UNIVERSITY OF
ILLINOIS LIBRARY
AT UR8 ANA-CHAMPAIGN
AGRICULTURE
AGRIGIMTIW mv^,
DEC 1 1 1989
These staff papers are published at the discretion of their authors who are
solely responsible for the decision to publish as well as for the contents.
UNIVERSITY OF IUINOIS
AGRICULTURE LIBRARY
ILLINOIS
AGRICULTURAL ECONOMICS
STAFF PAPER
Department of Agricultural Economics
University of Illinois at Urbana-Champaign
305 Mumford Hall, Urbana, IL 61801
Series S, Rural Sociology
THE UTILITY OF DISAGGREGATING THE MIGRATION DECISION
MAKING PROCESS: A SUBSTANTIVE EXAMPLE
by
James D. Williams and David Byron McMillen
December, I978 78-S7
Digitized by the Internet Archive
in 2012 with funding from
University of Illinois Urbana-Champaign
http://archive.org/details/utilityofdisaggrOOwill
Introduction
With the turnaround phenomenon has come an increasing Interest in
traditionally less important and, particularly, noneconomic motivations
for migration. Certainly there is in nonmetropolitan America a great
variety of stimuli for recent net inmigration patterns which reflect both
economic and noneconomic "pull" factors. But, as Wilbur Zelinsky rather
eloquently notes, "the economic-sum-metropolitan-sprawl explanatory strategy
collapses when we confront those hundreds of remote, thinly settled, and
emphatically bucolic counties for whose recent demographic resurgence
there is no halfway plausible economic rationale" ( 1977:176).
The purpose of this paper is to demonstrate the usefulness of incor-
porating recent developments in migration decision making theory into the
design of surveys eliciting, from respondents, the salient evaluative
dimensions involved in the decision making process. While our suggestions
are generally applicable to any survey including "reasons for moving"
questions, we argue that there is a heightened need to improve upon the
standard "why did you move?" approach when surveying migrants who are
likely to have moved for nontraditional reasons. In particular, it is sug-
gested that investigators operational ize the evaluative dimensions for
migration behavior in terms of at least two decisions: (1) the decision
to leave a place of origin, or outmigrate, and (2) the choice of destina-
tion, or basis for inmigration.
While we intend to demonstrate the utility of this approach with
particular reference to nonmetropolitan "amenity area" inmigrants, we may
note that our suggestions could prove important in the future even for
surveys of all migrants. Indeed, if future flows increasingly come to
„2-
reflect rather nontraditional evaluative bases Ifor migration behavior, tl
national random sample surveys would benefit from the greater specificity
which we propose for operationalizing migration motivations.
Data
In the course of demonstrating the utility of operationally disaggre-
gating the migration decision making process, we will employ an analysis
of data from a recent midwestern study of inmigrants to rapidly growing
nonmetropolitan counties. In the next few paragraphs, we describe relevant
aspects of the study design.
As of November, 1975, there were 866 nonmetropolitan counties in the
12 state North Central Region. On the basis of estimates published yearly
by the Bureau of the Census, we identified and selected all 75 nonmetro-
politan counties which had greater than 10 percent (1970 base population)
net migration between 1970 and 1975. This target group contained no coun-
ties in Iowa or Kansas, while Missouri and Michigan accounted for 24 and
21 counties, respectively. Forty-eight of the counties contained no urban
place in 1970, and 25 of the counties were adjacent to an SMSA in 1975.
Within these high net inmigration counties a survey population of 316, •
households with telephones was estimated from 1975 census estimates of hous<
holds and 1970 estimates of telephone coverage for the target counties.
For each county, all telephone exchange areas were identified and the most
recent directories (1976 or 1977) were obtained. From these directories
a systematic sample of 11,329 households was drawn using a sampling inter-
val of 1/28 excluding, as much as possible, double and business listings.
-3-
The sources of bias are those normally associated with telephone sur-
veys: households without telephones or with unlisted numbers. The aver,
telephone coverage of households for the target counties was 82.5 percent
in 1970. Only six counties, which accounted for less than 4 percent of
the survey population had phone coverage of less than 70 percent. Estimates
by the Bureau of the Census indicate that national phone coverage has in-
creased since 1970 and thus the 1970 phone coverage data may overestimate
the potential for bias. Available data indicate that unlisted numbers
are only a problem in large metropolitan areas and thus present virtually
no source of bias in this study. A further potential source of bias unique
to this study is the tendency for recent inmigrants to be excluded from
telephone listings. Only five inmigrant households were located which had
moved in in 1977, though the distribution of migrants by year of inmigration
is fairly regular for 1970-1976.
In order to maximize the probability of obtaining an inmigrant on any
es
given call, the sample names, address/, and phone numbers were matched with
the appropriate 1970 telephone directory. This matching, performed at the
Library of Congress, yielded two strata: (1) expected resident (matched)
households, and (2) expected inmigrant (unmatched) households. Problems
arising with common surnames, intra-county migrants, and redistricting of
telephone exchange areas were handled by treating all ambiguous cases as
unmatched and placing them in the expected migrant stratum.
Within the survey population of households, three respondent types we
identified, and quotas established, for subsequent disproportionately strati-
fied sampling: (1) continuous residents of the counties since April, 1970:
(2) inmigrants since April 1970 who had moved from an SMSA county; and
-4_
(3) inmigrants since April 1970 who had moved ftrom a non-SMSA county.
Resident status and migrant type were determined from a series of initial
screening questions. The various selection rules and probabilities of selec-
tion yielded interviews with 500 metropolitan migrants and 208 interviews
with nonmetropolitan migrants. The resident sample is not used in this
analysis.
Heads of households were the primary respondents, though spouses were
interviewed after several unsuccessful attempts at contacting the head. We
are thus studying household rather than individual migration. Only persons
reporting the current location as their usual place of residence were inter-
viewed and thus seasonal residents were excluded. The very few households
which came through the expected resident stratum and which turned out to be
inmigrants (out and back in during the interval 1970-197 7) are excluded
from this analysis.
The refusal rate on the screening section was 3.7 percent, and on the
main interview it was 9 percent for the metro migrants and 3 percent for the
nonmetro among contacted households. Interviewing was conducted in the
spring and early summer of 1977. Interviews lasted from 30 to 50 minutes
and interviewers reported that respondents generally were very cooperative.
Indeed, the low refusal rate and interviewer perceptions suggest considerable
ease in interviewing in these predominantly rural areas.
In the subsequent analysis, the two migrant substrata (metropolitan
and nonmetropolitan origin) have been combined. As the numbers of completed
interviews are the result of complex sampling and not simple random sampling,
the two migrant substrata have been weighted to reflect estimated proportir
-.5-
i
representation in the population. Weighting of! the two migrant groups has
been performed in such a way as to maintain the the number of total inter-
views. The metro-nonmetio odds are altered from the interviewed ratio of
about 5:2 to an estimated 4:3. This rather minor adjustment does not, in
our opinion, require extensive alterations in the formulae for significance
testing in the subsequent analysis. Our statistical analysis will treat
the data as if it were the result of simple random sampling. Use of weighted
data causes slight discrepancies in Table frequencies.
Rationale
In migration research, motivation has been investigated indirectly
by inferring motives from individual and household characteristics and from
contextual factors, and directly investigated by enumerating evaluative
dimensions, or reasons given by respondents who have migrated. Both approach
es involve problems. Of particular concern to this research we should note
that stated reasons may involve rationalization, or people may respond in
socially acceptable ways, not know why they moved, or give such vague
answers as to be useless (Lansing and Mueller, 1967). In spite of these
problems, though, Rossi concluded that reason analysis is "particularly
applicable for human actions which involve a conscious choice among alter-
natives ..." (1955-124).
Implicit in any reason analysis is the assumption that the relative
importance of each of the members of a set. of reasons obtained from a sam-
ple of respondents reflects the relative importance of that factor as a
cause of migration. The classification of reasons into some coding scheme
represents the researcher's efforts to obtain a simplified schema which
will maximize within-category homogeneity and across-category heterogeneity.
In a final tabulation, one might see "economic" reasons contrasted with
-6-
!
"social" reasons and depending on the proportions in each category, the
data may be used to suggest which set of reasons captures the greatest:
share of the "explanation of migration."
To date there has been little systematic concern with the wording of
questions designed to elicit the evaluative dimensions for migration
decision making. However, it is clear that some scholars, and geographers
in particular, view migration decision making as involving more than one
decision. If more than one decision is involved, then there is more than
one behavior to be explained and the causes of each need not be the same.
Thus, the reason structures obtained from a sample of respondents for
different migration-related decisions may themselves differ, and reflect
the differing bases of causation for the different behaviors involved.
For intraurban mobility in particular. Brown and Moore (1970) have
suggested that migration involves, for a significant number of migrants,
at least two" decisions: (1) The decision to leave an area of origin, and
(2) the decision of where to move (see also Roseman, 1977). The causal
bases of the~ first help explain outmigration while the causal bases of the
second decision help explain immigration when evaluated at point of origin
and point of destination, respectively. ►
Wolpert (1965), in conceptualizing migration decision making, implies
that these two decisions are not necessarily separate but rather suggests
that the individual tends to simultaneously evaluate the present residence
in the context of alternative residences. His concept of place utility
encompasses both an evaluation of the current residence and an evaluation
of alternative residences. Place utility tends to be operationalized through
satisfaction measures. Similar approaches are evidenced in the work of Rossi
-7-
(1955) and Speare,et al. (1974) as well as Brown and Moore (1970).
Migration is but one of numerous possible responses to the disequili-
brium which results from dissatisfaction at a place of residence. The
individual might alternatively restructure the environment or alter desires
and expectations. The stresses which cause a consideration of migration
among alternative behaviors, are, in turn, a function of variables familiar
to migration researchers, such as changes associated with life or career cycle
development.
The extent to which migration is viewed as a viable option in response
to stress is importantly related to the individual's capability to evaluate
alternative residences, and thus to the formation of place utilities. With-
out an alternative location as a reference point of comparison, we may
surmise that the individual is not likely to leave the current location.
Thus, the factors which impinge upon the destination selection and evaluation
process may also affect the initial decision to leave an area of origin.
Development of the concept of place utility, however, has been greatest
with reference to the process of destination selection. Wolpert, assuming
"intendedly rational" behavior wherein individuals engage in an evaluation
process which can be flawed, writes that "...the utility with respect to
. . . alternative sites consists largely of anticipated utility and optimism
which lacks reinforcement of past rewards. This is precisely why the stream
of information is so important in long distance migration — information about
prospects must somehow compensate for the absence of personal experience"
(1965:162).
-8-
The concept of search space describes a sujbset of places within an
i
awareness space (Brown and Longbrake, 1970). Awareness space contains
the places about which a potential migrant has some information, no matter
how limited. The search space contains only those seriously evaluated,
or those for which place utilities are formed. Thus we see that informa-
tion sources may determine the number of places in the ultimate search
space by determining the places in the initial awareness space.
DaVanzo and Morrison have recently introduced the phrase "location
specific capital" as a "generic term denoting any or all of the diverse
factors that 'tie' a person to a particular place" (1978:8) . They find,
in their analysis, empirical support for the hypothesis that "when a per-
son who has migrated moves again, he or she should favor some former
place of residence as the destination because the person has location
specific capital there" (1978:8). Thus, location specific capital is
suggested to determine the direction of migration. When viewed also as
a general influence on the extensiveness of the awareness space, which in
turn affects the decision to leave, then location specific capital may also
be a determinant of the degree of migration.
In their work, DaVanzo and Morrison were attempting to explain return
migration. Two other central hypotheses of their research are: (1) That
location specific capital depreciates over time, and; (2) that if a person
miscalculates net place utilities, moves, and finds the move to be an unwise
investment, the migrant then has superior information about the place recent-
ly left and will tend to return rather quickly. They are, thus, discussing
hypotheses about factors which influence awareness and search space as well
as place utility formation.
-9-
The last several paragraphs provide a brief and much simplified over-
view of concepts and approaches to be used in subsequent analysis. At this
point let us turn to direct evidence suggesting the utility of enumerating
the evaluative dimensions, or reasons for both out- and inmigration
behaviors.
Demonstration
In the current study, respondents were asked questions designed to
elicit reasons for leaving the place of origin and criteria for destination
selection. For the former, respondents were simply asked why they decided
to leave (origin city name inserted) . We elicited up to three reasons and
these were recorded verbatim. About 26 percent of respondents gave more
than one reason and for these a subsequent question asked which reason the
respondent felt was the most important one. The following data refer to
one "main" reason for leaving. Reasons related to destination selection
are based upon a question asking the respondent why s/he picked "this"
place instead of some other. Again, we report data for only one cited
reason.
The open-ended responses to the reason questions were later coded into
an initial 62 category scheme allowing for considerable specificity of
responses. In order to assure reliable results, the coding of all reasons
ouesM~«.a was performed independently three times. Where inter-coder dis-
crepancies occurred, differences were arbitrated and necessary changes made,
The most obvious approach to demonstrating the importance of each ques-
tion is to simply examine the marginal distributions in order to observe
differences in evaluative dimensions reported by respondents. The distri-
butions of responses to both questions are presented in Table 1.
-10-
Optimally, one should apply exactly the sie classification scheme
to both sets of reasons in order to make appropriate comparisons. However,
precisely because we are dealing with different behaviors, we must note
that it was not completely possible to apply identical classification
schemes. For about 15 percent of the sample, retirement was given as
the reason for leaving the place of origin. In contrast, retirement is
not an appropriate response to a question asking why a respondent chose
the particular destination. In three cases, however, retirement was men-
tioned as the reason for choosing the destination. These cases have been
receded to the "other-other" response category eliminating retirement as
a possible basis for destination selection.
It is quite apparent from the distributions that the evaluative dimen-
sions for the two decisions differ. In particular, we may note that nearly
half (47.6 percent) of all respondents chose their destination on the basis
of location specific capital in a variety of forms while only about 18 per-
cent decided to leave for tie-related reasons. Clearly the causal bases
of in- and outraigration would appear to differ.
From a somewhat more formal perspective, the marginals in Table 1 ex-
press the net results of some relationship between the evaluative dimensions
for the two decisions. The greater the relationship between evaluative
criteria for the two decisions, the less is the need for separate operatlon-
al%\P^ 1V° imp°rtantly Afferent types of migrants may result in a
high^relationship between reasons for leaving and for destination selection.
The first and most obvious of these is the migrant who reports very similar
bases for leaving and for destination selection. In our categorization
scheme, this would be, for instance, a person who reports an employment-relate
reason for both behaviors. In a statistical sense, we need not have asked
both questions for persons responding to both questions in similar ways.
-11-
1
The second source of a relationship between criteria for the two
behaviors is attributable to the nature of the migration decision making
process. For some migrants, the basis for initially deciding to leave an
origin minimizes the process of search space formulation so that no des-
tination selection process can be separated from the decision to leave the
area of origin. The prime example is the person who reports having left
in order to move "back home." We would expect (and find) that this mi-
grant's basis for selecting a destination is that the place is "home."
Thus, for some migrants, the response to the question on leaving simul-
taneously, and rightfully, determines the response to the question on
destination selection resulting in a boost in the relationship between the
two evaluative dimension sets of responses. For these migrants, all impor-
tant information related to the entire decision-making process being inves-
tigated here is contained in the response to why the person left their
origin and we need not have asked about destination selection.
We may further illustrate this point by examining a bivariate table
relating gross categories of reason for leaving to reason for picking the
destination as presented in Table 2 (see Technical Appendix for discussion
of significance testing.) Results of siunificanee testine for this and
subsequent tables are summarized in Table 4. For Table 2f for instance, wo
fit a classic independence model based on the expected frequencies general
using the observed marginals. This is symbolized as (1) (2) in Table 4.
Since the chi-square is large and significant, we know that there is a strong
relationship between reasons for leaving and basis for selecting a destina-
tion in Table 2.
-12-
Froui Table 2, we may observe that much of the relationship between the
two sets of reasons derives from the influence of the category combinations
of "employment-employment" and "ties-ties.1' Embedded in these cells, however,
are both sources of relationship just Identified. The reasons given by
these respondents for leaving their origins provide a clue as to whether
they appear in identical reason categories because of identical evaluative
dimensions in a two-step decision-making process, or whether they simultan-
eously chose a destination given a certain reason for leaving.
Consider, first, those who report employment-related reasons as both
the basis for leaving and choosing a destination. These migrants are moving
for relatively traditional reasons. Based upon a reason for moving ques-
tion, transfers and searches for new or better employment accounted for
nearly half of all interstate moves among respondents in the U.S. Annual
Housing Surveys, 1974-1976 (Long and Hansen, 1978). Within the employment-
employment cell, however, are both sources of the equivalence of responses.
The person who is transferred, for instance, is quite different in terms
of decision-making process than the person who reports leaving to look
for better employment. Specifically, the transferee, and for that matter
the person who left because of finding a new or better job, has not engaged
in any destination selection process separable from the decision to leave
as operational ized here. Regardless of what process of search space formu-
lation may precede temporally the decision to leave, for our purposes we
would expect a tendency for equivalence in responses among transferees and
those who left because they found another job. In contrast, the person
who left in order toJLiniLa better job engages in a conceptually, given our
- 1 3-
classification scheme* distinct process of destination selection.
Those migrants suggesting that they Initially decider! to leave their.
origins because of location specific capital at an already chosen destina-
tion are conceptually airailar Co transferees and those who report, having
found a better job as the reason for leaving, I'frK destination choice is
inseparable linked to the reason for leaving and it makes Little sense to
a.sk separately about destination selection, /is a: result, we find about IH
percent of those leaving for tie reasons also choosing their destination
for tie reasons, as seen in Table 2.
For the rest of this paper we will want to focus on the process of
destination selection, and so should re.str.ixt our analysis to only chose
for whom a separation of the decisions to leave and choose a destination
seems reasonable. For convenience.., we mist. Ignore, some variability and
specify ideal types. We have deleted from furthest' analysis all persons
who reported reasons for leaving for which destination selection has no
separate meaning. The categories involved are indicated by asterisks in.
Table 1 and include persons leaving because of a transfer, or because new
employment was located » or because of a desire to maximlsse some farm of:
location specific capital. In. ail, we have deleted 247 households or about
35 percent of the sample, For these people especially there is no empirical
reason, to ask a separate question eliciting the evaluative basis for des-
tine t i o n s e 1 e c t i o n ..
The removal of these "simultaneous" decision-makers reduces appreciably
the level of the relationship between reasons for the. two behaviors, though
the disjunctures in categorizations tend to make statistical interpretation
problematic. The data are presented in Table 3 and the relationship regains
significant as shown in Table 4. We now have, left in the table tvo types of
i&
migrants of further relevance .to this paper; we have those who have the
sane bases for both decisions said w»a have those with different bases..
For 143 cases in table 3, or about 32 percent of these norj.sij»ulta£a.eou"S
decision makers we technically need not have asked both questions (ceils
arid othe r-ot.h« r ) ..
eraployraent~esn^ioyment and environment "-e.ttV<lro.iiment/ Fox the rest, Che
evaluative dimensions, our key to causal liases of the two behaviors, are
different with respect to out-arid ixuaigratio-ia. Thus,, for about 45 percent
of all interviewed migrants, there is clear empirical justification for
asking both questions.
Substantively., the patterns in Table 3 are quite reveal log. Those
initially motivated to leave for lob reasons tend to choose a destination on
the basis of job- related criteria (57 percent).. Thus., there lis a tendency
for more traditionally motivated Kd.giran.ts to make both decisions osa the
basis of similar evaluative criteria and thus less statistical need for
enumerating the reasons for both behaviors. But, there remains, among those
whose leaving was employment motivated, an additional 43 percent who selected
their destination on the basis of a different criterion, especially ties
or location specific capital (32 percent). From date; not displayed,, we find
that their ties are generally in the fern: of a job or business ia the area.
In contrast to employment motivated migrants* retirees, especially
important to recent patterns ol nonmetropolitan faJcLgration., tend to have
selected their destinations on the basis of location specific capital in
a variety of forms, including family and friends and prior residence as well
as property and vaction experience. Those motivated to leave because of
environmental reasons most often suggest destination selection on the basis
of environmental reasons (4. 3 percent) but also draw heavily upon location
- 1 rv
specific capital (41 percent) .
If i« quite clear i. hat. Kith the possible exception ot" those k&o left
for job related reasons, destination selection is import antly a function o:l
location specific capital, awoag migrants by wkoffi we can reasonably suggest
at least two decisions were joade* Ettrcheruio-fre for the two suoal important
types of iumigraatB in terms of reasons for Leaving* those who responded with
environmental reasons ox who cited Eetirtaneatjwe would hav/e underestimated
the role of location specific, capital in the decision staking, process had \:c
not also asked the basis for destination select: ion.
As demonstrated In the detailed categorization scheme of ^Cahle 1»
location specific capital has ht;en utilized by these vaigras.ts in a variety
of forms. Soashe. have chosen their destination in order to be closer to family
or friends,, others ytnply stated that they h«d experience with the area thro^g'i
previous residence, and many seem fcu have had ox received property in the
area. As suggested by these reepom.es 5 s migrant need wot 'he we. ever n?ig r-n ted
before., or lived in the t:rea tefore, in. order to have acquired location
specific capital in the destination area. For instance, friends or fassiiy \?,&y
have migrated to the ares at sox&e earlier tisae and served a* the lio.k to a
potential migrant, vacation contact also need not entail prior ssigraot
status or prior long-term residence, The importance oi vacation contact,
especially among re tiroes, in shaping the process of search ©pace fotiBfitian
has been documented by Sly C.197'.) in a study of Florida iauRigraa.ts. He
found that nearly three-anar ters of the respondents had visited Florida prior
to Moving there, and wast vt the visits were in the fori* of vacations.
These cOKimeii-ts simply reinforce oar contention that 'Dav'anxo and
Morrison's concept of location specific capital is relevant la the decision
-16-
making process of a great variety of types of migrants J those moving for r
the first time, those who have moved before and do not return to a prior
residence in a subsequent move, as well as return migrants.
Tie-related responses to the reason questions suggest that the respondent
has drawn upon some form of location specific capital in the migration decision
making process. While we are concentrating upon destination selection, we
may note that those who gave tie responses as their reason for leaving, have,
in a sense, "cashed in" on location specific capital closer to the presumed
outset of the decision making process. The 126 households suggesting ties
as a reason for leaving (Table 1) plus the 213 households suggesting ties only
as the basis for destination selection (Table 3) account for about 48 percent
of all inmigrant households. Though the subsequent analysis could be performed
with respect to reasons for leaving, let us concentrate on the role of location
specific capital in the process of destination selection.
Let us first define the utilization of location specific capital as the
proportion suggesting ties as the basis for destination selection. We would
anticipate that the utilization of location specific capital presumes the
existence of location specific capital in some form. But, location specific
capital need not be cashed in in the sense of being the reason for selecting
the destination area. There may be numerous migrants with friends and relatives
in the area, or with prior residence, who selected their destination on the
basis of employment or other non- tie- re la ted reasons. If we can objectively
measure the existence of location specific capital, then we can investigate
the relationship between having and drawing upon location specific capital.
In line with DaVanzo. and Morrison's work, we have chosen to investigate
-17-
only one form of location specific capital - prior residence. Among several
questionnaire items related to contacts prior to moving, respondents were
asked if they had ever lived in "this" area prior to inmigrating. We may
thus form a "dummy" variable where those who are return migrating are defined
to have one unit of location specific capital in the form of prior residence.
They account for about 30 percent of nonsimultaneous decision households. We
may now investigate the relationship between two "dummy" variables, having
location specific capital in the form of prior residence, and using it by
responding a tie related reason for choosing the destination.
The relationship is best defined by a slope li»e, which in this special
case is simply the difference in the percentages reporting tie reasons
between those with and without prior residential experience in the area. This
relationship is graphed in Figure 1. The significance test as summarized in
Table 4 shows that there is a significant relationship between the two variabl*
The slope of the line in Figure 1 may be interpreted in a variety of
meaningful ways. Among other interpretations, it is a rate of return on one
unit of location specific in the form of prior residence and where returns
involve any type of tie-related reason for picking the destination. We could
also think of it as a "cash- in" rate for prior residence, or, alternatively,
as the salience of prior residence to destination selection on the basis of
location specific capital.
We may note from Figure 1, that the rate of return is positive and
substantial as expected. We may further note that even among those without
prior residence, the level of utilization of location specific capital is
substantial (about 40 percent).
-18-
Since we are focusing on destination selection, we are currently using
only one of the two reason questions. Figure 1 requires only a question about
why the respondent chose the destination, and a question asking about prior
residence. Technically, we have yet to demonstrate the utility of both
reason questions for this particular substantive problem of returns to
location specific capital. Statistically, we need to demonstrate an interaction
effect between reason for leaving, choosing a destination on the basis of ties,
and the existence of location specific capital in the form of prior residence.
Substantively, the interaction effect provides knowledge about differing
r
rates of return to prior residence capital for migrants initially motivated to
leave their origins for different reasons. For instance, we may determine
whether prior residence is more salient for those who are retiring, or for
those who left because of employment reasons. The person motivated to leave
for job related reasons, however, we have suggested tends to chose a destinatior
on the basis of job related criteria. This seems reasonable. The job related
outmigrant then should tend to cash in on location specific capital in destina-
tion areas to a lesser degree. The various slope lines are presented in Figure
We may note, first, that the line for those who left for employment reason:
is the lowest suggesting that overall, they tend the least to select destina-
tions on the basis of location specific capital. This simply restates the
findings in Table 3. However, it would appear that the highest rates of
return on location specific capital in the form of prior residence are among
retirees and those initially motivated to leave because of environmental reasom
Those leaving for employment and other reasons have very low rates of cashing
in on prior residence.
The significane test presented in Table 4, however, is a bit problematic.
-19-
Th e expected frequencies to be tested against observed frequencies have
been generated on the basis of all possible bivariate relationships in
this model
the cross-classification and/is symbolized by (12) (13) (23). The proba-
bility level of .069 suggests that we can, though only barely, fit the
observed frequencies without taking into account the interaction of all three
variables. That is, the expected frequencies on the basis of all two-way
relationships are not different from the three-way observed frequencies at
the five-percent level, but are different at the 10 percent level. The slopes
in Figure 2 then, are not significantly different from each other at the lower
probability level, but are at the 10 percent level.
In part, the lack of differences between the rates of return may be a
function of imprecision in our linking all tie reasons for picking the
destination to only one form of location specific capital - prior residence.
Prior residence almost certainly entails the acquisition of location specific
capital in diverse forms. The return migrant may respond to family or friends
left behind in an earlier move, may have housing to return to, a business left
behind, or simply want to "go back home." As a first step in furthering our
understanding of the role of location specific capital in the migration behavior
of these migrants, let us investigate the nature of some of these diverse forms
of location specific capital reasons for picking the desination.
We have broken down the gross category of tie reasons for destination
selection into four subcomponent sets of responses. We have, from Table 1,
combined those responding a desire to be closer to a business or job and those
suggesting property ties into a category we now call "economic ties." We
have also combined vacation contact responses with other ties into a category,
-20-
"vacation and other" ties. Those expressing a desire to be closer to family
or friends, and those wanting to "go back home" are maintained as separate
response categories. We may now analyze the salience of prior residence to
destination selection by examining the different rates of return for the four
types of tie responses. Figure 3 presents these results.
Only two bases for destination selection reveal positive rates of return
to location specific capital in the form of prior residence: a desire to go
back home, and the family and friends factor. As would be expected, returns
to prior residence are most manifest in the form of responses suggesting a
desire to return home. It would appear that among these migrants, the more
global expression of wanting to go back home is more closely linked to return
migration than is the desire to be with family and friends left behind in a
prior move. Of course, part of the desire to return home may be a function of
family and friends left behind and so we probably are not investigating mutually
exclusive factors. Results of significance testing (table 4) confirm the
significance of differences in the slope lines.
It is clear from these data that we should pursue the linkage between
having lived in the area before, and choosing that area because of a desire
to return to a former residence. Once again we shall ask, in this purifeid
specification, if those initially motivated to leave their origins for
different reasons cash in on prior residence at differing rates. Relevant
slope lines are presented in Figure 4.
It should be noted first that some migrants without prior residence, but
who suggest a desire to return home continue to exist as evidenced by the non-
zero left intercepts for all lines. The explanation for this may lie in
-21-
differential subjective perceptions and evaluations on what constitutes the
respondent's relevant geographic referrent area. It would appear that some
have responded that they have not lived in "this" area before for purposes
of answering that question, but do see themselves as moving "back home".
Unless these responses are random error, which they could be given the small
number of cases involved, we might infer that the territorial scope of "back
home" is larger than the scope of area included in a respondents view of a
current residential environment.
In any event, the results in terms of differential slopes are quite
revealing. As evidenced in Table 4, we do need the interaction effects in
order to reproduce the three-way table underlying Figure 4. That is, the
interaction effect is significant at the .025 level. It is retirees who have
tended most to cash in on prior residence in selecting their destinations. To
a lesser extent, though, the other types of outmigrants have also impor-
tantly drawn upon prior residential experience. Even those initially motivated
as a function of employment factors have drawn upon location specific capital
in the form of prior residence.
Since the interaction term is significant and the slopes are different,
we conclude that knowing the reason for leaving (in conjunction with the
basis for immigration and the prior residence variable) provides truly
additional information about the relationship between having, and using
at least one form of location specific capital. Thus, asking respondents
about their reasons for leaving and basis for selecting a destination is
fully justified in the context of the substantive example presented here.
22
Summary and Discussion
Since location specific capital is a generic term, the previous
analysis could readily be expanded to incorporate other forms of ties
which respondents may or may not suggest as important to destination
selection. For instance, one could generate rates of return for family,
friends, prior vacation experience, or property ownership given a
questionnaire including items asking about the existence of such contacts
prior to inmigration. Rates of return on each of these forms of location
specific capital could be compared and interpreted as to the relative
salience of each in destination selection. We have also ignored an
analysis of rates of return on location specific capital in the decision
to leave. Our findings, however, clearly suggest that had we asked only
the respondent's reason for leaving the place of origin, we would have
seriously underestimated the importance of location specific capital in
the total migration decision making process.
The suggestion that survey researchers include two operationalizations
of evaluative criteria, or reasons, is a conservative approach since it
implies a discrete, two-stage decision making process. For some migrants,
however, we might find a much more continuous decision making process
involving a narrowing down of the awareness space into a viable search
space and ultimately to a single destination. Survey researchers, of course,
must make operationalizations on the basis of a discrete process but still
we might benefit greatly from more than two questions. For instance, we
have no idea how our sample members would have responded to a question
designed to elicit reasons for moving to a nonmetropolitan area, in general.
To ask this question presumes that a decision point was encountered at
which a respondent chose to eliminate all metropolitan places from the
awareness space.
23
The importance of disaggregating the migration decision making
process into more than a simple move or not framework implies that the
migrant has been able to make choices affecting his or her behavior.
That is, separate operationalizations are particularly applicable to the
voluntary, relatively unconstrained migrant. This point has important
implications.
On the basis of the reasons given for leaving their origins, we
must suspect that many of these inmigrants to nonmetropolitan high growth
areas of the Midwest are voluntary migrants, and perhaps relatively
unconstrained. The modal response category was an environmental reason
for leaving the origin. Further analysis, however, demonstrated that those
who left for environmental reasons, as well as because of retirement,
drew heavily upon location specific capital in the form of prior resi-
dence when selecting their particular destinations. If those leaving
their origins for environmental reasons, or retirement reasons are a major
factor in nonmetropolitan growth nationally, then clearly it is important
to ask these migrants about destination selection criteria. In contrast,
we demonstrated that the employment-related motivated outmigrant tends
similarly to choose a destination on the basis of employment criteria and
so, perhaps it is less important to ask such migrants about destination
selection. We conclude that current nonmetropolitan growth, especially,
is importantly a function of migrants for whom we gain much Insight if we
disaggregate our operationalization of the migration decision making criteria.
This probably reflects the voluntary and relatively unconstrained nature
of their behavior.
24
If indeed voluntary migrants in the future increasingly come to
suggest nonemployment bases for .migration decision making, then we should
also expect to need disaggregated operationalizations in the future.
As yet, we simply don't know the nature of this possible trend. Long and
Hansen (1978) have attempted a time series investigation of reasons for
migrating, but conclude that it is simply not possible to investigate the
matter fruitfully given available data.
In summary, we feel that it is essential that future surveys of
migrants, especially those which are certain to focus on nonmetropolitan
"amenity" growth areas, include separate reason questions for in and out-
migration decisions. This is a minimum, for an accurate understanding of
the total decision making process. To other researchers, we submit that
the prudent approach for any survey of migrants attempting to elicit
"reasons for moving" is to ask criteria for both leaving and choosing a
destination.
Technical Appendix
The methodology used in this paper is log-linear analysis of cross-
classified data. Drawing on the works of Goodman (1978) and his computer
program ECTA, we present likelihood ratio chi-squares as tests of signi-
ficance. The advantage of this methodology is that it allows us to test
for interaction effects in three-way contingency tables.
The logic of log-linear analysis is similar to traditional chi-square
tests. We begin with a table of observed cell frequencies and from the
marginals of that table we calculate expected cell frequencies. We then
test the departure of the expected frequencies from the observed frequencies.
In the traditional x2 analysis, this would be a test for statistical
25
2
independence between two variables. The larger the value of the x, , the
more readily we can reject the null hypothesis of statistical independence
which is to say that the expected cell frequencies generated on the basis
of the marginals do not fit the observed frequencies.
Similarly, log-linear analysis generates expected cell frequencies
under a variety of assumptions, and provides a test for the fit between
2
expected and observed cell frequencies. A small value of the X indicates
a fairly good fit (subject to degrees of freedom of course) while a large
2
X indicates a poor fit under the assumptions of the model tested.
The basic model for log-linear analysis of a two variable cross-
classification is given by the identity
1 2 12
fij = ™l Tj Tij <«
where f.. refers to the observed frequencies of variables 1 and 2; n. is
the geometric mean of f . ; and the x parameters are the probabilities
that an observation appears in the subscripted cell of the superscripted
univariate or joint distribution. The model for testing statistical
independence in the bivariate case is given by the formula
F.. = n T1 T2 , where T12 = 1. (2)
ij i j ' ij
In this equation F is the expected frequency in the ij cell; n and the
x parameters are as defined above. The formula for the likelihood ratio
chi-square is
X2 = 2E f±j In (f^/F^) (3)
where f.. is the observed frequency in the ij cell and F. , is the
xj ^ ij
expected frequency in the ij cell. This Is equivalent to the G
statistic reported by Cohen (1975).
26
Extending this analysis to three variables, we have the identity
f „ 1 J. J$ 12 13 23 123 ns
fijk = T1Ti fj TkTij TikTjkTi3k- (4)
And, we might postulate the following model
v n 12 3 12 13 23 . 123 .. ,-*
Fijk = n Ti Tj Tk Tij Tik Tjk' Where Tijk = X- (5)
This model hypothesizes that there is independence between two variables
across the third, or that there is no three-way interaction between the
variables, or that the relationship between two variables is constant
2
across the levels of the third. A large and significant value of x
would lead us to reject this hypothesis and conclude that we cannot fit
the observed cell frequencies with a model devoid of the three-way
2
interaction effect. A small, insignificant x would suggest no interaction
in that we would have fit the table of observed frequencies rather well.
Remember, the statistical test is of how well the expected frequencies
based upon some assumption (s) fit the observed frequencies.
Using ECTA we obtain maximum likelihood estimates of the cell
frequencies and a likelihood ratio chi-square test of the departure of
the expected cell frequencies from the observed cell frequencies. Table
4 summarizes the results for relevant tables and figures. Though not
presented, it will be remembered that a cross-classification underlies the
figures and provides the cell frequencies for the log-linear analysis.
For the bivariate case, we have tested for statistical independence
between two variables. This test is symbolized as (1) (2) in Table 4
and corresponds to equation 2 above. The low significance levels in the
results for Tables 2 and 3 and Figures 1 and 3 suggest that the expected
frequencies on the basis of an assumption of independence do not fit the
observed frequencies and so the relationships are significant.
27
The three-variable tests correspond to equation 5 above and are
symbolized in Table 4 as (12) (13) (23). The three-way interaction terra
is all that is left out and so we are attempting to fit observed
frequencies on the basis of expected frequencies generated without
allowing for the three-way interaction. The significance levels for
Figures 2 and 4 simultaneously suggest how well we fit the table and the
significance of the interaction term. A significant interaction suggests
that the slope lines for any two variables differ across levels of a
third variable. In Figure 2, the interaction term is significant at the
.069 level while in Figure 4 it is significant at the .025 level and so
we may have greater confidence in the existence of differing slopes in
Figure 4.
28
References
Brown, Lawrence A. and David B. Longbrake
1970 Migration Flows in Intraurban Space: Place Utility Considera-
tions. Annals of the Association of American Geographers
60(2) -.368-384.
_J and Eric G. Moore
1970 The Intra-urban Migration Process: A Perspective. General
Systems XV: 109^122.
Cohen, Joel E
1975 Childhood Mortality, Family Size, and Birth Order in Pre-
Industrial Europe. Demography 12(1): 35-56.
DaVanzo, Julie and Peter A. Morrison
1978 Dynamics of Return Migration: Descriptive Findings From a
Longitudinal Study. Rand Paper P-5913, Palo Alto, CA.
Goodman, Leo
1978 Analyzing Qualitative/Categorical Data: Log-Linear Models
and Latent Structure Analysis. Abt Books. Cambridge, MA.
Lansing, John B and Eva Mueller
1967 The Geographic Mobility of Labor. Ann Arbor: University
of Michigan, Institute for Social Research, Survey Research
Center.
Long, Larry H. and Kristen A. Hansen
1978 Reasons for Interstate Migration: Jobs, Retirement, Climate
and Other Influences. Paper presented at the annual meeting
of the Southern Regional Demographic Group, San Antonio,
Texas, October.
Roseman, Curtis C.
1977 Changing Migration Patterns Within the United States. Resource
Papers for College Geography No. 77-2. Association of American
Geographers.
Rossi, Peter H.
1955 Why Families Move: A Study in the social Psychology of Urban
Residential Mobility. Glencoe, IL: Free Press.
Sly, David F.
1974 Tourism's Role in Migration to Florida: Basic Tourist-
Migration Relationship. Governmental Research Bulletin, The
Florida State University, Institute for Social Research,
Vol. 11.
29
Spear t , Alden, Jr., Sidney Goldstein, and William H. Frey.
1974 Residential Mobility, Migration, and Metropolitan Change.
Ballinger, Cambridge, MA.
Wo 1 pert, Julian
1965 Behavioral Aspects of the Decision to Migrants. Papers and
Proceedings, Regional Science Association, 15:159-177.
Zelinsky, Wilbur
1977 Coping with the Migration Turnaround: The Theoretical
Challenge. International Regional Science Review, 2(2):
175-178.
TABLE
DETAILED MOTIVATIONS
Reason for Leaving
Destination Selection
Criteria
All Reasons
1. Employment: job change; reassignment 182
Transfer
Look for new or better job
Found new or better job
Unemployment
Other (incl. military)
2. Ties: location specific capital
Moved closer to business or job
Owned or received property
Moved closer to family or friends
Moved back home; Lived in area before 23*
Vacationed in or visited area before
Other ties
3. Environmental
General anti-urban or pro-rural
Congestion; Wanted a smaller town
Polution; Environment
Climate
Crime
Schools
Recreational opportunities
Cost of living; Taxes
Liked or disliked area in general
Other environmental factors
% of
% of
% of
% of
N
Total
100.0
Catg.
N
Total
100.0
Catg.
710
710
it 182
25.6
100.0
148
20.8
100.0
58*
8.2
31.9
42
5.9
28.4
22
3.1
12.1
14
2.0
9.5
63*
8.9
34.6
64
9.0
43.2
12
1.7
6.6
0
27
3.7
14.8
28
338
9.9
47.6
18.9
126
17.7
100.0
100.0
33*
4.6
26.1
40
5.6
11.8
28*
3.9
22.2
70
9.9
20.7
31*
4.4
24.6
97
13.7
28.7
ire 23*
3.2
18.3
81
11.4
24.0
■e 1*
0.1
0.8
43
6.1
12.7
10*
1.4
8.0
7
1.0
2.1
216
30.4
100.0
176
24.8
100.0
93
13.1
43.1
45
6.3
25.7
31
4.4
14.4
6
0.8
3.4
4
0.6
1.9
12
1.7
6.8
6
0.8
2.7
6
0.8
3.4
13
1.8
6.0
6
0.8
3.4
16
2.3
7.4
12
1.7
6.8
5
0.7
2.2
24
3.4
13.6
12
1.7
5.6
15
2.1
8.5
22
3.1
10.2
19
2.7
10.8
14
2.0
6.5
31
4.4
17.6
4. Retirement
5. Other
Family; Life cycle
Housing
Health
Other
99
13.9 100.0
83
11.7
100.0
47
6.6
100.0
32
4.5
33.6
11
1.5
23.4
10
1.4
12.0
19
2.7
40.5
20
2.8
24.1
5
0.7
10.6
21
3.0
25.3
12
1.7
25.5
* see text for explanation
Table 2. Relationship between Criteria for Destination Selection anc
Reason for Leaving Origin (all households)
Destination
Reason
for
' leaving
selection
criteria
Employ-
ment
Ties (location
specific capita
1)
Environ-
ment
Retire
Other
Employment
125
(68%)
3
(2%)
17
(8%)
1
(1%)
3
(4%)
Ties (location
specific capita
.1)
36
(20%)
107
(84%)
87
(40%)
62
(62)
44
(54)
Environment
18
(10%)
16
(13%)
93
(43%)
30
(30%)
19
(23%)
Other
4
(2%)
1
(1%)
19
(9%)
7
(7%)
15
(19%)
Table 3. Relationship between Criteria for Destination Selection and
Reason for Leaving Origin (Households in which decisions we
not simultaneous)
Destination
Reason for leaving
selection
Employment
criteria
Employment
35
(57%)
Ties (location
specific capita
Ll)
20
(32%)
Environment
7
(11%)
Environment
Retire
Other
Other
0
(0)
14
(7%)
87
(41%)
93
(43%)
19
( 9%)
1
(1%)
62
(62%)
30
(30%)
7
( 7%)
2
(3%)
44
(55%)
19
(23%)
15
(19%)
2
Table 4. Likelihood Ratio x Values for Relevant Tables and Figures
Table 2
Table 3
Figure 1
Figure 2
Figure 3
Figure 4
Model Fit
y2
df
Sig.
(1) (2)
385.49
12
.000
(1) (2)
129.48
9
.000
(1) (2)
26.24
1
.000
(12) (13) (2
:3)
7.07
3
.069
(1) (2)
125.44
4
.000
(12) (13) (2
13)
9.32
3
.025
FIGURE 1
RETURNS TO LOCATION SPECIFIC CAPITAL (PRIOR RESIDENCE)
FOR NON-SIMULTANEOUS DECISION MAKERS
100
90
80
70
{•cent who
Iked
ictination
ii basis of
.c:ation
■jcific
:<>ital
10
100
90
80
70
60
50
• 40
30
20
10
NO PRIOR
RESIDENCE
PRIOR
RESIDENCE
i are 2. Returns to Location Specific Capital by Reason for Leaving
Origin (non-simultaneous decision makers)
100
rcent who
aked des-
laation on
isis of lo-
:tion spe-
:fic capital
20
10
0
7 100
90
Retire
80
70
60
50
40
30
20
10
0
Environmental factors
Other
Employment
NO PRIOR
RESIDENCE
PRIOR
RESIDENCE
Figure 3. Salience of Prior Residence to Utilization of Location S
Capital in Various Forms (Non-simultaneous decision make.
'rcent who
jzked des-
:aation on
jsis of
secific
Erms of
Lcation
secific
:pital
100
90
80
70
60
50
100
90
80
70
60
50
40 Return
30
20
Family and friends
10
Economic ties
Vacations and other
0
NO PRIOR
RESIDENCE
PRIOR
RESIDENCE
re 4
Salience of "Return" as a Form of Location Specific Capil
in Relation to Prior Residence bv Reason for Leaving
(Non- simultaneous decision makers).
100f
TlOO
krcent who
]Lcked des-
•Lnation on
lisis of a
i^sire to
iturn to
:ior area
if residence
Environmental factors
30 Employment
Other
NO PRIOR
RESIDENCE
PRIOR
RESIDENCE
'2-89
Live Music Archive
Librivox Free Audio
Metropolitan Museum
Cleveland Museum of Art
Internet Arcade
Console Living Room
Books to Borrow
Open Library
TV News
Understanding 9/11