Full text of DOI:10.1093/geront/40.4.469

Copyright 2000 by
The Gerontological Society of America
Poverty risks among elders are shaped in critical ways by their work history, demographic
characteristics, current marital status, and residential context. Using 25 years of data from
the Panel Study of Income Dynamics, we combined past occupation and work history of
elders and their spouses with information on current marital status and residence to estimate
discrete time event history models of first transitions into poverty after reaching age 55.
Education, work history, and preretirement wages contributed to men’s and some women’s
probability of becoming poor. Work history remained an important predictor of transitions
into poverty, even after controlling for preretirement wages and human capital.
Metropolitan residence was associated with a lower probability of making transitions into
poverty. This residential difference was not appreciably attenuated in three of four elderly
subgroups after measures of work history, preretirement wages, current life events, and
demographic characteristics of the elders were included in the models.
The Gerontologist
Vol. 40, No. 4, 469–479
Key Words: Residential differences, Event history analysis
Work History and U.S. Elders’ Transitions
into Poverty
Diane K. McLaughlin, PhD,1 and Leif Jensen, PhD2
Despite improved economic well-being of elders
as a group, poverty remains prevalent among se-
lected subgroups of elders—minorities, women, and
rural residents (McLaughlin & Jensen, 1993; Smeed-
ing, 1990; Warlick, 1985). O’Rand’s (1996) article in
this journal, and earlier work by Crystal and Shea
(1990), have outlined the links between social insti-
tutions and life course processes and how these ac-
cumulate over a lifetime to place some groups of
older people in relatively advantaged or disadvantaged
circumstances. We refine these ideas by incorporating
work history and residential context in a study of transi-
tions into poverty among older Americans.
& Holden, 1993; McLaughlin & Jensen, 1993). This
cross-sectional research has been complemented by
studies that have found that rates of transition into
poverty during older ages are higher for women,
Blacks, those with lower educational attainment, and
those living in nonmetro areas (Coe, 1988; Holden,
Burkhauser, & Feaster, 1988; McLaughlin & Jensen,
1995). The same factors are found to inhibit the es-
cape from poverty among elders (Jensen & McLaugh-
lin, 1997). These studies have provided strong evi-
dence that residence does affect poverty, even when
controls for important characteristics such as race,
education, and age are included. However, most of
these studies have not included work history in ex-
plaining poverty or poverty transitions.
Burkhauser, Holden, and colleagues (Burkhauser,
Butler, & Holden, 1991; Holden et al., 1988) have
examined transitions into poverty after retirement for
always-married couples compared with eventual
widows or eventual widowers. Higher total preretire-
ment income and a higher percentage of income
contributed by the wife decrease the risk of becom-
ing poor for intact couples and eventual widows. Al-
though useful, these studies do not tell us how work
history, which determines both pre- and postretire-
ment income, influences poverty. We suggest that
occupation and work history affect poverty risks be-
yond the wages earned. Occupations that demand
more complex thinking, decision making, and intel-
lectual challenges may not only pay better, they may
better prepare people for making decisions and plan-
ning for retirement. Some occupations, or coverage
by a union, may also affect the nonpay benefits of
employment such as retirement and health care
plans. We explore these more complex relationships
by linking poverty transitions back to the work his-
Elders’ Sociodemographic Attributes, Recent Life
Events, and Transitions into Poverty
Prior research on the economic well-being of indi-
vidual elders has examined the likelihood that elders
are poor, will become poor, or will suffer a drop in
income. Cross-sectional studies have provided clear
links between poverty and being from a minority
group, having lower educational attainment, living in
rural or nonmetropolitan (nonmetro) areas, being fe-
male, and living alone (Glasgow, 1988; McLaughlin
Support for this research was provided by National Institute on Aging
Grant RO1 AG11240; by the Population Research Institute, The Pennsyl-
vania State University, which has core support from the National Institute
of Child Health and Human Development (P30HD28263); and by The
Pennsylvania State University Agricultural Experiment Station Research
Projects (3644 and 3501). We are indebted to Henk Meij and Don Gensi-
more for computational assistance with file construction. We alone are
responsible for errors in logic or analysis.
1Address correspondence to Diane K. McLaughlin, 110C Armsby
Building, University Park, PA 16802. E-mail: dkk@psu.edu
2Department of Agricultural Economics and Rural Sociology and Popu-
lation Research Institute. The Pennsylvania State University, University
Park, PA.
Vol. 40, No. 4, 2000
469

tory of elders, while controlling for educational at-
tainment and preretirement wages.
tween individual characteristics, work experiences,
and work history prior to retirement and economic
well-being after retirement. The important implica-
tion is that education and work history directly and
separately affect elders’ postretirement income and
thus their poverty risks. It is less clear whether these
factors remain important predictors of post-retire-
ment well-being after controlling for preretirement
income.
Changes in marital status, whether before or dur-
ing older ages, influence the risks of becoming poor
differently for men and women. These differences are
partly rooted in men’s and women’s disparate labor
force participation and wages and in noncovered
spouses’ access to pension and Social Security bene-
fits. Studies of eventual widows revealed that 52.6%
of women who eventually are widowed experience
poverty (Holden, Burkhauser, & Myers, 1986). The
events most often associated with falls in income-to-
needs ratios of eventual widows are the death of the
spouse and declines in non-wage income (Bound,
Duncan, Laren, & Oleinick, 1991; Burkhauser, Holden,
& Feaster, 1988; Smith & Zick, 1986). Share of assets
in housing, wife’s contribution to household income,
and higher preretirement income protected even-
tual widows from poverty (Burkhauser et al., 1991;
Holden & Burkhauser, 1986; Holden et al., 1988).
For eventual widowers, the wife’s death reduced the
risk of poverty (Burkhauser et al., 1988).
This literature has identified sociodemographic at-
tributes and recent life events that correspond with
drops in income among elders, as well as a strong re-
lationship between higher preretirement income lev-
els and lower poverty risks. Generally, these studies
have not attempted to incorporate information about
actual labor force experiences and work history be-
fore retirement. Individuals with similar preretirement
incomes could have arrived at those incomes through
quite different career paths. These differential career
paths may influence the ability to save and plan for
retirement and, ultimately, affect poverty risks.
Recent studies of earnings attainment have argued
that structural characteristics of labor markets can
constrain or enable individuals’ labor market suc-
cess. Although this approach has been applied
broadly to studies of working-age people, it has sel-
dom been used to examine poverty transitions or in-
come levels of elders. Using the broad concept of
place of residence that captures differences in labor
market structure and operation, prior research has
documented higher poverty prevalence, higher rates
of transition into poverty, and lower probability of
leaving poverty among nonmetro elders (Jensen &
McLaughlin, 1997; McLaughlin & Jensen, 1995). This
pattern is argued to result from the more limited em-
ployment opportunities available in nonmetro areas
(Colclough, 1988; Glasgow, Holden, McLaughlin, &
Rowles, 1993; McGranahan, 1988), different occu-
pational and industrial structures (Tienda, 1986),
lower pay for similar jobs (McLaughlin & Perman,
1991), fewer opportunities for upward mobility (Do-
eringer, 1984; McGranahan, 1988), seasonality of em-
ployment (Lucas, 1971; Tomaskovic-Devey, 1987),
and low unionization rates and few pension (Erickson,
1981) or health benefits. Prior studies have docu-
mented differences in the poverty prevalence, type,
and amount of income received by elders in metro
and nonmetro areas (McLaughlin & Jensen, 1993),
providing indirect evidence of the differences in em-
ployment opportunities and how they influence el-
ders’ economic well-being.
Status Attainment, Cumulative Advantage, and
Structural Explanations for Elders’
Economic Well-Being
A smaller set of studies has argued that an elder’s
postretirement economic well-being is determined
by labor force experiences. Studies of status attain-
ment and status maintenance into older ages reveal
that men’s postretirement economic well-being is de-
termined by the same factors that influence eco-
nomic well-being before retirement (Campbell &
Henretta, 1980; Henretta & Campbell, 1976). More
recently, this line of research has been expanded to
argue that local opportunity structures affect the sta-
tus and income attainment processes.
Those who do well during their working life do
even better once they retire—advantages experi-
enced during younger ages accumulate to increase
relative well-being during older ages (Crystal & Shea,
1990; Crystal, Shea, & Krishnaswami, 1992; Henretta &
Campbell, 1976; Leon, 1985a, 1985b). O’Rand (1996)
took this argument further to suggest that researchers
need to incorporate more information on the lifetime
experiences of individuals in understanding their
ability to avoid poverty at older ages.
Prior studies that have found higher rates of transition
into poverty among nonmetro elders have not included
detailed work history information or preretirement in-
come. Thus, it is important to determine whether higher
nonmetro poverty and transitions into poverty can be
explained by lower wages or poorer employment histo-
ries of workers in nonmetro areas. Bringing these lines
of research together requires that we estimate models
of transitions into poverty that include work history
and an indicator of preretirement wage levels. Find-
ing that both work history and education influence
poverty risks with preretirement income controlled
for indicates that there are other aspects of employ-
ment and characteristics of occupations that may
provide resources beyond those reflected in preretire-
ment income alone. Additionally, it has long been ar-
gued that the higher poverty prevalence of nonmetro
residents is due to poorer human capital and work
opportunities. Incorporating these factors into models
estimating transitions of elders into poverty will de-
termine whether these factors completely explain the
differential in poverty transition probabilities of metro
and nonmetro elders.
Despite a focus on income levels rather than pov-
erty, these studies have documented a direct tie be-
470
The Gerontologist

Data and Methods
likely to be understated compared with an analysis
using monthly income information, but the factors in-
fluencing transitions into poverty would change little.
We also argue that persons reporting annual incomes
below poverty face much more severe income re-
strictions than would persons who report only 1
month in poverty, making this a more stringent crite-
ria for becoming poor.
Poverty transitions have been tied to the marital
experiences of older men and women. Because of
this, we conducted our analysis on four distinct sub-
groups of elders. These four groups were male heads
of households, female heads of household who have
experienced no change in marital status while they
were in the PSID, women who have experienced
changes in marital status, and wives who remained
wives through 1992. This distinction is important be-
cause the poverty experiences of men and women
are quite different and because women are more
likely to become widowed than men. To show how
these groups differ from each other and in metro and
nonmetro areas, Table 1 gives the weighted number
of people in each group by residence and the per-
centage in poverty at age 55 or when first observed
for this study.
We examined only first observed transitions into
poverty after the person reached age 55, so those
who were poor at age 55 or when first observed after
age 55 were eliminated from the analysis. Second and
higher transitions into poverty had much higher prob-
abilities of occurring in a given year than did first tran-
sitions into poverty (McLaughlin & Jensen, 1995), and
the effects of covariates differed for first and later tran-
sitions into poverty (analysis not shown).
We restricted our analysis to elders who were
heads or wives in households headed by persons
aged 55 and over. The PSID contained very few el-
ders (5.6% of the weighted sample) who were neither
heads nor wives. These persons were excluded be-
cause they frequently lived in households with younger
household heads who contributed significant amounts
of income, thus any transitions into poverty reflect
the experiences of the younger head combined with
any contributions made by the elder. In addition,
there was little employment information available in
the PSID for those who were not heads or wives.
There were 3,438 sample individuals in the PSID
To examine transitions into poverty among older
men and women, we analyzed data from the Family–
Individual Response and Nonresponse Files of the
1968–1992 waves of the Panel Study of Income Dy-
namics (PSID; Hill, 1992). Individuals in families have
been surveyed annually since 1968 with an emphasis
on income, poverty, and work-related activities. The
PSID data are collected annually, and the panel is one
of the longest running studies available for examining
income; thus, poverty transitions and geographic iden-
tifiers are available, making comparisons by residence
possible. Enough older persons were included in the
original sample or have aged during the panel to pro-
vide an adequate sample size of those aged 55 and
over. When appropriate weighting procedures are
used, the PSID sample provides representative esti-
mates of the U.S. (nonimmigrant) population (Beck-
etti, Gould, Lillard, & Welch, 1988). Although the
lack of data on immigrants to the United States in the
first 20 years of the PSID is regrettable, ancillary anal-
ysis of 1990 U.S. Census Public Use Microdata sam-
ple data indicates that only 2.7% of the 1990 popula-
tion aged 53 and older had immigrated to the United
States since 1965. These immigrants do not consti-
tute a large percentage of the population in this age
group.
We created a person-year file of transitions into
poverty in which individuals contribute 1 year of
data for each year they are observed in the study and
are at risk of becoming poor. Among those in the
PSID who were heads or wives and aged 55 and
older, 7.6% were poor at age 55 or when first ob-
served, and so were excluded from the analysis. Of
those in nonmetro areas, 11.8% were poor when first
observed compared with 5.6% of metro persons (see
Table 1). We selected age 55 as the starting point so
that we could examine transitions into poverty as
people approach and move through retirement ages.
Less orderly transitions to retirement and continued
labor force participation past “typical” retirement age
make the role of work history and current work effort
particularly interesting for examining factors that in-
fluence transitions into poverty among older Ameri-
cans. We use annual, rather than monthly, poverty
ratios because the PSID reports income information
annually. The resulting transition probabilities are
Table 1. Number of Persons and Percentage Poor by Marital Status Category and Residence Among Persons at the Beginning of the
Study Period, Age 55 or When First Observed in the Panel Study of Income Dynamics
Nonmetropolitan
No. of persons % poor
Metropolitan
Total
No. of persons
% poor
No. of persons
% poor
Men
511
71
336
310
1,228
8.5
20.8
19.6
6.4
1,025
170
669
630
2,494
3.7
11.2
10.1
2.4
1,536
241
1,005
940
5.3
14.1
13.3
3.7
Female heads with no marital status change
Women with changes in marital status
Wives with no change in marital status
Total
11.8
5.6
3,722
7.6
Note: Data are weighted using year-specific individual weights.
Vol. 40, No. 4, 2000
471

who met our selection criteria between 1968 and
1992 (weighted Ns). They contributed 39,952 non-
poverty person-years (weighted) to the file.
Current sociodemographic attributes of family heads,
disability status, current work effort, occupation, and
work history were measured as indicated in Table 2.
The mean values for the person-year data are also pre-
sented in Table 2. We discuss only those measures
that are not self-explanatory. The value of the house
is included as a proxy for the full set of financial as-
sets of the elders. Although the 1984, 1989, and
1994 waves of the PSID included a supplement that
focused on financial and other assets held, these data
were not available in the 16 survey years prior to
1984. Thus, we used only the information on the
value of owned housing available in every year of the
PSID as a proxy for financial assets. We also included
two recent life events that have been found to be im-
portant for transitions into poverty—a decline in
hours worked and changes in marital status. Changes
in marital status are indicated by a dummy variable
in the year following the event. This is consistent
with findings in earlier research that most of the larg-
est income declines following widowhood or divorce
occur in the year immediately after the death or di-
vorce (Zick & Smith, 1991). Disability status was
measured as a time-varying covariate so that the sep-
arate influence of current disability status in any
given year could be assessed.
The variables measuring work history are particu-
larly important in this study, although the structure of
the PSID made the determination of these somewhat
complex. The measures we used included occupation,
years of work experience, union coverage, and prere-
tirement wages. Because many of our sample mem-
bers were no longer employed, we measured occu-
pation, union coverage, years worked full time since
age 18, and preretirement wages based on the previ-
ous employment data available, with priority given to
the work information reported when the respondent
last worked full time. If we found no record of the last
full-time employment, we used the information re-
ported when the respondent last worked. The infor-
mation for occupation, wages, and union coverage
was retained and assigned to subsequent person-
years in which the respondent did not work. A mea-
sure of the number of years the respondent worked
full time since age 18 captured lifetime work effort.
The occupational categories corresponded to those
used by Featherman and Hauser (1978) and included
professional and managerial workers; clerical, sales,
and service workers; and operative and craft occupa-
tions. Those respondents who did not have a re-
ported occupation were divided into two groups:
those who had worked full time since age 18 and
those who had never worked full time since age 18.
This captured some work history information for those
who were not observed working during the PSID (this
generally included those who were older than age 65
in 1968 or who were never in the labor force). The
reference group in this set of occupation-related vari-
ables was laborers (except farm); farmers, farm man-
agers, and farm workers; and other occupations not
mentioned.
Measurement
A transition into poverty was the dependent variable
of interest. Our definition of poverty closely followed
the official U.S. Bureau of the Census definition. Fami-
lies (or individuals in families) are in absolute poverty if
the family’s annual cash income is less than a mini-
mum threshold needed for subsistence as determined
by the U.S. Department of Agriculture’s economy
food plan (Orshansky, 1965). In 1979, for example,
the poverty threshold was $4,389 for a two-person
family with a head aged 65 or older.
Admittedly, this poverty measure has been criti-
cized for a number of reasons (Citro & Michael,
1995; Ruggles, 1990). First, it has been criticized be-
cause it is an absolute rather than a relative mea-
sure of poverty (Ruggles, 1990). A relative measure of
poverty bases the poverty level on the economic
well-being of the population rather than on a mini-
mum-needs level. We also estimated models using
transitions into relative poverty (measured as 50% of
the median income-to-needs ratio for the popula-
tion), but the differences were negligible so these re-
sults are not presented.
Other criticisms stem from the measure’s not in-
corporating the value of fungible non-cash benefits
(food stamps), the housing service value of owner-
occupied housing and the spending down of assets,
and not adjusting for geographic cost-of-living differ-
entials. Ruggles (1990) and Citro and Michael (1995)
discussed each of these measurement issues at length,
but Citro and Michael (p. 72) recommended that it is
more appropriate to define resources as disposable
income from all sources, including any income from
assets. They recommended against adjusting income
for spending down of assets and for housing service
value because it is not fungible income. We did ad-
just income for the cash value of food stamps as re-
ported in the PSID, to create new poverty transition
indicators. Models estimated using the food-stamp–
adjusted transition indicators were similar to those in
Table 3, so they are not shown.
Although it would be interesting to incorporate
geographic cost-of-living differentials, such data are
not readily available for rural and nonmetro areas at
the county level or for multiple years (and the he-
donic housing price approach recommended by Citro
and Michael, 1995, is not feasible with current data).
The appropriateness of the basket of goods on which
cost-of-living differentials are based and their repre-
sentativeness of the regular purchases of elders, the
poor, or persons in different geographic regions can
also be questioned. One criterion for making adjust-
ments to income is whether the adjustments make
the data more or less accurate. Until accurate cost-of-
living differentials are determined, it is not clear that
adjustments using the limited data available improve
the accuracy of the income data.
The context within which persons live and work
472
The Gerontologist

Table 2. Variable Definitions and Descriptive Statistics (Based on Weighted Person Year Files)
Elders With Different Marital Status Experiences
Men in
Male-Headed
Households
Female Heads
With No Change in
Marital Status
Women With
Changes in
Marital Status
Wives
0.044
Sociodemographic attributes
Race (Black ϭ 1; 0 otherwise)
Marital status
0.055
0.045
0.081
Married ( ϭ 1; 0 otherwise)
Widowed ( ϭ 1; 0 otherwise)
Single (single ϭ 1; 0 otherwise)
Divorced or separated (divorced or separated ϭ 1;
0 otherwise)
0.916
0.033
0.015
NA
0.716
0.138
0.554
0.307
0.029
NA
NA
NA
0.036
0.262
0.146
0.361
0.110
0.396
NA
0.161
Disabled (disability that limits work ϭ 1; 0 otherwise)
Education
Less than high school (11 or fewer years of
education ϭ 1; other ϭ 0)
0.334
0.387
0.421
0.265
High school (12 years of school completed;
reference group)
More than high school
0.362
0.304
0.302
0.311
0.352
0.227
0.524
0.211
Hours worked (hours worked in current year)
House value (CPI-adjusted house value, $10,000,
1987 dollars)
1,135 (1,149)
696 (823)
708 (1076)
451.94 (863.87)
16.994 (25.466)
6.636 (9.675)
10.170 (19.276) 19.980 (28.539)
Recent life events
Worked fewer hours (worked fewer hours than prior
year ϭ 1; 0 otherwise)
0.341
0.292
0.263
0.177
Change in marital status
Newly widowed
Newly divorced
0.008
0.004
NA
NA
0.040
0.005
NA
NA
Work History
Professional or managerial
0.317
0.207
0.376
0.077
0.180
0.403
0.111
0.188
0.205
0.278
0.280
0.132
0.146
0.392
0.103
0.048
Clerical, sales, service occupation
Craftsmen, operatives occupation
Laborers, farmers, farm managers, and other
occupations (reference)
Worked full-time but no occupation reported
Never worked full-time and no occupation reported
Covered by union ( ϭ 1; 0 otherwise)
Years head worked full-time since age 18
Years wife worked full-time since age 18
Wife never worked full-time
Preretirement wage (CPI-adjusted last hourly wage of
head, 1992 $)
0.018
0.005
0.267
0.070
0.048
0.112
0.086
0.019
0.195
0.263
0.048
0.037
38.59 (9.51)
13.59 (13.11)
0.575
25.36 (14.02)
NA
32.09 (19.91)
7.62 (13.70)
0.670
39.07 (9.14)
14.03 (13.12)
0.611
NA
23.85 (33.98)
NA
10.16 (13.55)
NA
14.42(28.33)
NA
26.54 (39.48)
10.70 (19.99)
Preretirement wage (CPI-adjusted last hourly wage of
wife, 1992 $)
Residence
Metropolitan (metro residence in current year ϭ 1;
0 otherwise)
South (lives in South ϭ 1; 0 otherwise)
Other controls
0.653
0.281
0.671
0.218
0.716
0.288
0.622
0.265
Mean age
63.18 (7.26)
66.54 (8.15)
64.70 (8.80)
61.81 (6.66)
No. of persons contributing data
No. of person years (weighted)
No. of events (weighted)
1,605
17,140
245
327
3,032
122
810
10,666
291
769
9,114
96
Notes: Standard deviations for continuous variables in parentheses. Measures of the number of years the person was in a nonpoverty
spell and for 10-year birth cohort were included in the analysis. Descriptive statistics on these are available from the authors. The Panel
Study of Income Dynamics includes persons who reach age 55 over a 20-year period; we included an indicator of birth cohort in the
models. We expected that cohort would have an effect separate from age because of the increasing economic well-being of younger
compared to older cohorts of elders and the differential experiences of these cohorts with employment and retirement, Social Security
coverage, and marriage patterns (Browne, 1995; Gendell & Siegel, 1996). NA ϭ not applicable; CPI ϭ Consumer Price Index.
influences their ability to earn adequate incomes and
to accumulate assets. We used metro and nonmetro
residence in each year, based on the 1970 metro des-
ignation of the individual’s county of residence. This
measure slightly overstated the proportion of elders
in nonmetro areas today because of county reclassifi-
cation (McLaughlin & Jensen, 1998). Because we
were interested in examining whether nonmetro resi-
Vol. 40, No. 4, 2000
473

Table 3. First Observed Transitions Into Absolute Poverty for Persons Aged 55 or Over
Elders With Different Marital Status Experiencesa
Female Heads
of Households
With no
Marital Status
Change
Women With
Changes in
Marital Status
Wives With no
Changes in
Marital Status
Male Heads of
Households
Intercept
Ϫ3.230***
Ϫ1.3522*
Ϫ2.176***
Ϫ3.543***
(0.531)
(0.665)
(0.432)
(0.825)
Sociodemographic Attributes
Left censoring indicator
Ϫ0.679**
(0.224)
0.921***
(0.196)
Ϫ0.966**
(0.318)
0.677
Ϫ0.843***
(0.218)
0.930***
(0.173)
Ϫ1.194*
(0.534)
1.009**
(0.378)
Race (Black)
(0.367)
Marital status
Married
Ϫ0.630*
(0.303)
(ref.)
NA
Ϫ0.295
(0.183)
(ref.)
Ϫ2.02*
(0.994)
0.154
(0.207)
0.084
(0.143)
NA
Widowed
Single
(ref.)
NA
NA
0.055
0.884*
(0.369)
0.197
(0.345)
0.361
(0.486)
Ϫ0.457
(0.399)
0.435**
(0.149)
Divorced or Separated
Disabled
NA
0.429
(0.266)
(0.224)
Education
Less than high school
0.702***
(0.163)
Ϫ0.438
(0.260)
Ϫ0.390***
(0.098)
Ϫ0.001
(0.006)
0.416
(0.262)
Ϫ0.968*
(0.426)
Ϫ0.980***
(0.224)
Ϫ0.058*
(0.024)
0.356*
(0.150)
Ϫ0.883**
(0.302)
Ϫ0.600***
(0.109)
Ϫ0.020*
(0.010)
0.459
(0.238)
Ϫ0.172
(0.420)
Ϫ1.180***
(0.321)
0.005
More than high school
Hours worked
House value
(0.008)
Recent Life Events
Worked fewer hours
0.606***
(0.169)
1.022**
(0.320)
0.760***
(0.170)
0.657
(0.404)
Change in marital status
Newly widowed
1.037*
(0.464)
1.335*
(0.579)
NA
NA
1.006***
(0.244)
1.814**
(0.555)
NA
NA
Newly divorced
Work history
Occupation
Laborer, farmer and other occupations
Professional or managerial
(ref.)
Ϫ1.244***
(0.268)
Ϫ0.353
(0.204)
(ref.)
0.424
(0.464)
Ϫ0.391
(0.326)
0.229
(ref.)
Ϫ0.587*
(0.271)
0.072
(0.183)
Ϫ0.186
(0.184)
(ref.)
Ϫ1.163
(0.647)
Ϫ0.410
(0.373)
0.298
Clerical, sales, service
Craftsmen, operatives
Ϫ0.686***
(0.187)
(0.421)
(0.422)
Worked full time but no occupation
reported
Ϫ0.129
0.273
Ϫ0.773*
Ϫ1.423**
(0.354)
(0.371)
(0.308)
(0.440)
Never worked full time and no
occupation reported
Ϫ0.580
(0.618)
Ϫ0.894***
(0.200)
0.010
(0.007)
Ϫ0.026***
(0.007)
Ϫ0.320*
(0.155)
Ϫ0.011
(0.006)
NA
0.888*
(0.444)
Ϫ0.039
(0.365)
Ϫ0.010
(0.009)
NA
Ϫ0.048
(0.341)
Ϫ0.416*
(0.192)
0.001
(0.004)
Ϫ0.021**
(0.008)
Ϫ0.270
(0.152)
Ϫ0.054***
(0.011)
NA
Ϫ0.654
(0.472)
Ϫ0.525
(1.360)
0.014
(0.011)
Ϫ0.037**
(0.013)
Ϫ0.460
(0.256)
Ϫ0.010
(0.007)
Ϫ0.052*
(0.023)
Covered by union
Years head worked full time since age 18
Years wife worked full time since age 18
Wife never worked full time
NA
Preretirement wage head (1992 dollars)
Preretirement wage of wife (1992 dollars)
Ϫ0.029
(0.015)
NA
(Table continues on next page)
474
The Gerontologist

Table 3. First Observed Transitions Into Absolute Poverty for Persons Aged 55 or Over (Continued )
Elders With Different Marital Status Experiencesa
Female Heads
of Households
With no
Marital Status
Change
Women With
Changes in
Marital Status
Wives With no
Changes in
Marital Status
Male Heads of
Households
Residence
Metropolitan residence
Ϫ0.363**
(0.139)
0.100
(0.150)
431.6
Ϫ0.875***
(0.231)
0.538*
(0.251)
229.9
Ϫ0.427**
(0.135)
0.337*
(0.136)
534.4
Ϫ0.274
(0.223)
0.141
(0.234)
168.7
9,114
96
South
Model chi-square
No. of person years (weighted)
No. of events weighted (unweighted)
17,140
245
3,032
122
10,666
291
(455)
(206)
(405)
(146)
Notes: Data are from the 1968–1992 Panel Study of Income Dynamics and are weighted using year-specific weights. Standard errors
are in parentheses.
Other controls in the model include the number of years in the nonpoverty spell, dummy variables for 10-year birth cohorts when
those are significant, a dummy variable for years of economic decline, and dummy variables for 5-year age groups, except for single-year
age dummies from ages 61 to 65. Coefficients for these controls are available from Diane K. McLaughlin. Among the age dummies, men
aged 62 were more likely to become poor, and women with changes in marital status aged 61 to 64 were less likely to become poor, as
were wives aged 55–60 or age 61. Cohort dummies were not significant in the men’s or wives’ models, but female heads born from 1890
to 1899 were less likely to make a transition into poverty, as were women with changes in marital status born between 1910 and 1919.
Full models results are available from the authors. NA ϭ not applicable; ref. ϭ reference group or omitted category in the set of marital
status measures.
a The Panel Study of Income Dynamics is based on a probability sample that requires weights be used to obtain unbiased parameter
estimates (Sudman, 1976; Survey Research Center, 1987). Original sampling weights were adjusted for nonresponse in later years. Be-
cause the event history analysis takes advantage of data from the response and nonresponse files, we weighted the data using the individ-
ual weight for each year of data (e.g., if a person contributed data in 1975, then the 1975 individual weight was used to weight that per-
son-year of information). Using this strategy ensured proper weighting of respondents and nonrespondents alike.
* p Ͻ .05; ** p Ͻ .01; *** p Ͻ .001, for a two-tailed test.
dence during late working age and old age influences
poverty risks, the 1970 nonmetro designation is ap-
propriate. Because respondents move, we did exam-
ine patterns of moving to or from nonmetro areas.
Out of the roughly 39,000 person-years, there were
344 in which people who had lived in nonmetro ar-
eas when they were first observed later moved to
metro areas and 1,559 in which those living in metro
areas when first observed moved to nonmetro areas.
Variables measuring these moves were included in
preliminary models but were not statistically signifi-
cant. We included a county wage rate as a proxy for
cost-of-living differences, but it was not statistically
significant in any model and so was eliminated from
the final models. To capture the influence of the di-
rection in which the national economy was headed,
we used a national composite index of coincident
economic indicators for June of each year. We tested
models with the actual measure but found that a
dummy variable that identified years during which
the national economy was in decline was more pre-
dictive of transition into poverty. It was only signifi-
cant in the men’s model, however.
not make adjustments for the nonindependence of
observations in the person-year data. Such an adjust-
ment was not necessary because only records that
correspond with a transition into poverty enter the
calculation of the likelihood function, thus the num-
ber of observations does not affect the standard errors.
Results
Transitions Into Poverty: The Importance of
Work History
Transitions into poverty for older male household
heads are shown in the first column of data in Table 3.
This full model allowed us to examine how work his-
tory influenced poverty risks while controlling for so-
ciodemographic attributes, recent life events, and pre-
retirement wage. We controlled for the number of
years the person had been above poverty since age 55
by using single-year dummy variables, 10-year birth
cohort dummies, and age of the respondent (mea-
sured in 5-year age groups for ages 55–60 and 66–70
and in single-year dummy variables for ages 61–65;
the reference group was age 70 and up). Because
some respondents were not observed until after they
had reached age 55, we included a left-censoring
variable to indicate that the nonpoverty spell was in
progress when the person was first observed.
Using the person-year file, we estimated discrete
time logistic regression hazard models of the first ob-
served transition into poverty after reaching age 55
(Allison, 1982, 1984; Yamaguchi, 1991). We used
discrete time models because the income data and
independent time-varying covariates were based on
annual reports. We give the coefficients and standard
errors in Table 3. Following Petersen (1986), we did
Characteristics from each of the variable groups
influenced older men’s chances of becoming poor.
Focusing on the work history measures, having been
Vol. 40, No. 4, 2000
475

employed in a professional or managerial occupation
significantly decreased the probability of becoming
poor such that these persons were 29% (eϪ1.244) as
likely to become poor as laborers or farm workers
(the reference group). Those who worked in a crafts
or operative occupation were half as likely to make a
transition into poverty, whereas men who held a job
that was covered by a union were 41% as likely to
become poor. Having a wife who never worked full-
time decreased poverty risks, as did each year of full-
time work reported for the wife. This suggests a non-
linear relationship of wives’ employment and poverty
transitions, with households where wives never worked
less likely to become poor as were those where the
wife worked more years full time. Preretirement
wage of the head was not statistically significant. The
work history measures were significant even after
controlling for preretirement wages, and education
also remained statistically significant. Having less
than a high school education doubled the probability
of becoming poor compared with graduating from
high school even after controlling for work history
and wages. Counter to findings in other studies, be-
coming newly widowed or divorced increased the
likelihood that these older male household heads be-
came poor. A head who worked fewer hours was 1.8
times more likely to become poor in that year than a
head who worked the same or more hours.
After controlling for sociodemographic attributes,
recent life events, and work history, metro men were
70% as likely to become poor as nonmetro men. The
higher probability of becoming poor for nonmetro
residents found in prior research was not explained
by differences in preretirement wages or work history
as measured in this study. We also tested two-way in-
teractions between nonmetro residence and educa-
tion and occupation. None of those coefficients were
statistically different from zero. We found little evi-
dence of any multiplicative effects of nonmetro resi-
dence for older men making a transition into poverty.
number of events. Focusing on the work history vari-
ables, only the variable indicating that the woman
did not work full time and had no occupation in-
creased the probability of becoming poor. Looking at
current work patterns, each additional hour worked
reduced the probability of becoming poor, and a re-
duction in work hours in a given year made female
heads 2.8 times more likely to become poor. Having
more than a high school education made these fe-
male heads 38% as likely to become poor as women
with a high school education. Single women were
2.4 times more likely to make a transition into pov-
erty in a given year than were widows.
Even after controlling for work history and wages,
female heads from metro areas were 42% as likely to
become poor as female heads in nonmetro areas.
Nonmetro ϫ Education and Nonmetro ϫ Occupation
interactions revealed that metro women with less than
high school educations, professional or managerial
occupations, and clerical or sales occupations were
especially less likely to make a transition into poverty
than their nonmetro counterparts. Thus, metro resi-
dence was especially protective for female heads with
these characteristics compared with nonmetro female
heads with the same attributes.
The third model in Table 3 identifies characteris-
tics associated with transitions into poverty of older
women who experienced changes in marital status.
These are women who were married and became
widowed or divorced or who were not married and
married during the survey period. Again looking at
the work history variables, women in households
where the head reported working in a professional or
managerial occupation, who worked full time but had
no reported occupation, or whose job was unionized
were less likely to make a transition into poverty.
This also was true if, when the woman was a wife,
she reported working full time. A higher preretire-
ment wage reported by the household head reduced
these older women’s poverty risks, as did more cur-
rent hours worked by the head. A reduction in hours
worked compared with the prior year increased the
probability of becoming poor, such that women in
households where the head reported working fewer
hours this year than last were twice as likely to make
a transition into poverty.
Women living in metro areas were 65% as likely
to become poor, and those residing in the South had
poverty risks 1.4 times those of women who lived in
other regions of the United States. Having more than
a high school education decreased, and having less
than a high school education increased, the likeli-
hood of making a transition into poverty. Newly wid-
owed women were 2.7 times more likely to fall into
poverty, and newly divorced women were six times
more likely to make a transition into poverty the year
following the divorce. None of the Nonmetro ϫ Edu-
cation or Nonmetro ϫ Occupation interactions were
statistically significant in explaining the transitions
into poverty of this group of older women.
Women’s Models of Transitions Into Poverty
Women’s economic well-being was tightly tied to
their marital status. Because of this, we estimated
models separately for female household heads who
experienced no marital status changes over the course
of the study and for women who did experience mar-
ital status changes. Models for wives who remained
married throughout the study were estimated using the
wives’ characteristics and only a few of the spouses’
characteristics to determine if the women’s attributes
were predictive of transitions into poverty. Models
for wives were also estimated using the head’s at-
tributes, but they are not shown because these mod-
els were essentially the same as those for the elderly
men who headed households (in 92% of the person-
years the men were married).
Beginning with the model for female heads with
no marital status change, fewer coefficients were sta-
tistically significant in this model than in the men’s
model, possibly because of the smaller sample and
The final model in Table 3 shows the estimated
coefficients for women who remained wives through
476
The Gerontologist

1992. As indicated earlier, this model includes mea-
sures of the wife’s own work and educational charac-
teristics, as well as a few indicators of the spouse’s
work effort, to assess how well women’s own charac-
teristics predicted their transitions into poverty as part
of a married couple. Among the work history vari-
ables, wives who had worked full time but had no re-
ported occupation and those with some full-time
work experience were less likely to make a transition
into poverty. The preretirement wage of the wife was
also associated with lower risks of becoming poor. A
wife who still worked, as evidenced by her reporting
hours worked in the current year, was less likely to
make a transition into poverty. Metro wives were nei-
ther more nor less likely to become poor than their
nonmetro counterparts. Testing for Metro ϫ Educa-
tion and Metro ϫ Occupation interactions revealed
that metro women in clerical and service occupa-
tions, or who reported no occupation whether they
worked full time or not, were less likely to make a
transition into poverty than their nonmetro counter-
parts with similar characteristics.
ferent groups of control variables were added to the
models. The first model includes the basic controls
for age, birth cohort, years nonpoor, economic de-
cline, and left-censoring. Metro residence had essen-
tially the same coefficient in each of the models, and
although the coefficient in the wives’ model was
smaller it was not statistically different. The coeffi-
cient for metro became smaller in all but the model
for female heads when controls for demographic
characteristics (race, marital status, education, house
value, hours worked, and working fewer hours in that
year) were added. None of these changes were statis-
tically significant, using a two-tailed test. Adding pre-
retirement wage to the model caused only minor
changes in the metro coefficient; adding the full set
of work history variables resulted in the metro coeffi-
cient in the wives’ model becoming insignificant, but
there were no statistically significant changes in the
other models.
Preretirement wage became insignificant when the
work history variables were added to the men’s and
the female heads’ models. Head’s preretirement
wage remained statistically significant in the model
for women with changes in marital status. The head’s
preretirement wage was not statistically significant in
the wives’ model, but the wives’ own preretirement
wage was significant and associated with a lower
probability of making a transition into poverty. We
found that even after controlling for a variety of fac-
tors, including measures of work history and prere-
Two of the questions addressed in this research are
whether the metro and nonmetro differences in pov-
erty risks can be explained by differences in the work
histories of people living in these areas and whether
information on work history adds explanatory power
once controls for preretirement wages are included.
Table 4 shows the estimated coefficients for metro
residence and preretirement wage of the head as dif-
Table 4. Comparison of Change in Coefficients for Metropolitan Residence and Preretirement Wage of the Head
3:
4:
2:
Metropolitan Plus
Demographic Controls
and Wage of Head
Full Model
(Column 3 Plus Work
History Variables)
1:
Metropolitan Plus
Demographic Controls
Variable
Metropolitan
Male heads of households
Metro
Ϫ0.736
(0.130)
—
Ϫ0.492
(0.137)
—
Ϫ0.440
(0.138)
Ϫ0.020
(0.006)
Ϫ.374
(.138)
Ϫ0.011
(0.006)
Wage of head
Female heads, no marital status change
Metro
Ϫ0.773
(0.199)
—
Ϫ0.880
(0.224)
—
Ϫ0.838
(0.224)
Ϫ0.033
(0.014)
Ϫ0.875
(0.231)
Ϫ0.029
(0.015)
Wage of head
Female heads, change in marital status
Metro
Ϫ0.722
(0.122)
—
Ϫ0.575
(0.132)
—
Ϫ0.501
(0.010)
Ϫ0.051
(0.010)
Ϫ0.427
(0.135)
Ϫ0.054
(0.011)
Wage of head
Wives with no marital status change
Metro
Ϫ0.565
(0.208)
—
Ϫ0.536
(0.213)
—
Ϫ0.521
(0.214)
Ϫ0.010
(0.007)
Ϫ0.034
(0.016)
Ϫ0.274
(0.223)
Ϫ0.010
(0.007)
Ϫ0.052
(0.023)
Wage of head
Wage of wife
—
—
Notes: Standard errors are in parentheses. All models include an intercept, left-censoring indicator, age dummies, cohort indicators,
an indicator of years in which economic decline occurred, and dummy variables for number of years the person had been in that non-
poverty spell. Demographic controls added include measures of race, marital status, disability status, education, house value, and work-
ing fewer hours in that year. Variables in the full model are the same as those in Table 3.
Vol. 40, No. 4, 2000
477

tirement wages, nonmetro elders (except wives) were
more likely to make a transition into poverty, with
nonmetro female heads the most disadvantaged.
perhaps as part of a retirement process, increases the
probability of making a transition into poverty. This in-
dicates the tenuous economic status of elders who
rely on current work effort to stay out of poverty.
The limited explanatory power of women’s own
sociodemographic and work history variables in these
models may occur because their economic well-being
was determined more by their complex marital histo-
ries and by their former spouse’s occupations and
work histories than by their own (especially for the
widowed or divorced women; Holden & Kuo, 1996).
Some of these formerly married women may have
been able to retain income sources and assets gener-
ated during their marriages. In addition, women’s own
information on occupation and work history may look
nothing like that of their former spouses, which may
weaken the explanatory power of work history and
education. The limited influence of work history and
education on poverty risks for women may also reflect
the very limited diversity of labor market opportuni-
ties available to women in these cohorts. Depending
on the women’s marital status, models incorporating
former husbands’ characteristics, as well as those of
the women, might do a better job of identifying the
probability of making a transition into poverty.
Another explanation may lie with the special char-
acteristics of women with no marital status change.
Many women who were widowed, divorced, or sin-
gle when they reached age 55 (or started the study at
ages above 55) were already poor as a result of their
prior marital and labor force histories (Holden & Kuo,
1996; Zick & Smith, 1986). Many of these women
were not at risk of becoming poor in our study.
For those older widowed, divorced, and separated
women who do become poor, the transition into
poverty may result from declines in interest income
or the gradual erosion of assets rather than any partic-
ular event that triggers a drop in income. Such gradual
changes are more difficult to capture using individual
human capital and work history characteristics.
Even after controlling for work history differences
that we hypothesized would explain higher transi-
tions into poverty among nonmetro elders, those
higher poverty transitions remained. Work history, as
measured in this study, human capital, and changes
in work hours and marital status do not explain the
residential difference in transitions into poverty. And
residence interacted with education and occupation
for female heads to further reduce metro women’s
probability of making a transition into poverty. Occu-
pation ϫ Residence interactions were also protective
of metro wives, reducing transitions into poverty rela-
tive to their nonmetro counterparts. Thus, we found
some evidence of differential effects of occupation
and education for women’s transitions into poverty in
metro and nonmetro areas.
Discussion
Work history helped explain transitions into pov-
erty after controlling for human capital and preretire-
ment wages. These measures remained significant
while preretirement wage became insignificant in the
models for two of the four groups of elders examined.
More work history measures were statistically signifi-
cant in the models for men and women with changes
in marital status. Few work history variables were sig-
nificant in the models for female heads or wives,
however. In addition, three of the four models showed
that metro residence remained a significant determi-
nant of transition into poverty even after controlling
for work history, preretirement wage, and human
capital. Metro residents were less likely to make a
transition into poverty after reaching age 55 than
their nonmetro counterparts.
These results suggest that using education or prere-
tirement wage as a proxy for work history does not
adequately capture the influence of work experiences
on transitions into poverty. The importance of educa-
tion and occupation or hourly wage combined with
the influence of union coverage and years worked full
time by the spouse in the men’s model supported the
idea that personal and labor market advantages con-
tinue in older age to help prevent moves into poverty.
Older men at the lower end of the status spectrum—
those who had less than a high school education and
held nonunion laborer jobs—were 16 times more
likely to become poor than older men at the upper
end of the status spectrum—those with more than a
high school education and held union, professional, or
managerial jobs. Each of these attributes of older men—
education, work history, and residence—should be in-
corporated to fully understand how a lifetime of ex-
periences translates into higher probability of making
a transition into poverty.
This research verifies the importance of including
work history when attempting to understand falls into
poverty among older individuals. As hypothesized,
those in professional or managerial occupations while
of working age were less likely to become poor in old
age, but the influence of education, and for women
with changes in marital status, pre-retirement wages
remained. Life experiences affect poverty risks in com-
plex ways and do not funnel exclusively through pre-
retirement earnings. The influence of professional oc-
cupations may occur because these individuals are
likely to have had high earnings throughout their ca-
reers, or it may result from the cognitive demands
and decision-making skills required in these jobs that
then give the individuals improved capacity for re-
tirement planning. The advantages of union coverage
for older men suggest that population subgroups with
less union coverage may face poorer prospects. For
both men and women, current employment protects
against becoming poor, while decreasing hours worked,
The persistence of the main effect of residence in-
dicates that other factors that operate differently
across residence and that would influence transitions
into poverty need to be identified. One such factor is
job mobility. We measured last reported full-time oc-
cupation and wages when available, but this did not
478
The Gerontologist

well-being of widows considered? (DP-92-802). Madison, WI: Institute
for Research on Poverty.
capture a succession of job or employer changes dur-
ing a career that may have influenced earnings, pen-
sions, and benefits, nor did it capture bridge jobs as
workers left the labor force (Quinn & Kozy, 1996).
Early labor force exit or long spells of unemployment
due to plant closings or downsizing were also not
measured in this study. These processes are likely to
have varied by residence and may have helped ex-
plain differential transitions into poverty and some of
the remaining metro/nonmetro gap. What is clear
from these models is that better information on lifetime
employment and earnings patterns, assets accumu-
lated, and on the patterns by which men and women
leave the labor force are essential to improving the
ability to explain elders’ transitions into poverty.
Holden, K. C., Burkhauser, R. V., & Feaster, D. J. (1988). The timing of falls
into poverty after retirement and widowhood. Demography, 25, 405–414.
Holden, K. C., Burkhauser, R. V., & Myers, D. A. (1986). Income transi-
tions at older stages of life: The dynamics of poverty. The
Gerontologist, 26, 292–297.
Holden, K. C., & Kuo, H. D. (1996). Complex marital histories and economic
well-being: The continuing legacy of divorce and widowhood as the
HRS cohort approaches retirement. The Gerontologist, 36, 383–390.
Jensen, L., & McLaughlin, D. K. (1997). The escape from poverty among
rural and urban elders. The Gerontologist, 37, 462–468.
Leon, J. (1985a). The effects of labor market segmentation on economic re-
sources in retirement. Social Science Research, 14, 351–373.
Leon, J. (1985b). A recursive model of economic well-being in retirement.
Journal of Gerontology, 40, 494–505.
Lucas, R. A. (1971). Minetown, milltown, railtown: Life in Canadian commu-
nities of single industry. Toronto, Ontario, Canada: University of Toronto.
McGranahan, D. A. (1988). The role of rural workers in the national econ-
omy. In D. L. Brown, J. N. Reid, H. Bluestone, D. A. McGranahan, &
S. M. Mazie (Eds.), Rural economic development in the 1980’s: Pros-
pects for the future (pp. 29–47). Washington, DC: Economic Research
Service.
References
McLaughlin, D. K., & Holden, K. C. (1993). Nonmetropolitan elderly
women: A portrait of economic vulnerability. Journal of Applied Ger-
ontology, 12, 320–334.
McLaughlin, D. K., & Jensen, L. (1993). Poverty among older Americans:
The plight of nonmetropolitan elders. Journal of Gerontology: Social
Sciences, 48, S44–S54.
Allison, P. D. (1982). Discrete time methods for the analysis of event histo-
ries. In S. Leinhardt (Ed.), Sociological methodology (pp. 61–98). San
Francisco: Jossey Bass.
Allison, P. D. (1984). Event history analysis: Regression for longitudinal
data. Beverly Hills, CA: Sage.
Becketti, S., Gould, W., Lillard, L., & Welch, F. (1988). The Panel Study of
Income Dynamics after fourteen years: An evaluation. Journal of Labor
Economics 6, 472–92.
Bound, J., Duncan, G. J., Laren, D. S., & Oleinick, L. (1991). Poverty dy-
namics in widowhood. Journal of Gerontology: Social Sciences, 46,
S115–S124.
Browne, I. (1995). The baby boom and trends in poverty, 1967-1987. So-
cial Forces, 73, 1071–1096.
Burkhauser, R., Butler, J. S., & Holden, K. C. (1991). How the death of a
spouse affects economic well-being after retirement: A hazard model
approach. Social Science Quarterly, 72, 504–519.
Burkhauser, R., Holden, K. C., & Feaster, D. (1988). Incidence, timing, and
events associated with poverty: A dynamic view of poverty in retire-
ment. Journal of Gerontology: Social Sciences, 43, S46–S52.
Campbell, R., & Henretta, J. (1980). Status claims and status attainment:
The determinants of financial well-being. American Journal of Sociol-
ogy, 86, 618–629.
Citro, C. F., & Michael, R. T. (Eds.). (1995). Measuring poverty: A new ap-
proach. Washington, DC: National Academy Press.
Coe, R. D. (1988). A longitudinal examination of poverty in the elderly
years. The Gerontologist, 28, 540–544.
Colclough, G. (1988). Uneven development and racial composition in the
Deep South, 1970-1980. Rural Sociology, 53, 73–86.
Crystal, S., & Shea, D. (1990). Cumulative advantage, cumulative disad-
vantage, and inequality among elderly people. The Gerontologist, 30,
437–443.
Crystal, S., Shea, D., & Krishnaswami, S. (1992). Educational attainment, oc-
cupational history, and stratification: Determinants of later-life economic
outcomes. Journal of Gerontology: Social Sciences, 47, S213–S221.
Doeringer, P. B. (1984). Internal labor markets and paternalism in rural ar-
eas. In P. Osterman (Ed.), Internal labor markets (pp. 271–289).
Cambridge, MA: MIT Press.
Erickson, R. A. (1981). Corporations, branch plants, and employment sta-
bility in nonmetropolitan areas. In J. Rees, G.J.D. Hewings, & H.A.
Stafford (Eds.), Industrial location and regional systems: Spatial organi-
zation in the economic sector (pp. 135–153). Brooklyn, NY: Bergin.
Featherman, D. L., & Hauser, R. M. (1978). Opportunity and change. New
York: Academic Press.
Gendell, M., & Siegel, J. S. (1996). Trends in retirement age in the United
States, 1955-1993, by sex and race. Journal of Gerontology: Social Sci-
ences, 51B, S132–S139.
Glasgow, N. (1988). The nonmetro elderly: Economic and demographic
status. Washington, DC: Economic Research Service.
McLaughlin, D. K., & Jensen, L. (1995). Becoming poor: The experiences
of elders. Rural Sociology, 60, 202–229.
McLaughlin, D. K., & Jensen, L. (1998). The rural elderly: A demographic
portrait. In R.T. Coward & J.A. Krout (Eds.), Aging in rural settings: Life
circumstances and distinctive features (pp. 15–43). New York: Springer.
McLaughlin, D. K., & Perman, L. (1991). Returns versus endowments in
the earnings attainment process for nonmetropolitan men and women.
Rural Sociology, 56, 339–365.
O’Rand, A. (1996). The precious and the precocious: Understanding cu-
mulative advantage and cumulative disadvantage over the life course.
The Gerontologist, 36, 230–238.
Orshansky, M. (1965). Counting the poor: Another look at the poverty pro-
file. Social Security Bulletin, 28, 3–29.
Petersen, T. (1986). Estimating fully parametric hazard rate models with
time-dependent covariates: Use of maximum likelihood. Sociological
Methods & Research, 14, 219–246.
Quinn, J. F., & Kozy, M. (1996). The role of bridge jobs in the retirement
transition: Gender, race, and ethnicity. The Gerontologist, 36, 363–372.
Ruggles, P. (1990). Drawing the line: Alternative poverty measures and
their implications for public policy. Washington, DC: Urban Institute.
Smeeding, T. M. (1990). Economic status of the elderly. In R. H. Binstock
& L. K. George (Eds.), Handbook of aging and the social sciences (3rd
ed.; pp. 362–381). San Diego, CA: Academic Press.
Smith, K. R., & Zick, C. D. (1986). The incidence of poverty among the re-
cently widowed: Mediating factors in the life course. Journal of
Marriage and the Family, 48, 619–630.
Sudman, S. (1976). Applied sampling. New York: Academic Press.
Survey Research Center. (1987). User guide to the Panel Study of Income
Dynamics. Ann Arbor, MI: Institute for Survey Research.
Tienda, M. (1986). Industrial restructuring in metropolitan and nonmetropol-
itan labor markets: Implications for equity and efficiency. In M. S.
Killian, L. E. Bloomquist, S. Pendleton, & D. A. McGranahan (Eds.), Sym-
posium on rural labor markets research issues (pp. 33–70). Washington,
DC: Economic Research Service, U.S. Department of Agriculture.
Tomaskovic-Devey, D. T. (1987). Labor markets, industrial structure, and
poverty: A theoretical discussion and empirical example. Rural Sociol-
ogy, 52, 56–74.
Warlick, J. L. (1985). Why is poverty after 65 a woman’s problem? Journal
of Gerontology, 40, 751–757.
Yamaguchi, K. (1991). Event history analysis. Newbury Park, CA: Sage.
Zick, C. D., & Smith, K. R. (1986). Immediate and delayed effects of wid-
owhood on poverty: Patterns from the 1970’s. The Gerontologist, 26,
669–675.
Zick, C. D., & Smith, K. R. (1991). Patterns of economic change surround-
ing the death of a spouse. Journal of Gerontology: Social Sciences, 46,
S310–S320.
Glasgow, N., Holden, K. C., McLaughlin, D. K., & Rowles, G. D. (1993).
The rural elderly and poverty. In Rural Sociological Society Task Force
on Persistent Rural Poverty (Ed.), Persistent poverty in rural America
(pp. 259–291). Boulder, CO: Westview Press.
Henretta, J. C., & Campbell, R. T. (1976). Status attainment and status
maintenance: A study of stratification in old age. American Sociologi-
cal Review, 41, 981–992.
Hill, M. S. (1992). The Panel Study of Income Dynamics, a user’s guide.
Newbury Park, CA: Sage.
Received June 28, 1999
Accepted January 28, 2000
Holden, K. C., & Burkhauser, R. (1986). Pensioner’s annuity choice: Is the
Decision Editor: Vernon L. Greene, PhD
Vol. 40, No. 4, 2000
479