retirement and interstate migration

17
Retirement and Interstate Migration Author(s): Albert Chevan and Lucy Rose Fischer Source: Social Forces, Vol. 57, No. 4 (Jun., 1979), pp. 1365-1380 Published by: Oxford University Press Stable URL: http://www.jstor.org/stable/2577275 . Accessed: 15/06/2014 00:38 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. . Oxford University Press is collaborating with JSTOR to digitize, preserve and extend access to Social Forces. http://www.jstor.org This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AM All use subject to JSTOR Terms and Conditions

Upload: albert-chevan-and-lucy-rose-fischer

Post on 20-Jan-2017

213 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Retirement and Interstate Migration

Retirement and Interstate MigrationAuthor(s): Albert Chevan and Lucy Rose FischerSource: Social Forces, Vol. 57, No. 4 (Jun., 1979), pp. 1365-1380Published by: Oxford University PressStable URL: http://www.jstor.org/stable/2577275 .

Accessed: 15/06/2014 00:38

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

Oxford University Press is collaborating with JSTOR to digitize, preserve and extend access to Social Forces.

http://www.jstor.org

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions

Page 2: Retirement and Interstate Migration

Retirement and Interstate Migration*

A L B E R T C H E V A N, University of Massachusetts LUCY ROSE FISCHER, University of Massachusetts

ABSTRACT Studies of the relationship of migration to employment have focused on

migration during the early stages of work careers. In this study, migration of the recently retired is observed and found related to two factors unique to the last stage of work careers-retirement and concern for climate. Analysis of 1970 Census data shows that previous lifetime migration as well as the usual social class variables continue to exert an influence on migration. Migration of retirees to the sunshine states and to other states is predicted and differences noted between the two streams. An additive model of effects best describes the relationships observed. A broader understanding of the impact of retirement on migration is gained through a couple analysis in which joint patterns of retirement are observed.

It is by now a truism that migration is related to age. The young migrate often, the old, infrequently. During the period 1965 to 1970, 4 percent of persons age 60 and over changed their state of residence as compared to the more than 10 percent changing state among those under 60. Neverthe- less, 4 percent of the population 60 and over represents over a million people. As Goldscheider has suggested, although there is a general pattern of greater stability among older people, the question yet remains: Who among the elderly migrate? Our major purpose is to establish those factors which account for the redistribution of the elderly between states.

As the mechanism whereby members of the labor force adjust to economic opportunity, migration for the elderly would seem only weakly related to labor force attachment. Goldstein has shown that migration of the elderly may be a consequence of detachment from the labor force. Migration rates for the elderly in the labor reserve in the period 1955 to 1960 were more than double the rates of those elderly who remained in the labor force. These results indicate that migration may be related to retire- ment adjustments the elderly make in living arrangements and place of residence. Barsby and Cox, in their aggregate analysis of migration, found a sharp rise in interstate migration rates around the retirement years. Their

*Without the acumen of George Myers this paper would have been relegated to oblivion. The comments and support of Gordon Sutton are also appreciated. ? 1979, University of North Carolina Press. 0037-77321791570417066$01.60

1365

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions

Page 3: Retirement and Interstate Migration

1366 / Social Forces / vol. 57:4, june 1979

findings and Goldstein's suggest, but by no means confirm, that those undergoing the retirement transition are most responsive to factors asso- ciated with migration. Swayed by their lack of conclusive results Barsby and Cox wrote, "It is clear that full understanding and exploitation of the labor market attachment-migration nexus requires either more powerful statistical techniques or superior data, or both" (81). Atchley, on the other hand, maintains that "When the effects of retirement alone are examined, it becomes clear that retirement has little impact on migration" (101). To settle these issues we will compare the migratory behavior of a large na- tional sample of retirees and non-retirees.

For most labor force members during the later years, migration would probably disrupt jobs and career, as well as community and family attachments. In the absence of labor market considerations, migration after retirement can enhance retirement by introducing the retired worker to a new and more congenial social, community, and climatic situation. The growth of retirement communities is based on this premise. Florida, Ari- zona, and California, defined in this study as the sunshine states, are the predominant destination for elderly migration streams. About half of all retired migrants went to one of these sunshine states. During the 1960-70 decade net migration of the elderly to these three states was about 555,000 (Bureau of the Census, c). Retirement may thus be seen as a precipitating event for migration.

Aside from the impact of retirement, there are various possible in- terpretations for migration of the elderly. Social status is an important issue to be considered. Interstate migration may be expensive-especially with relocation from a northern to a sunshine state. Moreover, income in retire- ment is closely related to levels of income before retirement (cf. Henretta and Campbell), through savings, pension, social security benefits, and other assets. In effect, extension of the advantages of upper and middle class membership into old age is a reasonable assumption. One manifesta- tion of the privileges afforded by a high social ranking would be migration, particularly to the sunshine states.

In addition to economic opportunities social class has other mean- ings. The willingness to start a post-retirement life elsewhere, to innovate, may be related to the social class background of the individual. High levels of educational attainment would be instrumental in generating a willing- ness to consider alternatives to remaining in the area where retirement oc- curs. Well-educated individuals could also conduct a more effective search in choosing among various destinations. In fact, Bultena and Wood have shown that elderly migrants to Arizona and Florida appear disproportion- ately drawn from the higher social and economic levels of the population.

Beyond the effects of social class, a lifetime history of interstate mi- gration might be responsible for broadening the scope of post-retirement adjustments. Those who have once migrated have given evidence of their

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions

Page 4: Retirement and Interstate Migration

Retirement & Migration / 1367

willingness to shift environments. A special influence on the migration of the elderly is climate. The fact that so many and such a large proportion of elderly migrants go to the sunshine states indicates that climate and locale are important determinants of retirement migration. In short, a compre- hensive model of migration of the elderly would tend to be an additive one in which each factor has independent effects, logically as well as sta- tistically. Such factors as retirement, social class, lifetime migration, and climate, in addition to traditional variables such as age, race, sex, and marital status require investigation.

Data

Data are from the 5 percent sample of the 1970 Census of Population and Housing made available by the Bureau of the Census (a). The 5 percent sample in the 1970 Census contains key retrospective questions for 1965 including state of residence, work status, and occupation, as well as a health related question on work disability in 1970. All persons age 60 and over and working in 1965 were extracted from the 1 percent Public Use Sample.

The definition of a retired person is critical for this study. Four ques- tions are used to define a retiree. Persons reported at work in 1965, without a job in 1970, not looking for work during the four weeks preceding the census, and with their last year of work falling between 1964 and 1970 are considered retirees of the last five years. Those respondents reported at work in 1965 and still with a job or seeking work in 1970 are non-retired. By removing from consid- eration individuals who had retired before 1965 and who may have also migrated before 1965, we can more directly examine retirement migration than if these people had been included. Our definition does not preclude the possibility that people who appear retired in 1970 may return at some time to full or part time work. Indeed 4 percent (but only 4%) of older indi- viduals not working in 1965 were working again in 1970. Retirement by stages is common and what may seem to be an all or nothing matter may not be quite that simple (Atchley). Presumably, some of the working el- derly have followed this pattern, for a fourth of them report some Social Security or Railroad Retirement income. For persons not working in 1970 the question on looking for work is central to defining retirement. It is probable that a direct question on retirement would be no better at predict- ing reentry into the labor force than the question on seeking employment. Situational determinants of reentry, such as personal health and job op- portunities, are largely unpredictable circumstances. We note that labor market dynamics make it more difficult for an older person to find a job once he has retired or become unemployed. In short, from an operational viewpoint it is reasonable to consider retirement to have occurred if an

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions

Page 5: Retirement and Interstate Migration

1368 / Social Forces / vol. 57:4, june 1979

individual completely stopped working between 1965 and 1970 under the conditions specified above.

The question on work status in 1965 was not answered by 6 percent of persons 60 and over and these cases are omitted from the sample.1 Other excluded cases are those with an unknown occupation in 1965, those with a discrepancy between the 1965 work status and year last worked, those with an unknown year last worked, substituted cases, and those whose 1965 place of residence was abroad or unknown. Filtering the 109,- 403 cases reported working in 1965 reduces the sample by about 3 percent to 106,023 cases.2

Retrospective questions are generally subject to large response er- rors. However, an evaluation of the 1965 work status question by Miller found the data yield a reasonable approximation to the 1965 labor force (as measured in the April 1965 Current Population Survey).3

Occupations are recoded to an occupational prestige score using the Hodge-Siegel-Rossi scores (National Opinion Research Center) for the 441 census occupations. The average January temperature (Fahrenheit) in the major urban centers of each state during the period 1941 to 1970 (Bureau of the Census, b) is used to represent the winter temperature for the entire state. This value was appended to each record extracted and permits an analysis of the effect of climate upon migration. Income is confined to unearned income.4

The dependent variable in this analysis, migration during the period 1965 to 1970, has three categories, nonmigrant, sunshine migrant and mi- grant elsewhere. As an alternative to two categories of nonmigrant and migrant, this coding scheme seems warranted in view of the highly specific nature of elderly migration. California, Florida, and Arizona are chosen as the sunshine states: the first two because they rank highest among the states in the number of elderly inmigrants and the last two because they rank highest in the rate of elderly inmigration (Bureau of the Census, c). We are uncertain of the date of migration, but we are certain that if migra- tion occurred it did so within five years of retirement. The large differential in migration between the elderly who remain in the labor force and those who retire strongly suggests that retirement usually precedes migration. Rates of migration would be considerably higher for those remaining in the labor force if there were migration in anticipation of retirement.

Social, Demographic, and Climatic Differentials

Table 1 presents basic comparisons of retirees and non-retirees for nine variables. Above all, it is clear that the change in status from worker to retiree is the major influence upon the migration of the elderly. The rate of migration for retirees is more than three times the rate for non-retirees;

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions

Page 6: Retirement and Interstate Migration

Retirement & Migration / 1369

Table 1. PERCENT DISTRIBUTION OF PERSONS BY MIGRATION STATUS 1965-70 AND RETIREMENT STATUS, SOCIAL, DEMOGRAPHIC, AND CLIMATIC VARIABLES

Retirement Status

Ret i red Non-Ret i red

Total Sunshine Migrant Total Sunshine Migrant Variable Migrant Migrant Elsewhere N Migrant Migrant Elsewhere N

Total 7.2 3.4 3.8 38815 2.2 0.7 1.5 67208

Lifetime Migration

Yes 10.4 4.1 6.3 17948 3.5 0.9 2.6 29777 No 4.5 2.8 1.7 20867 1.2 0.5 0.7 37431

Unearned Income, 1969

$2,500+ 9.7 5.5 4.2 12541 3.3 1.4 1.9 7643 Under $2,500 6.0 2.4 3.6 26274 2.1 0.6 1.5 59565

Years of Education Completed

9 or mnore 9.5 4.8 4.7 19316 2.6 o.8 1.8 40084 8 or less 4.9 2.0 2.9 19499 1.4 o.4 1.0 27124

Occupational Prestige Score, 1965

45 or more 10.9 5.7 5.2 10839 3.0 0.9 2.1 21887 Under 45 5.7 2.5 3.2 27976 1.8 o.6 1.2 45321

Temperature, Degrees

26 or less 9.8 5.8 4.0 12763 2.3 0.9 1.4 21701 27 or more 5.9 2.2 3.7 26052 2.1 0.5 1.6 45507

Age, 1965

Under 60 8.6 4.0 4.6 11058 2.2 0.6 1.6 40373 60 and over 6.7 3.2 3.5 27757 2.1 0.7 1.4 26835

Race

White 7.6 3.7 3.9 35615 2.3 0.7 1.6 61731 Black and other 2.9 0.6 2.3 3200 1.0 0.1 0.9 5477

Sex

Male 7.2 3.7 3.5 23981 2.3 0.7 1.6 44938 Female 7.2 2.9 4.3 14834 2.0 0.6 1.4 22270

Marital Status, 1970

Married 7.6 4.1 3.5 25027 2.1 0.7 1.4 46647 Not married 6.4 2.1 4.3 13788 2.5 0.7 1.8 20561

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions

Page 7: Retirement and Interstate Migration

1370 / Social Forces / vol. 57:4, june 1979

sunshine migration is about five times greater for retirees. Differentials between retirees and workers persist for the categories of all variables. Levels of interstate migration for some groups of retirees are comparable to the level of migration in the younger population although, to be sure, there are selected groups among the young who have much higher rates than retirees. None of the listed variables taken by itself effectively discriminates between migrants and nonmigrants among the elderly remaining in the labor force; for relationships are either muted or nonexistent. Retirement allows the social, demographic, and climatic factors to operate freely, and as a result differentials emerge for some variables. Lifetime migration prior to 1965, climate, occupational prestige, educational attainment, unearned income, and race appear as strong influences on the migration of retirees. Differentials for age, sex, and marital status are much weaker.

The destination of migrants is influenced by their characteristics. Retirees who are married, of high social status, living in a cold climate, resident in their state of birth in 1965, and male, are more likely to migrate to the sunshine states than elsewhere; their opposite numbers go else- where more frequently than they go to the sunshine states.

Most older workers were in their state of birth in 1965, and pre- sumably were also there at retirement. Previous lifetime migration affects the amount and direction of migration after retirement. Retirees who in the past have been most reluctant to leave their home state, or who returned to their home state before retirement, display the same reluctance to leave after retirement. In contrast, retirees not resident in their state of birth have weaker ties to the state where they last worked. With work ties sundered these retirees have that much less to hold them and other places, including their home state, may beckon more strongly. Indeed, 31.4 percent of mi- grating retirees who are not in their state of birth on retirement migrate back to their home state. Return migration is not a phenomenon exclu- sively of retirement as Lee has shown, but it probably has a greater sense of finality among the retired than for other migrants.

The popular myth of flight from the cold northern winter climate appears borne out by the results for temperature. Climate signifies more than temperature differences. There are probably differential opportunities associated with climate and region. These take the form of housing costs, general cost of living, and the cost of carrying on outdoor activities, in- cluding traveling-all of which are influenced by climate. In a manner of speaking, climate may represent differentials in winter life style, including social life, which tends to be channeled by the climate. For the elderly retiree in colder climates the restriction placed on the use of leisure time may be as significant as the discomfort of low temperatures.

Income, education, and occupational prestige are similarly related to migration. High status retirees migrate more frequently than low status retirees and they are more likely to go to the sunshine states than else-

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions

Page 8: Retirement and Interstate Migration

Retirement & Migration / 1371

where. Low status retirees avoid the sunshine states and go elsewhere. Selectivity by social status is occurring at origin and at destination. This process implies that there is much greater physical separation of the social classes than customarily occurs with neighborhood segregation.

What is not apparent in Table 1 is that the weak relationship between age and migration occurs in the face of two other relationships, that of health to age at retirement and of health to migration. Persons in poor health retire earlier and migrate less than those in good health. A ques- tion on work disability was asked in the 1970 Census of persons aged 14 through 64 and in the comments which follow we focus on the 10,985 cases retiring before age 65. Among persons retiring at ages 55-59, 55.8 percent reported a work disability, as compared to 43.8 percent at ages 60-62 and 38.0 percent at ages 63-64. The relationship of age to migration is contin- gent on the health of retirees, as is evident from a comparison of migration by disability status. Among retirees not reporting a work disability, 12.6 percent retiring at ages 55-59 were interstate migrants, 11.9 percent at ages 60-62, and 8.0 percent at ages 63-64. Comparable percentages for the disabled were 6.6, 6.4, and 5.5 respectively. Poor health leads to early retirement, reduces migratory inclinations, and dampens the relationship between age and migration.

Multivariate Analysis

The preceding discussion has taken a standard cross-tabulation approach to the explanation of migration among the elderly. Some of the more im- portant factors affecting that migration were introduced by this means. However, it is not apparent from this presentation that an additive model describes the relationship between the dependent and independent vari- ables; nor is it readily apparent which variables are most useful in pre- dicting migration of the elderly. The task of sorting out the independent variables is made difficult by the fact that the dependent variable, migra- tion, is not in a form which conforms to the requirements of the more common multivariate statistical techniques. Migration is a highly skewed, nominally scaled variable with three categories. Given these conditions the log-linear techniques presented by Goodman and explicated by Davis offer a useful solution.

Log-linear models can test the additivity hypothesis through logic similar to the analysis of variance (Davis). Means are provided for testing the effects of single variables or variables in any combination by construct- ing models of effects. Models hypothesizing any order of interaction may be constructed for testing the presence of terms in a prediction equation. A likelihood-ratio chi square statistic is used to test the frequencies resulting from the hypothesized model of effects against the actual frequencies.

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions

Page 9: Retirement and Interstate Migration

1372 / Social Forces / vol. 57:4, june 1979

When the appropriate terms are entered the prediction equations are anal- ogous to a multiple regression. Log-linear regression starts with a baseline model which assumes the dependent variable is unrelated to other vari- ables. Adding terms relating the dependent variable to the independent variable allows tests of the effects of the independent variables. A com- parison of the regression model with the baseline model yields a coefficient of multiple determination (Goodman, b). This coefficient is obtained with the general proportionate reduction in error formula:

R2 = [X2(H") -X2(H')] / X2(H")

H" represents the baseline model and H' represents a regression model with additional terms.

A parsimonious explanation of elderly migration calls for something less than ten independent variables. Log-linear models are an efficient technique for finding out when "there is absolutely nothing going on of any statistical interest" (Davis, 200). Therefore, our first task is to reduce the number of variables by constructing a series of hypotheses (Hi, H2, etc.) and by eliminating those variables which contribute little to predicting migration. Inspection of Table 1 shows that at least age, sex, and marital status are likely candidates for elimination. The top panel of Table 2 con- tains models of the relationship between migration, retirement, education, age, sex, race, and marital status. Each model contains a six-way term controlling for the relationships between the independent variables. Low R2 values indicate that neither age (H4), race (H5), sex (H6), nor marital status (H7), singly or in combination explain much about migration. On the other hand retirement (H2) and education (H3) have considerably higher R2 values and explain a great deal. Together, in H8, the four variables of lesser importance explain little more than education alone. Education is used in this comparison because it is the least influential of the remaining independent variables.

Age, race, sex, and marital status are no longer considered in the bottom panel of Table 2. Climate, lifetime migration, unearned income, and occupational status are added to retirement and education. Model H9 serves as the baseline model in the rest of this analysis and a six-way interaction term is again used to control for the relationships between the independent variables. All variables make a significant contribution to the prediction of migration. In common with all chi-square tests, the likelihood-ratio statistic is sensitive to sample size.5 The coefficient of mul- tiple determination in this case is a more useful measure.

Judging from R2 in models H10 through H15, retirement is the most important factor in migration of the elderly. All the two-way effects but retirement are contained in model H18. This model's R2 of only .597 leaves much room for improvement and thereby reemphasizes the need to in- clude retirement in any discussion of elderly migration. Second to retire-

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions

Page 10: Retirement and Interstate Migration

Retirement & Migration / 1373

Table 2. LOG-LINEAR MODELS OF ELDERLY MIGRATION, 1965-70

Model Terms Hypothesis X2 Df* R2

Hi [1] [2 3 8 9 10 11] Migration is independent of other variables 2,579.40 126 --

H2 [1 2] [2 3 8 9 10 11] Only retirement directly affects migration 934.26 124 .638

H3 [1 3] [2 3 8 9 10 11] Only education directly affects migration 2,280.22 124 .116

H4 [1 8] [2 3 8 9 10 11] Only age directly affects migration 2,516.33 124 .024

H5 [1 9] [2 3 8 9 10 11] Only race directly affects migration 2,379.96 124 .077

H6 [1 10] [2 3 8 9 10 11] Only sex directly affects migration 2,559-19 124 .oo8

H7 [1 11] [2 3 8 9 10 11] Only marital status directly affects migration 2,472.10 124 .042

H8 [1 8] [1 9] [1 10] Age, race, sex, and marital [1 11] [2 3 8 9 10 11] status directly affect

migration 2,199.30 118 .147

H9 [1] [2 3 4 5 6 7] Migration is independent of other variables 4,057.70 126 --

H10 [1 2] [2 3 4 5 6 7] Only retirement directly affects migration 2,412.55 124 .405

Hll [1 3] [2 3 4 5 6 7] Only education directly affects migration 3,758-51 124 .074

H12 [1 4] [2 3 4 5 6 7] Only income directly affects migration 3,329.12 124 .180

H13 [1 5] [2 3 4 5 6 7] Only occupational prestige directly affects migration 3,740.79 124 .078

H14 [1 6] [2 3 4 5 6 7] Only lifetime migration directly affects migration 2,929.68 124 .278

H15 [1 7] [2 3 4 5 6 7] Only climate directly affects migration 3,733.06 124 .080

H16 [1 3] [1 4] [1 5] Education, income and [2 3 4 5 6 7] occupational prestige

directly affect migration 3,025.94 120 .254

H17 [1 2] [1 6] [1 7] Retirement, lifetime migration [2 3 4 5 6 7] and climate directly affect

migration 982.87 120 .758

H18 [1 3] [1 4] [1 5] [1 6] All variables except retirement [1 7] [2 3 4 5 6 7] directly affect migration 1,636.01 116 .597

H19 [1 3] [1 4] [1 5] [1 6] Retirement and climate jointly [1 2 7] [2 3 4 5 6 7] affect migration and all other

variables directly affect migration 175.21 112 .957

H20 [1 2] [1 3] [1 4] [1 5] All variables directly affect [1 6] [1 7][2 3 4 5 6 7] migration 205.88 114 .949

*All models are statistically significant.

Variables (all variables coded as shown in Table 1) 1 = migratory status 7 = temperature, degrees 2 = retirement status 8 = age 3 = years of education completed 9 = race 4 = unearned income, 1969 10 = sex 5 = occupational prestige score, 1965 11 = marital status 6 = lifetime migration

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions

Page 11: Retirement and Interstate Migration

1374 / Social Forces / vol. 57:4, june 1979

ment in accounting for the migration of the elderly is lifetime migration (H14), followed by unearned income (H12). Each of the three remaining variables has a small influence on migration. The three social status vari- ables are combined in H16 and the resulting R2 of .254 is actually slightly less than the effect of lifetime migration by itself. Nonetheless, social status is useful in understanding elderly migration, for omitting the three social status variables in H17 results in an R2 of .758.

All two-way effects are included in H20, the additive model. This model with an R2 of .949, fits the data quite well. Higher order interactions account for the remaining unexplained variation. The presence of one or more significant higher order interaction terms would question and modify the additive model of H20. There are fifteen possible third-order inter- action terms with the dependent variable. The strongest of these is the interaction between retirement, climate, and migration. Retirees from cold weather states migrate to the sunshine states while non-retirees and re- tirees from warmer climates migrate elsewhere. Although the contribution of this term is statistically significant, R2 is increased to only .957 by its inclusion in model H19. Whatever variation remains from model H20 is well dispersed among higher order interaction terms. Some indication of this wide distribution is given by the analysis of variation in the top panel of Table 3. Half of all the remaining variation involves third-order effects, while fourth-, fifth-, and sixth-order interactions account for the rest. We

Table 3. ANALYSIS OF VARIATION IN ELDERLY MIGRATION FOR SELECTED MODELS

Percent

2 Distribution Source of Variation X Source Variationi of Variation

Total effect of all interactions with migration involving two or 2 more independent variables X (H20) 205.88 100.0

Total effect of all interactions All fourth, fifth, and with migration involving three sixth order interactions or more independent variables with migration 103.32 50.2

Total effect of all interactions with migration involving two 2 independent variables X (H20)-(model above) 102.56 49.8

Total effect of retirement X2(H9)-X2(H10) 1,645.15 100.0

Direct effect of retirement unmediated by other variables X2(H18)-X2(H20) 1,430.13 86.9

Indirect effect of retirement X2(H9)-X2 (H10)- through other variables [X2(H18)-X2(H20)] 215.02 13.1

Total effect of social status X2(H9)-X2(H16) 1,031.76 100.0

Direct effect of social status 2 2 unmediated by other variables X (H17)-X (H20) 776.99 75.3

Indirect effect of social status X2 (H9) -X2 H16)- through other variables [X2(H17)-X (H20)] 254.77 24.7

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions

Page 12: Retirement and Interstate Migration

Retirement & Migration / 1375

may conclude that although higher order interactions may be present, their effects are relatively weak and well dispersed. The additive model of H20 is a reasonable description of how elderly migration occurs. A good illustration of this point is given by the extreme cases of the seven dimen- sion cross-tabulation used in the analysis. Among the 601 retired persons of high social status, living in a cold climate before retirement, and born outside their 1965 state of residence, 27.6 percent migrated between 1965 and 1970. In contrast, of the 8,880 elderly who were opposite on each characteristic, 0.5 percent migrated (interstate) during the same period.

Higher levels of migration for the retired than for the non-retired may be a consequence of their differences in composition on the other five characteristics. An analysis of variation is used to decompose the effects of retirement into its direct and indirect or compositional effects in the middle panel of Table 3. Most of the effect of retirement, 87 percent, is direct and is unmediated by the other five independent variables. It is not completely clear that a similar composition of retirees and non-retirees on other variables would necessarily reduce migration. Non-retirees have not yet attained their full retirement income and retirees appear inordinately wealthier because of the way the income variable is used. However, retirees are generally of lower social status than non-retirees and compositional equivalence would increase the migration differential. Thus, the indirect effect of retirement stems from a balance of opposing compositional effects.

A large compositional effect for the social status variables would also challenge the additive model. Indirect and direct effects of the social status variables are shown in the bottom panel of Table 3. Indirect effects are substantial at 25 percent, but still leave a large direct effect from these variables.

Couple Retirement and Migration

Retirement and migration usually occur in a family context. More than two-thirds of our sample were married in 1970 and for these people the migration decision is potentially contingent on two work and retirement patterns. Most persons do not migrate when they retire and part of the reason for this may be bound up with the continued presence in the labor force of a spouse. Because the Public Use Sample is a household sample it is feasible to observe the joint work and labor force status of husband and wife. To that end 60,624 couples were extracted who met four criteria in addition to those used in the individual analysis: (1) either or both spouses working in 1965; (2) at least one spouse 60 or over in 1970; (3) insofar as the Public Use Sample allows, no indication of marrying between 1965 and 1970;6 and (4) both spouses living in the same state in 1965.

Rather than adding the effects of variables as in the regression

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions

Page 13: Retirement and Interstate Migration

1376 / Social Forces / vol. 57:4, june 1979

analysis, the effects of events are added in this analysis. Reasoning from what we already know about retirement and migration, we expect a sub- stantial impact when husband and wife retire within a short time of one another. Given equal rates of migration between the sexes, we also expect the impact of the husband's or wife's retirement to be similar.

Eight joint work and retirement patterns are presented in Table 4. Of the several comparisons available, the most striking is between couples with both spouses working at both dates and couples with both spouses retiring between 1965 and 1970 (patterns 1 and 4). Overall migration rates are more than seven times greater for retiring couples than for working couples and the differential for sunshine migration is even larger. Retire- ment of both spouses within a short time of one another may be part of a scheme in which migration and retirement plans are intermixed. Couples who do not coordinate their retirement may be less likely to think in these terms.

Comparatively high rates of migration are found among couples for whom the retirement of either one or both spouses left neither spouse in the labor force by 1970 (patterns 4, 7, and 8). Retirement of one spouse only, with the other remaining in the labor force, does not lead to high rates of migration (patterns 2 and 3). The effect of retirement under these circumstances is only slightly greater than no retirement at all (patterns 1, 5, and 6). Migration of the married is thus contingent on a dual pattern of withdrawal from the labor force and the retirement of one spouse is insufficient to initiate migration unless the other spouse also retires. Sex differences are present, for the retirement of the wife leads to slightly higher migration rates than the retirement of the husband (patterns 2 vs. 3

Table 4. PERCENT DISTRIBUTION OF MARRIED COUPLES MIGRATING 1965-70 BY WORK AND LABOR FORCE STATUS OF HUSBAND AND WIFE, 1965 AND 1970

Work Status, 1965 and Labor Force Status, 1970 Migratory Status

Total Sunshine Migrant Husband Wife Migrant Migrant Elsewhere N

Pattern All Couples 1965 1970 1965 1970 3.9 1.8 2.1 60,614

1 In In In In 1.7 0.5 1.2 11,827

2 In In In Out 3.8 1.4 2.4 3,890

3 In Out In In 2.6 1.3 1.3 3,015

4 In Out In Out 12.9 7.2 5.7 2,880

5 In In Out Out 2.2 0.6 1.6 22,608

6 Out Out In In 2.1 o.8 1.3 2,567

7 In Out Out Out 7.5 4.2 3.3 12,143

8 Out Out In Out 7.8 3.5 4.3 1,694

In = working in 1965 or in labor force in 1970.

Out = not working in 1965 or out of labor force in 1970.

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions

Page 14: Retirement and Interstate Migration

Retirement & Migration / 1377

and 7 vs. 8). These results for couples indicate that migration of the elderly cannot be separated from the family context. Characteristics of the hus- band and wife have a joint influence which the usual case mode of analysis misses.

Census data allow only a partial view of family influence on the decision to migrate. Retirees may remain in place to be near kin, or they may migrate for the same reason, either returning home or following in the wake of family members. We do observe from our data that in their new homes migrant couples are less likely to have a child in the household and are more likely to live alone than nonmigrant couples. Both tendencies are most pronounced among sunshine migrants. Conversely, unmarried mi- grants are more likely to have a child present and less likely to live alone than unmarried nonmigrants. Although they suggest the importance of family, neither set of observations addresses the possibly greater impact of kin in the area of residence.

Conclusions

Migration of the elderly exhibits several of the differentials identified with migration of younger populations. For instance, educational attainment, occupational prestige, and income are commonly associated with interstate migration (Shryock). Scant attention has been paid in previous research on elderly migration to the relationship between state of residence and place of birth, but this differential is not unique to the elderly either as Eldridge has demonstrated. Two factors set the migration of the elderly apart from the migration of younger age groups-retirement and climate. These in turn lead to a third major distinction between the migration of old and young-the orientation of elderly migrants to the sunshine states. Though there may not be a mass movement of most elderly toward the sunshine states, the stereotypical image of migrating retirees may not be entirely mythical.

Little sense can be made of the migration of the elderly unless retire- ment is taken into account. Retirement lends an even more interesting explanation to elderly migration in a married couples context. Knowledge of both husband and wife is needed to predict the relationship of migration to retirement. Without the couple-analysis, the full impact of retirement on migration is not clearly seen.

Migration of the elderly is predictable, but social class, previous lifetime migration, climate, and retirement are disparate elements in that prediction scheme. The variables which predict interstate migration of the elderly do not fit into any neat package. One way to combine some of these disparate elements would be to construct a "cosmopolitan" model of el- derly migration. Research by Bultena and Wood indicates that one possible

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions

Page 15: Retirement and Interstate Migration

1378 / Social Forces / vol. 57:4, june 1979

interpretation of social class and previous lifetime migration is a sense of cosmopolitanism-a readiness to change environments upon the part of some elderly. But, although this model may explain some of our findings, it does not provide a comprehensive interpretation of elderly migration. We are left, then, with an additive model in which interstate migration of the elderly can be seen as occurring at the intersection of those variables which weighed heavily in the log-linear regression. From this model we note that it is not by any means the case that the elderly have uniformly low rates of interstate migration. In the early years of retirement, and particularly among some segments of the elderly, the rate of interstate mi- gration attains a level considerably above the years preceding retirement.

The additive model is useful for speculating on future interstate migration of the elderly. Retirement income, educational attainment, and the occupational level of retirees are certain to increase in future cohorts of retirees. With this increase should come more interstate migration, much of it to the sunshine states. And as more older workers are found outside their state of birth at retirement, interstate migration of the elderly may also be expected to grow. An increase in early retirement, when individuals are in good health, would raise rates of migration in late middle age. Although climate is more or less a constant, the cost of countering cold winters is not and should not be discounted as a possible future influence on migration of the elderly. Efforts by commercial interests promoting sunshine retirement while playing on this point are potentially important. Such efforts in the past have indirectly as well as directly stimulated migra- tion by popularizing the idea of "sunshine" retirement. Operating counter to these trends would be the increase in southerly migration among current labor force members. On balance, an increase in elderly migration appears in store.

In dealing with interstate migration other forms of residential mobil- ity have been ignored. In particular we cannot say how retirement effects intrastate or intracounty mobility; nor can we tell which other factors are operative. All forms of mobility taken together involve substantial num- bers of older people. One byproduct of interstate migration is an apparent change in the social class composition of the elderly population, at least at the state level. This process may be either strengthened or moderated at the community level by other forms of residential mobility. If strengthened, the process could have important social and political consequences for the elderly.

Notes

1. Unknown responses are routinely allocated among the labor force categories, but allocation was not carried out for this question. The rate of no response for the 1965 work question was only slightly greater than for the 1970 labor force question, which was allocated. We do know that among persons 60 and over, 80 percent of those not answering the 1965 work status

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions

Page 16: Retirement and Interstate Migration

Retirement & Migration / 1379

question were not in the labor force in 1970 as compared to 37 percent among those who said they were working in 1965. Thus we are led to believe that most persons who failed to respond, although not all, indicated they were not working in 1965 by not bothering to answer the question. 2. Cases with allocated items are also candidates for exclusion from the sample. One implica- tion of Rockwell's investigation of allocation is that purging such cases may be sometimes advisable. To our knowledge this step has not been taken by any user of the Public Use Samples. Allocation for age was slightly greater in the older than the younger population (Bureau of the Census, c). Removal of cases with an allocation for age, sex, year last worked, or 1970 labor force status had no effect on the results. 3. This is not to say that the 1965 older labor force can be fully reconstructed, for an older population is subject to greater mortality than the younger segment of the labor force. A shorter migratory period would reduce the influence of mortality on the sample, but would also reduce the number of migrants and retirees. Linkage of retirement and migration would also be more problematic. Older labor force members are probably healthier and younger than the rest of the elderly population. It is unlikely that mortality seriously affects the relationship between migration and the variables investigated, although the migration rate will be on the high side because those who died were less likely to have migrated. By applying reverse life table survival rates to the survivors of the 1965 labor force it is possible to estimate the effect of mortality. Application of the average of 1965 and 1970 life table survival rates by age, race, and sex, indicates that about 15 percent of the 1965 labor force had died by 1970. 4. Unearned income is the sum of three identified sources of income: (1) Social Security and Railroad Retirement payments; (2) Public Assistance income; (3) all other unearned income. Unearned income is also a better predictor of migration among workers than is total income or earnings income. 5. The size of chi square is a linear function of sample size given a constant relationship. With a large sample, such as the one in this study, the risk of rejecting the null hypothesis when it is true is increased. 6. Census data identify only the year of first marriage. We assume that husband and wife migrated together. Our assumption is incorrect if husband and wife were both married more than once and their marriage occurred after migration.

References

Atchley, R. C. 1976. The Sociology of Retirement. Cambridge: Schenkman. Barsby, S. L., and D. R. Cox. 1975. Interstate Migration of the Elderly. Lexington,

Mass.: Lexington Books. Bultena, G., and V. Wood. 1969. "Normative Attitudes toward the Aged Role

among Migrant and Nonmigrant Retirees." Gerontologist 9(Autumn):204-8. Bureau of the Census. a:1972. Public Use Samples of Basic Records from the 1970

Census. Washington: Government Printing Office. . b:1975. Statistical Abstract of the United States. Washington: Government

Printing Office. . c:1976. "Demographic Aspects of Aging and the Older Population of the

United States." Current Population Reports: Special Studies; Series P-23, No. 59. Washington: Government Printing Office.

Davis, J. A. 1974. "Hierarchical Models for Significance Tests in Multivariate Con- tingency Tables: An Exegesis of Goodman's Recent Papers." In H. L. Costner (ed.), Sociological Methodology: 1973-1974. San Francisco: Jossey-Bass.

Eldridge, H. T. 1965. "Primary, Secondary, and Return Migration in the United States." Demography 2:444-55.

Goldscheider, C. 1971. Population, Modernization and Social Structure. Boston: Little, Brown.

Goldstein, S. 1967. "Socio-economic and Migration Differentials between the Aged in the Labor Force and in the Labor Reserve." The Gerontologist 7(March):31-41.

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions

Page 17: Retirement and Interstate Migration

1380 / Social Forces / vol. 57:4, june 1979

Goodman, L. A. a:1972. "A Modified Multiple Regression Approach to the Analysis of Dichotomous Variables." American Sociological Review 37(February):28-46.

. b:1972. "A General Model for the Analysis of Surveys." American Journal of Sociology 77(May):1035-86.

Henretta, J. C., and R. T. Campbell. 1976. "Status Attainment and Status Main- tenance: A Study of Stratification in Old Age." American Sociological Review 41(December):981-92.

Lee, A. S. 1974. "Return Migration in the United States." International Migration Review 8(Summer):283-300.

Miller, A. R. 1976. "Retrospective Data on Work Status in the 1970 Census of Population: An Attempt at Evaluation." Journal of the American Statistical Asso- ciation 71(June):286-92.

National Opinion Research Center. 1974. Codebook for the Spring 1974 General Social Survey. Chicago: National Opinion Research Center.

Rockwell, R. C. 1975. "An Investigation of Imputation and Differential Quality of Data in the 1970 Census." Journal of the American Statistical Association 70 (March):39-42.

Shryock, H. S. 1964. Population Mobility Within the United States. Chicago: Com- munity and Family Study Center, University of Chicago.

This content downloaded from 195.34.79.228 on Sun, 15 Jun 2014 00:38:38 AMAll use subject to JSTOR Terms and Conditions