| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
a Department of Sociology, Case Western Reserve University, Cleveland, OH
b Department of Sociology, Wake Forest University, Winston Salem, NC
Correspondence: Eleanor Palo Stoller, Case Western Reserve University, 10900 Euclid Ave., Cleveland, OH 44106-7124. E-mail: ups3{at}po.cwru.edu.
Decision Editor: Vernon L. Greene, PhD
 |
Abstract |
|---|
 TOP Abstract BackgroundMethodsResultsDiscussionReferences |
|---|
life course model of retirement migration. Respondents were unlikely to anticipate a return move unless ties to the back home community made such a move possible. Implications: Ties with children, both back home and in Florida, significantly influence consideration of a return move.
Key Words: Retirement migration • Counterstream migration
A dozen years have passed since Litwak and Longino 1987 introduced their life course model of retirement migration, arguing that the large streams of older migrants from the North to the Sunbelt are accompanied by much smaller counterstreams in the opposite direction. Demographers of aging have gathered inferential evidence consistent with the existence of counterstream migrants, yet researchers have yet to determine why some older migrants decide to move back to their former communities and others decide to remain in their new Sunbelt locations. In this study we addressed the anticipated likelihood of a return move among a sample of elderly people who relocated to Florida following retirement. We explored the impact of retirees' ties to their community of origin and to their new retirement community, answering the question: "Is moving back ‘going home’ or ‘leaving home’?"
|
Background |
|---|
|
TOPAbstractBackgroundMethodsResultsDiscussionReferences |
|---|
began reporting that there are counterstreams of older migrants returning from popular destination states like Florida to popular sending states like New York, Pennsylvania, and Ohio. The finding, first discovered in data from the 1970 Census, was confirmed (Longino 1985) with greater detail in 1980 Census data. That is, alongside the large stream of older migrants from New York to Florida, or from California to Arizona, there are much smaller counterstreams in the opposite direction, away from Florida or Arizona. Migration researchers expected to find these counterstreams as a natural feature of the migration landscape. Ravenstein 1889 had concluded in his Laws of Migration, promulgated more than a century ago, that a migration stream, if persistent, would beget a counterstream. The reasons seem obvious. Not everyone who moves from one place to another likes their new location, and a few of them return. Because migration studies, since Ravenstein's time, have been focused on labor force movement, the same principle is usually expressed in economic terms: Not everyone finds a job, and some return to their origin. It was thought that those with the least to offer employers would have the hardest time finding a job and would be most likely, therefore, to return. Counterstream migration consequently was said to be negatively selective, that is, to select migrants with less education and fewer job skills on average than those found in the streams. Counterstreams of elderly return migrants were also negatively selective. That is, migrants in the counterstreams were somewhat older, on average, and more often widowed and living dependently with relatives or others at their destination than were migrants in the stream.
Gerontologists were left with a puzzle to solve. Why do the counterstreams of older migrants develop? There are some measurement problems in the census data that get in the way of formulating a clear answer to the question. First, there are no migration history questions among census items, so it is not known whether or not people in the counterstreams are actually "returning" to the same state from which they originally came. This could be their first move. Many Floridians, for example, after they retire, move to the mountains of North Carolina (Longino 1990), a favorite summer vacation spot for residents of the Sunshine State. Clearly they do not qualify as return migrants.
There is some inferential evidence, however, from migrants' characteristics that implies that many in the counterstream are returning. Migrants in the larger streams to Sunbelt states are younger by 4–5 years, on average, than those in the counterstream (Longino 1985; Longino, Biggar, Flynn, and Wiseman 1984). This pattern would be consistent with the picture of counterstream return migrants. In addition, there is a higher proportion of women and widows in the counterstreams, a proportional gap that ranges from one quarter to one third (Longino 1985). By inference, therefore, one would expect to find in the counterstream women who were returning from a retirement location after being widowed. Also, consistently higher proportions (more than 20%) of those in counterstreams are not living in their own homes after their move but are living dependently with others at their destination (Longino 1985). So, although census data cannot separate out those who move after retirement and then return at a later date to their origin state, it seems reasonable, from indirect evidence, that such persons are surely found in the counterstreams and may even dominate some of them.
As a result of these speculations from census characteristics, Litwak and Longino 1987 theorized that functional health may be one of the factors affecting migration in old age. In this context, they argued that there are three types of moves in retirement. These moves may be sequenced in the sense that the average age of movers falling into the three types would be different. Their sequencing is not meant to imply that individuals have to make any of these types of moves or make them in any particular order.
The first type of move tends to come soon after retirement when economic, social, and health resources are adequate. The second follows diminished resources, such as widowhood, income inadequacy, and a decline in functional health. This move is toward relatives, particularly adult children. The third type of move is to an institution, often when family caregiving is overwhelmed by the escalating needs of the older relative. The first and third types of moves have long been documented in the research literature (Patrick 1980); it is the second type of move that has been most recently studied, and the one that may be most often implicated in counterstream migration.
Longino, Jackson, Zimmerman, and Bradsher 1991 documented the second type of move. In doing so, they noted that some research literature in the field of geography attributes the decision to move in old age to the strength of both origin and destination place ties (Behr and Gober 1982; Oldakowski and Roseman 1986). Speare, Avery, and Lawton 1991 found two variables in the Longitudinal Study of Aging (LSOA) data set that measured place ties. These were duration of residence and homeownership, and these variables predicted residential mobility in their statistical model. Longino and colleagues 1991 likewise found that place ties were an important feature of the mix of independent variables in their model that documented health-related moves. However, the two variables used to measure place ties did little to let the researcher observe the process by which attachments influence decisions to move. These observations, ideally, would require survey data about anticipated and actual moves.
It is not place ties alone, however, but also personal attachments that may motivate the second type of move, the move that occurs in the face of declining resources. Analyses of the 1990 and 1984 waves of the LSOA also lend support to these inferences from census data regarding person ties among counterstream migrants (Kovar, Fitte, and Chyba 1992). Among the 16% of respondents who reported making at least one move between 1984 and 1990, approximately one fifth said they moved in order to live closer to their family and another fifth said they moved because of health declines, either of themselves or of their spouses. Silverstein 1995 documented the fact that a decline in physical health of older parents increased the propensity for parents and children to become geographically more proximate to one another. Likewise, in an analysis based on the 1984–86 waves of the LSOA, Zimmerman, Jackson, Longino, and Bradsher 1993 demonstrated that declines in instrumental activities of daily living were associated with a greater likelihood of changes in residence. Further, they also reported that close proximity of a potential helper reduced the likelihood of a move. Bradsher, Longino, Jackson, and Zimmerman 1992, using the same data set, found that the circumstance of widowhood greatly increased the probability of a move when health declines. Therefore, personal attachments either at or near the present residence, and at the place of potential destination, must be assessed.
Research on declines in functional health and increased residential mobility, with the LSOA data set, has certainly contributed to an understanding of mobility in old age. However, there are two reasons why those data cannot directly address the issue of returning to one's former home after an initial retirement move. First, the sample is of older persons living in the community. It does not focus on migrants; only residential change is noted. Perhaps a quarter of persons aged more than 60 make any kind of move in a 5-year period, and fewer than 5% move across state lines (Longino 1995). In other words, most of the respondents in the LSOA data are not, nor will they become, interstate migrants. Second, even if it were possible to identify the few respondents making long distance moves, there is no information about the origin or destination of the move. Returning to one's former place of residence is nowhere indicated in the LSOA data. So, although they are a treasure trove of statistical models, the LSOA-based studies cannot directly answer questions about why counterstreams develop, nor why some retirees who have once moved return later to their former hometowns.
In the present study we attempted to directly address, for the first time, the dynamics of the decision to make a second interstate move, one that may place the migrant into a counterstream to his or her place of residence before the original retirement move. As in other studies, we predicted that diminishing resources, such as changes in health, income, and marital status, would sometimes trigger moves, but, we also made the unique prediction that place and person ties back home and in the retirement location would tip the scales toward moving or staying.
We developed logistic regression models to test this part of the life course migration theory. In these models, the outcome variable was the likelihood of an anticipated return move, and the predictor variables were diminishing resources and the place and person ties connecting the migrant to the community back home and to their retirement location. The value of this study is heuristic rather than confirmatory, because anticipated moves may not always occur. A full test of the life course model of counterstream migration awaits a longitudinal follow-up of our sample.
|
Methods |
|---|
|
TOPAbstractBackgroundMethodsResultsDiscussionReferences |
|---|
). The original study limited data collection to older persons claiming European ethnicity, because the focus was on the impact of symbolic ethnicity, a voluntary label that individuals may choose to project in particular situations (Gans 1979). The development of this research question reflected the admonition by Markides, Liang, and Jackson 1990(p. 112) that "the almost exclusive focus on disadvantaged ethnic groups has taught us a great deal about disadvantage at the expense of understanding how cultural factors influence behavior and identity independent of minority group status."
Sample
Design.
For this study, we operationalized retired Sunbelt migrants as people aged 60 years and older who moved to Florida after retiring from jobs (or after their spouses retired from jobs) in the northern United States. Temporary residents ("snowbirds") who spent less than 6 months of the year in Florida were eliminated during the sample screening.
The sample of Sunbelt migrants who claimed European American ethnicity was identified through telephone screening, with listings from telephone directories. European ancestry was operationalized as a positive (i.e., European) response to any of the following questions: (a) "The U.S. Census asks people to identify their ancestry. How would you answer this question?" (b) "When people ask you what your ethnic background is, what do you answer?" (c) "From what countries or parts of the world did your ancestors come?" The sample was selected in two parts, reflecting the dimensions of European American ethnicity incorporated into the original study. We identified a subsample of 205 Sunbelt migrants who claimed European American ethnicity through telephone screening, using a systematic sample with a random start of listings from telephone directories for the Florida community. In households with more than one respondent, we used a selection grid modified from a technique developed by Groves and Kahn 1979 to identify the potential respondent. An additional subsample of 394 migrants who claimed Finnish American ethnicity also completed the interviews. These Finnish Americans were identified by review of the same telephone directories by a panel of three experts familiar with the structure of Finnish names. We supplemented the Finnish American sampling frame developed from the telephone directories through membership lists of retiree organizations and through snowball sampling techniques, following techniques recommended by Kalton and Anderson 1989. Potential respondents were selected from this frame through systematic screening with a random start. Finnish Americans identified during the initial telephone screening were added to this second subsample. The response rate for the combined sample was 78.3%. A complete description of the sample design and respondent characteristics is available elsewhere (Stoller 1998).
Because of our interest in the anticipated probability of a return move, respondents who had already moved to assisted living environments in the Florida community were excluded from the sample for this analysis. Ultimately, the analyses reported in this article were based on data provided by 465 retired migrants (121 other European Americans and 344 Finnish Americans).
Characteristics.
The mean age of these respondents was 75.2 years (SD = 6.2). Of the respondents, 5.5% were aged less than 65 years, 13.8% were aged between 65 and 69, 22.8% were aged between 70 and 74, 28.9% were aged between 75 and 79, 25.9% were aged between 80 and 84 years, and 3.1% were aged 85 years and older. The majority of respondents, 56.1%, were women. Because we selected only one older person per household, the percentage of married respondents in our study is lower than the percentage of married elderly persons in the population. More than half, 55.5%, of our respondents were married, 1.5% were living with a partner, 33.8% were widowed, 5.0% were divorced or separated, and 4.1% had never married. The median level of education was 12 years; 23.6% had not graduated from high school, 36.2% were high school graduates, 22.3% reported some postsecondary education, and 17.8% had baccalaureate degrees.
Almost one quarter (24.3%) of the population in the county from which the sample was drawn were 65 years of age and older. More than half (54.5%) of the county's population was married. The median level of education of the county population was also 12 years; 21.2% had not graduated from high school, 30.1% were high school graduates, 26.6% reported some secondary education, and 22.1% had baccalaureate degrees (U.S. Census Bureau 1990).
The majority of respondents were able to manage instrumental activities of daily living without assistance. The task area with which respondents were most likely to need assistance was housework or yard work (20.1%). Fewer than 10% needed help getting to places out of walking distance (8.1%) or shopping (8.5%). Fewer than 5% needed help with cooking and fewer than 5% needed help with managing finances; dressing, bathing, or using the toilet; and getting around their house or apartment. The mean score on the Older Americans Resources and Services functional rating scale (Fillenbaum 1988) was 24.4 (SD = 7.9); the median score was 22, and the 25th and 75th percentiles were, respectively, 19 and 27. According to 1990 Census data, 15.5% of the county population aged 65 years and older reported a mobility or self-care limitation; 5.5% reported a mobility limitation only, 5.1% reported a self-care limitation only, and 4.9% reported both mobility and self-care limitations.
Measurement
Outcome Variable: Likelihood of Anticipated Counterstream Migration.—We used Cantril 1965 ladders to assess the likelihood of counterstream migration—of a return move to the "back home" community. Rungs were numbered from 0 ("absolutely would not consider moving back home") to 10 ("definitely will move back home"). Cantril ladders have the advantage of being self-anchoring and producing a theoretically interval-level measure (Palmore and Luikhart 1976). Frequencies for each rung are presented in Table 1 . Only 27.6% of the respondents gave any consideration to a return move.
|
Financial adequacy was measured by a composite indicator based on responses to two items: (a) "How well does the amount of money take care of your needs?" ("very well," "fairly well," "poorly") and (b) "Do you usually have enough money to take care of those little extras?" ("yes, we can buy pretty much what we want," "yes, we can usually afford most of what we want," "sometimes, but we have to watch our budget carefully," "no, we can't afford anything extra"). We combined these items to create a six-category index: (a) money takes care of needs poorly; (b) money takes care of needs (very or fairly well) but respondent has to watch budget very carefully or cannot buy anything extra; (c) money takes care of needs fairly well and respondent can usually afford most of what he or she wants; (d) money takes care of needs fairly well and respondent can buy what he or she wants; (e) money takes care of needs very well and respondent can usually afford most of what he or she wants; (f) money takes care of needs very well and respondent can buy what he or she wants. Scores on this ordinal indicator ranged from 1 to 6, with a median score of 5; 25% of the cases scored 3 or lower, and the top 25% of the cases exhibited scores of 6.
The third aspect of diminishing resources that pushed people toward counterstream migration was widowhood. Anticipation of a move back home should be lower among people who are still married than among people who have already experienced widowhood. As indicated above, 55.5% of our respondents were married and had not yet experienced widowhood.
Community Ties.—Ties to the back home community were of two types: place ties and person ties. Place ties were measured by property owned back home and visits back home. Thirty-four percent of respondents still owned property back home, and 73% had visited back home within the past year. People ties were measured by the number of close relatives and friends still living back home. Children, siblings, other relatives, and close friends were counted separately; the mean number of persons in each of these relationships is presented in Table 1 . One third (33.4%) of respondents had at least one child living back home, 31.2% had at least one living sibling back home, 19.5% had at least one close relative back home, and 25.1% identified at least one close friend living back home.
Ties to the Florida retirement community were grouped into the same two categories: place ties and person ties. Place ties were measured as property ownership and duration of residence in Florida. The vast majority (93.0%) of the respondents had purchased property in Florida. Duration of residence varied from less than a year (i.e., 0 years) to 40 years. The mean number of years living in Florida was 15.3 (SD = 8.8). We also included a global measure of satisfaction with the retirement location, using a Cantril ladder with the rungs numbered from 0 ("not at all satisfied") to 10 ("extremely satisfied"). The mean satisfaction score was 8.7 (SD = 1.9). Once again, people ties were measured by proximate children, siblings, other relatives, and close friends (Table 1 ). Slightly less than one fifth (18.1%) had at least one child living in the Florida area; 19.4% reported at least one proximate sibling, 22.7% at least one proximate other relative, and 80.6% at least one proximate close friend.
Analytic Strategy
Because of the skewed distribution of the outcome variable, we dichotomized the Cantril ladder assessing the likelihood of moving back home, distinguishing respondents who selected rungs below the midpoint of the scale (i.e., scores of 4 or less) from respondents who selected the midpoint or high rungs on the ladder (i.e., scores of 5 or more). The univariate distribution of this variable (see Table 1 ) is consistent with this decision. Several logistic regression models were estimated and compared. Model 1 included diminishing resource indicators only; Model 2 included resources plus ties to the back home location; Model 3 included resources, ties back home, and ties to Florida; and Model 4 included Model 3 plus interaction terms between the diminishing resource indicators and place ties and between the diminishing resource indicators and person ties.
Each interaction term was tested separately, but only significant interactions are reported here. Parameter estimates for the equations with interaction terms are available from the authors. Results for the first three models are presented in Table 2 . Results of the regressions involving interaction terms are also available from the authors.
|
|
Results |
|---|
|
TOPAbstractBackgroundMethodsResultsDiscussionReferences |
|---|
2 (3, n = 451) = 2.3, ns.
Model 2
When place and person ties back home were added to the model, robustness increased considerably; for the back home block, 2 (6, n = 451) = 32.9, p < .0001. People who made visits back home were significantly more likely to anticipate returning to their previous location to live at some time in the future. The coefficient for owning property back home was not significant in the multivariate equation. Among the person ties, both the number of children and the number of siblings back home had a significant positive impact on the probability of a return move. Controlling for ties to the back home community caused financial adequacy to reach significance. In particular, elderly persons with adequate financial resources were less likely to consider moving back home than were persons whose financial resources were more limited.
Model 3
When place and person ties back home and to Florida were added to the resource model, the full picture of the main effects emerged. Incorporating indicators of ties to Florida also resulted in a significant improvement in the fit of the model; for the Florida community block, 2 (7, n = 451) = 22.7, p < .005. Among the resource variables, only financial resources had an independent effect on anticipation of a return move back home. Elderly persons with adequate financial resources were less likely to anticipate counterstream migration than were persons who assessed their financial resources as limited. Neither self-assessed health nor marital status were significant predictors.
The findings from Model 2 regarding the impact of back home ties were replicated in Model 3. That is, visits back home were an important predictor of an expected return move, although owning property back home was not. In terms of person ties, having children and siblings living back home significantly increased the odds of a respondent's anticipating a return move.
The place and person ties in Florida had similar effects. Higher levels of satisfaction with life in Florida discouraged expected moves back home, although owning property in the area did not. Duration of residence in Florida did not significantly influence consideration of a return. Among person ties, only the proximity of children discouraged an anticipated return move. At least among these migrants, neither ties with kin (other than children) nor ties with friends in the retirement community had a significant impact on the anticipated likelihood of a return.
Model 4
In keeping with the rationale of the second type of move embedded in the Litwak and Longino 1987 model, the resource variables in Model 1 did not become significant predictors of an expected return move until there were back home ties that made such a migration feasible. The cross-sectional design of the current study did not enable us to trace the impact of declines in resources, but the model implied that ties to the home community should be most relevant to people facing loss of resources.
To explore this hypothesis, we created interaction terms between resource predictors and the back home indicators. Interaction terms involving each of the three resource variables (health, marital status, and financial resources) were entered separately. Within each set of interactions, we used a difference of chi-square test to check the overall contribution of the set of interactions. The individual significance of each interaction term was also tested. The order in which the interaction terms (or sets of interaction terms) were entered into the equation did not result in any differences in the results. Results of the tests involving interaction terms are available from the authors.
None of the marital status interactions were significant, either independently or considered as a block. Adding the financial interactions to Model 3 resulted in a significant improvement in the fit of the model, but only at the .07 level of significance. This slight improvement reflected the contribution of a significant interaction between financial adequacy and visits back home. This interaction suggested that people facing financial difficulties were more likely to consider a return if they also reported visiting back home. Adding the health interactions to Model 3 also resulted in a slight improvement in the fit of the model, but only at the .06 level of significance. This slight improvement reflects a significant interaction between health and having siblings living back home. Thus, poor health increased the importance of siblings back home in decisions regarding a possible return.
|
Discussion |
|---|
|
TOPAbstractBackgroundMethodsResultsDiscussionReferences |
|---|
life course model of retirement migration, that is, a move motivated by a desire to be closer to family as a person's health begins to deteriorate. Considered alone, resources such as good health, adequate financial resources, and the presence of a spouse had no direct effect on anticipation of counterstream migration. Although declines in resources can create pressure for a move back home, the potential move is considered within the context of support resources available at the time of the interviews.
Our respondents were unlikely to anticipate counterstream migration unless place ties in the back home community made such a move possible. Without maintenance of ties in the community of origin, counterstream migration could not be considered going back home. Litwak and Longino 1987 concluded that, of the various types of person ties, it is only ties to children that have a significant effect upon anticipating a return move. This assertion is amply supported in Models 2 and 3. In both models, the presence of children and the presence of siblings back home significantly improved the fit of the model. Maintaining other ties in the originating location also encouraged the likelihood that elderly migrants would consider a return move. For example, elderly persons who made trips back home were more likely to consider a return move. Ties with siblings become particularly significant among persons in poor health.
Support from sisters or brothers has received less attention in the family care literature than support from spouses or adult children. However, investigators have documented the exchange of both instrumental and emotional support among siblings (Cicirelli 1994), particularly in the absence of a spouse or geographically proximate adult child (O'Bryant 1988). Siblings can also be an important source of emotional support, particularly among frail older persons whose social world is contracting (Gold 1989). The significant interaction involving siblings may also serve as a proxy for other family characteristics that encourage a return move. A larger number of siblings back home can reflect large, extended kinship networks and/or families that have experienced little geographic mobility.
Ties to the new retirement location can also discourage consideration of a return move. The availability of adult children in the Florida retirement community suppresses the anticipated likelihood of counterstream migration. This finding is also consistent with the life course model, which describes the second move as a form of kinship migration.
Dissatisfaction with the Florida location also prompts thoughts of counterstream migration, and satisfaction with Florida retirement fosters greater attachment to the new location. Indeed, ties developed in the post-retirement years can make the Florida community feel more like home, particularly if ties to the community of origin diminish over time. This inverse relationship between satisfaction with the Florida community and the anticipated likelihood of a return move reflects traditional models of migration that emphasize "push" and "pull" factors (Lee 1966), particularly the congruence between location-specific opportunities or amenities and migrant preferences (Longino 1995). We would expect the fit between migrant and location to be especially important among respondents who are relatively healthy. These retirees are still enjoying the benefits of their amenity migration, the first move to Florida, and have not yet faced the declines in resources that trigger thoughts of the second move.
The cross-sectional nature of the current study precludes our ability to assess the impact of declining health, inadequate financial resources, or widowhood on eventual migration decisions. Perhaps such crises will prompt actual rates of a return move considerably higher than the anticipated rates, particularly among elderly persons with few available resources. Perhaps, over time, some migrants will develop strategies for generating support within the Florida community. We will need to follow these retirees over time to discover the eventual distribution of our respondents across these two outcomes.
|
Acknowledgments |
|---|
Received for publication May 5, 1999. Accepted for publication June 7, 2000.
|
References |
|---|
|
TOPAbstractBackgroundMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  J. F. Sergeant, D. J. Ekerdt, and R. Chapin Measurement of Late-Life Residential Relocation: Why Are Rates for Such a Manifest Event So Varied? J. Gerontol. B. Psychol. Sci. Soc. Sci., March 1, 2008; 63(2): S92 - S98. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. K. Smith and M. House Snowbirds, sunbirds, and stayers: seasonal migration of elderly adults in Florida. J. Gerontol. B. Psychol. Sci. Soc. Sci., September 1, 2006; 61(5): S232 - S239. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. P. Stoller and A. T. Perzynski The Impact of Ethnic Involvement and Migration Patterns on Long-Term Care Plans Among Retired Sunbelt Migrants: Plans for Nursing Home Placement J. Gerontol. B. Psychol. Sci. Soc. Sci., November 1, 2003; 58(6): S369 - 376. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. F. Longino Jr., A. T. Perzynski, and E. P. Stoller Pandora's Briefcase: Unpacking the Retirement Migration Decision Research on Aging, January 1, 2002; 24(1): 29 - 49. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
