ArticlePDF Available

Gradual Retirement, Sense of Control, and Retirees' Happiness

Authors:

Abstract and Figures

This study explores the factors that affect an individual's happiness while transitioning into retirement. Recent studies find that workers often view the idea of gradual retirement as a more attractive alternative than a "cold turkey" or abrupt retirement. However, there is very little evidence as to whether phasing or cold turkey makes for a happier retirement. Using longitudinal data from the Health and Retirement Study, this study explores what shapes the change in happiness between the last wave of full employment and the first wave of full retirement. Results suggest that what matters is not the type of transition (gradual retirement or cold turkey), but whether people perceive the transition as chosen or forced.
Content may be subject to copyright.
1
Shortened running title: RETIREMENT, CONTROL, AND HAPPINESS
Gradual Retirement, Sense of Control, and Retirees’ Happiness
*
Esteban Calvo
Sociology Department
Center for Retirement Research
Boston College
calvobra@bc.edu
Work phone: (617) 552-1762
Home phone: (617) 894-6661
Fax: (617) 552-0191
Kelly Haverstick
Center for Retirement Research
Boston College
haversti@bc.edu
Steven A. Sass
Center for Retirement Research
Boston College
sassst@bc.edu
The final, definitive version of this paper has been published in Research on Aging,
Volume 31, Number 1, January 2009 by SAGE Publications Ltd, All rights reserved. ©
The online version of this article can be found at:
http://roa.sagepub.com/content/31/1/112
2
Esteban Calvo, Kelly Haverstick, and Steven A. Sass. Gradual Retirement, Sense of
Control, and Retirees' Happiness. Research on Aging January 2009 31: 112-135.
doi:10.1177/0164027508324704
*
Please send any inquiries to Esteban Calvo, Center for Retirement Research at Boston
College, Hovey House room 206, 140 Commonwealth Avenue, Chestnut Hill, MA
02467, (617) 552-1762, calvobra@bc.edu. The authors would like to thank Paula
Errázuriz, Shari Grove, Estelle James, Natalia Sarkisian, Sarah Mack, Mauricio Soto,
Alicia H. Munnell, Robert Triest, Anthony Webb and John B. Williamson for their
comments and other forms of help in connection to this article. However, the authors
should be held responsible for any errors or inaccuracies that remain in this article. The
research reported herein was performed pursuant to a grant from the U.S. Social Security
Administration (SSA) funded as part of the Retirement Research Consortium (RRC). The
findings and conclusions expressed are solely those of the authors and do not represent
the views of SSA, any agency of the Federal Government, the RCC, or Boston College.
3
ABSTRACT
This study explores the factors that affect an individual‟s happiness while
transitioning into retirement. Recent studies find that workers often view the idea of
gradual retirement as a more attractive alternative than a „cold turkey‟ or abrupt
retirement. However, there is very little evidence as to whether phasing or cold turkey
makes for a happier retirement. Using longitudinal data from the Health and Retirement
Study, this study explores what shapes the change in happiness between the last wave of
full employment and the first wave of full retirement. Results suggest that what matters is
not the type of transition (gradual retirement or cold turkey), but whether people perceive
the transition as chosen or forced.
Keywords: happiness, retirement, gradual, phased, control.
4
Workers approaching retirement often say they want to retire gradually, rather
than going straight from full-time employment to complete retirement. Some surveys
report that more than half of all older workers prefer to exit the labor force this way.
(Hutchens and Papps 2005). Most policymakers also view gradual retirement favorably,
as a way workers can extend their careers and thereby improve retirement income
security (Gustman and Steinmeier 2007, however, question whether gradual retirement
would increase average working lives).
To accommodate worker preferences and
improve retirement income security, expanding opportunities for gradual retirement has
gained a prominent place on the policy agenda.
It is not clear, however, that retirees are indeed better-off if they retire gradually
as opposed to „cold turkey (De Vaus et al. 2007). Workers who want to retire gradually
are not basing their preference on their own personal experience. They have not retired
both wayscold turkey and in stagesand concluded that that they are happier in
retirement after a gradual transition. Moreover, individuals typically misestimate their
future happiness (Gilbert 2007).
In this study we seek to learn whether individuals are indeed happier if their
transition out of the labor force is gradual as opposed to abrupt. As a large share of
workers says they want to retire gradually, a second research goal is to learn whether
retirees who say they had control over their transition to retirement are happier than those
who felt forced to retire. We use happiness as the yardstick for evaluating work-to-
retirement transitions. Happiness has an important advantage over other yardsticks, as it
measures realized quality of life. Other criteria, such as income, wealth, social status, or
health, measure potential quality of life (Veenhoven 2006). Our study thus asks whether
retirees who exited the labor force gradually are “happier” than those who left cold
turkey. By happiness we mean the individual‟s general experience of different kind of
feelings. Some feelings we experience are pleasurable, such as enjoyment of life. Others,
such as sadness, are unpleasant. Our study thus asks whether the type of transition from
work to retirement affects the degree to which the feelings a person experiences are
generally pleasant or enjoyable.
5
I. LITERATURE REVIEW
A Gradual versus an Abrupt Transition
One out of three workers age 55 and over say they would stay in the labor force
longer if they could cut back their hours (Watson Wyatt 2004). And two out of three
workers age 50 to 70 say they plan to work in “retirement” (Brown 2003). This
preference for a gradual exit is understandable. These workers have spent thirty or more
years in the labor force, and retirement represents a sharp social, psychological, and
economic break with life as they know it. Thus, it is not surprising that workers prefer to
negotiate the transition in stages. A smooth transition allows older workers to continue
daily activities similar to those performed in middle-age. This should be helpful in
maintaining meaning and a sense of purpose in life, as well as adapting to aging (Atchley
1999; Rowe and Kahn 1998). Gradual retirement also could enhance opportunities to
remain active and socially engaged. Evidence suggests that remaining active and socially
engaged has a strong positive impact on health and well-being in retirement (Brummett et
al. 2001; Erikson, Erikson, and Kivnick 1986; Everard et al. 2000; Siegrist, Von dem
Knesebeck, and Pollack 2004).
On the other hand, previous research also finds that people are quite resilient.
Happiness is very stable across the life span, with large shocks often having but a short-
term effect on our affects and sense of well-being. Thus researchers argue that
individuals have a normal baseline level of psychological well-being that varies only
moderately in response to current events (Appley 1971; Costa, McCrae and Norris 1981;
Lykken and Tellegen 1996). This perspective suggests that the type of transition might
not have a meaningful effect on happiness in retirement.
Factors That Influence Happiness
Researchers have found that most retirees are happy in retirement. But they also
have found that the degree of satisfaction fluctuates substantially and is associated with
6
various factors. To identify the independent effect of gradual versus abrupt transitions to
retirement, we will need to control for factors identified in the literature as affecting
happiness in retirement.
One factor that has a significant effect on happiness is individuals‟ sense of
control over their life (Kunzmann, Little, and Smith 2002; Rodin 1986; Sweeney,
Anderson, and Bailey 1986). In terms of the work-retirement transition, individuals who
retired voluntarily are happier than those who were forced out of their job (Calvo 2006;
De Vaus et al. 2007; Gall, Evans, and Howard 1997; Gallo et al. 2006).
Social relationships are another factor found to have an impact on happiness
(Chan and Lee 2006; Glass et al. 2006; Vanderhorst and McLaren 2005). For example,
retirees who are married tend to be happier than those who are single (Bierman, Fazio,
and Milkie 2006; Demo and Acock 1996). The death of a spouse, relative, or close friend
and divorce or separation all diminish happiness (Cheng and Chan 2006; Lucas et al.
2003; Pinquart 2003). Because of the age of individuals in this study, death of a spouse
is a relevant factor that may affect individuals‟ happiness when transitioning into
retirement.
Good health is positively associated with happiness, and health tends to decline as
individuals age (Cohen 2004; Hilleras et al. 2001; Rohwedder 2006; Rowe and Kahn
1998). Using a longitudinal design, Kosloski et al. (2005) found that self-reported health
has a consistent effect on depressive symptoms, while the level of depressive symptoms
has no significant effects on self-reported health.
Aging alone does not seem to have much of an effect on happiness. After
controlling for the decline in health and the loss of social roles and loved ones that comes
with aging, neither longitudinal nor cross-sectional studies have found a substantial
relationship between age and happiness in retirement (Cheng 2004; Jorm 2000;
Kunzmann, Little, and Smith 2000; Pinquart 2001).
Researchers have found the effect of income and wealth on happiness somewhat
mixed. Cross-sectional studies provide evidence of a positive relationship between wealth
and happiness. But researchers analyzing longitudinal data have found the effect of
7
income and wealth on happiness to be generally small, except around the poverty
threshold (Arendt 2005; Arthaud-day and Near 2005; Diener and Biswas-Diener 2002;
Easterlin [1974] 2002; Inglehart and Klingemann 2000; Michalos 1985; Saris 2001).
However, the wealth of an individual approaching and entering retirement may have
more of an impact on this individual‟s happiness. One economic factor found to affect
happiness in retirement is pension type. Some recent research found that retirees are
happier with a defined-benefit pension as opposed to having a comparable amount of
wealth in a retirement account (Bender 2004; Panis 2003).
II. DATA AND METHODOLOGY
Data
This study uses data from the Health and Retirement Study (HRS), a nationally
representative, biennial, panel survey of older Americans and their spouses (University of
Michigan 2007). Many variables used in this project are from the RAND (2007) cleaned
version of the HRS. The HRS began in 1992 and data are available through 2004.
For the sample selection we begin with the age-eligible individuals from the initial
HRS cohort. These are 9,760 individuals born between 1931 and 1941 who responded the
HRS in 1992. The next task is to identify those individuals we can observe making the
transition from work to retirement.
Researchers use a variety of measures to characterize individuals as fully
employed, fully retired, or something in-between. Among the most common are self-
reported status, hours worked (per week or per year), change in earnings, and whether
they have claimed Social Security benefits (Chen and Scott 2006; Gustman and
Steinmeier 2000, 1984; Honig and Hanoch 1985; Haider and Loughran 2001; Ruhm
1990). To identify individuals in the HRS we can observe making the transition from
work to retirement we use two criteria, their usual hours of work per week and self-
reported retirement status. The usual hours worked per week variable is the sum of
usual hours worked per week at the main job and the usual hours worked per week at a
8
second job. We classify individuals as “fully employed” in the first wave of the HRS if
they then worked at least 30 hours per week and reported themselves “not retired” (as
opposed to “completely retired” or “partly retired.”) As illustrated in Figure 1, 5,744 of
the 9,760 individuals in the HRS cohort were fully employed in 1992. The final column
of this figure gives the total number fully retired by 2004, the last available wave. We
classify individuals as “fully retired” if they had zero hours of work and report
themselves “completely retired.” Of the 5,744 individuals who were fully employed in
the initial wave, 3,022 were fully retired in at least one subsequent wave. Once an
individual is fully retired we ignore all further labor-market activity. In other words, we
do not look at un-retirement dynamics.
Our final usable number of observations is 2,389. Of the 3,022 individuals who
made a full transition from work to retirement, 219 cases do not fit our definitions of
gradual and cold turkey retirement (the process of classifying gradual and cold turkey
retirees is described below) and 414 have missing happiness values and are dropped after
imputing values for the other variables. For each wave of the HRS, Figure 1 indicates the
number of individuals fully employed in 1992 that were: (1) fully retired, (2) not fully
retired, or (3) out of the HRS due to attrition. Since we select only individuals we
observe transitioning from full-time employment to full retirement, the sample used for
this study may not be representative of the general population.
[FIGURE 1]
We handle missing values in the variables used for the selection of the sample by
deriving the information from other variables whenever possible. In addition, we perform
a single imputation on these variables, and after selecting the sample, a Multiple
Imputation by Chained Equations (MICE) on the variables used for the analysis.
Imputing data allows us to include an additional 307 cases, and using MICE prevents us
from underestimating the standard errors, as each model is estimated over five imputed
datasets including a random component (Allison 2001; Royston 2004).
9
Dependent Variables
Happiness, the focus of our study, is a slippery concept that researchers define in
many different ways. We defined happiness as the degree to which a variety of feelings
that a person experiences are pleasant or enjoyable. Our definition of happiness focuses
on: (1) more or less stable feelings as opposed to temporary feelings, such as the sensory
delight of a chocolate; (2) an evaluation of one‟s feelings in general as opposed to the
evaluation of a specific domain of life, such as satisfaction with job; (3) an affective as
opposed to a cognitive notion of happiness, such as the degree to which we think we have
achieved our goals.
To measure happiness we use five yes-or-no questions, about pleasant and
unpleasant feelings, that the HRS asks of both working and retired respondents: “Now
think about the past week and the feelings you have experienced. Please tell me if each of
the following was true for you much of the time this past week. … Much of the time…
you were happy; you enjoyed life; you felt lonely; you felt depressed; you felt sad.” The
first two questions are measures of positive feelings and the last three measure negative
feelings. Previous research has established the validity and reliability of such self-report
measures (for example, see: Diener et al. 1999; Frey and Stutzer 2002; Layard 2005;
Perreira et al. 2005; Steffick 2000; Veenhoven 1991).
For each individual in the HRS who makes the transition from work to retirement,
we measure the change in each of the five happiness indicators (feelings of happiness,
enjoyment of life, loneliness, depression, and sadness). To do that we take the baseline
measure of each indicator in the last wave in which the individual was fully employed
and record the change in the first wave in which the individual is fully retired.
1
As the
1
Because we wanted to measure the change in happiness between the times when individuals were
employed and when they first retired, there are differences in the amount of time passing between the
measurements across individuals. We include a variable controlling for the number of years from full
employment to full retirement which does not have a significant effect on the dependent variable. Tables
including these variables are available upon request.
10
indicator variables are dichotomous, the value either remains the same or changes to the
opposite value. We record no change as 0, a “yes” to “no” change as “-1,” and a “no” to
“yes” change as “+1.” For example, the change in “enjoyment of life” would be -1 if the
individual answered “yes” in the last wave of full employment and “no” in the first wave
of full retirement; 0 if there was no change, and +1 if the individual said that they did not
enjoy life while working, but did in retirement.
Independent Variables
Measurement of Gradual/Abrupt Retirement. Our primary concern is whether
happiness in retirement is affected by a gradual as opposed to a “cold turkey” transition.
To do this, the first task is to identify those individuals in the HRS population who made
the transition from work to retirement. We then classify these individuals into those who
retired gradually and those who retired abruptly.
As illustrated in Figure 2, we distinguish between the two types of transitions
based on the respondent‟s self-reported status between being classified as “fully
employed” in the initial wave of the HRS (working 30 or more hours a week and
reporting themselves “not retired”) and “fully retired” (working 0 hours a week and
reporting themselves “completely retired”) in a subsequent wave. For an individual to be
classified as a cold turkey retiree, they must report themselves as not retired” in each
wave prior to being classified as “fully retired.” For an individual to be classified as a
gradual retiree, they must report themselves as “partially retired” in all intervening waves
between reporting themselves “not retired” and being classified as “completely retired.”.
We thus omitted from the sample individuals who report themselves in an intervening
wave as “completely retired” but work more than zero hours, as their employment status
in that wave and the nature of their transition are both ambiguous. We also omit from the
sample individuals who reversed direction and reported themselves “not retired” after
reporting themselves “partly retired,” as the nature of their transition was ambiguous.
The two gray rows in Figure 2 give the number of individuals in our 2,389 person
sample that had transitioned to full retirement at each wave. The gray row at the top gives
11
the number of workers who had retired „Cold turkeyand the bottom row gives the
number who had „Retired gradually.‟ By 2004, 1,733 individuals had retired „Cold
turkey(73%) and 656 had „Retired gradually‟ (27%). Figure 2 also reports for each
wave the number of respondents still „Not retired‟ (self-report as “not retired”) and the
number „Partially retired‟ (self-report as “partly retired”).
[FIGURE 2]
We do not consider hours of work in distinguishing between gradual and abrupt
retirements for several reasons. As we are primarily interested in individual perceptions
of the retirement process, self-reported status is the more relevant single criterion.
Moreover, the HRS only asks respondents if they were forced or wanted to retire if they
report themselves “completely retired” or “partly retired.” So we would not get this
information from individuals who work between 0 and 30 hours per week, but consider
themselves “not retired.” As the voluntary or involuntary nature of a worker‟s separation
from employment has been identified as an important factor contributing to happiness in
retirement, we would not be able to include this variable in our analysis had we classified
transitions as “gradual” based only on hours of work.
Measurement of Sense of Control. Additional independent variables were included
in the regressions. We create a set of three dummy variables measuring the respondent‟s
perception on whether retirement was something the respondent “wanted to do,” was
“part wanted, part forced,” or something the respondent was “forced into.” Included in
the regressions are “chose retirement” and “part wanted retirement,” so the estimated
effects are relative to “forced retirement.” Feeling forced into retirement is expected to
decrease happiness and enjoyment of life and increase loneliness, depression, and
sadness. The wave from which this variable was taken depends on the type of transition.
For those coded as cold turkey retirees, this variable is coded based on the individual‟s
response in the first wave of full retirement. For those coded as gradual retirees, this
variable is coded using, in this order: (1) if not missing, the value of the “wanted
12
to/forced into” variable in the wave immediately following the last wave of full
employment (2) otherwise, the first non-missing response to the “wanted to/forced into”
variable after the last wave of full employment and before the first wave of full
retirement.
Controls
Our analysis also controls for other factors identified in the literature as affecting
happiness at old age:
Death of spouse during the transition. If a respondent has a marital status of
“married” or “married, spouse absent” in his last wave of full employment and a marital
status of “widowed” in the first wave of full retirement, he or she is coded as having a
spouse who died.
Change in health. A change in health status, measured as a change in self-reported
health status. Self-reported health status is given on a scale from 1 to 5, with 1
corresponding to “poor” and 5 to “excellent” health. This variable measures the change
in self-reported health status from the last wave of full employment to the wave of full
retirement. A positive value for this variable indicates an improvement in health,
according to the respondent.
Defined-benefit pension coverage. This reports whether the respondent was
covered by a defined-benefit pension plan at their job at the last wave of full
employment.
Other control variables included are: the number of years between full
employment and full retirement, “unemployment” reported in the first wave of “complete
retirement,” and various socioeconomic and demographic variables such as: gender,
race/ethnicity, age, education, wealth and type of occupation. Although some of these
variables may not have a substantial impact on happiness, the type of work to retirement
transition, the sense of control workers have over the transition, and health levels are
known to vary by these characteristics (Burr et al. 1996; Flippen and Tienda 2000; Link
and Phelan 1995; Mirowsky and Ross 2007). Unemployment controls for a possible
13
misclassification of individuals as retired when they are actually unemployed. If the
labor force status is unemployed in the wave an individual is coded as fully retired, this
variable takes a value of one. For race/ethnicity, respondents are classified as: (1) white
and non-hispanic, or (2) non-white and/or hispanic. The difference in ages of the
individual in the last wave of full employment and at the first wave of full retirement is
included in order to control for the number of years between full employment and full
retirement. The educational attainment of respondents is categorized as high school or
less or more than high school. The measure of wealth is the natural logarithm of mean
wealth (excluding the house value) from the first wave to the wave an individual is coded
as fully retired. The variable is recoded to zero if mean wealth is negative and values in
each wave were adjusted by Consumer Price Index (CPI) to 2003 real dollars. Occupation
is a dummy variable where white-collar worker is the omitted category and blue-collar
workers are respondents who classified themselves in any of the following occupational
categories: (1) farming, forestry, or fishing; (2) mechanics or repair; (3) precision
production; (4) operators (machine, transportation, or handlers); and (5) armed forces.
Analytic Strategy
The panel nature of the HRS is extremely valuable for a study on the effects of the
work-retirement transition on happiness. Most of the research on happiness in old age
cited above uses cross-sectional designs, which can raise serious concerns about the
direction of causation (for a methodological discussion, see Charles 2004). This study
takes advantage of the longitudinal nature of the HRS to test whether the type of
transition out of employment affects an individual‟s happiness in retirement. We do this
by establishing a baseline level of happiness for all individuals when they were
employed. We then compare that baseline to their happiness when retired. By focusing
on the change in happiness, we expect to minimize the effect of genetic or personality
based inter-individual differences that may be causing individuals to self-select into one
14
specific type of retirement transition.
2
By contrast, cross-sectional studies using the level
of happiness could merely identify differences between „happy‟ and „unhappy‟
individuals, not what changes the happiness of individuals.
To explain changes in happiness we use two regression specifications. The first
set of regressions uses as the dependent variable the change in each of the five HRS
variables from the last wave when the individual fully employed to the first wave when
the individual was fully retired. As these changes can take on three possible values (-1,
0, or 1), we use ordered logistic regressions.
The second specification divides our sample, for each of the five HRS variables,
into those individuals that are initially “happy” and those that are initially “unhappy” and
use the change when retired as the dependent variable in a logistic regression. The
positive (“initially happy”) sample gives a clearer view of factors that tend to diminish
“Happiness” and the negative (“initially unhappy”) sample gives a clearer view of factors
that tend to increase “Happiness.” The logistic regression coefficients provide a better
view of the magnitude of the effects of the explanatory factors on the happiness
indicators.
However, caution should be taken when comparing the magnitude of the effect
of an explanatory variable between the two samples. Since these subsamples allow for a
different distribution for happiness for each of the two groups, the magnitudes of the
coefficients are not directly comparable.
III. RESULTS
Descriptive Results
2
While there may be concern that the two sets of retirement-related variables measuring the sense of
control and the type of retirement are collinear, this does not appear to be the case in this sample. The
correlations between each of the „perceived control over retirement‟ variableschose to retire, partly
forced to retire, and forced to retireand abrupt retirement are 0.05, 0.01, and -0.06 in the raw data.
15
Table 1 reports descriptive statistics for the five happiness measures for the
overall sample and for the gradual retirement and cold turkey groups, as well as
significance tests for differences between these two groups. It shows that there is little
change in the positive affect variables. Both “happiness” and “enjoyment of life” are
high when individuals are working, and show slight increases in retirement. The negative
affects are generally low, but also increase in retirement. The increases in the negative
affects, in feelings of loneliness, depression, and sadness, are also generally larger than
the increases in the positive affects.
The descriptive statistics, however, generally show no significant difference
between the group that retired gradually and the group that retired cold turkey. The one
indicator that did register a significant difference was sadness. Workers who retired
gradually registered a much larger increase in feelings of sadness.
[TABLE 1]
As shown in Table 2, there are various differences between individuals who
retired gradually and those who retired cold turkey. More cold turkey individuals report
a high degree of control over their retirement transition and say that they wanted to retire.
They were younger, healthier, wealthier, more educated, more likely to be white collar,
more likely to have a defined benefit pension, and less likely to report unemployment at
the same time that they report being fully retired. All of these differences should make
the cold turkey individuals happier than gradual retirees, independent of the way that they
retired.
[TABLE 2]
Regression Results
16
The results of the ordered logistic regressions for the change in our five happiness
measures are reported in Table 3 (coefficient estimates for all independent variables are
available upon request). The results indicate that cold turkey retirement (as opposed to
the default, gradual retirement) has no significant effect on happiness, enjoyment of life,
loneliness, depression, or sadness in retirement. The results also confirm earlier findings
on the effect of other factors reported in the literature on the effects of the death of a
spouse, voluntary as opposed to forced retirement, and health status. Because of the
difficulty in interpreting ordered logit coefficients, we focus on the direction and
statistical significance of the effect, not on the size of the coefficient.
The death of a spouse had a significant impact on all indicators of happiness
except depression. Those who lost their spouse are more likely to have a decrease in
happiness and enjoyment of life and are more likely to have a decrease in feelings of
loneliness and sadness.
Having control over the retirement decision (reporting that retirement was chosen
rather than forced) has a significant effect on all happiness indicators except enjoyment
of life, and the direction of the coefficient on enjoyment of life is consistent with the
other four measures. Respondents who said retirement was chosen were more likely to
have increases in happiness and enjoyment of life and less likely to have increases in
loneliness, depression, and sadness, than if their retirement was forced. Even individuals
who only felt partly forced into retirement were less likely to have an increase in
loneliness.
A change in self-reported health status has a significant effect on the change in
each of the five happiness indicators. An improvement in health is associated with
increased happiness and enjoyment of life, and decreased loneliness, depression, and
sadness. All of the coefficients for the change in health status are significant at the 1
percent level.
3
3
The coefficients on the health variable are generally smaller than the coefficients on control over the
retirement decision or the loss of a spouse. To gauge the indicated effect of health on happiness, however,
17
In contrast to previous studies, we find that wealth and having a defined benefit
pension plan have no significant effect on happiness. We attribute this difference to the
fact that we use a longitudinal design and focus on the change in happiness, while earlier
studies used a cross-sectional design to predict the level of retirement satisfaction. Earlier
studies also differ in that they used “satisfaction with retirementas the dependent
variable (Bender 2004; Panis 2003 Rohwedder 2006). This measure is not available at
our baseline, as all individuals are then fully-employed.
[TABLE 3]
The results of the logistic regressions on the divided sample are shown in Tables 4
and 5 (coefficient estimates for all independent variables are available upon request.)
Table 4 shows the results from the “positive sample” (where the value of the variable
when the individual was fully employed was “happy”) and Table 5 shows the results of
the “negative sample” (where the value of the variable when the individual was fully
employed was “unhappy”). In both samples, the type of transition to retirement had no
significant effect. The small size of the negative sample limits the power of analysis for
that sample. Nevertheless, the factors identified as significant in the other models are
generally significant in the positive sample models.
These logistic regression models provide a much clearer indication of the
magnitude of the effects of the various factors. The coefficients reflect the difference in
the probability that the value of the indicator variable will change (as opposed to
remaining the same) with one unit increase in the independent variable, holding all other
variables at their means. In the positive sample, where more of the results are statistically
one needs to multiply this coefficient by the change in self-reported health status, measured on a scale from
-4 to 4. As the mean change in health is -2, with a standard deviation of 1, indicated effect of a change in
health status is relatively small, relative to the effect of the loss of a spouse or even control over the
retirement decision.
18
significant, the coefficient for the loss of a spouse generally has the largest effect on the
happiness indicators. These results also show control over one‟s retirement to have a
large impact on happiness; in the case of the depression indicator, the effect is even larger
(in absolute value) than the loss of a spouse. Control over one‟s retirement is also the
only independent variable to have a statistically significant effect on all five indicator
variables in the positive sample.
[TABLES 4 AND 5]
IV. DISCUSSION AND CONCLUSION
Retirement is a transition between two significantly different stages in an
individual‟s life. A gradual transition gives workers time to shift their daily activities,
social relationships, and identity in a more deliberate manner than a cold turkey
transition. Previous theory and research suggests that this may help workers to make a
better transition to retirement.
Our study, however, finds no evidence of a difference in happiness that can be
traced to the type of transition to retirement. Although many workers see gradual
retirement as an attractive idea for the future, in practice this alternative seems to be less
stimulating than anticipated. Once we are ready to do the transition to retirement, we do
not seem to need much time to adjust to our new status
While the nature of our transitiongradual or abrupthas no effect on our
happiness in retirement, the sense of control workers have over the transition does have a
significant effect. Although our measure of sense of control involved the choice of
retiring or not retiring, not the choice of the mode of the transition, it is possible to
speculate about what would happen if workers had more choice about whether to retire
gradually or abruptly. Among the current barriers to retire gradually, we find: laws about
pension entitlement, expectations about the „normal‟ retirement age, and age
discrimination (Hutchens and Papps 2005). If the option to retire gradually were readily
19
available, people who choose this path should experience greater emotional rewards, as
they would be exercising greater control over the retirement process. But it would be the
ability to retire gradually if they want tonot the effect of the gradual transition per se
that would make people happier in old age.
Since we observe individuals every two years and focus our attention on
happiness in full retirement, future research will explore change in happiness within
shorter periods of time and explore happiness during the transition to retirement,
especially during the period of partial retirement. Future research will also attempt to
identify a latent variable driving all five happiness indicators, which would allow changes
in happiness to be measured on a continuous scale. This should introduce significantly
more variation into our measure of happiness and reduce the censoring problem created
by the dichotomous nature of the HRS happiness variables.
The current findings, however, can be used to inform both workers preparing for
retirement and policymakers interested in retirement issues. We provide evidence
suggesting that giving workers a sense of control over their retirement, not necessarily
creating gradual retirement paths, should have a more important place on the policy
agenda. We provide evidence that gradual retirement has no effect on happiness once
retired. But we also provide evidence that giving workers more control over the
retirement processperhaps by creating gradual retirement pathwaysshould have a
more important place on the policy agenda.
20
REFERENCES
Allison, Paul D. 2001. Missing Data Quantitative Applications in the Social Sciences,
07-136. Thousand Oaks, CA: Sage Publications.
Appley, Mortimer H., ed. 1971. Adaptation-level Theory. New York: Academic.
Arendt, Jacob N. 2005. “Income and "Outcomes" for Elderly: Do the Poor Have a Poorer
Life? Social Indicators Research 70(3):327-47.
Arthaud-day, Marne and Janet Near. 2005. The Wealth of Nations and the Happiness of
Nations: Why "Accounting" Matters. Social Indicators Research 74(3):511-48.
Atchley, Robert C. 1999. Continuity and Adaptation in Aging: Creating Positive
Experiences. Baltimore, MD: Johns Hopkins University.
Bender, Keith A. 2004. “The Well-being of Retirees: Evidence Using Subjective Data.”
Working Paper No. 2004-24, Center for Retirement Research at Boston College,
Chestnut Hill, MA.
Bierman, Alex, Elena M. Fazio and Melissa A. Milkie. 2006. “A Multifaceted Approach
to the Mental Health Advantage of the Married: Assessing How Explanations Vary
by Outcome Measure and Unmarried Group. Journal of Family Issues 27(4):554-
82.
Brown, Kathi. 2003. Staying Ahead of the Curve 2003: The AARP Working in
Retirement Study. AARP, Washington, DC.
Brummett, Beverly H., John C. Barefoot, Ilene C. Siegler, Nancy E. Clapp-Channing,
Barbara L. Lytle, Hayden B. Bosworth, Williams, Redford B. Jr. and Daniel B.
Mark. 2001. Characteristics of Socially Isolated Patients with Coronary Artery
Disease Who Are at Elevated Risk of Mortality. Psychosomatic Medicine
63(2):267-72.
Calvo, Esteban. 2006. “Does Working Longer Make People Healthier and Happier?”
Issue in Brief No. WOB#2, Center for Retirement Research at Boston College,
Chestnut Hill, MA.
21
Chan, Ying and Rance Lee. 2006. “Network Size, Social Support and Happiness in Later
Life: A Comparative Study of Beijing and Hong Kong. Journal of Happiness
Studies 7(1):87-112.
Charles, Kerwin K. 2004. “Is Retirement Depressing? Labor Force Inactivity and
Psychological Well-being in Later Life. Pp. 269-99 in Accounting for Worker
Well-being, edited by S.W. Polachek. San Diego, CA: Elsevier.
Chen, Yung-Ping and John C. Scott. 2006. “Phased Retirement: Who Opts For It and
Toward What End?” Issue 2006-01, AARP, Washington, DC.
Cheng, Sheung-Tak. 2004. "Age and Subjective Well-being Revisited: A Discrepancy
Perspective." Psychology and Aging 19(3):409-15.
Cheng, Sheung-Tak and Alfred C. M. Chan. 2006. Relationship with Others and Life
Satisfaction in Later Life: Do Gender and Widowhood Make a Difference?”
Journals of Gerontology: Psychological Sciences 61(1):46-53.
Cohen, Sheldon. 2004. “Social Relationship and Health. American Psychologist
59(8):676-84.
Costa, Paul T., Robert R. McCrae and Arthur H. Norris. 1981. Personal Adjustment to
Aging: Longitudinal Prediction from Neuroticism and Extraversion. Journals of
Gerontology 36(1):78-85.
Demo, David H. and Alan C. Acock. 1996. “Singlehood, Marriage, and Remarriage: The
Effects of Family Structure and Family Relationships on Mothers' Well-being.
Journal of Family Issues 17(3):388-407.
De Vaus, David, Yvonne Wells, Hal Kendig and Susan Quine. 2007. Does Gradual
Retirement Have Better Outcomes Than Abrupt Retirement? Results from an
Australian Panel Study. Ageing and Society 27(5):667-82.
Diener, Ed and Robert Biswas-Diener. 2002. “Will Money Increase Subjective Well-
being? Social Indicators Research 57(2):119-69.
Diener, Ed, Eunkook M. Suh, Richard E. Lucas and Heidi L. Smith. 1999. Subjective
Well-being: Three Decades of Progress. Psychological Bulletin 125(2):276-302.
22
Easterlin, Richard A. [1974] 2002. “Does Economic Growth Improve the Human Lot?
Some Empirical Evidence. Pp. 5-41 in Happiness in Economics, edited by R.A.
Easterlin. Northampton, MA: Elgar.
Erikson, Erik H., Joan M. Erikson and Helen Q. Kivnick. 1986. Vital Involvement in Old
Age. New York: Norton.
Everard, Kelly M., Helen W. Lach, Edwin B. Fisher and Carolyn M. Baum. 2000.
Relationship of Activity and Social Supports to the Functional Health of Older
Adults. Journals of Gerontology: Social Sciences 55(4):208-12.
Frey, Bruno S. and Alois Stutzer. 2002. What Can Economists Learn from Happiness
Research? Journal of Economic Literature 40(2):402-35.
Gall, Terry L., David R. Evans and John Howard. 1997. The Retirement Adjustment
Process: Changes in the Well-being of Male Retirees across Time. Journals of
Gerontology: Psychological Sciences 52(3):110-7.
Gallo, William T., Elizabeth H. Bradley, Joel A. Dubin, Richard N. Jones, Tracy A.
Falba, Hsun-Min Teng and Stanislav V. Kasl. 2006. “The Persistence of Depressive
Symptoms in Older Workers Who Experience Involuntary Job Loss: Results from
the Health and Retirement Survey. The Journals of Gerontology: Social Sciences
61(4):221-8.
Gilbert, Daniel. 2007. Stumbling on Happiness. New York: Vintage Books.
Glass, Thomas A., Mendes De Leon,Carlos F., Shari S. Bassuk and Lisa F. Berkman.
2006. Social Engagement and Depressive Symptoms in Late Life: Longitudinal
Findings. Journal of Aging and Health 18(4):604-28.
Gustman, Alan L. and Thomas L. Steinmeier. 2007. “Projecting Behavioral Responses to
the Next Generation of Retirement Policies.” Working Paper No. 12958, National
Bureau of Economic Research, Cambridge, MA.
------. 2000. “Retirement Outcomes in the Health and Retirement Study.” Working Paper
No. 7588, National Bureau of Economic Research, Cambridge, MA.
------. 1984. "Partial Retirement and the Analysis of Retirement Behavior." Industrial and
Labor Relations Review 37(April):403-15.
23
Haider, Steven and David Loughran. 2001. “Elderly Labor Supply: Work or Play?”
Working Paper No. 2001-04, Center for Retirement Research at Boston College,
Chestnut Hill, MA.
Hilleras, Pernilla K., Hedda Aguero-Torres and Bengt Winblad. 2001. “Factors
Influencing Well-being in the Elderly. Current Opinion in Psychiatry 14(4):361-5.
Honig, Marjorie and Giora Hanoch. 1985. Partial Retirement as a Separate Mode of
Retirement Behavior. The Journal of Human Resources 20(1):21-46.
Hutchens, Robert and Kerry L. Papps. 2005. Developments in Phased Retirement. Pp.
133-160 in Reinventing the Retirement Paradigm, edited by R.L. Clark and O.S.
Mitchell. New York: Oxford University.
Inglehart, Ronald and Hans-Dieter Klingemann. 2000. “Genes, Culture, Democracy, and
Happiness. Pp. 165-83 in Culture and Subjective Well-being., edited by E. Diener
and E.M. Suh., Cambridge, MA: MIT.
Jorm, A. F. 2000. “Does Old Age Reduce the Risk of Anxiety and Depression? A Review
of Epidemiological Studies across the Adult Life Span. Psychological Medicine
30(1):11-22.
Kosloski, Karl, Donald E. Stull, Kyle Kercher and Daniel J. Van Dussen. 2005.
Longitudinal Analysis of the Reciprocal Effects of Self-assessed Global Health
and Depressive Symptoms. The Journals of Gerontology: Psychological Sciences
60(6):296-303.
Kunzmann, Ute, Todd D. Little and Jacqui Smith. 2000. Is Age-related Stability of
Subjective Well-being a Paradox? Cross-sectional and Longitudinal Evidence From
the Berlin Aging Study. Psychology and Aging 15(3):511-26.
Kunzmann, Ute, Todd Little and Jacqui Smith. 2002. Perceiving Control: A Double-
edged Sword in Old Age. Journals of Gerontology: Psychological Sciences
57(6):484-91.
Layard, Richard. 2005. Happiness. Lessons from a New Science. New York: Penguin.
Lucas, Richard E., Andrew E. Clark, Yannis Georgellis and Ed Diener. 2003.
Reexamining Adaptation and the Set Point Model of Happiness: Reactions to
24
Changes in Marital Status. Journal of Personality and Social Psychology
84(3):527-39.
Lykken, David and Auke Tellegen. 1996. “Happiness is a Stochastic Phenomenon.
Psychological Science 7(3):186-89.
Michalos, Alex C. 1985. Multiple Discrepancies Theory (MDT). Social Indicators
Research 16(4):347-413.
Panis, Constantijn. 2003. Annuities and Retirement Satisfaction. Working Paper No.
03-17 DRU 3021, RAND, Santa Monica, CA.
Perreira, Krista M., Natalia Deeb-Sossa, Kathleen M. Harris and Kenneth Bollen. 2005.
What Are We Measuring? An Evaluation of the CES-D across Race/Ethnicity and
Immigrant Generation. Social Forces 83(4):1567-602.
Pinquart, Martin. 2003. Loneliness in Married, Widowed, Divorced, and Never-married
Older Adults. Journal of Social and Personal Relationships 20(1):31-53.
------. 2001. Age Differences in Perceived Positive Affect, Negative Affect, and Affect
Balance in Middle and Old Age. Journal of Happiness Studies 2(4):375-405.
RAND Center for the Study of Aging. 2007. RAND HRS Data, Version G. Retrieved
May 2007 (http://hrsonline.isr.umich.edu).
Rodin, Judith. 1986. Aging and Health: Effects of the Sense of Control. Science
233(4770):1271-76.
Rohwedder, Susann. 2006. “Self-assessed Retirement Outcomes: Determinants and
Pathways.” Working Paper No. 2006-141, Michigan Retirement Research Center at
University of Michigan, Ann Arbor, MI.
Rowe, John W. and Robert L. Kahn. 1998. Successful Aging. New York: Pantheon.
Royston, Patrick. 2004. Multiple Imputation of Missing Values. Stata Journal
4(3):227-41.
Ruhm, Christopher J. 1990. “Bridge Jobs and Partial Retirement. Journal of Labor
Economics 8(4):482-501.
25
Saris, Willem E. 2001. “The Relationship between Income and Satisfaction: The Effect
of Measurement Error and Suppressor Variables. Social Indicators Research
53(2):117-36.
Siegrist, Johannes, Olaf Von dem Knesebeck and Craig E. Pollack. 2004. Social
Productivity and Well-being of Older People: A Sociological Exploration. Social
Theory & Health 2(1):1-17.
Steffick, Diane E. 2000. Documentation of Affective Functioning Measures in the
Health and Retirement Study HRS/AHEAD Documentation Report No. DR-005,
Institute for Social Research at University of Michigan, Ann Arbor, MI.
Sweeney, Paul D., Karen Anderson and Scott Bailey. 1986. Attributional Style in
Depression: A Meta-analytic Review. Journal of Personality and Social
Psychology 50(5):974-91.
University of Michigan. 2007. Health and Retirement Study, 1992-2004.” Retrieved
May 2007 (http://hrsonline.isr.umich.edu).
Vanderhorst, Rob K. and Suzanne McLaren. 2005. "Social Relationships as Predictors of
Depression and Suicidal Ideation in Older Adults." Aging & Mental Health
9(6):517-25.
Veenhoven, Ruut. 2006. Happiness in Nations. Introductory Text. Rotterdam,
Netherlands: World Database of Happiness, Erasmus University Rotterdam.
Retrieved June 29, 2006
(http://www1.eur.nl/fsw/happiness/hap_nat/introtexts/intronat-contents.html).
------. 1991. “Is Happiness Relative? Social Indicators Research 24(1):1-34.
Watson Wyatt. 2004. Phased Retirement Aligning Employer Programs with Worker
Preferences. Watson Wyatt Worldwide, Washington, DC.
26
BIOGRAPHICAL NOTE
ESTEBAN CALVO is a doctoral student in Sociology and graduate research assistant at the
Center for Retirement Research at Boston College. His research interests sociology of
aging and the life-course, comparative public policy, health and mental health, and social
security reforms around the world. Much of his work aims to understand what factors
influence the happiness and health of older adults, and to evaluate policies that could
contribute to improved psychological and physical well-being. Mr. Calvo, with John B.
Williamson, are the authors of “Old-Age Pension Reform and Modernization Pathways:
Lessons for China from Latin America (Journal of Aging Studies).
KELLY HAVERSTICK is a Research Economist who joined the Center for Retirement
Research in June, 2006. Ms. Haverstick earned her doctorate in economics from Boston
College in 2004. Her research interests are in labor economics and econometrics. Ms.
Haverstick, with Alicia H. Munnell and Geoffrey Sanzenbacher, wroteJob Tenure and
Pension Coverage” (Center for Retirement Research WP#2006-18). She is currently
involved in a major research initiative to study state and local pension plans for which
she has coauthored several short publications.
STEVEN A. SASS is Associate Director of the Center for Retirement Research at Boston
College. He is the author of The Promise of Private Pensions: The First Hundred Years
(Harvard University Press, 1997) and, with Alicia Munnell, Social Security and the Stock
Market (Upjohn, 2006) and Working Longer (Brookings, 2008), and, with Wei Sun and
Anthony Webb, "Why Do Married Men Claim Social Security Benefits So Early:
Ignorance, Caddishness, or Something Else?" Sass earned his B.A. from the University of
Delaware and his Ph.D. from The Johns Hopkins University.
27
Table 1. Descriptive Statistics for Five Happiness Indicators and Latent Affect by Type of Retirement Transition
All
Phased
Cold Turkey
(N=2,389)
(N=656)
(N=1,733)
Variable
Metric
Mea
n
Std.
Dev.
Mean
Std.
Dev.
Mean
Std.
Dev.
Happiness
t1
0=no; 1=yes
0.87
4
0.331
0.861
0.346
0.879
0.326
Happiness
t2
0=no; 1=yes
0.88
3
0.321
0.875
0.331
0.886
0.318
Change in Happiness
-1=less; 0=equal; 1=more
0.00
9
0.404
0.014
0.408
0.007
0.402
Enjoyment of Life
t1
0=no; 1=yes
0.93
0
0.256
0.916
0.277
0.935
0.247
Enjoyment of Life
t2
0=no; 1=yes
0.93
3
0.251
0.918
0.275
0.938
0.241
Change in Enjoyment of Life
-1=less; 0=equal; 1=more
0.00
3
0.312
0.002
0.343
0.003
0.300
Loneliness
t1
0=no; 1=yes
0.10
7
0.309
0.107
0.309
0.107
0.309
28
Loneliness
t2
0=no; 1=yes
0.15
1
0.358
0.152
0.360
0.150
0.357
Change in Loneliness
-1=less; 0=equal; 1=more
0.04
4
0.394
0.046
0.384
0.043
0.398
Depression
t1
0=no; 1=yes
0.11
8
0.323
0.120
0.326
0.118
0.322
Depression
t2
0=no; 1=yes
0.15
4
0.361
0.165
0.371
0.149
0.357
Change in Depression
-1=less; 0=equal; 1=more
0.03
5
0.413
0.044
0.413
0.032
0.413
Sadness
t1
0=no; 1=yes
0.12
1
0.327
0.093
**
0.291
0.132
**
0.339
Sadness
t2
0=no; 1=yes
0.17
2
0.378
0.184
0.388
0.168
0.374
Change in Sadness
-1=less; 0=equal; 1=more
0.05
1
0.436
0.091
**
0.436
0.036
**
0.435
Notes:
t1
= last wave of full employment;
t2
= first wave of full retirement.
* p < .05; ** p < .01; *** p < .001 (two tailed t tests and chi-squared tests reported, denoting statistically significant differences
between the two types of transition groups.)
29
Table 2. Descriptive Statistics for the Independent Variables by Type of Retirement Transition
All
Phased
Cold Turkey
(N=2,389)
(N=656)
(N=1,733)
Variable
Metric
Mean
Std.
Dev.
Mean
Std.
Dev.
Mean
Std.
Dev.
Retirement Wanted
0=no; 1=yes
0.643**
*
0.479
0.581
0.494
0.667
0.471
Retirement Partly
Wanted/Forced
0=no; 1=yes
0.089
0.284
0.079
0.270
0.092
0.290
Retirement Forced
0=no; 1=yes
0.268**
*
0.443
0.340
0.474
0.241
0.428
Spouse Death
0=no; 1=yes
0.023*
0.151
0.034
0.180
0.020
0.139
Improved Health
-4 (deteriorated) to 4
(improved)
-0.222**
1.003
-0.320
1.027
-0.185
0.991
Any Defined Benefit
0=no; 1=yes
0.517**
*
0.500
0.375
0.484
0.570
0.495
Age
t1
50 to 71 years
59.56**
*
3.338
58.73
3.392
59.88
3.264
Age
t2
52 to 73 years
62.58**
*
3.417
64.02
3.434
62.04
3.250
30
Years Employment-Retirement
1 to 12 years
3.019**
*
1.849
5.284
1.884
2.162
0.834
Male
0=no; 1=yes
0.494
0.500
0.492
0.500
0.495
0.500
White Non-Hispanic
0=no; 1=yes
0.316
0.465
0.294
0.456
0.324
0.468
Log of Mean Wealth
0 to 18.765 (real $USD, 2003)
10.549*
3.180
10.28
4
3.593
10.64
9
3.004
More than High School
0=no; 1=yes
0.390*
0.488
0.378
0.485
0.395
0.489
Blue-Collar
0=no; 1=yes
0.299
0.458
0.305
0.461
0.297
0.457
Unemployed
0=no; 1=yes
0.005
0.068
0.003
0.055
0.005
0.072
Notes:
t1
= last wave of full employment;
t2
= first wave of full retirement. These summary statistics include both recorded and
imputed values.
* p < .05; ** p < .01; *** p < .001 (two tailed t tests and chi-squared tests reported, denoting statistically significant differences
between the two types of transition groups.)
31
Table 3. Ordered Logit Results on Five Indicator Variables (N=2,389)
Enjoy
Happy
Life
Lonely
Depressed
Sad
Cold Turkey Retirement
0.086
0.262
-0.058
0.236
-0.101
(0.19)
(0.23)
(0.19)
(0.18)
(0.17)
Retirement Wanted
0.481**
0.250
-
0.633***
-0.470**
-0.443**
(0.16)
(0.18)
(0.14)
(0.15)
(0.14)
Retirement Partly Wanted
0.095
0.263
-
0.917***
-0.150
-0.262
(0.25)
(0.29)
(0.24)
(0.22)
(0.22)
Spouse Death
-
1.110**
-
1.462***
2.493***
0.235
0.707*
(0.34)
(0.38)
(0.30)
(0.35)
(0.31)
Improved Health
0.262**
*
0.354***
-0.160**
-0.282***
-
0.231***
(0.06)
(0.07)
(0.06)
(0.05)
(0.05)
Any Defined Benefit
Pension
-0.037
0.016
0.128
0.059
0.042
(0.12)
(0.15)
(0.12)
(0.12)
(0.11)
Log Likelihood
-1295
-895
-1223
-1334
-1440
Note: Change from the last wave of full employment to the first wave of full
retirement. Regular ordered logit regression coefficients reported for the indicator
variables. The standard errors are in parentheses. All five models control for
demographic and socioeconomic variables.
* p < .05; ** p < .01; *** p < .001 (two tailed tests for all variables).
32
Table 4. Logit Marginal Effects on Five Indicator Variables for Positive Sample
Happy
Enjoy Life
Not-
Lonely
Not-
Depressed
Not-Sad
to
to
to
to
to
Not-
Happy
Not-Enjoy
Life
Lonely
Depressed
Sad
Cold Turkey Retirement
0.010
-0.013
0.010
0.027
0.01
(0.02)
(0.01)
(0.02)
(0.02)
(0.02)
Retirement Wanted
-
0.087***
-0.047***
-0.084***
-0.087***
-
0.086**
*
(0.02)
(0.01)
(0.02)
(0.02)
(0.02)
Retirement Partly Wanted
-0.020
-0.024**
-0.046**
-0.022
-0.030
(0.01)
(0.01)
(0.02)
(0.02)
(0.02)
Spouse Death
0.135*
0.125*
0.478***
0.046
0.117*
(0.06)
(0.05)
(0.08)
(0.04)
(0.06)
Improved Health
-
0.017***
-0.014***
-0.010
-0.024***
-
0.025**
*
(0.00)
(0.00)
(0.01)
(0.01)
(0.01)
Any Defined Benefit
Pension
0.028*
0.012
0.004
0.013
0.019
(0.01)
(0.01)
(0.01)
(0.01)
(0.01)
Log Likelihood
-665.1
-678.6
-774.6
-678.6
-774.6
N
2089
2221
2134
2106
2099
Note: The "positive sample" includes only individuals starting with positive affects at the last
wave of full employment. The coefficients reflect the difference in the probability of a negative
change in each indicator variable (as opposed to no change) from the last wave of full
employment to the first wave of full retirement, with one unit increase in the independent
variable and holding all variables held at their means. The standard errors are in parentheses.
All models control for demographic and socioeconomic variables.
33
* p < .05; ** p < .01; *** p < .001 (two tailed tests for all variables).
34
Table 5. Logit Marginal Effects on Five Indicator Variables for Negative Sample
Not-
Happy
Not-Enjoy
Life
Lonely
Depressed
Sad
to
to
to
to
to
Happy
Enjoy Life
Not-
Lonely
Not-Depressed
Not-
Sad
Cold Turkey Retirement
0.009
0.090
0.191
-0.034
0.098
(0.09)
(0.11)
(0.12)
(0.11)
(0.13)
Retirement Wanted
0.227***
0.258***
0.160
0.262***
0.240*
**
(0.06)
(0.08)
(0.08)
(0.07)
(0.07)
Retirement Partly Wanted
0.115
0.047
0.284*
0.201*
0.225*
(0.09)
(0.12)
(0.11)
(0.10)
(0.09)
Spouse Death
-0.242
-0.350
0.055
0.109
-0.286
(0.19)
(0.23)
(0.24)
(0.17)
(0.17)
Improved Health
0.041
0.032
0.086*
0.049
0.028
(0.03)
(0.03)
(0.04)
(0.03)
(0.03)
Any Defined Benefit
Pension
-0.044
0.118
0.039
0.059
0.092
(0.06)
(0.08)
(0.07)
(0.07)
(0.07)
Log Likelihood
-170.6
-84.4
-158.9
-160.9
-174.9
N
300
168
255
283
290
Note: The "negative sample" includes only individuals starting with negative affects at the last
wave of full employment. The coefficients reflect the difference in the probability of a negative
change in each indicator variable (as opposed to no change) from the last wave of full
employment to the first wave of full retirement, with one unit increase in the independent
variable and holding all variables held at their means. The standard errors are in parentheses.
All models control for demographic and socioeconomic variables.
* p < .05; ** p < .01; *** p < .001 (two tailed tests for all variables).
35
... Studies on retirement adjustment are often based on data from a narrow time period and thus are only representative with respect to that specific period (Calvo et al., 2009;Henning et al., 2021;Hershey & Henkens, 2014). Other studies included transitions taking place over longer time spans in long-time longitudinal studies such as the German Socioeconomic Panel (Pinquart & Schindler, 2007; or the Health and Retirement Survey (Calvo et al., 2012). ...
... Because of the later retirement age and the reduced options to retire early, in later samples, an increasing number of older adults may either lack the financial resources to retire when they want, or lack the physical resources to actually continue working, which directly affects the freedom of choice between work and retirement (Hofäcker & Radl, 2016). This may be problematic as control over retirement plays an important role in retirement adjustment (Calvo et al., 2009). Retirement adjustment may therefore have become more difficult for individuals. ...
... Only differences between groups in retirement satisfaction seem to have increased between 1996 and 2008. One reason for the increasing gap between blue-and white-collar workers that we presented in the introduction may be the historical decrease in control over the timing of retirement for blue-collar workers (Hess, 2018), given the important role of control on retirement adjustment (Calvo et al., 2009;Hershey & Henkens, 2014). Thus, changes in the pension system may have contributed to a higher social inequality in retirement. ...
Full-text available
Article
The context of retirement has changed over the last decades, but there is little knowledge on whether the quality of retirement adjustment has changed as well. Changes in retirement regulations and historical differences in resources may affect the quality of adjustment and increase inequalities between different socioeconomic groups. In the present study, we investigated historical differences in retirement adjustment by comparing cross-sectional samples of retirees from 1996, 2002, 2008, and 2014, based on the population-based German Ageing Survey. Adjustment was measured with three different indicators (perceived change in life after retirement, retirement satisfaction, adjustment difficulties). Retirement satisfaction was higher in later samples, but for the other two outcomes, there was no evidence for systematic increases or decreases in levels of retirement adjustment with historical time over the studied period. White-collar workers reported better adjustment than blue-collar workers did, and for two of three outcomes, this effect was stable over time. The white-collar workers’ advantage concerning retirement satisfaction, however, increased. We conclude that in Germany, at least for those who retire within the usual time window, adjustment quality has not changed systematically over the examined 18-year period. We only found mixed evidence for a growing social inequality in the retirement adjustment. However, as individual agency in choosing one’s retirement timing and pathway is increasingly restricted, social inequalities in well-being before retirement may increase.
... A much larger body of literature examines the question of how parttime employment transitions in late careers affect workers' health status into and after retirement. Except one Australian study (de Vaus et al., 2007), all studies are based on the HRS (Calvo et al., 2009;Dave et al., 2008;McDonough et al., 2017;Zhan et al., 2009). While one study found no health differences between those who had fully and partially retired (Calvo et al., 2009), all other studies found that part-time employment in late careers was associated with better health outcomes than full retirement. ...
... Except one Australian study (de Vaus et al., 2007), all studies are based on the HRS (Calvo et al., 2009;Dave et al., 2008;McDonough et al., 2017;Zhan et al., 2009). While one study found no health differences between those who had fully and partially retired (Calvo et al., 2009), all other studies found that part-time employment in late careers was associated with better health outcomes than full retirement. For instance, Dave et al. (2008) studied the causal effects of entering retirement on health outcomes. ...
Full-text available
Article
In this exploratory study, we examine how older workers' part-time employment and health are associated in four countries promoting this type of employment in late careers but with a different welfare regime: the United States, Germany, Sweden, and Italy. Using data from two large representative panel surveys and conducting multichannel sequence analysis, we identified the most typical interlocked employment and health trajectories for each welfare regime and for three different age groups of women and men. We found that there is more heterogeneity in these trajectories in countries with a liberal welfare regime and among older age groups. Overall, women are more strongly represented in the part-time employment trajectories associated with lower health levels. In countries with a social-democratic or corporatist welfare regime, part-time employment in late careers tends to be associated with good health. Our findings suggest that the combination of a statutory right to work part-time in late careers with a more generous welfare regimes, may simultaneously maintain workers’ health and motivate them to remain active in the labor force.
... Nevertheless, a majority of studies have found retirement to improve mental health (Sharpley & Layton, 1998;Van der Heide et al., 2013). All in all, this literature suggests that the impact of retirement depends greatly on situational factors, such as the characteristics of the job, control over the transition into retirement, and personal characteristics, such as one's health before retirement (Calvo et al., 2009;Henning et al., 2016). ...
... There is some research on flexible retirement arrangements. This literature particularly focuses on the impact of flexible retirement trajectories on post-retirement well-being and mental health (Calvo et al., 2009;De Vaus et al., 2007). Despite its intuitive appeal among public policy makers, the impact of phased retirement arrangements on health outcomes of older workers who are still in the workforce, remains unexplored. ...
Full-text available
Article
This study investigates the effects of phased retirement on vitality and how this effect differs for workers dealing with work, family and health strain and low levels of baseline vitality. We used two waves of the NIDI Pension Panel Survey, conducted in 2015 and 2018, in the Netherlands. Data from 1,247 older workers, of whom 10% opted for phased retirement, were analyzed. Vitality is assessed in three ways: a composite measure of vitality, and its subcomponents energy and fatigue. Conditional change regression models demonstrated that transitioning into phased retirement improved vitality and energy levels and reduced fatigue. Older workers with low energy levels at baseline showed greater improvements in energy after using phased retirement: this result was not evident for those with low vitality and high fatigue at baseline. Phased retirement improved vitality for workers with high work strain. Vitality for workers with family or health strain was not improved. Interestingly, the positive effects of phased retirement were equally visible among workers with and without adverse health conditions and caregiving responsibilities. Our study provides evidence on the benefits of phased retirement as a method to sustainably ensure healthy aging of not only vulnerable but all older workers.
... Specifying the predictors of retirement adjustment has been the subject of numerous examinations and qualitative reviews in the last few decades (e.g., Asebedo & Seay, 2014;Bender, 2012;Bonsang & Klein, 2012;Calvo et al., 2009;Reitzes et al., 1996;Reitzes & Mutran, 2004;Yeung & Zhou, 2017). The first qualitative review was provided by Beehr (1986), and it took 25 years for an updated review to appear (Wang & Shultz, 2010). ...
Full-text available
Article
While most people experience a positive transition to retirement, as many as one third of the population find the transition challenging. Previous research has identified a number of factors that predict adjustment outcomes – with finances, physical health, marital relationship, wider social participation, and exit conditions identified as being particularly key. This study aimed to examine their relative contribution to retirement adjustment by assessing the magnitude of the associations between each key predictor category and retirement adjustment outcomes, as well as to examine potential important moderating factors. A three-level meta-analysis (based on 915 effect sizes, k = 139, N = 78,632) revealed that social participation had the strongest positive association with adjustment (r = 0.23), followed by physical health (r = 0.22), marital relationship (r = 0.18), finances (r = 0.17) and exit conditions (r = 0.15), respectively. Additional analyses revealed substantial variation within each category (with effect sizes ranging from r = −0.03 to r = 0.43), suggesting that there is value in future research and theory to recognise substantive theoretical and empirical differences in defining retirement predictors. Less physical health symptoms and ease of maintaining social relationships were identified as the most important subfactors for successful adjustment. We discuss theoretical and practical implications of these findings in facilitating retirement adjustment.
... First, the older individuals rarely undergo upskilling given their age (50 and above). Existing studies in Europe and North America find that older workers are significantly less likely to participate in training compared to the younger counterparts, for various reasons including older workers' low motivation to take up training and employers' beliefs that training for older employees is ineffective (Calvo et al., 2009;Petkoska and Earl, 2009;Picchio, 2021;Warr and Fay, 2001;Zwick, 2011). Second, workers' primary skills harnessed through career jobs are crystalized skills that tend to remain stable until late 70s (Desjardins and Warnke, 2012). ...
Full-text available
Article
Many retirees are seeking to return to work after retirement. In light of growing unretirement, we explore the linkages between unretirement decisions and individuals’ skill-specific automatability, i.e., the degree to which machines can replace their skills. The automatability is measured by how routinized individuals’ skills are and how little non-routine analytical adaptability, non-routine social intelligence, non-routine physical adaptability, and non-routine interpersonal adaptability their skills encompass. Using data from the 2000-2016 Health and Retirement Study, we find that retirees whose skills are more automatable are less likely than others to return to work. Among unretirees, those whose skills are more automatable tend to be hired into jobs whose task-contents are also more automatable. This work provides empirical evidence that retirees’ high automatability poses work disincentives and discourages productive aging.
... 36 This seems to be true also for financial attitudes and behaviours. 37 Individuals who have high levels of self-control over their finances are more willing to save and more careful towards financial decisions in favour of longterm goals. 38 Research has also shown that individuals who are more willing to save are more confident in their retirement and, therefore, they feel more prepared to make the transition to retirement. ...
Full-text available
Article
Older workers who are confident about the changes accompanying retirement report higher well-being. We have developed an index to measure retirement confidence – the Retirement Confidence Index (RCI). A six-stage approach was used to develop the index items, including (i) a literature review to catalogue retirement confidence components; (ii) a consultation with a panel of experts to review the proposed indicators and combine components according to their meaning; (iii) normalisation of the selected components to make them comparable; (iv) weighting of the top-level dimensions using experts’ judgement; (v) linear aggregation of the dimension scores according to their corresponding relative weight; and (vi) correlation of the composite score with a self-report measure of retirement confidence. Based on the review of the literature, a list of nine sub-components (financial literacy, financial attitude and behaviour, financial control, financial anxiety, physical health, mental health, social connectedness, goal setting for retirement and future uncertainties) was compiled. Subsequently, these components were grouped into four broad dimensions. Correlations between these dimensions (social, financial awareness and skills, health and well-being, and retirement awareness and planning dimensions) and the corresponding self-reported measures were as high as r = 0.555, r = 0.603, r = 0.591 and r = 0.569, reflecting 30.8%, 36.3%, 34.9% and 32.3% shared variance with the corresponding self-reported indices, respectively. The Retirement Confidence Index provides the foundation for future research to measure retirement confidence, with the aim of identifying deficient RCI dimensions and directing efforts to targeted policies to ensure older workers are confident about retirement.
Article
Australia’s retirement system is often heralded as world-class. Yet research investigating the superannuation system has reported highly gendered outcomes even in the university sector, which has generous superannuation provisions and established career management practices. This study investigates gender differences in retirement transition goals and barriers in the sector. Data from university employees who are 50 years or older show that gender differences exist. Women are more likely than men to be frustrated in their attempts to transition into retirement as desired and are more likely to desire a reduction in workdays per week. Men are more likely to prefer to continue working as currently or to reduce responsibility. Economic reasons are more likely to be the barrier for women, while organisational reasons are more likely to be the barrier for men.
Full-text available
Article
This article examines three dimensions of mothers' well-being (personal happiness, self-esteem, and depression) across four diverse family structures (first-married, remarried, divorced, and continuously single-parent families). Using a nationally representative sample of 2,781 mothers, the results indicate small but statistically significant differences across family structures. Mothers in their first marriage enjoy the highest well-being, mothers in stepfamilies fare nearly as well, and divorced and continuously single mothers have the lowest well-being. Most of the differences persist when relevant variables are controlled. Multiple regression analyses indicate that the strongest predictors of mothers' well-being are measures of family relations, especially children's well-being, marital happiness, marital stability, and low levels of marital conflict. Implications of the findings are discussed in terms of the relative importance for mothers' well-being of family structure, sociodemographic variables, and family processes.
Full-text available
Article
This study looks for evidence of an adulthood trajectory of perceived control over one's own life, and education's role in shaping it. Vectors from a 1995-2001 U. S. survey of adults imply a much steeper trajectory than did previous cross-sectional studies, peaking in late middle age rather than early adulthood. They also show a trend toward larger increases in younger cohorts. Education influences the trajectory, particularly the level at which it begins. Estimates show an increase of .60 standard deviations with a four-year increase in education. The sense of control also increases more outside of school the higher the level of education.
Article
Publisher Summary This chapter discusses the association of income and happiness. The basic data consist of statements by individuals on their subjective happiness, as reported in thirty surveys from 1946 through 1970, covering nineteen countries, including eleven in Asia, Africa, and Latin America. Within countries, there is a noticeable positive association between income and happiness—in every single survey, those in the highest status group were happier, on the average, than those in the lowest status group. However, whether any such positive association exists among countries at a given time is uncertain. Certainly, the happiness differences between rich and poor countries that one might expect on the basis of the within-country differences by economic status are not borne out by the international data. Similarly, in the one national time series studied, for the United States since 1946, higher income was not systematically accompanied by greater happiness. As for why national comparisons among countries and over time show an association between income and happiness that is so much weaker than, if not inconsistent with, that shown by within-country comparisons, a Duesenberry-type model, involving relative status considerations as an important determinant of happiness, is suggested.
Chapter
This chapter explores why employers might permit phased retirement only after employees officially retire. This issue is addressed based on interviews with close to 1,000 establishments regarding their phased retirement policies. Employers were asked whether they would permit an older worker to reduce hours, and, if so, whether they favoured reduction in hours before or after official retirement. Results show that many employers do not indicate a strong preference; rather, they seem open to informally arranged reductions in hours, both before and after official retirement. Statistical methods were used to analyze what types of employers might permit hour reductions to occur before and/or after official retirement. The findings suggest that the preference for retire/rehire is at the individual rather than the establishment level, often due to pension and other benefit plan inducements. Government policy could enhance work/retirement flexibility by clarifying the meaning of what constitutes retirement under tax and labour law.
Article
Using a life-course/opportunity-cost framework, we study racial differences in labor force behavior among African American and white men aged 55 to 69. A multifaceted measure of labor force behavior is examined within a longitudinal framework. We perform the analyses with a merged sample of the 1984 and 1985 Survey of Income and Program Participation, and we find that the most stable status is not working, followed by full-time, part-time, and unemployed statuses. Results from multivariate logistic regression change models show race-specific effects of age, health, and not-working status on several labor force status and attrition contrasts. Researchers have much to gain by continuing to consider racial differences in late-life labor force behavior and by focusing on contemporaneous and lagged measures of life-course variables.
Article
Contents: Preface Acknowledgments Continuity Theory How Did Continuity Theory Arise? Continuity Theory as Theory Elements of Continuity Theory Development versus Aging in Later Adulthood Case Examples Internal Continuity Continuity of the Self Self-Confidence Emotional Resilience Personal Goals Beliefs about the Effects of Retirement Summary External Continuity Living Arrangements, Household Composition,and Marital Status Income Adequacy Modes of Transportation Patterns of Activity: Stability, Continuity, and Change over Time How Activities Fit Together to Form Lifestyles Summary Adaptive Capacity Proactive Coping and Motivation for Continuity How Did Respondents Cope? Coping with Specific Changes: Retirement, Widowhood, and Functional Limitations Functional Limitation and the Self Patterns of Coping withFunctional Limitations General Patterns of Adaptation Factors Linked to Negative Outcomes in Later Life Summary Goals for Developmental Direction Continuity of Personal Goals Disposition toward Continuity Spiritual Development The Theory of Gerotranscendence The Study of Goals for Developmental Direction in Later Life Conclusion Assessing Continuity Theory Evidence on the Assumptions and Propositions of Continuity Theory Continuity Strategies Are Generally Effective Methodological Issues Related to the Study of Continuity Theory Future Research Using Continuity Theory Appendixes A. Tables B. The Ohio Longitudinal Study of Aging and Adaptation C. The 1995 Study Questionnaire D. Worksheets Used to Examine Longitudinal Patterns References Index
Article
The sociological study of the mental health of racial-ethnic minorities depends on the measurement quality of the instruments used to evaluate mental health. A commonly used instrument in research on mental health disparities, the Center for Epidemiologic Studies Depression Scale (CES-D), has not been thoroughly validated for use in the multiethnic and foreign-born populations currently living in the U.S. Using data from the National Longitudinal Study of Adolescent Health, this analysis provides the first multiethnic evaluation and psychometric analysis of the CES-D by acculturation level among youth ages 12-20. Correcting for the measurement problems contained in the CES-D improves the ability to detect differences in depression across ethnocultural groups, and to identify relationships between depression and other outcomes.
Article
Predictors of loneliness were investigated in married, widowed, divorced, and never-married older adults. Contacts with adult children, siblings, friends, and neighbors showed a stronger negative relationship with loneliness in unmarried than in married adults. However, divorced and widowed adults were more likely to profit from contact with adult children, whereas never-married and childless unmarried respondents profited most from contacts with siblings, friends, and neighbors. A better functional status was associated with less loneliness in divorced, widowed, and nevermarried adults, but not in married adults. Furthermore, unmarried men showed higher levels of loneliness than unmarried women, whereas only small sex differences in loneliness were found in married respondents. Sex differences in the loneliness of divorced and never-married adults were eliminated by controlling for sex differences in contact with children, siblings, and friends. However, widowers were lonelier than widows even after controlling for sex differences in these contacts.