expected benefit
of engaging in each of the health risk behaviors presented above using 7-point likert
scales (1 = not at all risky/ no benefits at all, 7 = very risky/ many benefits; Appendix C).
The reliabilities of the perceived risk and expected benefits scales for active (α =.79, α
=.72) and inactive (α =.81, α =.74) health risk behaviors were adequate. Correlations
among perceived risk, expected benefits, and their respective health risk behavior are
provided in Table 2-5.
Table 2-5. Correlations Between Health Risk Behavior, Perceived Risk, & Expected
Benefits
Expected
Perceived Risk w/
Perceived Risk
Benefits
Expected Benefits
Heavy Drinking
-.30***
.35***
-.36***
Number of Sexual
-.08***
.10†**
-.48***
Partners
Smoking
-.08***
.20***
-.28***
Illegal Substance Use
-.36***
.43***
-.54***
Reckless Driving
-.34***
.45***
-.39***
Soda Consumption
-.23***
.24***
-.19***
Not Exercising
-.11***
-.01**
-.10†**
Not Eating
-.09†**
.03**
.04**
Fruits/Veggies
Not Wearing Seatbelt
-.35***
.32***
-.27***
Unprotected Sex
-.23***
.25***
-.44***
Not Wearing Sunscreen
-.37***
†p<.10 *p<.05 **p<.01 ***p<.001
.13***
15
-.31***
Chapter Three
Study 1 Results
Social Class Differences in Perceived Risk & Expected Benefits
A 2 (Social Class: lower social class, upper social class) X 2 (Behavior: active,
inactive) X 2 (Judgment: perceived risk, expected benefits) mixed model analysis of
variance (mixed ANOVA) was used to establish that there are social class differences in
perceived risk and expected benefits. The first factor is between-subjects and the second
and third factors are within-subjects. Separate analyses were run for each social class
indicator, with an income of ≥$50,000, 1 or more years of college, and SSS ≥ 5
signifying upper social class.
The pattern of results was similar for all three indicators. The main effects of
Judgment (all F’s >1223.00, p’s <.001, ηp2’s =.79) and Behavior (all F’s >72.74, p’s
<.002, ηp2’s =.18) were significant; however, there was no main effect of social class (all
F’s <1.25, p’s >.26, ηp2’s <.004). Overall, participants rated the health risk behaviors as
being more risky than beneficial and indicated that active health risk behaviors are more
risky and more beneficial than inactive health risk behaviors.
As hypothesized, the main effects were qualified by a significant Social Class X
Judgment interaction (all F’s >3.81, p’s <.05, ηp2’s >.01; except SSS F(1,333)=1.98,
p=.16, ηp2=.01); however, the nature of this interaction was not as predicted. Lower social
class participants rated the health risk behaviors as more risky and less beneficial than
those in upper social classes (Figure 3-1). Descriptive statistics for this interaction are
presented in Table 3-1. The Behavior X Judgment interaction was also significant (all F’s
>19.33, p’s <.001, ηp2’s =.06). Participants rated active and inactive health risk behaviors
16
as more risky than beneficial; however, active health risk behaviors were rated as
relatively more risky than inactive health risk behaviors (Figure 3-2).
Table 3-1. Descriptive Statistics for the Class X Judgment Interaction by Indicator
(Study 1)
Education
Risk
Income
Benefits
Risk
SSS
Benefits
Risk
Benefits
M
SD
M
SD
M
SD
M
SD
M
SD
M
SD
Lower
5.00
1.02
1.95
.88
4.94
.96
2.08
.92
4.88
.93
2.02
.87
Upper
4.67
.95
2.36
1.08
4.76
1.03
2.19
1.07
4.81
1.06
2.26
1.10
6
5
Upper Class
4
Lower Class
3
2
1
Perceived Risk
Expected Benefits
Figure 3-1. Social Class X Judgment Interaction (Study 1)
Descriptive statistics for this interaction are presented in Table 3-2. The Social Class X
Behavior (all F’s <1.44, p’s >.23, ηp2’s <.001) and Social Class X Behavior X Judgment
(all F’s <.56, p’s >.45, ηp2’s <.002) interactions were not significant.
17
6
5
Active
4
Inactive
3
2
1
Perceived Risk
Expected Benefits
Figure 3-2. Behavior X Judgment Interaction (Study 1)
Table 3-2. Descriptive Statistics for the Judgment X Behavior Interaction (Study 1)
Perceived Risk
M
5.54
4.15
Active Risk Behavior
Inactive Risk Behavior
SD
1.12
1.25
Expected Benefits
M
2.17
2.11
SD
1.18
1.20
Mediation & Active vs. Inactive Risk Behaviors
Structural equation modeling (SEM) was used to evaluate whether perceived risk
and expected benefits mediated the relationship between social class and health risk
behavior and if perceived risk and expected benefits differentially predicted active vs.
inactive health risk behavior. SEM is a particularly effective technique when the
structural relations among constructs can be specified and the number of constructs is
limited. Consistent with this idea, the hypothesized model examines whether perceived
risk and expected benefits mediate the relationship between social class and different
categories of health risk behavior, namely those that derive risk from action vs. inaction.
In contrast to more traditional techniques (e.g., regression and multivariate analysis of
variance), SEM allows the researcher to specify and simultaneously test direct and
18
indirect effects of variables on one another (Kline, 2010) and to understand the relations
among sociodemographic variables (i.e., sex, race) rather than simply controlling for
them. Additionally, SEM allows researchers to compare alternative models that are
generated a priori or that are devised post-hoc based upon limitations of the hypothesized
model (Kline, 2010).
Model Specification
Hypothesized Model. The hypothesized model is depicted in Figure 1-1 and
includes three exogenous variables: social class, sex, and race. Social class is a
continuous latent variable measured by annual household income, educational attainment,
and SSS. Sex and race are categorical observed variables. In addition, there are six
endogenous variables: active and inactive health risk behaviors, perceived risk for active
and inactive health risk behaviors, and expected benefits for active and inactive health
risk behaviors. Inactive health risk behavior is a continuous latent variable measured by
six indicators. Active health risk behavior is a continuous latent variable measured by
five indicators. Perceived risk and expected benefits are continuous observed variables,
which are calculated by averaging across the perceived risk and expected benefits ratings
of health risk behaviors in each category (active vs. inactive).
As can be seen in Figure 1-1, perceived risk and expected benefits partially
mediate the relationship between social class and each category of health risk behavior.
Both perceived risk and expected benefits are hypothesized to predict active and inactive
health risk behavior. However, perceived risk is expected to be a stronger predictor of
inactive health risk behavior whereas expected benefits is expected to be a stronger
predictor of active health risk behavior. In Figure 1-1, this nuance is depicted using solid
19
vs. dotted lines, with solid lines denoting stronger paths and dotted lines denoting weaker
paths. Sex and race are included in the model because risk-taking is known to differ
systematically by sex (e.g., Byrne et al., 1999) and because women and minorities are
more concentrated in lower social classes (Braveman et al., 2005). Thus, unanalyzed
associations are posited among social class, race, and sex. Consistent with research by
Weber and colleagues (2002), perceived risk and expected benefits mediate the
relationship between sex and each category of health risk behavior. Prior research has not
demonstrated a link between race, perceived risk, and expected benefits; therefore, we
assumed that except for the association with social class, race is unrelated to other
variables in the model.
Alternative Model. In addition to the hypothesized model, we offer an alternative
model derived from research using the DOSPERT scale which suggests that regardless of
category, expected benefits is a stronger predictor of health risk behavior than perceived
risk (Weber et al., 2002; Hanoch et al., 2006). The alternative model is depicted in Figure
1-2 and includes the same three exogenous variables as the hypothesized model.
However, the endogenous variables – perceived risk, expected benefits, and health risk
behavior – differ from the hypothesized model in important ways. In the hypothesized
model, health risk behavior, perceived risk, and expected benefits are split into two
categories based upon whether the behavior derives risk from action or inaction. By
contrast, the alternative model conceptualizes health risk behavior as a single latent
variable measured by eleven indicators. Perceived risk and expected benefits are
continuous observed variables calculated by averaging across the perceived risk and
expected benefits ratings of the eleven health risk behaviors.
20
As can be seen in Figure 1-2, perceived risk and expected benefits mediate the
relationship between social class and health risk behavior. Expected benefits is expected
to be a stronger predictor than perceived risk. In the figure, this pattern is depicted using
solid vs. dotted lines, with a solid line denoting a stronger path and a dotted denoting a
weaker path. As in the hypothesized model, sex and race are included due to their
associations with risk-taking and social class. For the reasons presented above, perceived
risk and expected benefits mediate the relationship between sex and health risk behavior
and except for the association with social class, race is unrelated to other variables in the
model.
Parameter Estimates & Fit Indices
Robust Maximum Likelihood Estimation (MLR) was used to estimate free
parameters in both models due to non-normality. In order to evaluate how well the
hypothesized and alternative models fit the data the following fit indices were used: (1)
chi-square goodness-of-fit index (e.g., Brown, 2006; Hu & Bentler, 1998); (2)
comparative fit index (CFI; Bentler, 1990); (3) Tucker-Lewis Index (TLI; Tucker &
Lewis, 1973); and (4) root- mean-square error of approximation (RMSEA; Steiger,
1990). Published rules for significance levels of fit indices were followed (Bentler, 1990;
Browne & Cudeck, 1993; Hu & Bentler, 1998; Yu & Muthén, 2002). Specifically, the
chi-square goodness-of-fit index should be non-significant at α=.05, minimum TLIs and
CFIs of .90 are required for model acceptance, and values of .95 or greater are accepted
as an indication of good model fit. Additionally, RMSEAs of less than .06 were accepted
as indicator of a good-fitting model. Indirect pathways between social class and each
category of health risk behavior were tested using the Mplus “Model Indirect” command.
21
Model Modification
Hypothesized Model. Several modifications were made to the hypothesized
model in order to obtain a satisfactory fit. The modified model is depicted in Figure 3-3.
Figure 3-3. Modified Hypothesized Model
The most significant modifications concerned the three latent variables of social class,
active health risk behavior, and inactive health risk behavior. Latent variables that are
misspecified can lead to poor overall model fit and unreliable parameter estimates. It is
necessary, therefore, to build each measurement model (latent variable) individually
before integrating them into the larger structural model. As evidenced by the fit indices
for the social class measurement model (Table 3-3), annual household income,
educational attainment, and SSS did not form a cohesive factor – a finding foreshadowed
by the weak correlations among these indicators (p’s <.37; Table 2-2).
Table 3-3. Fit indices for Social Class Latent Factor (Study 1)
χ2 Test of Model Fit
Social Class
RMSEA
.32
χ2(1)=35.53, p<.001
(CI:.23-.41)
22
CFI
TLI
SRMR
.59
-.23
.10
Specifically, annual household income had a weak, negative loading and was
uncorrelated with perceived risk and expected benefits (Table 3-4); consequently, it was
dropped from the model. Because three indicators are needed to form a latent factor,
removing annual household income resulted in an unidentified measurement model.
Table 3-4. Correlation Between Perceived Risk, Expected Benefits, & Sociodemographic
Variables
Income
Edu
SSS
Sex
Race
Perceived Risk
Heavy Drinking
Number of Sexual Partners
Smoking
Illegal Substance Use
Reckless Driving
Soda Consumption
Not Exercising
Not Eating Fruits/Veggies
Not Wearing Seatbelt
Unprotected Sex
Not Wearing Sunscreen
Expected Benefits
Heavy Drinking
Number of Sexual Partners
Smoking
Illegal Substance Use
Reckless Driving
Soda Consumption
Not Exercising
Not Eating Fruits/Veggies
Not Wearing Seatbelt
Unprotected Sex
Not Wearing Sunscreen
†p<.10 *p<.05 **p<.01 ***p<.001
-.03
-.01
.04
-.01
-.10†
-.01
.01
-.01
.06
-.01
-.01
-.07
-.08
-.08
-.04
-.16**
-.15**
-.15**
-.16**
-.08
-.11*
-.17**
-.10†
-.08
-.02
-.13*
-.11*
-.04
-.08
-.05
-.02
-.05
-.04
-.17**
-.23***
-.19***
-.16***
-.25***
-.12*
-.10†
-.15***
-.12*
-.25***
-.20***
.03
-.03
.02
.05
-.07
-.11*
-.07
-.08
-.05
-.07
.04
.02
.09†
-.05
-.03
.19***
-.01
-.05
-.09
-.04
.04
.01
.15**
.18**
.08
.02
.25***
.16**
.15**
.12*
.08
.08
.11*
.12*
.10†
.03
.04
.16**
.06
.01
.12*
.05
.05
.05
.19***
.28***
.10†
.09†
.19***
.14**
.06
.15**
.21***
.19***
.07
-.08
-.05
-.08
-.05
-.03
-.10†
-.01
-.14*
-.14*
.01
-.06
Educational attainment was a relatively stronger indicator and was more strongly
correlated with perceived risk and expected benefits (Table 3-4); thus, it was retained and
SSS was dropped. Educational attainment, therefore, served as the sole proxy for social
23
class. Notably, however, exchanging educational attainment for SSS produced a similar,
albeit weaker, pattern of results in the overall model.
As mentioned above, the measurement models for active and inactive health risk
behaviors also underwent slight modification. Soda consumption and smoking were
removed from the active latent factor, and unprotected sex and seatbelt use were removed
from the inactive latent factor due to cross-loading (Figure 3-4), and negative or near zero
loadings. Consequently, four indicators were retained to represent the active latent factor
(i.e., substance use, heavy drinking, reckless driving, and number of sexual partners) and
three indicators were retained to represent the inactive latent factor (i.e., exercise, fruit
and vegetable consumption, and sunscreen use). Additionally, the error variances for
heavy drinking and reckless driving, and exercise and fruit and vegetable consumption
were allowed to correlate.1
Figure 3-4. Loading Plot for Health Risk Behaviors (Study
1)
24
Perceived risk and expected benefits for active and inactive health risk behaviors
were modified to reflect changes in the active and inactive latent factors mentioned
above. Specifically, perceived risk and expected benefits were calculated by averaging
across perceived risk and expected benefits ratings for the remaining active and inactive
health risk behaviors. Furthermore, the error variances for perceived risk and expected
benefits were allowed to correlate. Support for this modification comes from prior
research demonstrating that perceived risk and expected benefits tend to be negatively
correlated (Peters & Slovic, 1996). Slovic and colleagues argue that this pattern arises
because people largely rely on affective information when making risk-benefit
judgments. That is, people rate an activity as having high risk and few benefits when it is
viewed unfavorably and as having low risk and many benefits when it is viewed
favorably (Alhakami & Slovic, 1994). Other researchers also imply a correlation between
perceived risk and expected benefits. For instance, Weber and colleagues (2002) argue
that the trade-off between perceived risk and expected benefits predicts risk-taking.
Finally, race was removed from the model because it was uncorrelated with all
but one of the remaining variables in the model – namely, fruit and vegetable
consumption (Tables 2-5 & 3-4). Recall that race was originally included in the model to
ensure that it did not confound social class. Critically, however, race was not correlated
with education (r =-.03; Table 2-2), the only remaining social class indicator. Sex, by
contrast, was highly correlated with many variables in the model including perceived
risk, expected benefits and health risk behavior; consequently, sex was retained.
Alternative Model. The alternative model was also modified in order to ensure
that it was comparable to the hypothesized model. The modifications mirror those made
25
to the hypothesized model where appropriate. Specifically, educational attainment was
included as the sole proxy for social class and race was dropped from the model.
Furthermore, the four indicators dropped above – soda consumption, smoking,
unprotected sex, and seatbelt use – were dropped from the health risk behavior latent
factor and the perceived risk and expected benefits calculations. In addition, perceived
risk and expected benefits were allowed to correlate. The modified model is depicted in
Figure 3-5.
Figure 3-5. Modified Alternative Model
Model Performance
Fit indices for the hypothesized and alternative models are presented in Table 3-5.
The hypothesized model demonstrated superior fit across all five fit indices compared to
the alternative model. Furthermore, the hypothesized model exceeded the threshold for
acceptable model fit for all fit indices except the chi-square test of model fit. The
26
Table 3-5. Fit indices for the (modified) Hypothesized & Alternative Models
χ2 Test of Model Fit
Hypothesized Model
Alternative Model
χ2(51)=70.60, p=.04
χ2(51)=151.04, p<.001
RMSEA
.03
(90% CI:.01-.05)
.08
(90% CI:.06-.09)
CFI
TLI
SRMR
.95
.92
.04
.54
.41
.07
alternative model, by contrast, failed to meet the threshold for acceptable model fit for
any of the fit indices. This suggests that the hypothesized model is a relatively better
representation of the data than the alternative model. To discern the nature of the
relations among variables in the model, however, we must examine the parameter
estimates for each path. Path estimates for the hypothesized and alternative models are
depicted in Figures 3-6 and 3-7. Given that the hypothesized model is a better
representation of the data, we focus our attention on the pattern of results for the
hypothesized model below.
Consistent with the analyses presented above, lower social classes and women
perceived the behaviors as more risky and less beneficial than upper social classes and
men. This pattern persisted regardless of the category of health risk behavior. Replicating
prior research (e.g., Blais & Weber, 2006; Weber et al., 2002), greater perceived risk and
fewer expected benefits were associated with lower levels of risk-taking. Again, this
pattern held regardless of the category of health risk behavior. As hypothesized, however,
expected benefits was a relatively stronger predictor of active health risk behavior than
perceived risk whereas perceived risk was a relatively stronger predictor of inactive
health risk behavior than expected benefits. Interestingly, this pattern was more
pronounced for inactive than active health risk behavior. Perceived risk and expected
benefits were both statistically significant predictors of active health risk behavior (albeit
27
Figure 3-6. Path Estimates for the Modified Hypothesized Model
Figure 3-7. Path Estimates for the Modified Alternative Model
28
expected benefits appeared to be the stronger predictor). By contrast, perceived risk was
the only significant predictor of inactive health risk behavior.The indirect effects from
education to active health risk behavior through perceived risk and expected benefits
were significant (Table 3-6).
Table 3-6. Indirect effects for the (modified) hypothesized model
Education to Active Risk Behavior
Perceived Risk
Expected Benefits
.07***
.03***
Education to Inactive Risk Behavior
Perceived Risk
Expected Benefits
.09***
.01***
Sex to Active Risk Behavior
Perceived Risk
Expected Benefits
.08**8
.09***
Sex to Inactive Risk Behavior
Perceived Risk
Expected Benefits
*p<.05 **p<.01 ***p<.001
.08***
.01***
Perceived risk and expected benefits fully mediated the relationship between education
and active health risk behavior. Lower education was associated with greater perceived
risk and fewer expected benefits than higher education and in turn, less active health risk
behavior. At first, this finding seems to contradict prior research demonstrating that
people in lower social classes engage in more health risk behavior than those in upper
classes. However, the descriptive statistics for our sample reveal that upper and lower
social classes reported similar mean levels of active and inactive risk behavior (Table 37). The discrepancy in the pattern of results obtained for means vs. correlations is not
clear, but may be due to the fact that the pattern of means is based upon a subjective split
of the data such that an income of ≥$50,000, 1 or more years of college, and SSS ≥ 5
29
signified upper social class. Different criteria for placing participants into groups (e.g.,
above vs. below the poverty threshold, bachelor’s degree or higher) may have produced a
different pattern of mean-level results.
Table 3-7. Descriptive Statistics for Active & Inactive Health Risk Behavior by Indicator
Education
Lower
Upper
M
SD
M
Income
Lower
Upper
SSS
Lower
Upper
SD
M
SD
M
SD
M
SD
M
SD
.82
2.07
.89
2.10
.75
2.02
.82
2.15
.80
Active Risk
2.07 2.08 2.09
Inactive Risk
5.01 1.23 4.81 1.18 4.98 1.19 4.85 1.23 5.00 1.19 4.83 1.22
The pattern of results was more complex for inactive health risk behavior .
Perceived risk partially mediated the relationship between education and inactive health
risk behavior, but expected benefits did not. Lower education was associated with higher
perceived risk and, in turn, lower inactive health risk behavior (e.g., less exercise, lower
fruit & vegetable intake). The direct effect from education to inactive health risk behavior
remained significant and was opposite the indirect effect such that lower education was
associated with greater inactive health risk behavior. Notably, the direct effect was more
than twice the size of the indirect effect.
The indirect effects from sex to active health risk behavior through perceived risk
and expected benefits were also significant (Table 3-6), replicating prior research
showing that perceived risk and expected benefits mediate sex differences in risk
behavior. Women had greater perceived risk and lower expected benefits than men and in
turn, lower levels of active health risk behavior. Similarly, the indirect effect from sex to
inactive health risk behavior through perceived risk was significant, suggesting that
perceived risk mediates the relationship between sex and inactive risk behavior. Women
30
had greater perceived risk than men and in turn, lower levels of inactive health risk
behavior. There was no indirect effect through expected benefits.
Summary of Findings
Study 1 demonstrated social class differences in perceived risk and expected
benefits such that participants from lower social classes rated active and inactive health
risk behaviors as being more risky and having fewer benefits than participants from upper
social classes. Although inconsistent with our first hypothesis, this pattern was apparent
regardless of the social class indicator used for comparison and occurred at the group and
individual levels of analysis. Consistent with our second hypothesis, perceived risk and
benefits mediated the relationship between social class and health risk behavior, but the
nature of the relationship depended upon whether the risk behavior was active or inactive.
Consistent with our third hypothesis, a moderated-mediation pattern was observed such
that expected benefits was a stronger predictor of active health risk behavior whereas
perceived risk was a stronger predictor of inactive health risk behavior.
31
Chapter Four
Overview of Study 2
The goal of Study 2 is to help clarify the relationship between social class,
perceived risk, and expected benefits. An important limitation to Study 1 is that it cannot
speak to the causal relations among these variables and consequently, it is difficult to
discern whether the pattern of results obtained in Study 1 is due to social class or some
other variable (e.g., vigilance, power). In Study 1, we demonstrated social class
differences in perceived risk and expected benefits for income and education, and a
similar but marginal effect for SSS. Income and education cannot be manipulated. In
Study 2, therefore, we examine whether shifting the reference point that people use to
make SSS judgments alters perceived risk and expected benefits in a manner similar to
that observed in Study 1.
Previous research has shown that shifting the reference point that people use to
make SSS judgments alters their subsequent behavior. For example, Hensel (2014)
showed that people who were made to feel as though they were lower social class were
more likely to help a fellow student than people who were made to feel as though they
were upper social class. This finding is consistent with other research showing that
people in lower social classes are more empathetic, prosocial, and ethical than people in
upper social classes (Kraus, Côté, & Keltner, 2010; Piff, Kraus, Côté, Cheng, & Keltner,
2010; Piff, Stancato, Côté, Mendoza-Denton, & Keltner, 2012).
Although speculative, shifting the reference point that people use to make SSS judgments
may alter perceptions and expectations by conjuring feelings of relative
deprivation/provision (Wilkinson, 1999) or scarcity (Mullainathan & Shafir, 2013).
32
Consistent with this idea, we examine whether shifting the reference point that
people use to make SSS judgments impacts perceived risk and expected benefits of the
health risk behaviors used in Study 1. In order to ensure that shifts in perceived risk and
expected benefits are due to the manipulation and not to other factors, Study 2 was
conducted in the lab with undergraduate students rather than online where our ability to
control participants’ environment is diminished. Consistent with Study 1, we expected
that people in the low SSS condition would perceive both active and inactive health risk
behaviors to be more risky and have fewer benefits than people in the high SSS
condition. It may also be the case, however, that the manipulation will interact with
participants’ actual SSS. Research has shown that in some instances people’s chronic and
momentary construal of information interact to produce nuanced patterns of behavior
(Cross, Hardin, & Gercek-Swing, 2011). However, this has never been demonstrated with
SSS. To address this issue, we included a set of exploratory analyses that test for an
interaction between the manipulation and participants’ actual SSS.
33
Chapter Five
Study 2 Method
Participants & Overview of Procedure
Participants were 113 undergraduate students (89 females) recruited from the
University of Toledo psychology research participation pool. Upon entering the lab,
informed consent was obtained and participants underwent the SSS manipulation
(described below). Then, they completed the perceived risk and expected benefits
questionnaire used in Study 1. Socio-demographic variables were collected in a
prescreening session. In exchange for their participation, students received partial credit
for their introductory psychology course.
SSS Manipulation
SSS was measured using the MacArthur Scale (Goodman et al., 2001). The
MacArthur Scale consists of a ladder with ten rungs. Typically, it is anchored on the top
by “those who are best off in society” and on the bottom by “those who are worst off in
society”. In order to manipulate the reference point that participants used to make SSS
judgments (Hensel, 2014), participants in the low SSS condition were asked to think
about people who are better off than them in society and to generate a list of the ways in
which those people are better off than them. They were then asked to complete the
MacArthur Scale. However, it was only anchored at the top with “those who are best off
in society”. Participants in the high SSS condition were asked to think about people who
are worse off than them in society and to generate a list of the ways in which those people
are worse off than them. They were then asked to complete the MacArthur Scale.
However, it was only anchored at the bottom with “those who are worst off in society”.
34
Perceived Risk & Expected Benefits
Perceived risk and expected benefits were measured using the perceived risk and
expected benefits questionnaire administered in Study 1 (Appendix C).
Socio-demographic Questions
Participants completed the same socio-demographic questions administered in
Study 1 during a prescreening session.
35
Chapter Six
Study 2 Results
Manipulation Check
An independent samples t-test was run on participants SSS ratings in order to
assure that the SSS manipulation was successful. As expected, participants in the low
SSS condition rated their SSS significantly lower than those in the high SSS condition,
t(111)=-3.64, p<.001.
Primary Analyses
A 2(Condition: low SSS, high SSS) X 2(Behavior: active, inactive) X 2
(Judgment: perceived risk, expected benefits) mixed model analysis of variance
(ANOVA) was run to determine whether manipulating social class alters perceived risk
and expected benefits for the set of health behaviors. The first factor was between
subjects and the second and third factors were within-subjects. Sex was entered as a
covariate. The main effects of Judgment (F(1,110)=78.17, p<.001, ηp2=.42) and Behavior
(F(1,110)=16.00, p<.001, ηp2=.13) were significant. As expected, participants rated the
behaviors as having high levels of risk and few benefits, but active health risk behaviors
were rated as more risky than inactive health risk behaviors. The main effect of Condition
was not significant (F(1,110)=78.17, p<.001, ηp2=.42).
The main effects were qualified by a significant Judgment X Behavior interaction,
F(1,110)=3.69, p=.05, ηp2=.03. Participants rated active and inactive health risk behavior
as having a similar number of benefits; however, they rated active health risk behaviors
as more risky than inactive health risk behavior (Figure 6-1). Descriptive statistics for this
interaction are presented in Table 6-1. The hypothesized Condition X Judgment
36
interaction (F(1,110)=.15, p=.70, ηp2=.001) as well as the Condition X Behavior
(F(1,110)=.05, p=.83, ηp2<.001) and Condition X Judgment X Behavior (F(1,110)=.09,
p=.76, ηp2=.001) interactions were not significant.
6
5
4
Active
Inactive
3
2
1
Perceived Risk
Expected Benefits
Figure 6-1. Behavior X Judgment Interaction (Study 2)
Table 6-1. Descriptive Statistics for the Judgment X Behavior Interaction (Study 2)
Perceived Risk
M
5.37
4.78
Active Risk Behavior
Inactive Risk Behavior
SD
1.01
.85
Expected Benefits
M
2.07
2.06
SD
1.08
1.14
This finding contrasts with our hypothesis that people in the low SSS condition would
rate the behaviors as having higher risk and fewer benefits than those in the high SSS
condition. This may not be surprising given the marginal effect observed in Study 1. As
discussed above, however, the manipulation may have interacted with participants’ actual
social class. We test this in the set of exploratory analyses below.
Exploratory Analyses
Exploratory analyses were undertaken in order to test whether the manipulation
interacted with participants’ actual social class (i.e., annual household income and
37
parents’ educational attainment). High social class was defined as an annual household
income of ≥$50,000 and one or more parents with at least a bachelor’s degree. We ran
two 2(Social Class: Low Income/Education, High Income/Education) X 2(Condition:
Low, High) X 2(Judgment: Perceived Risk, Expected Benefits) X 2(Behavior: Active,
Inactive) mixed ANOVAs. Sex was entered as a covariate. Mirroring the results reported
above, the main effects of Judgment (F’s >44.55, p’s <.001, ηp2’s >.37) and Behavior
(F’s >10.14, p’s<.002, ηp2’s >.12), and the Judgment X Behavior interaction were
significant (F’s >2.96, p’s<.08, ηp2’s >.03). The main effects of Social Class (F’s <2.33,
p’s >.13, ηp2’s <.02) and Condition (F’s <.88, p’s >.35, ηp2’s <.01), and the remaining 2-,
3-, and 4-way interactions (F’s <1.68, p’s>.20, ηp2’s <.02) were not significant. For
interested readers, the expected Social Class X Condition X Judgment interaction for
education is depicted in Figure 6-2. A similar pattern of results was obtained for income.
Overall, the null results in the main and exploratory analyses for Study 2 fail to replicate
results obtained in Study 1.
38
6
6
5
5
4
4
3
3
2
2
Low SSS
High SSS
1
Perceived Risk
1
Expected Benefits
Perceived Risk
Figure 6-2. Class X Condition X Judgment Interaction (Study 2)
39
Expected Benefits
Chapter Seven
Discussion
The pathways by which social class affects health risk behavior are not wellestablished. The current studies are among the first to examine how psychological factors
– namely, perceived risk and expected benefits – affect social class differences in health
risk behavior. We hypothesized that 1) people in lower social classes would perceive
active and inactive health risk behavior to be less risky and have greater benefits than
those in upper social classes, 2) perceived risk and expected benefits would mediate
social class differences in active and inactive health risk behavior, and 3) perceived risk
would be a stronger predictor of inactive health risk behavior whereas expected benefits
would be a stronger predictor of active health risk behavior. We discuss findings related
to each of these hypotheses below.
Social Class Differences in Perceived Risk & Expected Benefits
Study 1 provides evidence for social class differences in perceived risk and
expected benefits; however, the nature of this difference was not as expected. Lower
social classes perceived health risk behaviors as being more risky and having fewer
benefits than upper social classes. The reason for this pattern is unclear. Notably,
however, it was observed for active and inactive health risk behaviors, at the individual
and group levels of analysis, and regardless of the social class indicator chosen for
comparison. Though this finding seems to contradict prior research on sex and culture
differences in risk perception (Blais & Weber, 2006; Weber et al., 2002; Weber & Hsee,
1998), there are several possible explanations for our results. First, people in upper and
lower social classes may call upon different kinds of information when making risk
40
judgments or weight similar information differently. Support for this perspective comes
from qualitative studies demonstrating that upper and lower social classes think and
communicate about health in different ways. In one study, for instance, Chamberlain and
O’neill (1998) showed that people in lower social classes have more limited definitions
and expectations of health, view health-promoting behaviors as less effective, and discuss
greater situational pressures to engage in health risk behavior (see also Fiske & Markus,
2012). Although speculative, it may be the case that lower social classes are considering
the social and situational consequences of health risk behavior (e.g., inability to fulfill
obligations, social acceptance) whereas people in upper social classes are considering the
mental and physical consequences of health risk behavior (e.g., disease, depression,
death). Second and relatedly, people in lower social classes may be more aware of risks
than those in upper social classes. Consistent with this idea, Kraus and colleagues (2011)
have demonstrated that people in lower social classes are more vigilant than those in
upper social classes and therefore, are more sensitive to threat and more attentive to
contextual information. This suggests that lower social classes may generally be more
aware of the risk and benefits associated with an action.
Given the unexpected pattern of results obtained in Study 1 and the limitations
associated with correlation designs, we attempted to replicate this pattern experimentally
in Study 2 by manipulating the reference point that participants used to make SSS
judgments. Unfortunately, the pattern did not replicate in Study 2. This may not be too
surprising given that SSS was the weakest predictor of social class differences in
perceived risk and expected benefits in Study 1, yielding only a marginal effect. It is
unclear why SSS was a weaker predictor; however, it may be because there is a mismatch
41
in the concreteness of SSS judgments and the perceived risk and expected benefits
judgments. SSS is often a stronger predictor of abstract judgments, such as global selfrated health and life satisfaction, than OSS (Adler & Rehkopf, 2008). It may be the case,
therefore, that SSS is a stronger predictor of global outcomes whereas OSS is a stronger
predictor of concrete outcomes. That said, there are a number of other possible reasons
for the null results obtained in Study 2. The manipulation, for instance, may not have
been strong enough to alter participants’ perceptions of risk and associated benefits. Few
studies have attempted to shift SSS experimentally and as mentioned earlier, we are
aware of only one study that has successfully employed the manipulation used in Study 2
(i.e., Hensel, 2014). Manipulations that place participants in positions of higher vs. lower
social status, such as those employed by Sharif and colleagues (2013), may be more
effective for altering perceived risk and expected benefits, as previous research
demonstrates they are effective at altering behavior (i.e., test performance, spending
patterns). It may also be the case that social class shapes perceived risk and expected
benefits over time in a manner similar to culture. If so, we may be unable to capture the
relationship between social class, perceived risk, and expected benefits using single-shot
laboratory manipulations. Evidence for this idea comes from research demonstrating that
social class has a cumulative effect on health and well-being. Pensola and Martikainen
(2003) showed, for instance, that people residing in lower social class environments
during childhood and adulthood have poorer health outcomes than either 1) those who
transitioned to from a lower social class during childhood to a higher social class during
adulthood or 2) those who transition from a higher social class during childhood to a
lower social class during adulthood. Finally, the sample used in Study 2 may account for
42
the null results. Study 1 was conducted on a sample of adults that were more diverse in
terms of income, education, and SSS than the undergraduate sample used in Study 2.
There may be important distinctions in what social class means for these groups, as
college students have yet to obtain their full level of education and income.
Perceived Risk & Expected Benefits as Mediators
Overall, Study 1 supports our second hypothesis – that perceived risk and
expected benefits mediate the relationship between social class and health risk behavior.
Though, the pattern differed somewhat for active and inactive health risk behaviors.
Perceived risk and expected benefits fully mediated social class differences in active
health risk behavior such that lower social class led to higher perceived risk and fewer
expected benefits and in turn, lower active health risk behavior. This finding may seem to
contradict prior literature demonstrating that lower social classes engage in more active
health risk behavior than upper social classes (Fone, 2013; Mackenbach & Bakker 2002).
In our sample, however, upper and lower social classes reported engaging in similar
levels of active health risk behavior. This finding may be the result of sampling error or
the criteria used to split participants into groups. However, this seems unlikely given that
a similar pattern was demonstrated for inactive health risk behaviors (discussed below)
and as mentioned earlier, regardless of the social class indicator chosen for comparison.
That said, additional research is necessary to rule out these possibilities
The relationship between social class and inactive risk behavior is more complex.
Perceived risk (but not expected benefits) partially mediated social class differences in
inactive health risk behavior such that lower social class led to higher perceived risk and
in turn, lower inactive health risk behavior. However, the direct effect from social class
43
remained significant and acted in the opposite direction such that lower social class led to
more inactive health risk behavior. It is unclear what other variables may underlie the
direct effect. It may be the case, for instance, that structural constraints partially mediate
social class differences in inactive health risk behavior. It is difficult for lower social
classes, regardless of their risk perceptions, to engage in healthy behavior if they have
limited access to supermarkets or recreational resources like city parks (Kwarteng et al.,
2013; Virtanen et al., 2007). Additionally, the struggle lower social classes experience in
making ends meet may direct their focus to the short-term consequences of behavior. The
consequences of inactive health risk behavior are not immediately apparent, but occur
slowly over time. A focus on short-term consequences, therefore, may lead to greater
inactive health risk behavior. As evidence, Orbell and colleagues (2004) demonstrated
that the extent to which people consider the long-term consequences of their actions is
related to their intentions to engage in healthy behavior.
Perceived Risk & Expected Benefits Predict Active & Inactive Risk Behaviors
Findings from Study 1 most clearly support our third hypothesis – that perceived
risk and expected benefits differentially predict engagement in active vs. inactive health
risk behavior. Both perceived risk and expected benefits were significant predictors of
active health risk behavior. Higher perceived risk and fewer expected benefits were
associated with less active health risk behavior. Critically, however, expected benefits
was a relatively stronger predictor of active health risk behavior than perceived risk. By
contrast, perceived risk was a relatively stronger predictor of inactive health risk behavior
than expected benefits. In fact, findings suggest that expected benefits was unrelated to
inactive health risk behavior. These findings are consistent with prior research on the
44
Behavioral Inhibition and Behavioral Activation Systems (BIS and BAS, respectively),
demonstrating that the BIS is sensitive to punishment and motivates inaction or
avoidance whereas the BAS is sensitive to reward and motivates action (Verbruggen et
al., 2012).
Sex Differences in Risk Behavior
In addition to the findings related to our three primary hypotheses, we
successfully replicated prior research demonstrating that perceived risk and expected
benefits mediate sex differences in health risk behavior (Weber et al., 2002) and extended
these findings by showing that the pattern of results differs depending upon the category
of health risk behavior. Echoing the pattern obtained for social class, perceived risk and
expected benefits mediated sex differences in active health risk behavior whereas only
perceived risk mediated sex differences in inactive health risk behavior. The effect from
sex to inactive health risk behavior was similar in size to that from social class to inactive
health risk behavior; however, the effects from sex to active health risk behavior was
notably larger than those from social class to active health risk behavior. Although
speculative, this may be rooted in biological differences between men and women that
are not applicable in the case of social class. Evidence for this idea comes from studies
demonstrating that males are more likely to take risks across cultures (Weber & Morris,
2010). That said, we are not suggesting that socialization is unimportant in shaping
perceived risk and expected benefits. Indeed, the existence of cross-cultural and classbased differences suggest that socialization processes are a key component in shaping
these judgments. Rather, we are suggesting that sex differences may arise due to a
combination of social and biological factors.
45
Limitations
There are several limitations to the current studies. First, Study 1 employed a
correlational design which limits our ability to discern causal relationships. It is possible
that an extraneous variable underlies the observed pattern of results. Study 2 employed an
experimental design; however, it only tested a portion of the hypothesized model and
focused on a different social class indicator than was used in Study 1. Additional work is
needed to 1) develop stronger social status manipulations, 2) determine whether the null
results obtained in Study 2 replicate in these paradigms, and 3) establish causal relations
among perceived risk, expected benefits, and each category of health risk behavior.
Second, participants in Study 1 were asked to self-report the frequency with
which they engaged in active and inactive health risk behavior. Self-report measures are
susceptible to a variety of biases and errors that may not arise using other methods of
measurement. Research has shown, for instance, that self-report measures are susceptible
to social desirability concerns and problems with recall. It is important, therefore, that
these results be replicate using other methods such as direct observation. That said, there
are advantages to self-report measures that are critical to the current studies. Specifically,
self-report measures allow researchers to estimate the frequency of private behaviors,
which typically cannot be observed by researchers – for instance, sexual encounters or
illicit drug use.
Third, the health risk behaviors used in the current studies may differ along
dimensions other than action-inaction. For instance, active health risk behaviors also tend
to have immediate consequences whereas inactive health risk behaviors tend to have
distant consequences. Future research is needed to isolate these factors. Researchers
46
should also consider other ways of conceptualizing health risk behaviors. For example,
health risk behaviors are commonly placed into categories based upon whether they are
public or private, socially acceptable, and prevent disease or promote health.
Understanding the independent and interactive effects of these factors will be important
for advancing theory on the link between risk perception and health risk behavior.
Fourth, the MTurk sample obtained in Study 1 limits our ability to generalize
results. The sociodemographic characteristics of the sample differ in important ways from
the US population (Census, 2013). Furthermore, a limited number of participants reported
engaging in active health risk behaviors in Study 1. A larger sample may be required in
order to gain a full understanding of how perceived risk and expected benefits affect
active health risk behavior. Research would benefit from replicating the observed pattern
of results in a larger, nationally representative sample.
Fifth, the health risk behaviors used in the current studies do not differentiate
between the initiation of risk behavior and the maintenance of risk behavior; though, the
items used in the health behavior questionnaire may represent maintenance to a greater
extent than initiation. Critically, the factors that lead to initiation have been shown to
differ in important ways from the factors that underlie maintenance (Rauscher, Hawley,
& Earp, 2005). Future research would benefit from understanding the extent to which the
observed pattern of results differs depending upon whether the outcome is initiation vs.
maintenance. Additionally, research would benefit from identifying factors that give rise
to social class differences in perceived risk and expected benefits. For instance, studies
that aim to understand whether differences in perceived risk and expected benefits are
47
driven primarily by affect vs. cognition may be particularly useful in targeting health
messages.
Basic & Applied Contributions
Our findings contribute to the risk-taking literature in a variety of ways. First, the
current studies are the first to demonstrate social class differences in perceived risk and
expected benefits. Although the pattern of results were not as hypothesized and
challenging to explain in the broader context of social class-risk effects, they are
consistent with prior research on vigilance, demonstrating that people in lower social
classes are more vigilant than those in upper social classes (e.g., Kraus et al., 2011) and,
therefore, more sensitive to threat. Future work is needed to examine whether vigilance
mediates the relationship between social class, perceived risk, and expected benefits
Second, we provide additional evidence for the distinction between active and inactive
risk behaviors. The active-inactive distinction is prominent throughout the psychological
literature; however, this distinction has only recently been applied to risk behavior. Our
findings support those obtained by Keinan and Bereby-Meyer (2012) and suggest that
different mechanisms may underlie active and inactive risk behaviors. Third and
relatedly, we demonstrate that the relations among risk-taking, perceived risk, and
expected benefits differ depending upon the category of risk behavior under
consideration. No study in this area has address this topic and the findings here directly
impact theory and research related to understanding the relationships between perceptions
of risk behaviors and the actual behaviors (Weber et al., 2002; Weber & Morris, 2010).
Future research would benefit from generalizing the relations among perceived risk,
48
expect benefits and active vs. inactive risk behaviors to domains other than health – for
instance, ethics or finance.
The current studies also advance research and theory on health disparities by
demonstrating that perceived risk and expected benefits mediate social class differences
in health risk behavior and by specifying when these mechanisms are most impactful.
Our findings suggest that public health policies and interventions that increase the
perceived risk and decrease the expected benefits of health risk behavior may be critical
for reducing class-based health disparities and that the category of health risk behavior
should determine whether the focus of public health campaigns should be directed toward
perceived risks for a behavior or expected benefits for the behavior. In either case,
perceived risk and expected benefits can be easily incorporated into pre-existing models
of behavior change used in public health, including the theory of planned behavior (TPB;
Azjen, 1991; 2002), the health belief model (HBM; Janz, 1984), the informationmotivation-behavioral skills model (IMB; Fisher, 2002), and the trans-theoretical model
of change (TTM; Prochaska & DiClemente, 1983).
49
References
Adler, N. E., Epel, E. S., Castellazzo, G., & Ickovics, J. R. (2000). Relationship of
subjective and objective social status with psychological and physiological
functioning: Preliminary data in healthy White women. Health Psychology, 19,
586-592.
Adler, N. E., & Rehkopf, D. H. (2008). US disparities in health: descriptions, causes, and
mechanisms. Annual Review of Public Health, 29, 235-252.
Adler, N. E., & Snibbe, A. C. (2003). The role of psychosocial processes in explaining
the gradient between socioeconomic status and health. Current Directions in
Psychological Science, 12, 119-123.
Ajzen, I. (1991). The theory of planned behavior. Organizational behavior and human
decision processes, 50, 179-211.
Barke, R. P., and Jenkins, H. C. (1993). Politics and scientific expertise: Scientists, risk
perception, and nuclear waste policy. Risk Analysis, 13, 425-439.
Barr, R. G., Somers, S. C., Speizer, F. E., & Camargo, C. A. (2002). Patient factors and
medication guideline adherence among older women with asthma. Archives of
Internal Medicine, 162,1761-1768.
Bentler, P. M. (1990). Comparative fit indexes in structural models.
PsychologicalBulletin, 107, 238–246.
Blais, A-R., & Weber, E. U. (2006).A domain-specific risk-taking (DOSPERT) scale for
adult populations. Judgment and Decision Making, 1, 33-47.
Bollen, K. A. (1989). Structural equations with latent variables. New York, NY: Wiley.
50
Braveman, P. A., Cubbin, C., Egerter, S., Chideya, S., Marchi, K. S., Metzler, M., &
Posner, S. (2005). Socioeconomic Status in Health Research: One Size Does Not
Fit All. Journal of the American Medical Association, 294(22), 2879-2888.
Brewer, N. T., Weinstein, N. D., Cuite, C. L., & Herrington, J. E. (2004). Risk
perceptions and their relation to risk behavior. Annals of Behavioral Medicine, 27,
125-130.
Brown, T. A. (2006). Confirmatory factor analysis for applied research. New York, NY:
Guilford Press.
Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A.
Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 136–162).
Beverly Hills, CA: Sage.
Businelle, M.S., Kendozer, D.E., Reitzel, L.R., Costello, T.J., Cofta-Woerpel, L. et al.
(2010). Mechanisms linking socioeconomic status to smoking cessation: a
structural equation modeling approach. Health Psychology, 29(3), 262-273.
Byrnes, J. P., Miller, D. C., Schafer, W. D. (1999). Gender differences in risk taking: a
meta-analysis. Psychological Bulletin, 125, 367–383.
Caraher, M., Dixon, P., Lang, T. & Carr-Hill, R. (1998). Access to healthy foods: Part I.
Barriers to accessing healthy foods: Differentials by gender, social class, income
and mode of transport. Health Education Journal, 57, 191 - 201.
Cohen, S., Doyle, W.J., & Baum, A. (2006). Socioeconomic Status Is Associated With
Stress Hormones. Psychosomatic Medicine, 68, 414-420.
Darmon, N., & Drewnowski, A. (2008). Does social class predict diet quality? The
American Journal of Clinical Nutrition, 87, 1107-1117.
51
Elo, I. T. (2009). Social class differentials in health and mortality: patterns and
explanations in comparative perspective. Annual Review of Sociology, 35, 553572.
Figner, B., & Weber, E. U. (2011). Who takes risks when and why? Determinants of risktaking. Current Directions in Psychological Science, 20, 211-216.
Fiske, S. T., & Markus, H. R. (2012). Facing social class: How societal rank influences
interaction. Russell Sage Foundation.
Genovese, J., & Wallace, D. (2007). Reward sensitivity and substance abuse in middle
school and high school students. The Journal of GeneticPsychology, 168, 465469.
Goldman, D., & Smith, J. P. (2002). Can patient self-management help explain the SES
health gradient? Proceedings of the National Academy of Science, 99,10929–
10934.
Goodman, E., Adler, N. E., Kawachi, I., Frazier, A. L., Huang, B., & Colditz, G. A.
(2001). Adolescents’ perception of social status: Development and evaluation of a
new indicator. Pediatrics, 108, e31.
Hanoch, Y., Johnson, J. G., & Wilke, A. (2006). Domain specificity in experimental
measures and participant recruitment: an application to risk-taking behavior.
Psychological Science, 17, 300-304.
Haught, H. & Rose, P. (under review). Income and education differentially predict
engagement in prevention vs. detection behaviors.
52
Hsee, C. K., & Weber, E. U. (1997). A fundamental prediction error: Self-other
discrepancies in risk preference. Journal of Experimental Psychology: General,
126, 46-53.
Hu, L., & Bentler, P. M. (1998). Fit indices in covariance structure modeling: Sensitivity
to underparameterized model misspecification. Psychological Methods, 3, 424–
453.
Ickovics, J. R., Beren, S. E., Grigorenko, E. L., Morrill, A. C., Druley, J. A., & Rodin, J.
(2002). Pathways of risk: Race, social class, stress, and coping as factors
predicting heterosexual risk behaviors for HIV among women. AIDS and
Behavior, 6(4), 339-350.
Isaacs, S. L., & Schroeder, S. A. (2004). Class – the ignored determinant of the nation’s
health. The New England Journal of Medicine, 351, 1137-1142.
Kawachi, I., & Berkman, L. F. (2003). Neighborhoods and Health. Oxford University
Press.
Kline, R. B. (2010). Principles and practice of structural equation modeling (3rd ed.).
New York: Guilford Press.
Kraus, M. W., Côté, S., & Keltner, D. (2010). Social class, contextualism, and empathic
accuracy. Psychological Science, 21, 1716-1723.
Krokstad, S. A, Kunst, E.,& Westin, S. (2002). Trends in health inequalities by
educational level in a Norwegian total population study. Journal of Epidemiology
and Community Health, 56, 375–80.
Kwarteng, J. L., Schulz, A. J., Mentz, G. B., Zenk, S. N., Opperman, A. A. (2013).
Associations between observed neighborhood characteristics and physical
53
activity: findings from a multiethnic urban community. Journal of Public Health,
3, 1-10.
Kynavez, G. (2010). Reward seeking as a predictor of drug use in youth: effect of gender
and social environment. Addiction Journal, 3, 1-8.
Lynch, J.W., Kaplan, G.A., & Salonen, J.T. (1997). Why do poor people behave poorly?
Variation in adult health behaviors and psychosocial characteristics by stages of
the socioeconomic life course. Social Science and Medicine, 44, 809-819.
MacDonald, T. K., Fong, G. T., Zanna, M. P., & Martineau, A. M. (2000). Alcohol
myopia and condom use: can alcohol intoxication be associated with more
prudent behavior? Journal of Personality and Social Psychology, 78, 605-619.
Marmot, M. G, Kogevinas, M., & Elston, M. A. (1987). Social/economic status and
disease. Annual Review of Public Health, 8, 111–35.
Moore, B. J., Glick N., Romanowdki, B., & Quinley (1996) Neighborhood safety, child
care, and high costs of fruit and vegetables as barriers to increased activity and
healthy eating and linked to overweight and income. FASEB Journal, 10:A562.
Muthén, L. K.., & Muthén, B. O. (2007). Mplus User's Guide (Sixth Edition). Los
Angeles, CA: Muthén & Muthén.
Oakes, J. M., & Rossi, R. H. (2003). The measurement of SES in health research: Current
practice and steps toward a new approach. Social Science and Medicine, 56, 769784.
O'Connor, R., & Colder, C. (2005). Predicting alcohol patterns in first-year college
students through motivational systems and reasons for drinking. Psychology of
Addictive Behaviors, 19, 10-20.
54
Ouimet, M. C, Morton, B. G, Noelcke, E. A., Williams, A. F., Leaf, W. A., Preusser, D.
F., & Hartos, J. L. (2008). Perceived risk and other predictors and correlates of
teenagers' safety belt use during the first year of licensure. Traffic Injury
Prevention, 9, 1-10.
Piff, P. K., Kraus, M. W., Côté, S., Cheng, B. H., & Keltner, D. (2010). Having less,
giving more: the influence of social class on prosocial behavior. Journal of
personality and social psychology, 99, 771.
Piff, P. K., Stancato, D. M., Côté, S., Mendoza-Denton, R., & Keltner, D. (2012). Higher
social class predicts increased unethical behavior. Proceedings of the National
Academy of Sciences, 109, 4086-4091.
Powell, Lisa M., S. Slater, F Chaloupka. (2004) “The relationship between community
physical activity settings and race, ethnicity and socioeconomic status”. EvidenceBased Preventive Medicine, 1,135-144.
Preston, S. H, & Elo, I. T. (1995). Are educational differentials in adult mortality
increasing in the United States? Journal of Aging Health,7, 476–96.
Rose, F. R., & Mossler, D. G. (2013). BIS/BAS and college alcohol use: motivation,
consequences, and attention. Journal of Sciences, 2, 1-10.
Steiger, J. H. (1990). Structural model equation and modification: An interval estimation
approach. Multivariate Behavioral Research, 25,173–180.
Stephan, Y., Boiche, J., Trouilloud, D., Deroche, T., & Sarrazin, P. (2011). The relation
between risk perceptions and physical activity among older adults: a prospective
study. Psychology & Health, 26, 887-897.
55
Teese, R., & Bradley, G. (2008). Predicting recklessness in emerging adults: a test of a
psychosocial model. Journal of Social Psychology, 148, 105-126.
Tucker, L. R., & Lewis, C. (1973). A reliability coefficient for maximum likelihood
factor analysis. Psychometrika, 38, 1–10.
Virtanen, P., Kivimaki, M., Vahtera, J., & Koskenvuo, M. (2006). Employment status
and differences in one-year coverage of physician visits: different needs or
unequal access to services?. BMC Health Services Research, 6, 123.
Weber, E. U., Blais, A.-R., Betz, E. (2002). A domain-specific risk-attitude scale:
Measuring risk perceptions and risk behaviors. Journal of Behavioral Decision
Making, 15, 263–290.
Weber, E. U., & Hsee, C. K. (1998). Cross-cultural differences in risk perception but
cross-cultural similarities in attitudes towards risk. Management Science, 44,
1205–1217.
Weber, E. U., Hsee, C. K., & Sokolowska, J. (1999). What folklore tells us about risk and
risk-taking: Cross-cultural comparisons of American, German, and Chinese
proverbs. Organizational Behavior and Human Decision Processes, 75, 170-186.
Weber, E. U., & Johnson, E. J. (2008). Decisions under uncertainty: Psychological,
economic, and neuroeconomic explanations of risk preference. In: P. Glimcher, C.
Camerer, E. Fehr, & R. Poldrack (Eds.), Neuroeconomics: Decision making and
the brain (pp. 127-144). New York: Elsevier.
Weber, E. U., & Milliman, R. (1997). Perceived risk attitudes: Relating risk perception to
risky choice. Management Science, 43, 122–143.
56
Weber, E. U. & Morris, M. W. (2010). Culture and judgment and decision making: The
constructivist turn. Perspectives on Psychological Science, 5, 410-419.
57
Appendix A
Think of the ladder as representing where people stand in society. Some people are better
off – they have more money, more education, and better jobs. Other people are worse off
– they have less money, less education, and worse jobs. The higher up on the ladder you
are, the closer you are to the people at the top and the lower you are, the closer you are to
the people at the bottom.
Think about yourself. Please use an ‘X’ to indicate on which rung of the ladder you
would place yourself.
People in society who are the best off
People in society who are the worst off
58
Appendix B
Please read carefully and answer as accurately and honestly as possible. Indicate the
frequency with which you engage in each behavior.
1. On days when you drink, what is the average number of drinks that you consume per
day?
0 drinks
1-2 drinks
3-4 drinks
0
1
2
5-6
7-8
drinks
drinks
3
4
9-10 drinks
≥11 drinks
5
6
2. How many sexual partners have you had in the past month?
0
partners
1-2
partners
3-4
partners
5-6
partners
7-8
partners
9-10
partners
≥11
partners
0
1
2
3
4
5
6
9-10
cigarettes
5
≥11
cigarettes
6
3. On average, how many cigarettes do you smoke per day?
0
cigarettes
0
1-2
‘cigarettes
1
3-4
cigarettes
2
5-6
cigarettes
3
7-8
cigarettes
4
4. How many times have you used an illegal substance (e.g., marijuana) during the past
month?
0 times
0
1-2 times
1
3-4 times
2
5-6 times
3
7-8 times
4
9-10 times
5
≥11 times
6
5. How frequently do you drive more than 10 mph over the speed limit?
Never
1
Always
2
3
4
59
5
6
7
6. On average, how much soda do you consume per day?
None
12oz or less
13-24oz
25-36oz
37-48oz
≥48oz
1
2
3
4
5
6
7. On average, how many days per week do you exercise for at least 30 minutes?
0 days
1 day
2 days
3 days
4 days
5 days
6 days
7 days
8. On average, how many servings of fruits and vegetables do you eat per day?
None
0
Less than
½ cup
1
½-1cup
1-2 cups
2-3cups
3-4 cups
2
3
4
5
4 or more
cups
6
9. How frequently do you wear your seatbelt?
Never
1
Always
2
3
4
5
6
7
10. How many times have you had unprotected sex in the past month?
0 times
0
1-2 times
1
3-4 times
2
5-6 times
3
7-8 times
4
9-10 times
5
≥11 times
6
11. How frequently do you use sunscreen when going outside for more than 1 hour?
Never
1
Always
2
3
4
60
5
6
7
Appendix C
Please read carefully. Using the scale provided, indicate how risky you perceive each
behavior.
Not at all
risky
1
Very risky
2
3
4
5
6
7
1. Consuming five or more servings of alcohol in a single evening.
______
2. Having multiple sexual partners.
______
3. Smoking cigarettes.
______
4. Using illegal drugs.
______
5. Driving more than 10mph over the speed limit.
______
6. Consuming more than 12oz (1 can) of soda per day.
______
7. Not exercising 4 or more days per week.
______
8. Eating fewer than 5 servings of fruits and vegetables per day.
______
9. Not wearing your seatbelt when being a passenger in the front seat.
______
10. Engaging in unprotected sex.
______
11. Exposing yourself to the sun without using sunscreen.
______
61
Please read carefully. Using the scale provided, indicate the number of benefits you
associate with each behavior.
No
benefits at
all
1
Many
benefits
2
3
4
5
6
7
1. Consuming five or more servings of alcohol in a single evening.
______
2. Having multiple sexual partners.
______
3. Smoking cigarettes.
______
4. Using illegal drugs.
______
5. Driving more than 10mph over the speed limit.
______
6. Consuming more than 12oz (1 can) of soda per day.
______
7. Not exercising 4 or more days per week.
______
8. Eating fewer than 5 servings of fruits and vegetables per day.
______
9. Not wearing your seatbelt when being a passenger in the front seat.
______
10. Engaging in unprotected sex.
______
11. Exposing yourself to the sun without using sunscreen.
______
62