| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
BRIEF REPORT |
Correspondence: Address correspondence to William H. Haas III, PhD, Professor of Sociology, One University Heights, University of North Carolina at Asheville, Asheville, NC 28804-8508. E-mail:Haas{at}UNCA.edu
Abstract
Purpose: We examine the methodological and economic policy implications of three operationalizations of retirement migration. Design and Methods: We compared the traditional age-based definition of retirement migration and two retirement-based definitions, based on degree of labor-force participation and retirement income, by using the 2000 U.S. Census Public Use Microdata Sample. Results: The age-based definition overestimated the number of retired migrants, although the ranking of host and donor states remained relatively stable; nevertheless, states revealed different rates of change in inmigration and outmigration and income streams declined. Implications: Retirement-based definitions are more useful and precise for those researchers considering the economic implication of retirement migration.
Key Words: Economic policy • Methodology • Migration • Retirement
At the conceptual level, consensus on the definition of migration has not been problematic. Researchers agree that migration involves a residential move across important political boundaries (county, state, or nation). The conceptualization of retirement, however, has been problematic. Unfortunately, the initial age-based proxy measure for retirement migration, which is the age of 60 years and older, includes people who are still working or who have returned to work after retirement. Researchers initially hoped that this age-based measure would be useful to state planners, who, under the Older Americans Act, used the same age measure for program inclusion. They would be able to use the migration studies to determine how many of their target service population had moved from out of state during the previous 5 years.
Clarifying the meaning of retirement migration is important for two reasons. First, from a methodological standpoint, retirement migration researchers need a common understanding of their principle concept, and a more valid way of operationalizing it. We hope that this article will raise questions and start that important discussion.
Second, and perhaps more important, for those concerned with public policy and community development, different definitions produce different estimates of migration flow and stream size and their resulting impact on host and donor communities. Knowledge of the relative size of these migration flows is important for a variety of policy discussions, ranging from the use of health care (Cowper et al., 2000; Haas & Crandall, 1988; Marshall, Longino, Tucker, & Mullins, 1989) and social support (Aday & Miles, 1982; Glasgow, 1995; Joseph & Cloutier, 1991) to the political consequences of a changing electorate (Button, 1992; Haas, 1990; Rosenbaum & Button, 1989).
It is the economic impact of retired migrants (Serow, 2003), however, that typically takes center stage for discussions of the policy implications of the phenomenon. What is the financial loss for donor states and what is the financial gain for host states (Longino & Crown, 1990)? Work on this topic has been funded by federal agencies (Bennett, 1993; Glasgow, 1990; Serow & Haas, 1992). Publications have been issued (Fagan, 1988; Severinghaus, 1990) and conferences held concerning the economic benefits for host states and how to attract retired migrants. Hence, a more precise definition of retired migrants would be valuable to the community of scholars who are focused on the economic impact of retired migrants in their host states.
Interest in the economic implications of retired migrants rests on two assumptions. The first assumption is that income from Social Security, private pensions, and certain forms of equity income will follow these recently transplanted retirees. This phenomenon is sometimes referred to as the mailbox economy. Goods and services purchased in the receiving community using externally derived mailbox-economy funds stimulate local economies. In addition, retired migrants' Medicare and health insurance benefits are derived from outside of the local economy and help to finance the local health care delivery system. The second assumption is that migrating retirees do not compete for jobs in the local labor force. These two assumptions highlight the importance of incorporating retirement income and employment status in definitions of retirement migration.
Failure to adequately operationalize retirement migration undermines efforts to measure the economic implications of later-life migration flows. An age-based definition aggregates those in the workforce together with migrants who are no longer working. This is problematic because the two groups are likely to have different economic implications for receiving communities. The most obvious difference is that the former may compete with long-term residents for a limited supply of jobs, whereas the latter would not. At the same time, those who retire prior to age 60, though not identified by an age-based definition, participate in the mailbox economy, so that their arrival should generate an infusion of funds producing expanded demand for goods and services.
In this study, we explore the implications of three different operational definitions of retirement migration, all of them concerned with only a national-level analysis. The resulting research questions that guide our study are as follows.
Methods
Data
In this study, we analyze individual-level data from the 5% Public Use Microdata Sample (PUMS) of the 2000 U.S. Census. The PUMS data are a stratified subsample of those households and individuals selected to receive the census long form, itself a sample, and is representative of the U.S. population as a whole. The size of the sample is a central advantage of these data. Long-distance mobility is a relatively rare event among older adults. Only 4.6% of the older population makes such a move in any 5-year period (Longino, 2001).
Measures
We defined migrants as individuals who at the time of the 2000 Census said that they lived in another state 5 years earlier. Ekerdt and DeViney (1990) offer four possible indicators of retirement that are relevant to conceptualizing and operationalizing retirement: (a) self-definition as retired; (b) separation from one's career; (c) labor-force exit or declines in labor-force participation; and (d) receipt of pension income, including Social Security, public or private pensions, and retirement savings plans. The census PUMS files provide a measure for only the third and fourth indicators of retirement. Nonetheless, incorporating these two indicators of retirement would refine estimates of retirement migration beyond what is now possible by using the age-based proxy measure alone. We examine estimates based on three operational definitions of retired migrants.
We selected the age of 50 and older as a lower age limit for retirees who make long-distance moves. We considered it very unlikely that both labor-force reduction and retirement income would coincide at ages younger than 50.
Results
First, to what extent does the traditional age-based definition of retired migrants overestimate and underestimate those who are retired based on labor-force participation and receipt of retirement income? In regard to the overestimation, 75% (1,575,409 ) of the estimated 2,096,841 adults aged 60 and older who reported interstate moves received retirement income and had reduced labor-force participation, whereas 70% (1,468,636) received retirement income and were not in the labor force. These results suggest that a substantial number of movers aged 60 or older may not be retired. The difference in the number of migrants between retirement-based Definition 1 (reduced labor-force participation) and retirement-based Definition 2 (no labor-force participation) is rather small. The more restrictive definition, no labor-force participation, comprises more than 93% of the individuals in Definition 1.
Next, how many early retirees are underestimated by the 60-and-older definition? The second and third data columns in the second row of Table 1 present weighted estimates for the number of migrants aged 50 to 59 that are underestimated by the traditional age-based definition. As is apparent, there are a substantial number of retired migrants in their 50s, if we consider that reduced labor-force participation uncovers 222,282 additional migrants, whereas the more conservative parameter (not in the labor force) identifies an additional 187,651 migrants not counted by the 60-and-older definition. Here, a slightly larger difference exists between the two retirement-based definitions.
|
Are migrants who report an employment disability analytically distinct from the others? For both retirement-based definitions, an estimated 24% of retired migrants aged 50 and older reported an employment disability. Migrants with an employment disability remained in the analyses because they fit the stricter definition by Eckert and DeViney (1990) of retirement income based on previous employment. Disability benefits are conceptualized as retirement income and facilitators of the retirement process.
Second, what are the economic implications of the three operational definitions of retirement migration? We computed the total income from all sources reported for 1999 across individuals identified by each of our definitions of retirement migration. Figures are weighted to represent estimates of total income transferred across states by retirement migration flows, variously defined. The total state-to-state income transfer accompanying the traditional age-based migrants into all states for the nation as a whole is $59.5 billion. This includes some fully employed migrants. Definition 1, using reduced labor-force participation, saw a total income transfer of $49 billion, 17.6% less than that of the traditional age-based definition. For those who are not in the labor force (Definition 2), the national income transfer is $45.2 billion or 24.1% less than that of the traditional age-based definition.
Disability income is lower than Social Security retirement income and it also precludes income from employment. One fourth of the retirees between the ages of 50 and 60 years received disability income. It is not surprising, therefore, to find that the inclusion of younger retirees in the alternate definitions does not increase the income transfer, but reduces it.
Third, we examine the impact of the three definitions of retirement migration on the ranking of sending and receiving states. Table 2 displays the rank ordering of the host states. The top 5 states (Florida, Arizona, California, Texas, and North Carolina) retain the same rank across all three definitions. Nevada, Georgia, and Pennsylvania shuffle among the 6th-, 7th-, and 8th-ranked states on the basis of the two retirement-based definitions. There is more shifting in rankings for the 9th through 15th states, although the percentage difference between these states is very small. What is more important is that the top 10 states account for approximately 54% of all the migrants and the top 15 states contain 65%, regardless of the definition.
|
|
Among the top 5 host states, the number of migrants they received was reduced between the age-based definition and Definition 1 by –11.04% and –12.73%. The migrant inflow into California (–19.86%) and Texas (–19.83%) was reduced even more so. The percentage change between the age-based definition and Definition 2 finds Florida and Arizona with a reduction of inmigrants of –16.78% and –18.14%, respectively, with North Carolina having a bit higher reduction of –19.81%. Again, the more restrictive retirement-based definition reduces California and Texas inmigrants by a larger percent (–26.69% and –25.32%, respectively). The next 10 states also exhibit this larger reduction than Florida, Arizona, and North Carolina experienced. This might suggest that later-life migration involving these states has less of a labor-force component than does migration involving other states.
The donor states also demonstrate this kind of variation, albeit the pattern is not as clear among the top five states as it was with the host states. Considering the difference between the age-based definition and Definition 1, we see that Texas (–19.43%), Massachusetts (–17.52%), and California (–16.82%) exhibit the largest percentage reductions. The percentage change for the more restrictive no labor-force participation retirement-based definition witnesses California (–25.02%) with the largest reduction, followed by Michigan (–24.51%) and Massachusetts (–23.32). Again, it appears that some states may have a greater labor-force participation component than others in their later-life outmigration.
Implications
Retired migrants traditionally have been defined as individuals aged 60 and older who lived in another state 5 years prior to the census. However, as others have argued, age alone is not a very precise proxy for retirement (Walters, 2002). Our analysis explores the utility of retirement-based operationalizations that specifically seek to measure retirement status in a manner consistent with that of Ekerdt and DeViney (1990).
As we expected, our three operational definitions generated three different estimates of the number of retired migrants. Introducing the indicators of actual retirement generated smaller estimates of the numbers of retired migrants by 14.3% and 21% less than using age alone. Hence, the age-60 threshold overestimates the number of retired movers. The two retirement-based estimates of migration do not differ very much. The estimate based on no labor-force participation is 92% of the estimate based on reduced labor-force participation.
In a consideration of the second mailbox-economy assumption, that retired migrants do not compete in the local labor force, perhaps the second retirement-based definition has higher utility. As the size of the migration flows are reduced on the basis of the two retirement-based definitions, the income streams are also reduced by 17.6% and 24.1%, respectively. It is not safe to assume that migrants older than age 60 receive all of their income from outside the state through retirement-income transfer. If the mailbox economy is important to migration researchers, then the 60-and-older definition should not be used. The more restrictive alternative definition should be used instead.
These different estimates are relevant for receiving states, particularly for counties and local communities that are planning for the economic impact of older migrants relocating to their area. Nevertheless, when the distribution of these thrice-defined migrants among the receiving and sending states was compared, there was considerable overlap. The same proportion went to the top 10 and the top 15 states, collectively, and the actual rankings of the top 5 leading states were identical.
Still, it was clear that the retirement-based definitions had differential impacts on the states. Among the top 5 host states, the two retirement-based definitions created a larger percentage reduction in migrants going to Texas and California, compared with Florida, Arizona, and North Carolina. Thus, the proportion of the traditional age-based migrants that report the Eckert and DeViney (1990) indicators of retirement is larger in Florida, Arizona, and North Carolina. The traditional age-based definition inflates the estimate of retirement-based migrants more so for Texas and California, and so the impact of a mailbox economy would be overestimated. Texas and California seem to be demonstrating more labor-force-based migration among those individuals aged 60 years and older.
When using age-based measures of migration, we would suggest that it is more accurate to use the term later-life migration. The term retirement migration, which incorporates restrictions on labor-force participation and retirement income into definitions, will be more precise and more useful to those assessing the economic impact on host and donor communities.
It is possible that some later-life migrants who still engage in the labor force are geographically positioning themselves for retirement. This maybe peri-retirement migration that in a future time will allow the host community to reap the rewards of the mailbox economy, albeit, at the moment, the later-life migrants may be competitors with the indigenous labor force.
Limitations
Using the census data unfortunately means embracing their limitations. When the census asks "Where did you live 5 years ago?" it ignores multiple moves during that period. The American Community Survey, which is replacing the decennial census long form, will ask about residence 1 year ago, permitting a more detailed analysis of moves. Another possible limitation is that our respondents may not be reporting earnings from the so-called underground economy arising from cash transactions for work or services rendered, from sales of handicraft items, or regular and organized flea-market and garage-sale operations.
Obviously there are program and policy implications of using one or another of the definitions compared in this article. For example, money circulates more than once in the economy when it comes from sources other than local employment, so using the tighter definitions would help planners to make more accurate estimates of the impact of retirement migration on a local economy. Unfortunately, the geographical codes in census microdata require units to contain a minimum of 100,000 people. Under the more rigorous definitions of retirement migration, there may be too few cases on the county or county-group level to offer reliable estimates of income transfer to the planner, except where there are large concentrations. Longino (2006) demonstrates that in some states retirement inmigration focuses tightly on a select number of high-amenity counties. Hence, the economic impact may not be statewide.
Consistent with prior research, our research assumed that retirement migration is an individual-level phenomenon. Retirement migration may be driven by household-level dynamics, so future research should consider the interaction between the retirement statuses of all adult members of a household.
Footnotes
This research was supported by the National Institute on Aging under Grant R03-AG023813-01. D. Bradley extends additional thanks for support from Eastern Carolina University's Faculty Senate in the form of a Research and Creative Activity Grant. 
1 Department of Sociology, University of North Carolina at Asheville.
2 Department of Sociology, East Carolina University, Greenville, NC.
3 Department of Sociology, Wake Forest University, Winston-Salem, NC.
4 Former Professor of Economics and Director of the Center for the Study of Population, Florida State University, Tallahassee.
Decision Editor: Linda S. Noelker, PhD
Received for publication December 5, 2005. Accepted for publication June 28, 2006.
References
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]
|
|
|
|
|
|
C. F. Longino Jr.,, D. E. Bradley, E. P. Stoller, and W. H. Haas III Predictors of Non-Local Moves Among Older Adults: A Prospective Study J. Gerontol. B. Psychol. Sci. Soc. Sci., January 1, 2008; 63(1): S7 - S14. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
