| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
Correspondence: William J. Serow, PhD, Center for the Study of Population, The Florida State University, Tallahassee, FL 32306-2240. E-mail: wserow{at}coss.fsu.edu.
Vernon L. Greene, PhD
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionAppendixReferences |
|---|
Key Words: Demography • Statistical estimation • Rural areas
In recent years, there has been increased interest in the role that retirees can play in the economic development of localities. Studies of particular localities (Bennett 1993
; Carlson, Junk, Fox, Rudzitis, and Cann 1998; Day and Barlett 2000; Hodge 1991; Serow and Haas 1992; Stallmann, Deller, and Shields 1999) have tended to confirm the general expectation (Deller 1995; Fagan and Longino 1993; Glasgow 1990) that such movement can provide an infusion of external funds into a locality's existing economic base. Via the multiplier mechanism, retirees should constitute an important stimulus to employment and income growth in the locality of destination. It is hardly surprising, then, that many localities have embarked on attracting retirees as a nonpolluting source of sustained economic growth. After all, mobile retirees are usually comparatively affluent individuals who will spend at the local level, but who will require comparatively little in the way of local-level publicly provided services.
Although there is extensive literature on the determinants and consequences of elderly migration (Longino 1995, provides a book-length summary), there has been surprisingly little attention paid to longer term issues associated with the phenomenon. None of the extant empirical literature on the topic of retirement migration as a local economic development tool considers the long-term implications of the strategy, particularly those implications in the absence of the final support-seeking move suggested by Litwak and Longino 1987.
If recruiting retirees is to become a viable development strategy for rural areas, it is necessary to understand both the local-level characteristics that make an area attractive as well as the long-term consequences of such a development course. To accomplish this, it is necessary to determine which localities have been successful in attracting retirees over an extended period of time. Although the U.S. Department of Agriculture (USDA) has identified a set of retirement counties, their typology dates only to the 1970s and removes counties from this designation if they subsequently are classified as metropolitan (Bender et al. 1985; Cook and Mizer 1994). The number of USDA-identified retirement counties declined from 515 in the 1970s to only 190 in the 1980s (Reeder 1998). Therefore, this article is a brief empirical attempt to identify rural areas that have consistently attracted older migrants and to ascertain the social, demographic, and geographic characteristics of the areas of destination that differentiate them from otherwise (initially) similar areas.
Although considerable literature exists on the determinants of destination for older migrants (Serow, Friedrich, and Haas 1996), very few studies analyze such determinants for more than a single point in time (typically the 5-year migration interval available from decennial census data). The research reported here covers all intercensal periods subsequent to World War II. Furthermore, two distinct models of analysis have emerged in the existing literature. In some studies, the unit of analysis is the state (in the U.S. context), where aggregate, place-level data are employed to measure the comparative attractiveness of states of destination (Newbold 1996). Other studies focus on a more locally defined area where individual-level survey data are employed to measure why the particular area was in fact chosen by newly arrived residents (Carlson et al. 1998; Haas and Serow 1993). This article follows the first model, but the focus is on comparative attractiveness of counties, rather than larger geographic entities. This emphasis is unique for this type of study. However, aggregate data do not promote the study of the impact of proximity to family and friends, an influence often found to be of great importance in location decisions (Gober and Zonn 1983; Silverstein 1995).
The ultimate intention of this research, beyond the scope of this article, is to identify the degree to which economic change (in the form of income and employment change) is associated with the varying models of elderly migration. One would expect that a policy of encouraging retirement migration would have positive income and employment effects for all elements of the locality's population, not just for those of retirement age. By ascertaining the empirical association between migration as a source of population aging and overall economic well-being among the selected localities, it is possible to evaluate the efficacy of retirement migration as a development strategy.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendixReferences |
|---|
. I chose an initial 50% of population as rural to eliminate from further analysis those localities that were already urban in character at the beginning of the period. Both the theoretical and empirical literature on the topic recognize the importance of a rural, or at least nonurban, environment if retirement migration is chosen as the development strategy (Haas and Serow 1997). However, because I chose to consider only predominately rural counties, a significant portion of retirement migration (often directed to relatively small urban areas) was excluded from the analysis. These counties were followed over the successive censuses from 1950 through 1990, identifying those that had experienced net in-migration among persons aged 60 and older at a rate greater than that of all counties in the initial universe. This was done for each intercensal period (1950–1960, 1960–1970, and so on) to ascertain which counties had consistently experienced elderly in-migration at a rate substantially greater than the overall level. Effectively, all rural southeastern counties were sorted according to their dispersion around the mean rate of retirement inmigration. A total of 837 counties (or combinations of counties and cities) met the previously stated criterion regarding share of population classified as rural in the 1950 census. (County definitions in the Commonwealth of Virginia are unique among all states in that cities are, by definition, legally separate entities from surrounding counties. To deal with this issue I combined Virginia's cities with their respective county(ies) and treated this combination as the unit for which initial determination of the criterion was made as well as for subsequent analysis.)
I also decided to concentrate analysis of migration behavior solely on the White population because the ultimate focus was amenity-oriented retirement migration. Perhaps because Blacks accounted for less than 5% of interstate older movers between 1975 and 1980, there has been very little study of retirement migration behavior among Blacks (Longino and Smith 1991, is the sole exception). Also, there is as yet no evidence to support the notion of any sizable amenity-oriented retirement migration among Blacks. With the exception of a few pockets of Native American populations in North Carolina, there were very few older "other race" (non-White, non-Black) persons residing in the rural Southeast in 1950.
From successive decennial censuses of population, I computed age- and race-specific proportions of population surviving from beginning to end of the decade; this was done for the entire population of the continental United States. For example, the survival ratio was computed as Whites aged 60 to 64 years in 1960 divided by Whites aged 50 to 54 years in 1950 and similarly for older ages and subsequent years. These national survival rates were applied to the initial (i.e., 1950, 1960, 1970, and 1980) White populations of each county in the study universe for ages 50–54 and in 5 year increments to 70–74, with the open-ended category of 75+. This procedure implicitly assumed no significant variation across counties in the sex ratio (males per 100 females) by age or in the extent of international migration among older Whites. The former assumption probably would have been violated only in those counties with large proportions of their elderly residing in nursing homes; there are few older Whites among immigrant streams to the United States, so the latter assumption was essentially trivial. This process produced what is termed the expected end of decade (1960, 1970, 1980, and 1990) White population of each county aged 60–64, 65–69, ..., 85+. The expected populations, by age, were compared with the census enumerations, by age, for the appropriate year. This residual net migration method decomposed the growth of the older population into two components: that due to aging in place and that due to net migration. (In addition to migration, differences also arose from violation of the above assumptions as well as from census enumeration and reporting errors; there was no way of ascertaining the degree of such errors at the county level, but there was also no reason to suspect systematic bias.) Similar estimates have previously been carried out for earlier periods (Bowles, Beale, and Lee 1975; Bowles and Tarver 1965) using a slightly different methodology; for purposes of comparability over time, I used only the estimates produced by the technique described above. For initial purposes, these estimates of net migration were accumulated across all (end-of-decade) age groups aged 60 and older.
Once the level of migration was estimated, the next step was to establish more formally the place-level characteristics associated with a county's status as a retirement migration destination. Questions such as the role of differentials in natural resource endowments, economic base, demographic structure, and political variables (including tax rates) needed to be considered.
More specifically, I posited associations between retirement migration and the following economic, demographic, and geographic measures. These were expressed in terms of the analysis of estimated migration between 1950 and 1960; the geographic variables remain invariant for subsequent analyses of 1960–70, 1970–80, and 1980–90 migration whereas the demographic and economic variables apply to the beginning of each decade for these analyses.
The geographic variables measured the extent to which location matters in the choice of a retirement destination. The airport variable was a proxy for an area's accessibility for the comparatively affluent retirees who tend to comprise most of the older interstate movers (Serow 1996). The importance of Florida in the retirement migration process was treated in two ways: (a) as a dummy variable in regression analysis of all 837 rural Southeastern counties, and (b) as a means of disaggregating the entire universe. The demographic variables measured the role that initial population size and age structure as well as ethnic similarity (recall that the analysis was limited to older Whites) played in the location decision. The economic variables captured cost of living and relative taxation effects; also included here was the only comparable measure of "safety" available prior to 1971, the share of 1962 local government expenditures devoted to police protection. Subsequently, the crime rate (number of crimes known by the police per 100,000 persons) as reported in 1975 and in 1985 was used for this purpose.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendixReferences |
|---|
|
. The total White population aged 60 and older as well as the older White population residing in the counties in the study universe are shown for each Southeastern state in Table 2 . These data are given for both the beginning (1950) and final dates of the period. Overall, this project incorporated localities with 62% of the region's older White population in 1950, declining slightly to 57% by 1990. Delaware, Florida, and Maryland were the only states where less than half the older White population was included in the analysis in either (and both) 1950 and 1990. However, these states (along with Georgia and, marginally, South Carolina) were the only ones where the share of population residing in the initially rural portions of the state grew at a faster pace over the entire 1950–1990 period than did the population of the more urban areas.
|
|
|
; Fuguitt and Tordella 1980). The other pattern is the concentration of retirement areas in mountain and coastal locations: (a) five of six North Carolina localities, (b) three Georgia mountain counties (proximate to the North Carolina mountain counties), (c) two coastal counties in South Carolina, and (d) single coastal counties in Alabama and Mississippi. Other core retirement areas include the county incorporating Pinehurst in North Carolina and single counties adjacent to the metropolitan fringes of Augusta (Georgia) and Memphis and Nashville (Tennessee).
Determinants of Migration
These locational findings are the initial phase of the results presented. The results of ordinary least-squares regression analyses between migration proportions of elderly Whites and the foregoing set of independent variables are summarized in Table 3 (see Table A1 1, Note 2). To isolate the nontrivial role played by Florida in the retirement migration process for the rural Southeast, I show two specifications for the model: the first includes a separate dummy variable that is set equal to one when the county in question is located in Florida; the second removes all 55 rural Florida counties from the universe.
|
Consider first the demographic variables. Although overall population size seems irrelevant to a county's status as a retirement destination, the extent of similarity between a locality's population and potential elderly migrants seems to matter quite a bit. Both the initial share of a county's population composed of older persons and the initial share composed of Whites were always positively related to in-migration and statistically significant.
In terms of geographic correlates, the initial level of urbanization was positively associated with in-migration during the 1950s but negatively thereafter. One might surmise that the first localities to emerge as retirement destinations may have possessed sufficient amenities to attract the first post World War II generation of retirement migrants, but other considerations and subsequent urban development may have overcome this initial advantage. This observation is supported by the initially important role of commercial air service, which completely disappears after the initial decade. A county's location on either the Atlantic or the Gulf coast was a consistently important predictor of the locality's attractiveness to potential in-migrants.
The social and economic variables generally do not exert the sort of influence that one might expect both from a theoretical perspective as well as from the literature on destination choice among elderly migrants. Median rent, chosen as a proxy for county-level living costs, was consistently and positively related to migration rates from 1960 onwards. Similarly, local property taxes per capita, which should also serve as a repellent force to potential migrants, was occasionally statistically significant but always in the opposite direction! In like fashion, crime rates for 1970 and 1980 were also positively associated with migration, but this may well reflect the well-known effect of these rates not adequately reflecting the population at risk (areas with substantial tourism typically had a relatively high incidence of crime partially due to the failure to count tourism days in the denominator of the crime rate as conventionally measured). As for taxes and living costs, there are two plausible explanations: (a) these variables may actually be tapping into an income-type effect; (b) these issues may be relatively unimportant in the location decision when the universe is restricted (as here) to rural places within the same overall region. Perhaps difficult to measure variables such as climactic preferences or proximity to family and friends are really the driving force.
When the analysis was restricted to counties outside the state of Florida, the overall explanatory power of the model lessened considerably, especially during the 1960s when only 6% of overall variance was explained. Otherwise, application of the same set of variables as described above (lacking the Florida dummy) explained about one third as much variance as did the model incorporating Florida counties, with the value of R2 notably higher in the most recent decade. In general, the variables behaved in precisely the same fashion as was the case when Florida's counties were included in the analysis. Specifically, both period initial demographic structure and coastal location persist as important measures of association for in-migration of elderly Whites to rural Southeastern (non-Florida) counties. These relationships are quite robust, with little variation in relative importance since 1950. For the three most recent decades, the median gross rent made the largest contribution to the model's explanatory power, albeit operating in the direction opposite from that hypothesized (see Table A1 1, Note 4). If indeed rental values are serving as a proxy for the level of income, these results suggest that retirees are attracted to places with coastal location whose existing populations have consistently achieved some measure of prosperity and are not dissimilar from the retirees themselves.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendixReferences |
|---|
Because this article analyzes migration at the county rather than the state level, it is able to provide insight for those localities considering retiree recruitment as an economic development strategy. It specifically suggests that many of the characteristics that attract retirees may not be amenable to change by local governmental or civic action. The model developed here clearly can be better specified with the direct inclusion of variables measuring local income, a measure of climate other than coastal location, and a better measure of accessibility such as presence of a four-lane divided highway or the like. Nonetheless, it appears that some level of economic prosperity is a precondition for attractiveness to retirees. Consequently, retiree recruitment might logically be secondary to other strategies for local economic development. Prosperity suggests the availability of local infrastructure—recreation, culture, health care, and the like. These results suggest that development of local infrastructure should precede retiree recruitment, rather than assuming that the direct and indirect economic effects of retiree spending will provide the funds for this infrastructure. Development of these "tourism" amenities may well lead to a longer run pattern of retirement migration. Other related strategies also emerge. If retiree recruitment is chosen as a development strategy, those responsible for carrying it out might advertise existing retirement locations in out-of-state media. Similarly, they might encourage and facilitate retirement-oriented housing development in coastal or mountain areas, especially those areas proximate to smaller metropolitan centers.
|
Acknowledgments |
|---|
Received for publication April 23, 1999. Accepted for publication April 11, 2000.
|
Appendix |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendixReferences |
|---|
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionAppendixReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  W. Duncombe, M. Robbins, and D. A. Wolf Place Characteristics and Residential Location Choice Among the Retirement-Age Population J. Gerontol. B. Psychol. Sci. Soc. Sci., July 1, 2003; 58(4): S244 - 252. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. A. Robert Community Context and Aging: Future Research Issues Research on Aging, November 1, 2002; 24(6): 579 - 599. [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
Elderly In-Migration since 1960.