we WIN, THE BG’QRT BASE. AND
mum ECGLQSML W

Thais for tire Degree of Pi Bi
HECHEQQN 3'3" F5 BWSEW
WES 303% E'YREPE 7
3-3172.

 

LIBRARY

1293 70326/51/é///£//// Milclfligisttfe ‘

 

This is to certify that the
thesis entitled
MIGRATION, THE EXPORT BASE, AND
COMMUNITY EEOLOGICAL STRUCTURE

presented by

James Joseph Byrne

has been accepted towards fulfillment
of the requirements for

Ph. D . Sociology

degree in

 

 

 

Date iflifl

0-7639

 

ABSTRACT

MIGRATION, THE EXPORT BASE, AND COMMUNITY ECOLOGICAL STRUCTURE

By
James Joseph Eyrne

Studies by Lowry and others have shown that net migration in

' the United States has a high positive correlation with change in
employment level. It is unclear, however, to what extent employment
opportunities act as a “pull" for’migration, and to what extent
increased.employment is merely an outgrowth of migration. In this

three part investigation, a theoretical framework is developed for
analyzing net migration which.uses the concepts of the community ”export
base” and the ”multiplier effect.” These concepts are theoretically
and operationally independent of the migration process, offering
improved opportunities for understanding and prediction.

The overall conceptual scheme attempts to explain net migration
in terms of differential manufacturing growth, and differential manu-
facturing growth in terms of community ecological characteristics.
Communities with certain types of ecological characteristics are
theoretically destined to expand because of their ability to attract
new industry, while those with other characteristics are destined to
stagnate. Specific ecological characteristics derived from the theory
and examined are: population size, population potential, median income,
population density, distance to nearest larger Standard.Metropolitan
Statistical Area, percentage employment in service industries, and

climate. Units of analysis are Standard Metropolitan Statistical Areas.

James Joseph Byrne

In Part I of the study, existing methods of export base measure-
ment, particularly the ”index of surplus workers, " are criticized, and
an mpirical test of the validity of the index is carried out. The
indent shows no correlation with net migration for 139 Standard Metro-
politan Statistical Areas, 1950-60. It is concluded that the index is
an inadequate measure of an area's export base; it measures not only
export activity, but import activity as well. _

In Part II, an alternate scheme of export base measurement is
presented, in which local growth in particular export activities,
especially in nanufacturing payroll, is linked to net migration. Strong
correlations with migration are found, so in Part III the seven
ecological variables listed above are linked to manufacturing growth
in a multiple correlation analysis. Population size and distance are
the only variables with strong statistical relationships to change in
manufacturing payroll. Jointly, change in manufacturing payroll,
population size, and a climate indent statistically explain 80 percent
of the net migration variance. It is concluded that the export base
approach is a potentially effective means of explaining net migration,
but that the ecological variables used in the study are not effective
in explaining the relationship between manufacturing growth and net

migration.

MIGRATION , THE EXPORT BASE, AND COMMUNITY ECOLmICAL STRUCTURE

By

James Joseph Eyrne

A THESIS

Suhnitted to
mchigan State University
in partial fulfillment of the requirements
for the degree of

DCXITOR OF PHILOSOPHY

Department of Sociology

1972

.ef ACKNMEDGMENTS

The investigator is grateful to all who helped to make this shidy
possible. Special thanks to Dr. Herbert H. Karp, whose theoretical
insights and helpful advice were continually used and much appreciated,
as well as to Professors J. Allan Beagle, K. Dennis Kelly, William
Evens, Pichard C. Hill, and James Zuiches for their useful criticisms
and suggestions. Appreciation is expressed to Professor Otis Dudley
Duncan for use of his data on population potential in the United States
for 1950, and to Professor Robert Carroll for advice on computing a
distance measure. The investigator also wishes to thank the
Agricultural Experiment Station at Michigan State University for
financial support during the period in which the research was conducted,

in the form of a research assistantship on a related research project.

11

TABLE OF CONTENTS

Chapter

I.

III.

V.

INTRODUCTION

SIGNIFICANCE OF THE PROBLEMLAND THE DISSERTATION

THEORY AND LITERATURE REVIEW

Development of the Export Base Logic. . .
Ekport R36 MatthIOg a o a a a a a a o
Other Ecological Analyses of Migration. .

The Export Base:

An Alternative Approach

Humrmentaoaeaaaaaoaaaa

Industrial nobility:
METHODOLOGY

UNITS OF ANALYSIS . . . . . . . .

DATA .
ANALYSIS OF THE DATA

A TEST OF THE SURPLUS WORKERS INDEX
TEE MIGRATION VARIABLE. . . . . .
OPERATIONAL HYPOTHESES. . . . . .

RESULTS 0 O O O O O O O O O O O O

DRAWBACKS OF THE SURPLUS WORKERS INDEX.

A REVISED EXPORT BASE APPROACH.

PROCEDURE........
OPERATIONALRYPOTHESES.
RESULTS.........
DISCUSSION. . . . . . . .

ECOLOGICAL FACTORS ASSOCIATED

Conceptual Scheme .

to

WITH.MANUEACTURING'SHIFTS

OPERATIONAL HYPOTHESES. . . . . . . . . . . . . . . .

iii

10
13

19

26
26
26
30
32

315.9

uz
#2
b3
1&5
#9
51
52

iv

Chapter

RESst 0 O O O O O O O O O I O O O O O O 0

Additional Multiple Correlation

Effects of Wage Inflation . .
DISCUSSION. . . . . . . . . . .

VI. SUMMARY.AND ENALUATION. . . . . .
CONCLUSIONS..........
FUTURE RESEARCH POSSIBILITIES .
BIBLIOGRAPHY . . . . . . . . . . . . . .

Analyses.

Page

53

61
63
67

7O
73

LIST OF TABLES

Table Page
1. Simple Correlation Matrix and Basic Statistics for
Part1....................... 37
2. Summary of Results for Hypotheses 1 to h . . . . . . . 37

3. Simple Correlation Matrix and Basic Statistics for
PartII.....o................. “'6

4. Multiple Correlation of Export Base Variables with
NetMigmtion................... “'8

5. Sum of Results for Hypotheses 5 to 9 . . . . . . . . . U8

6. Simple Correlation Matrix for Ecological Variables
”duetnigntionaaaooaeaoassesses. 5""

7. Basic Statistics and Summary of Results for Hypotheses
10 to 16, Concerning Relationships of Ecological
Variables to Manufacturing Payroll Shifts . . . . . . 55

8. Multiple Correlation Analysis of Seven Ecological
V‘ri‘bles With Payroll Shift. 0 a a a a a a o a a a o 57

9. Multiple Correlation Analysis of Seven Ecological
Variables and Payroll Shift with Net Migration. . . . 57

10. Multiple Correlation Analysis of Population Size and
Distance with Manufacturing Shifts. . . . . . . . . . 59

11. Multiple Correlation Analysis of Population Size and
Disunce With Net Migration a a a a o a a a o a a a a 59

12. Multiple Correlation Analysis of Distance, Population
Size, and Payroll Shift with Net Migration . . . . . 62

13. Multiple Correlation Analysis of Population Size and
Payroll Shift with Net Migration. . . . . . . . . . . 62

14. Multiple Correlation Analysis of Climate, Population,
and Manufacturing Payroll Shifts with Net Migration . 6h

15. Multiple Correlation Analysis of Manufacturing Payroll
19“? and Manufacturing Payroll Shifts with Net Migration 6h

V

CHAPTER I
INTRODUCTION

Demograplv has often been praised as one of the most advanced
of the social sciences, since its methodology is far more sophisticated
than that of most other areas of social research, and there exists a
much greater degree of quantification among its variables. There is one
area, however, in which demograplw has lagged badly: in the development
of comprehensive and integrated theory. As Rupert Vance (1952) has
noted: "There is one area where demography is getting rather poverty-
stricken and frayed at the edges. In the realm of high theory we have
been living off our capital and borrowing from our associates. " The
situation has changed little in the twenty years since Vance spoke
these words. This failure of demograplv is reflected in the disjointed
nature of its theoretical formulations, which often consist of poorly
defined concepts, unstated assumptions, empirical generalizations,
fractured typologies, isolated models, and sets of propositions
unrelated to specific observables. The “demographic transition, " for
mmple, which is one of the nearest approaches to demographic theory,
is really only an empirical generalization. It gives one no logical
basis for inferring that changes in birth and death rates of p-esently
industrializing countries will be similar to those of Eirope at a
comparable stage of development.

There are a number of reasons for this unfortunate state of
affairs, not the least of which my be the very sophistication of
techniques, which may have resulted in a tendency to bypass systematic

l

2
theory development. The development of migration theory has been no

more promising than in other areas of demographic theory. There has
been a proliferation of concepts vis-a-vis migration, but they have not
been combined into a comprehensive framework. Each researcher has
limited himself to certain variables and ignored others. We have
the concepts of intervening opportunities,- of the Ple/D Hypothesis,
and so forth, but they have not been refined and related to a broad
theoretical framework, or discarded as appropriate. The general pattern
of migration theory has consequently been a collection of empirical
generalizations and typologies. Ravenstein (1885: 199) set the pattern
early with his laws of migration, which were an impressive contribution
for his time, tut which have since been little improved-upon. These
laws specified that various characteristics of migration are univer-
sally prevalent. Ehuamples of such laws as developed by Ravenstein
are: ”Each main current of migration produces a compensating counter-
current,“ and "Migrants proceeding long distances generally go by
preference to one Of the great centers of commerce and industry."

Stouffer (19%) continued the tendency toward empirical generali-
zations, though he presumably injected a greater degree of precision
by introducing the notion of proportionality: ”The number of persons
going a given distance is directly proportional to the number of
opportunities at that distance and inversely proportional to the
number of intervening opportunities. " He likewise made other useful
refinements.

Recently Lee (1969: 288-9) has built upon the earlier work of
Bavenstein and others by developing a more comprehensive set of

generalizations, including the following: "The volume of migration
within a given territory varies with the degree of diversity of areas

3
included in that territory, " and ”The volume of migration varies with
the diversity of people. "

Work in this area Ms thus runained at the level of empirical
generalizations, albeit increasingly sophisticated ones. Such
generalizations are worthwhile, in that they provide evidence for the
relevance of particular concepts and specify relations among them.

As long as they remain unconnected to a formal body of theory, however,
they are of limited value, since the specific conditions under which
they operate are unknown and they cannot be used for prediction or
deductive explanation.

At the present stage of theoretical development, it would be
unrealistic for anyone to attempt the development of a mil-fledged
migration theory. The present study does make a few halting steps in
that direction, by mining internal migration in the United States
from a single theoretical framework, with variables derived from that
theory. The theoretical framework used will be taken from the
economics literature: that of the colmmnity ”export base” and ”mlti-
plier effect. "

It has long been recognized that economic factors play a
crucial role in determining the patterns of internal migration among
different regional areas of an industrial economy. Ravenstein (1885:
198), in his analysis of migration laws, commented that: ”It does
not admit of doubt that the call for labour in our centres of industry
and commerce is the prime cause of those currents of migration which
it is the object of this paper to trace." Later writers, including

Stouffer, have given at least implicit recognition to the importance
of economic factors in explaining migration.

L;

Certain economic variables have empirically been found to be
closely related to net migration: Lowry (1966), for example, discovered
a high correlation (r = .981) between net migration and change in
employment level among Standard Metropolitan Statistical Areas (SHSAs) .
The difficulty with studies like that of Lowry is that the presumed inde-
pendent variable is measured over a time interval concurrent with that
of the dependent variable. Lowry, for example, obtained the above corre-
lation by relating net migration, 1950-60, with net change in
civilian nonagricultural employment, 1950-60. For purposes of
prediction, his results are of limited value, since it would be as
difficult to predict changes in employment level as it would be to
predict changes in net migration. Moreover, it cannot be ascertained
to what extent employment opportunities are attracting migrants, and to
what extent migrants are moving for noneconomic reasons but managing
eventually to find sane type of employment at their destination. It
is thus not clear to what extent migration is acting as an independent,
rather than a dependent, variable. No one has yet been able to develop
an explanatory framework for migration based upon a comprehensive array
of interrelated variables that are independent of the migration process.
This has been due both to demography's atheoretical tendencies discussed
above and to the lack of precise knowledge as to what variables and
concepts are relevant. The latter in turn has been traceable to a paucity
of empirical studies linking migration to spatial ecological characteristics.
Several studies, discussed below, have attempted to isolate such
characteristics, but these studies are few in number, and until it
is known which ecological factors are consistently associated with net
migration, a concise theoretical scheme cannot be developed.

5

Ecologically, net.migration may be viewed as an equilibrium-
maintaining variable (flawley, 1950: 332). It responds to demographic
imbalances which result, among other things, from differential economic
growth among various regional areas within an economic system. The
stronger potential for economic expansion in one area as opposed to
another creates a migratory “pull.”

There has been a sustained search among economists and demographers
for a theory which would explain the tendency for certain areas of
an economy to experience faster economic growth than others. The most
common analytical framework for assessing economic growth potential
has been that of the community export base. This approach regards the
amount of local income derived from exports to other areas as the
crucial element in explaining economic growth and hence net migration.
This logic, while basically sound, suffers from a crudeness of
measuring techniques, resulting partially from a lack of suitable data
at the community level.

In view of the present status of migrathurtheory and export base
methodology, the plan and structure of the proposed study are threefold:
(1) an analysis and testing of the deficiencies of present export
base methodology: (2) development of an alternative approach to export
base analysis, centering upon growth in the secondary sector of the
econonw: and (3) linkage of a number of comunity ecological characteris-
tics to secondary sector growth. A conceptual framework will be
developed to account for industrial mobility and net migration in
the United States for the decade of the 1950s and to aid in the

selection of relevant ecological correlates of this movement.

6

SIGNIFICANCE OF THE PROBLEM AND THE DISSERTATION

0f the three basic variables in the dancgraphic equation, migration
is unquestionably the hardest to predict. Mortality and fertility both
occur in rather well-defined patterns by age and sex, and it is easy to
specify the population at risk. Rrring war or other major disaster, mor-
tality can be anticipated and provided for by an ongoing society so that
it does not exert a disruptive impact. The fertility level is more
variable, and variations in this variable, arch as the occurrence of
a baby boom, can be more of a disorganizing influence. Fertility
changes do not affect most social institutions until mpars after
they occur, however, so there is time to anticipate and provide for
than to some extent. Migration, by contrast, is highly variable,
cannot be well predicted, and can exert an immediate strain on the
society both at origin and destination. Migration is very age- and sex-
specific, so it can cause severe distortions in the age-sex structure
at origin and destination, with resulting dislocation of the social
structure. Similarly, the sheer volume of migration can cause numerous
social problems and make assimilation difficult, particularly if the
amount of migration is not anticipated in advance.

In view of the above, it is most important to find ways of
anticipating the amount and direction of migration. By shedding light
on the ecological characteristics of a given commity which are linked
to migration, the dissertation nay enhance the predictability of
migration. This would permit more accurate population projections and
a higher probability of being able to prepare in advance for social
dislocations caused by fast growth rates and a large migrant population.

 

 

 

7

The dissertation may also contribute to the advancement of
migration theory. It will be the first study in which ecological
characteristics have been selected for a multiple correlation analysis
with migration on the basis of a specific guiding theoretical framework.
In so doing, the study builds upon previous investigations which have
already shown most of the variables in the present study to be impor-
tant in migration analysis.

Methodologically, the study should advance export base analysis
by treating the export base as an intermediate set of variables
between migration and ecological influences, rather than as a structural
characteristic of the local labor force. By finding those ecological
factors which specifically increase local export income, one is
presumably finding those factors which ultimtely produce increased
net migration.

Substantively, the dissertation links net migration variability
directly to growth in the secondary sector of the econony, rather than
to employment growth in general. By delineating specific types of
employment which are associated with migration, one can better understand
the nature of the process, and eventually, refined theories may be
developed.

THEORY AND LITERATURE REVIEW

The export base concept is based upon the logic that an economic
unit derives its income for potential expansion from the sale of locally
produced goods and services to outside areas: to the extent that a

giva: product is being produced and sold in greater amounts than required
for local consumption, it is being exported in return for money income

8

to the area. Production of the product is therefore regarded as a
“basic” activity. Goods and services produced for local consumption
are in the nature of housekeeping activities which do not contribute
to the area's income and are therefore ”service" or 'nonbasic”
activities.

A region's growth potential is presumably a direct function
of the size of its export business, since the ratio of basic to service
workers in an area is assumed to remain fairly stable over the long run.
An increase in export income gives rise to an increased demand for
services, which is reflected in an expansion in service industries

and attraction of migrants to fill the jobs thus created.

eve o e s c

The export base concept arose within the framework ofinter-
national trade, but was early recognized as applicable to smaller
economic units. Cities in particular became targets for economic
analysis through the export base technique because of their unified
economic structure, their interdependence with other cities and with
rural areas, the need for urban planning, and the availability of
population and economic statistics for tha.

Weimer and Hoyt (1939) built upon scattered beginnings and
presented the first extensive development of the export base
theoretical framework. Since that time, numerous refinements and
criticisms have been made, such as a rejection of Hoyt's early notion
that there is a one-to-one or other fixed relationship for all cities
between the base and service sectors of the economy. A systematic discus-

sion of the historical development of the urban base idea was given by

9
Richard Andrews (1953): later discussions of export base logic and
methodology my be found in Pfouts (1960) and North (1961).

The export base approach, or selected aspects of it, has been
subjected to severe criticism by a variety of writers. mumenfeld
(1955) condemned it for several reasons, including its failure to
consider the role of imports, its apparent inapplicability to large
comaunities (which have a consistently smaller proportion of export
workers than small cmunities), its tendency to denigrate the role of
service industries (which are usually seen as a mere byproduct of
export activities), and its use of employment totals when it is really
money income that is important. In the absence of more sophisticated
techniques, however, and due to the paucity of detailed regional
economic data, the export base approach has remained practically the
only method of coping with econanic growth measurement and analysis.

A refinemmt of the export base concept was the treating ofth
potential as a function of both the export base and of what D. C.
North (1961) calls the 'mltiplier-accelerator" effect. The latter
idea recognizes that growth potential is dependent not only upon the
sheer size of the export base, hit upon the disposition of income
derived from exports. To the extent that export income flows directly
out of the local economy in the purchase of imports it amkos no
contrihition to economic growth: to have a beneficial effect, it
met circulate through the local economy, so a well-developed tertiary
sector is as important as a strong export sector.

A recent study by Karp and Kelly (1971) used this refinement of
the export base logic, overcoming most of Blumenfeld's criticisms.

The multiplier effect takes the role of imports into account, as

10

explained above: moreover, by treating services as equal in importance
to exports, it makes the export base logic more applicable to large cities,
which have a consistently higher percentage of service industries than
small coamamities. Blumenfeld's most important remaining criticism is
the use of employment totals rather than dollar income as a basis for
measurement: this pertains to export base methodology and will be
examined in Part I of the dissertation.

The Karp and Kelly stuck was also important in that it was the
first one to link a community's export base specifically to migration
patterns: this aspect of the study will also be discussed further below.

W

A variety of techniques have been proposed for measuring the
export base of an area. The two methods generally used are the
location quotient and the minimum requirements technique. The location
quotient compares the local percentage of workers in a given industry
to the national percentage, according to the following formula:

 

e1 Where: e1 = local industry employment

6t °t = local total employment

E1 E1 = national industry employment
T Et = national total uployment

The extent to which the formula exceeds a value of one (i.e. extent
to which the local percentage of employment in the industry is greater
than the national percentage) indicates the degree to which the
industry serves an export mnction for the local econony. The under-
lying assumption is that each area consumes its pro rata share of the

nation's output of goods from each industry.

11
The minimum requirements approach is somewhat different: it is
described by Harp and Kelly (1971:61) as follows:
. . . a sample of area units is arranged according to class
size and the percentage of the labor force employed in each
of a number of industry categories is calculated for each of
the areas in the sample. The figure for the area in each class
with the minimum percentage employed in each category is obtained,
and these minima are summed to obtain a figure that represents
the gross nonbasic component of an area's employment in that
particular size class. The remainder represents the gross
export or basic employment for’an area in that particular size class.

Pratt (1968) criticised the logic of the minimum requirements
technique by showing that there is no sound reason for using minima
rather than averages in apportioning workers to basic vs. service
categories. He showed that the use of minima assumes an economy in
which everyone exports goods and services, while no one imports:
moreover, the values of the minima are totally dependent upon the level
to which industrial categories are disaggregated. Even with these
drawbacks, the minimum.requirements technique was shown to make the
same stringent and questionable assumptions of the location quotient.
These assumptions include the dubious one that productivity and consumption
levels are constant throughout the national economy. An unconvincing
attempt to rebut these criticisms was made by Ullman (1968).

A variant of the location quotient, the index of surplus workers,
developed by hattila and Thompson (1955), appears to be the best available
indicator of migration potential. This index reflects the number of
workers in a given industry above the area's pro rata share: these
are presumed to be export workers. The apparent advantage of this
formula for migration analysis is that it assigns each industry a
weight proportional to the number of workers in that industry, rather
than assigning all industries an equal weight. As Hattila and

Thompson (1955:221) pointed out, the latter weighting system assumes

12
that ”all local industries are equally 'important' (i.e. afford
equal employment opportunities) to the comunity. " This is clearly not
the case.
By computing surplus workers for each industry, then totaling them
across industries, a measure of the area's overall export base is
obtained, called the absolute form of the ”index of total surplus

workers, " as represented by the following equation:

“2&1- : .1]

where S represents the absolute number of total ”surplus workers"

 

in the local economy, and the remaining teas are defined as in the
formla above.

It is entirely possible that the index of total surplus workers
in its absolute form is highly correlated with total local employ-
ment, since the potential range of variation in employment totals for
each industrial category is directly proportional to the size of
the local labor force. Total local employment will therefore be
included in the present study as a separate variable, permitting an
examination of its relationship with the index of surplus workers.

In the event that the two are highly correlated, a ”relative“ form

of the index of surplus workers, also developed by httila and Thompson,
can be used. The latter is simply the absolute form divided by total
local Glployment: in its "total" surplus workers fom, it would be

as follows:

 

s=z[_2_-_:§?:-Ex ]

13

The relative form should only be used if there is a high correlation
between the index of surplus workers (expressed in absolute form) and
total local employment, since the relative form treats surplus workers
as a percentage of total local employment. As Sakoda (1966: 5-6)
pointed out:

If there is no relationship between frequency in a category

and the total number, the frequency itself can be used as an

index. There would be no point in calculating a percentage.

As a latter of fact, the use of the percentage would result

in over correction so that where a correlation did not exist

between frequency and N, a negative correlation would exist

between A/N and N.

The advantage of the index of surplus workers for calculating
the comamnity export base is only apparent, however, and it will be
the task of Part I (chapter III) of the dissertation to carry out a
logical and empirical critique of this technique. The inadequacy of
both the location quotient, as represented by the index of surplus
workers, and the minimum requirements method, means that no suitable index

yet exists for measuring the community export base.

MW
m researchers have tried by one means or another to link

ecological characteristics of areas to net migration patterns. Among
these are Duncan and Reiss (1950); Oghlrn (1937): Goodrich m (1936);
Hawley (19%), Price (1950): Thomas (1960): and the 0.8. Bireau of the
Census (19“7).

In the last twenty years, more sophisticated techniques of
analysis have been introduced, such as those discussed by Bogue and
Harris (195%. The first investigators to link migration patterns
to the ecological characteristics of areas through multiple regression
and correlation techniques were Bogue gt_._a_IL,_ (1957). In that stuck,

 

11+
migration data for 1935-40, as reported in the 1940 census, were
correlated with 16 independent variables, including such factors as
population size, percent nonwhite, level of living, sex ratio, net
reproduction rate, median age, and percent increase in different
residence categories. The units of analysis were ”state subregions, "
forerunners of State Economic Areas. Regression equations were computed
for several metropolitan and nonmetropolitan areas: the equations
succeeded in accounting for “9% to 92% of the variance, depending upon
type of area. Some important variables were percent unemployed, sex
ratio, and median age. This study showed that economic factors are
an important factOr in migration patterns: ecological characteristics
in general were also shown to be significant, including social character-
istics.

A later study by Anderson (1953) also linked migration patterns
to social and economic characteristics of local areas. His analysis
covered metropolitan areas in the northern and eastern United States
only, for 1935-110. Four variables (percent unemployed, mean rent,
logarithm of population size, and location) successfully explained
about two-thirds of the total variance in each migration rate. Two
of the four factors were economic in naimre and one of them--percent
unemployment—was the same variable found by Bogue Min. to be significant.

Watson (1959) correlated a large number of social and economic
characteristics of SMSAs with gross and net migration, 1909-50. He
found it important to distinguish between economic opportunity and
level of living in analyzing areal economic variables. Migration varied

directly with economic opportunity and inversely with level of living.

The variables best explaining net migration in a regression analysis were:
economic structure of the SMA, maturity, ecological location, and

 

15
amount of economic opportunity. Like Bogue gulp he found both social
and economic variables significant: his scheme was much more effective
at explaining gross migration (81% of the variance) than net migration
(49$ of the variance). His measure of economic structure of the SHA
was percent employed in mnufactmring, which he found to be negatively
related to net migration (r = .141). As in previous studies, there was
no established theory to guide kitson's choice of variables, so he was
forced to rely on intuition and casting a wide empirical net.

A dissertation by Blanca (1962) attempted to locate, again through
multiple correlation, the determinants of interstate migration in the
United States. Some of the independent variables employed were:
unemployment level, changes in unemployment, ”prospective unemployment, "
wage rates, geographic distance between states, level of education,
racial prejudice, climate, and movements of military personnel.

"Pro spective unemployment” is ”the annual rate of change in unemploy-
ment which would be expected to occur if workers were not able to
migrate between states." Changes in this factor accounted for 85% of
the variance in the regional rate of total civilian migration among
states: the remaining factors were unimportant.

Blanco' s study suffers from the handicap mentioned in the
introduction: since the independent variable and the dependent
variable are measured concurrently, for 1950-57, predictability is not
possible and the causal relationship is unclear. Also, the use of
states as units of analysis may have clouded the relationships with
some variables, because states are heterogeneous agglomerations of
mller units, rather than relatively integrated labor nrket areas.

16

The later study by Lowry (1966) attanpted to build upon Blanco's
results by refining her migration models and applying than to SMSAs,
since SMSAs are the census unit most closely reflecting labor market
areas. Lawry's variables included changes in the rate of natural
increase, size of the local military establishment, school enrollments,
employment level, and median incane. In a multiple regression analysis,
he found change in employment level to be the variable most highly
associated with net migration, as previously mentioned. Overall, his
regression analysis accounted for an impressive 97% of the variance.

As discussed previously, Lawry's results, like Blanco's,
are not susceptible to clear interpretation because the independent
variable is not operationally distinct from the dependent variable. In
addition, Lam and Blanco used net migration 332.9; as the dependent
variable, rather than the absolute number of migrants. This may not be
the most desirable procedure: if the number of migrants is not highly
correlated with populatbn size, use of rates has the flaw outlined
by Sakoda (1966) which was previously quoted.

A geographer, Herbert Kariel (1963) did a multiple correlation
analysis of migration among those United States counties with a favorable
net migration balance for the 1950-60 decade, linking net migration
to four factors: change in the number of employees in mnufacturing,
median family income, proportion of professional, technical, and
kindred workers, and size of the uployed civilian labor force.

He found all four variables significantly related to net migration,
but only the last variable, size of labor force, statistically

explained a substantial portion (Mi) of the overall variance.
Combined, the four factors explained 147% of the net migration variance.

17
The counties with the highest residuals were in the warmer areas of
the country, which presunmbly reflected an "amenities" factor at
work in attracting migrants, in addition to the economic variables used
in the analysis. Kariel (1963:220) came to the general conclusion that:
The amount of population growth due to net migration of
an area appears to be more closely related to economic than
to noneconomic factors. There is, nevertheless, some
evidence to suggest that amenities, such as those which
are associated with population agglomerations or a climate which
is considered to be ”desirable, " tend to attract people to areas.
In using the number of employees rather than payroll data
as his manufacturing variable, it is possible that Kariel missed
capturing net migration variance due to the multiplier effect. In
terms of creating job opportunities in service activities, the hiring of
new nmnufacturing employees at twenty thousand dollars each, for
example, would have a much greater impact than hiring them at eight
or four thousand each, since they would be likely to spend much more
money locally with the former salary than with the latter ones.
Recently, others have extended ecological analysis and correlation
methods to the study of sull urban places (Zuiches, 1970) and to State
Econanic Areas in West Virginia (Rutman, 1970). Zuiches found such
factors as college enrollment and military labor force, which would
reflect a strong local export base, to be positively related to the
amount of in-migration , though the independent variable with the
strongest relationship to in-migration was simply size of place.
The recent study by Karp and Kelly (1971), mentioned above,
made a significant improvement in experimental design over those of

Blanco and Lowry. The authors managed to account for a substantial
portion of migration variance using determinants that were theoretically

18
and operationally independent of migration. In the first half of
this two part stuck, 116 migration streams among seventeen large SHSAs,
1955-60, were correlated with six independent variables, some of which
were derived from Zipf's model and the earlier work by Watson. They
included distance, population of the destination SHSA, percentage of
families under three thousand dollars in income, percent unemployment,
and percent in manufacmring, plus an index of climate. The last three
variables were measured for both origin and destination. In a stepwise
regression analysis, six of the nine variables accounted for 71% of
the variance. Clinte had the highest single correlation (r = .57),
and the economic factors were less important than size and distance.

In the second part of the study, the export base (minima:
requirements technique) was used to assess potential for economic
growth, along with several potential measures of the local mltiplier
effect, including populatim size, population potential, distance to
the nearest larger SMSA, and percent employment in services. Using
these variables, the authors were able to explain 60% of the net
migration variance, 1955-60, among the M largest SIBAs.

The logic in using these factors to assess the multiplier effect
is as follows: isolated communities with poor access to a larger
cammnity tend to develop a relatively diversified and self- sufficient
econom, due to high transportation costs. As a result, income derived
fran the export base tends to be spent locally rather than for imports,
compared to other communities. This contributes to growth in the service

sector and further export base expansion.
The most recent export base-migration study was done by Karp

(1971). He again examined variation in migration patterns as a response

19
to the potential for economic expansion. Using a different measure
of the export base (index of surplus workers), Karp tested three
hypotheses: the size of SMSA migration streams is proportional to:
(1) population size, (2) extensity of the export base, and (3)distance.
A significant relationship was found for the export base and distance:
together they explained 59% of the variance in migration streams. He
likewise found net migration associated with both these factors, the
latter, distance, explained 51% of the net migration variance.

0 6: An t v A ch to Hea t

The export base is not a fundamental ecological characteristic.
It is rather an economic mechanism through which ecological influences
presumably operate. A strong export sector is created through
industrial mobility to, and industrial growth in, a given local area, or
through other economic shifts. One must look for more basic area
characteristics, such as population size, climate, or location, to
explain why a strong export base comes into existence. Hence, the
export base could not provide a complete ecological explanation of
migration in any case, even if some acceptable means of measuring
it were available. Ecological characteristics must be specified which
give rise to a strong export base. Part II (chapter IV) of the
dissertation seeks to isolate those seyents ofthe economy which
reflect its export strength. Part III (chapter V) then attempts to
link manufacturing growth, theoretically the most important export
mechanism, to specific ecological characteristics.

The primary assumption of Part II is that at any single point

in time SHSA population size is an accurate reflectiond‘ the strength

20
of its underlying export base. There is no particular type of employment
structmre which is especially conducive to subsequent in-migration.
Specifically, the assumption of the index of surplus workers, that a
heavy concentration of workers in one industry or group of industries
produces a strong export base, is rejected. Detroit, for example, with
its extreme specialization in the manufacturing and exporting of auto-
mobiles, is in no stronger a position to attract migrants than is San
Francisco, with its highly diversified economy. It is instead assumed
that each SMSA has attained its present size, rather than being larger
or smaller, because it serves particular functions in the national
econoay, due to ecological or historical factors, which have dictated
the extent of its growth.

Migratory "pulls" subsequent to arm given point in time are
induced by increases in export activities subsequent to tint same
point in time. This implies that to explain net migration one must
first isolate those variables which define the export sector of the
economy, then search out ecological factors which induce changes in such
variables. For example, net migration is not necessarily correlated with
present employment in manufacturing for a given area (in recent decades
the correlation has been negative), but net migration Lg likely to be
correlated with subsequent increases in manufacturing in the area,
regardless of present manufacturing level.

It is here mintained that increased activity in the secondary
sector of the econom is crucial, since this is the sector predominantly
concerned with experts. By contrast, the primary sector has such a
sophisticated technology and high productivity that it involves only
a shall and decreasing proportion of total output. Moreover, it

21
requires large amounts of land and is inapplicable to metropolitan
areas, the present unit of analysis. The tertiary sector, on the other
hand, is labor intensive almost by definition and forms an ever-increasing
share of total employment and output. Services are performed
primarily for a local market, however, rather than for export. Numerous
exceptions exist, such as the performing of services for tourists or
retirees, whose income is generally derived from outside sources,

$7

or the locating of state or national government activities in the

local area, such as military bases or conservation projects. Changes in

F’Frr'r-"m
1

these latter types of service activities will be isolated insofar as
possible and treated, along with changes in the secondary sector, as
expert activities. All other services will be regarded as local
market activities, which do not lead to differential economic growth.
The objective of Part II is thus to isolate those community
variables which contribute to export base variance. These factors
will then be entered in a multiple correlation and regression analysis
with net migration to determine their overall and joint predictive
power. The factors isolated will be the four mentioned above. It
is felt that measures of nmnufacturing, retirees, governmental activities,
and military personnel come as close as presently possible to isolating
the export sector of the economy on a comparative basis among Sl‘lSAs.
These factors may be criticized on grounds similar to those of
Blanco's and Lawry's models. They are not ecological characteristics
but changes in ecological characteristics: they would be as hard to
predict as migration itself. This is a valid criticism, but if linked

to migration, these factors do at least represent a refinement of
Lowry's model. The first factor, for example, links migration

22
specifically to manufacturing growth, rather than to employment growth
in general. Manufacturing is much more capital intensive than the
rest of the economy and hence less mobile. For the short term, most
types of manufacbrring cannot increase locally in response to an
inflow of migrants, due to the need for heavy capital investment,
construction of new facilities, and so forth. Industrial mobility is
rather the result of relatively long term planning, so that if
manufacturing is related to migration over a period of just a few years,
it is almost certainly acting as a "pull" for migration, rather than as
an outgrowth of it. By contrast, the tertiary sector can increase
almost overnight in response to a heavy inflow of migrants, because of
two factors: (1) retail services are needed by the migrants and such
services are well known for their ability to spring into (and out of)
existence almost overnight: (2) most of the migrants themselves must
find work, and if it is unavailable in manufacturing, they are likely
to take jobs as salesmen or other service personnel, at sharply
reduced rates of pay if necessary. The tertiary sector, except for
these segments isolated as serving export functions, may therefore
be viewed as a vehicle for absorbing those elements of the labor force
displaced by increasing productivity in the secondary sector, as

LIMP-tad (1955) argued-

t o Sch e
Since secondary sector growth is expected to be the variable with
the biggest impact on export growth and net migration, Part III of
the dissertation will focus upon ecological characteristics associated
with industrial expansion. Einpirically, there have been two general
trends in industrial mobility in recent decades: (1) local dispersion,

23
from central city to rural fringe areas:(2) national dispersion, from
the northeastern sections of the country to the south and west.

The basic hypothesis of this part of the dissertation will be
that both of the above trends have been governed by the goal of
minimising production and distribution costs (influenced in large
measure by an increasingly sophisticated, well-developed, transporta-
tion and comnications technology). Production costs may be further
divided into the costs of land and labor (the reuining two factors of
production, capital and entrepreneurship, are disregarded, since they
tend to exist in an essentially national, highly fluid, nrket). Low
labor costs should be reflected in low median income: low land costs
should be reflected in low population density, following the argument
of the early ecologists that competition for space increases land
value. More specifically, low SHSA population density is likely to
be associated with extensive areas of open land and possibly a large
rural fringe area. Extensive space is required for building now,
often highly automated, factories, employee parking lots, and storage
areas.

Measuring the costs of distritmtion is somewhat more complex
than measuring the costs of production. Industry has increasingly
concentrated in the hands of a few giant firms, each one producing for
an enormous national market rather than for a regional one. such a
huge nrket makes production in one central location less efficient than
production in a number of smaller plants at scattered locations. Spatial
dispersion has twin advantages: (1) it mkes the firm less dependent
upon the vagaries of any one location in terms of governmental regula-
tions, labor market size, taxes, and similar factors which can determine

2b
the fate of a corporation without being under its control:(2) it
actually brings the firm closer to the market, because it tends to even
the distribution between population and factories.

National dispersion, in terms of a model which treats reduced
distribution costs through reduced market distance as the primary
motivator, implies that the ideal ecological conditions for secondary
sector growth would be found in a city: (1) with a large pgpgl;§ign_giggj
(2) remote from other large population concentrations (long M to
larger SHSAs and low W3 (3) where a well-developed
manufacturing establishment is not already in existence (a high percentage
of W within the local labor force): (h) prefereably
in a pleasant climate (discussed below). These five ecological variables,
along with population density and.median income, are hypothesized to
show a high correlation with secondary sector growth.

Previous studies, such as Bogue gt_a_L,_ (1957) and Watson (1959),
have hypothesised, and some have found, migration to be directly
related to level of living. Often, level of living is not distinguished
from economic opportunity, and it is assumed that a high level of
living implies a high degree of economic opportunity. In the earlier
stages of industrialisation, when production is centered in a single
region, this may be true. In a decentralisation phase of industrial
expansion, the opposite is more likely, however. In trying to reduce
costs, industry may deliberately move to areas with a low median income:
this appears to be the case in the recent trend of moves to the southern
states, such as the mnving of the furniture industry to North Carolina.
Hence, job openings may occur most often in low income areas, and since

it does little good to move to a high income area unless one can find

25
a job, migntion may be inversely proportional to median incme. It is
so hypothesised in this study.

Climate is included partially because of Ullnmn's (1951:) argument
that ”amenities" have becme an important factor in regional growth.
Retirees and tourists, as well as certain types of "footloose” service
industries, such as research and development, tend to seek warm climates:
heavier industries eventually follow as local populations increase.
Additionally, there are definite cost advantages associated with a
warm climate, such as reduced heating costs and the resulting ability
to spread out large ‘mildings over an extensive area. Fuchs (1962),
for example, has argued that this latter advantage was important for
settlement of the aviation industry in California . Finally, top
corporation executives may make mobility decisions based upon their own
preference for a pleasant location. Ullman (1995:129) noted that:
"Officials of a large utility in Chicago have told me that one of their
biggest problems in luring industry to Chicago is the unwillingness of
executives to live in the city. " During recent decades, the factors of
decentralization, increasingly efficient transportation technology
(including freeway building), and climate should presumably have rein-
forced each other in industrial mobility.

Since changes in the numbers of retirees, military personnel,
and government workers are typothesised to influence net migration
considerably less than manufacturing growth, no specific hypotheses
will be advanced linking them to ecological characteristics. It is
noted in passing, however, that hypotheses could be generated by future
research, linking them to migration. Such hypotheses could most usefully
be formulated within a general theoretical framework of the type developed
in this project to explain manufacturing growth in local communities.

‘._._ .'a- .‘fl

:ltfi...

CHAPTER II
METHODOLOGY
UNITS OF ANALYSIS

SMSAs are the units of analysis in all three parts of the
dissertation. Those SMSAs are included which were originally defined by the
United States Bureau of the Bldget (1961), with one exception: some of
the data used in the study are not directly available for MSAs, but-
must instead be aggregated from county data. Since SHSAs in New England
are cemposed of city and township units rather than counties, data
cannot be obtained for them, and they have been omitted from the study.
Also, only SHSAs in the contiguous 158 states are included.

There are compositional differences between the 1950 SHAs and
the 1960 SMSAs. Some SMAs were divided into two or more separate
SMSAs, outlying counties were sometimes added to SMSAs, and new SMSAs
were added, due to population growth. In order that the 1950 data be
strictly comparable with the 1960 data, adjustments have been made so

that the 1950 data reflect the same areal units as the 1960 data.
DATA.

Data have been gathered from several sources, bit primarily from
ta‘mlations of the United States Census, 1950 and 1960, as indicated in
the United States Bureau of the Census (1952: 1957: 1962: 1963) reports.
Because the study is a comparative one among SHSAs, with information
required for vast numbers of persons, census data is the only feasible
source. The main drawback in using such data is that they were originally

26

 

27
gathered for purposes other than testing the motheses of the present
investigation: several of the measures are far from pure. A prominent
example is the use of the number of persons over 65 as a surrogate for
retiree income. Many persons over 65 continue to work, while others
may have no income. Also, there is much variation in inch size among
the elderly. Nevertheless, it is expected that this crude indicator,
the best available, will give an acceptable estimate of the impact of
retirees on a comparative basis. A thorough description of the amnner
in which the census data were originally aggregated, with a careful
specification of sources, omissions, and inclusions for each variable,
is given by the United States Bireau of the Census (1957:1V-XIXI).

nt V e

There are three possible means of getting migration data from
census records, each with its own set of advantages and handicaps. The
agidua]. method requires population data fraa two censuses, 1950 and 1960,
plus accurate fertility and mortality data. Births are subtracted and
deaths added to the difference in population size between the two censuses:
the resulting figure is net migratim, as obtained frm the following
equation:

NM = P2 - P1 -B + D

where P1 is population size at the earlier census, P2 is population size
at the later census, B represents births, D represents deaths, and NM is
net migration. From this method, only get figures can be obtained, and
all the errors of both censuses, as well as fertility and mortality errors,
are retained in the migration estimate. Multiple moves are not picked
up by this method, nor indeed, by the other two methods. The second method

is derived from a question on the census fem concerning state of birth.

28
Data from this method can of course be aggregated only by states, a
rather gross category, and migration patterns for short time periods
cannot be analysed. The third method is place of residence at 3 fixed
m. This is similar to the state of birth method, except that
it covers a shorter time period. Both of the latter methods are
vulnerable to lapses in the respondent's manory, and are not easily
aggregated into migration figures for SMSAs. The residual method
is the most suitable method for the present stuck, so it is the one used. E“
All three methods are more thoroughly discussed by Shyrock (1961+).

..A_

Some problems are encountered by the use of the residual method:
in Chapter IV, the problem is the relatively long length (ten years) of
the migration interval. An adjustment of the data is required, which
will be discussed in that chapter. A second problem is the fact that
the residual method does not allow a breakdown of net migration into
its components of in-migration and out-migration. Since there is
considerable evidence (Lowry, 1966: Iansing and Mueller, 1967: and
Beale, 1970) that the factors influencing in-migration may differ from
those influencing out-migration, it would be preferable to analyze changes
in these components separately.

A third problem with the residual method is the one mentioned
above: the errors of both censuses, as well as in vital events registration,
are retained in the net migration estimate. Presumably, Ms are large
enough that the errors tend to cancel each other out and hence do not
represent serials distortions.

The net migration variable used is simply the absolute amount of
population change due to migration as measured in persons. This variable
will then be correlated with population size: if the correlation is
high (about .70 or above), then net migration mtg will be used as the

29
dependent variable. Otherwise, the first variable will be used.

This procedure is necessary because there is no inherent justifi-
cation for using a migration rate, since population size does not
represent the population at risk for immigration. It therefore should
not be controlled unless it is empirically shown to be highly related
to net migration, according to the logic of Haenszel (1967) and
Sakoda (l966:5), discussed in chapter I.

An underlying assumption of the investigation, focusing upon 0.91
migration and linking it to manufacturing growth, is that net migration
represents a process somewhat different than gross migration. In an
industrial economy, an individual may migrate for arw number of reason,
but he is unlikely to remain at his destination permanently unless he
is able to find work (provided he is not independently wealtmr, a
confirmed pauper, or drawing a pension). Net migration is thus sharply
limited by economic growth, while gross migration is not. The latter,
which Watson (1959) found to have a high correlathn with both its
components of in-migration and out-migration, tends to represent
"population turnover, " which is affected not only by growth but by other
influences, such as local age structure, and the local size of certain
organizations such as the military services which have a tendency to
transfer personnel frequently to different locations.

In attributing the majority of differential economic growth among
metropolitan areas primarily to manufacturing growth, it is assumed that
rural to urban migration is the basic factor in net migration totals,
while urban to urban migration is the basic factor in gross migration
totals. Though urban-to-urban migration forms the largest set of flows, SMSAs
to a considerable extent merely exchange residents, since there are large

streams in both directions (Miller, 1967: Bacon, 1971).

 

30
ANALYSIS OF THE DATA

Partial and multiple correlation will be the data analysis methods.
Since all variables are measurable on interval scales, the Pearson product
moment correlation will be used in calculating the correlations. It is
assumed that all correlations are linear.

Data will be analysed separately within each of the three parts
of the study. In Part I, zero order correlations among various pairs
of variables will be examined. In Part II, a multiple correlation
equation will reveal the ability of the four selected export base
variables to jointly explain variation in net migration. Partial and
multiple correlation in Part III will indicate the ability of the
selected.ecological variables to explain variation in manufacturing
growth can reduce the latter's correlation witthet migration.

Since it is just a statistical technique for revealing patterns
of common variance, multiple correlation cannot of course produce a
causal interpretation for the observed relationships among variables.
It can, however, indicate whether the theoretical framework presented
in chapter I is consistent with empirical fact. The mere size of
correlation coefficients likewise has no causal implications, as
Braschler (1971) paints out in criticizing wendt and Varian (1970),
because correlations may be spurious. A good statistical fit does not
imply that a cause and effect relationship exists. There is no way of
absolutely proving or disproving any theory at a single stroke, but
repeated statistical analyses can shed light on the theory's strong
points and inadequacies, and it is for this purpose that multiple

correlation methods are used in the present study. Discussions of the

31
advantages and limitations of multiple correlation analysis are scattered
throughout the literature. Among the significant articles are those

by Cohen (1968), Gordon (1968), Blalock (1963), and Fennessey (1968).

CHAPTER III

A TEST OF THE SURPLUS WORKERS INDEX

The general hypothesis of this chapter is that existing methods
of measuring an area's export base are inadequate. Pratt has already
shown that one of the two major methods of measuring the export base--
the minimum requirements technique--is unsatisfactory. This chapter
provides an empirical test of the other major method-~the location
quotient, as represented by the index of surplus workers--to show that
it too is inadequate. In its absolute form, the index reflects total
labor force size: in its relative form, it reflects degree of speciali-
sation in the local economy, which in turn tends to be an inverse measure
of population size.

As indicated previously, the method will be to correlate the
index of surplus workers with net migration. If export base theory is
correct, an accurate measure of the export base should be highly correla-
ted with net migration, since a strong export base is conducive to
economic growth: it creates a demand for labor which can only be filled
under normal circumstances by migration of workers from other areas. If
a given export base measure is not strongly correlated with net migration,
one of two things is implied: (1) export base theory is invalid: (2)the
measure is a poor one. In the present case, the latter explanation is

much the more likely, for reasons cited in chapter I.

THE MIGRATION VARIABLE

If migration is measured over too short a time interval, it may

not reflect an area's mu potential for attracting migrants, since
32

33

substantial lag effects are possible. If the migration interval is
too long, extraneous variables become increasingly important. Given
the lack of a registration system for migration in the United States,
it is not possible to determine the most probable lag interval between.
the builddup of economic potential and the migration response. A study
in British labor mobility (Makower, Marschak, and Robinson, 1939)
indicated a lag of six to eighteen months. Assuming this amount
of lag, and the likelihood that economic growth would tend to be a
self-sustaining process occurring over a period of years, rather than
an instantaneous one, a five year migration interval seems appropriate.
Unfortunately, net migration estimates are available for SMSAs only
for ten year intercensal periods (United States Census Bureau, 1962),
obtained by the residual method.

Five year net migration figures are available for State Econaaic
Areas, however, from place of residence data on the 1960 census (United
States Census Bureau, 1963). Persons five years of age and older are
classified by place of residence in 1960 and place of residence in 1955
for each State Economic Area. This information provides a matrix of
migration flows, from which net.migration can be calculated. Since these
data are for’l955-60, they are not directly useful, though, because
export base data are available only for a census year (1950). The
desired migration period is therefore 1950-55.

Since 139 SMSAs are coterminous with metropolitan State Economic
Areas, it is possible to obtain estimates of net migration, 1950-55, for
such SMSAs (which represent nearly two-thirds of all SHSAs) by subtracting
the five year State Economic Area totals from the ten year SMSA totals.

The export base can then be calculated from 1950 census data. The 139

m 34 e... ‘3

3h
SMSAs in this case are obtained by matching State Economic Areas with
SMSAs, both as defined for the 1960 population census. (United States
Bureau of the Budget, 1961: United States Bureau of the Census, 1963)
SUSAs excluded by this process include those which: (1) are

located in.New England, since in this area SMSAs are builtwup from towns
while State Economic Areas are brilt up from counties: or (2) are located
on the border of two or more contiguous states, since a state boundary
divides such SMSAs into separate State Economic Areas: or (3) contain

more than one State Economic Area because their area is heterogeneous

 

economically.

As a check on the importance of the time interval, migration
data are also analyzed for the 1950-60 ten year interval. This should
provide insight into the degree to which selection of one specific
migration interval rather than another affects the results. Also, to
check on Lowry' s findings concerning the high positive relationship
between net migration and employment growth, a variable is included
which measures each SMSA's growth in employment, 1950-60, relative
to the overall growth in United States employment during that period.
(United States Department of Comerce, 1965-66). Seven variables are
thus included in the analysis.

OPERATICNAL HYPOTHESES

There are four specific vaotheses to be tested in this chapter:

mothesis l: The absolute form of the index of surplus workers
is directly proportional to total local employment.

Total local employment will be calculated as the sum of employment
totals in each of eleven broad industrial categories used in labor force

35
data for the 1950 census. Data are reported for SMSAs by the United States
Bureau of the Census (1953) for nine major industrial categories: agri-
culture: mining: construction: nanufactrring: transportation, communica-
tions, and public utilities: wholesale and retail trade: finance, insur-
ance, and real estate: ‘msiness and personal services, excluding private
households: and professional and related services. No additional
categories, entertainment and recreational services, and public adminis-
tration, are aggregated from county data listed in the United States

Bureau of the Census (1952). The ”industry not reported" category, as

trek—.11

well as the categories of "forestry and fisheries” and ”personal services
in private households, " are omitted from the analysis. These categories
are not readily available for SMSAs, and being mull, are unlikely to
significantly affect the SHSA employment structure.

W: The relative form of the index of surplus workers
is inversely proportional to population size.

Population size is reported in the United States Brreau of the
Census (1953) .

mpgmesig :2: The relative form of the index of surplus workers
is unrelated to net migration.

W: Net migration is directly proportional to change
in employment level.

Data on employment level changes, 1950-60, are found in the Office
of Brsiness Economics (1965-66). Verification of this impothesis confirms
a relationship between net migration and general employment growth,
replicating Lawry's findings. Succeeding hypotheses will link net

migration specifically to measures of the community export base.

36
RESULTS

The zero-order correlation matrix is given in Table 1. There is
a substantial correlation of .60, betwaen the 1950 export base and net
migration, 1950-55. The correlation between the export base and total
local employment is .99, however. This indicates that in its absolute
form the export base is so strongly affected by labor force size that
it is nothing more than an alternative indicator of the latter variable,
strongly confirming Hypothesis 1 that the absolute form of the index is
directly proportional to total local employment. When the influence of
total uployment is removed by using the relative form of the index, the
latter's correlation with net migration plummets to —.01, a completely
random linear relationship. This finding confirms Hypothesis 3, that
the relative form of the index of surplus workers is unrelated to net
migration. A similar conclusion is reached for Hypothesis 3 if a ten
year'migration interval is used, rather than the five year one, though
the initial correlation is lower (.35), due to the longer interval.
Hypothesis 2, that the relative form.of the index is inversely proportional
to population size,is not supported by the findings, since the two have
an essentially random -.09 correlation. Population size is highly
correlated with total local employment (.99). and in controlling for the
latter, the former is also controlled.

Total relative employment growth, 1950-60, correlates highly
(.93) with net migration, 1950-60, confirming Lowry's results on the
close relationship between employment growth and net migration, and thus
confirming Hypothesis h as well. A summary of the above findings

for the first four hypotheses is presented in Table 2.

 

37
Table 1. Simple Correlation Matrix and Basic Statistics for Part I

 

 

 

 

Variable Variable Number Standard

Number‘ 1 2 3 h __ .5 '6' 2_ Mean Dgziation
1 1.00 - - - - - - 28752.8 63068.5
2 .99 1.00 - - - - - 168012.6 “03325.4
3 -.02 -.09 1.00 - - - - 0.18 0.06
a .99 .99 -.09 1.00 - - - “49584-5 1027955.“
5 .00 .0“ .03 .06 1.00 - - lth3.“ 67768.1
6 .35 .38 .02 .uo .93 1.00 - “8.8 150.8
7 .6“ .67 -.Ol .68 .76 .93 1.00 #0182.“ llu677.5

 

‘Variables are as follows: (1) absolute form of the index of total
surplus workers, 1950; (2) total local employment, 1950: (3) relative form
of the index of total surplus workers, 1950; (h) population size, 1950:
(5) change in employment total, 1950-60, relative to total United States
employment total; (6) net migration, 1950-60; (7) net migration, 1950-55.

Table 2. Summary of Results for Hypotheses l to h

 

 

 

 

Hypothesis Hypothesis
Number Supported
1 Yes
2 No
3 Yes
h Yes

 

38
It is evident from these findings that either economic growth

potential is not an accurate predictor of migration or that the

index of total surplus workers is not a suitable measure of an area's
potential for economic growth. For reasons given in the next section,
the latter is much the more likely explanation.

DRAWBACKS OF THE SURPLUS WORKERS INDEX

According to Mattila and Thompson's line of reasoning, the total
surplus workers formula represents the number of export workers in
an area's econoaw. There is an inherent weakness in the fonnla,
however: for each surplus worker in a local industry, there must be
a corresponding "deficient" worker in some other local industry. Overall,
total surplus workers must equal total ”deficient” workers, so that the
formula really measures the degree of specialisation (that is, the
degree of deviation of the area's employment profile from the profile
of the nation as a whole), rather than local export strength, for cities
of roughly comparable population size. An increase in export strength
(that is, an increase in surplus workers) can occur only at the expense
of a decrease in self sufficiency, since the combined sum of basic and
service workers must equal total local uployment, th denominator of
the index.

Algebraically, the requirement that total surplus workers in
any area must equal deficient workers can be shown as follows, using
only the asmmptions that Eei = °t and 2521 = Et' with S equal to the
combined total of surplus workers and "deficient" workers:

mere—Bi]
Zei - 2E1
Et

0)
I

°t

 

I‘A’

= e - —— e (since E is constant for
1:. El
t all E11)t

= 0

Thus, if industries with less than their pro rata share of workers
(”deficient” workers) are included in the calculations, total surplus
workers invariably sum to zero. Effectively, this assumes that an
increase in service workers weakens the community export base, when in
fact the opposite is true, as implied by the ”multiplier effect. "

A simple Impothetical example will illustrate the above weakness
of the index. Suppose that a local economy has 5000 workers, divided
into four industrial categories, as shown below, along with the corresponding

figures for the hypothetical national econom:

e11 = 1200 E11 = 40.000
.12 = 200 1212 = 130.000
.13 = 3000 E13 = 68,000
011; = 600 311, = 12,000
.15 = 5000 )5;t = 250,000

Surplus workers in each of the four industries are as follows: 51 = 1500,
$2 :2 -21+00, 53 = 161:0, and Sh = 360. Total surplus workers thus equal
M0 4- 1640 + 360 = 2M0. Likewise, there are 2M0 ”deficient" workers
in the local economy: in this case, they all happen to be in industry 2.
Suppose now that the local economy manages to greatly increase
its production in industry 2, to the extent that 5000 new worker are
attracted from other geographic areas and employment in this industry
skyrockets locally from 200 to 5200. Recalculating the surplus workers

 

40

formulas gives the following totals: $1 = J+00, 82 = 0, S = 280, and

3
31+ = 120. Total surplus workers now equal #00, and there are l+00
deficient workers. Thus a dramatic increase in industry 2 employment,
which has eliminated the 2M0 "deficient“ workers in that industry, has
acimally caused the local number of total surplus workers to decrease
from 21400 to 400, even though employment in all industries remained
constant, except for industry 2, in which employment increased.
Mathematically, this peculiar result occurred because the increase

in industry 2 employment caused a corresponding increase in total

local employment, so that all formulas had to be re-calculated on a .
new larger base. Logically, such a result challenges the validity
of the location quotient as a reflection of an area's export base, since,
as explained below, an increase in service workers actully tends to
strengthen, not hinder, the local economy. 7

Pratt found the minimum requiranents technique deficient from
both a logical and empirical standpoint, for reasons discussed in
chapter I. The present study reveals that, as a measure ofthe comnity's
overall export base, the location quotient suffers from an equally
severe internal contradiction: since it is based on employment totals,
it assumes that an increase in service workers weakens the local
expert base, when in fact the opposite is true. An increase in surplus
workers actually strengthens the export base, because it helps insure
that export income will be spent locally rather than for imports.
Export dollars only promote economic expansion to the extent that they
remain in the local econonw to create new jobs rather than flowing

back out in the purchase of imports.

41
Ironically, to the extent that its assumptions of equal consumption
(and productivity levels throughout the economy are true, the total surplus
workers index measures not only the amount of export activity bit the
a t a as w . Since the method must logically
assume that each worker in every industry is supported ty a proportionate
number of workers in every other industry, the absence of such workers
in the local area (a high total surplus workers index) implies that
they are located in some other area: hence their services must be
imported. For each city, the amount of exporting consequently equals
the amount of importing (“surplus" workers equal "deficient" workers),
and the index of total surplus workers is meaningless as ameasure of
the local export strength. It appears that am measure based on
employment totals, rather than dollar flows representing the value of
exports and imports, is subject to this fatal flaw.

Hattila and Thompson (l955:226)found that there is mob variation
among major United States cities in the ratio of surplus to service workers,
ranging from a low of 1:1.99 in Chicago to a high of 1:4.lb7 in
Philadelphia. As Blumenfeld (1955:121) asks: ”Are we to believe
that one ”basic" worker supports two ”nonbasic” workers in Chicago
and four and one-half in Philadelphia?! Actually, the implication is
that Chicago does a great deal more exporting W than
Philadelphia: the index of total surplus workers measures both variables.
Part II of this project therefore addresses itself to an alternative means

of analyzing an area's export base and linking it to migration.

CHAPTER IV
A REVISED EXPORT BASE APPROACH

Since no suitable index of an area's export base exists, and
since the more complicated input-output matrix approach has not been
perfected by economists, the only choice left is to find another means
of assessing the community export base.

The export base will be treated here as a function of four
variables: changes in gggggamgptgl employment, the number of militgzy
personnel and mg. and the level of manufacturing activity (Euoll shift).1
Increases in these variables create money income for a community, which,
if spent locally, leads to a still stronger export base and an ever-
expanding cycle of economic growth. The four indicators of these export
base variables selected for use are: change in the number of employees
in public administration, change in labor poyroll, change in the number
of persons over 65 years of age, and change in the number of military

personnel.
PROCEDURE

A multiple correlation analysis will be carried out which will
include the four independent variables and net migration. In this and
the succeeding chapter, the full sample of 188 SMSAs, as defined by the
Bureau of the Budget (1961) is available for use, since netvmigration

can appropriately be measured over a full ten year interval.

 

1In the tables reporting the results of data analysis, the under-
lined words will be used to refer to the corresponding variable.

#2

43

One variable, size of labor payroll in 1947, has a certain
amount of missing data. Phyroll figures are not published for'any
county in which aggregate totals may reveal information concerning the
operating budget of any individual manufacturing establishment. Figures
for counties which are dominated by one or a small number of establish-
ments are thus withheld from publication. Fortunately, only a handful
of SMSAs are so affected. In two SMSAs, payroll information is not
given for the county in which the central city of the SMSA is located.
These two, Durham, North Carolina, and Iawton, Oklahoma, are omitted
from the calculations, reducing the overall semplesize to 186. In four
other SHSAs, Amarillo, Denver, Knoxville, and Richmond, data are withheld
only for one of the peripheral counties. Rather than omit these SHSAs
for lack of complete information on a single item, an estimate is made
of payroll size in each of the affected counties. This estimate is
based on the assumption that payroll per>emp1oyee is the same in the
county without published data as in the rest of the SHEA. The following

formula is used:

Employees; _ Payrollx
EmployeesY - PayrollY

where X refers to the entire SMSA and Y refers to the SMSA minus the
county in question. Luckily, complete figures on the number of employees

are published for every county with missing payroll data.
OPERATIONAL HYPOTHESES

Five hypotheses will be tested in this chapter:
fizngfihggig_fi?: Net migration is directly proportional to manufac-
turing growth.

 

2Hypotheses are numbered consecutively throughout the dissertation.

 

Ill-Ii]
lull

M

Manufacturing growth will be measured by change in labor payroll,
1947-58. This time period is used because the Census of Manufactures
was conducted in 1947 and 1958, rather than in 1950 and 1960. Payroll
data will be used because export income rather than export employment
is the basis of export base theory.

Eyppthesis 6: Net migration is directly proportional to increases
in the number of retirees residing in the community.

Retirees will be measured in terms of the absolute number of
persons 65 years or older in the SMSA in 1950, subtracted from the number
in 1960. All retirement income is effectively export income, since
retirees perform no local economic functions, as a group.

Extent of tourism is probably a significant export base variable,
but no satisfactory census data exist for measuring it. In any event,
it is likely to be highly correlated with the number of retirees,
since retirees and tourists favor the same type of environment, such
as Florida. Tourism will therefore not be included as a separate variable
in the study.

gupppppp;p_z: Net migration is directly proportional to changes
in the amount of governmental activity in the community.

Governmental activity should be reflected in the number of
workers employed in public administration in the community. This
category includes federal, state, and local government employees
involved in predominantly governmental functions, such as administrative
or legislative activities. It should capture the substantial governmental
growth on all levels that has been occurring in recent decades, and
which has undoubtedly benefitted some areas more than others.

Hypothesis 8: Net migration is directly proportional to change

in the size of the United States military establishment in the community.

45

Changes in this variable should be reflected by changes in the
absolute number of military personnel in each SMSA, 1950 to 1960.

Number of military personnel is determined by subtracting the size of the
civilian labor force from the total labor force.

h s s : Changes in the level of manufacturing payroll, government
activity, military personnel, and retirees jointly account for a substan-
tial portion of the statistical variance in net migration.

It is not to be expected that net migration can be completely
explained in a statistical sense by this or any other set of variables.
It is well-established that migrants move for a wide variety of reasons
(for example, Lansing and Mueller, 1967), and that patterns of migration
vary widely among different areas. Since most migrants must fit into
the economic structure at their destination in order to ranain there
permanently, however, it is reasonable to expect that net migration, as
opposed to gross migration, would vary to a considerable extent with
those factors which govern economic growth. This twpothesis will there-
fore be regarded as confirmed if the above four export variables do
not show excessively high correlations with each other, if each variable
explains a substantial amount of net migration variance independent of
the other three, and if jointly they explain statistically the mjority

of the net migration variance.
RESULTS

Since each of the variables except manufacturing payroll is
expressed in terms of absolute numbers of persons, a sixth variable,
population size of each SMSA in 1950, is included in the analysis, to
determine how closely the independent variables are correlated with

simple population size.

46

As indicated in Table 3, the simple correlation between net
migration, 1950-60, and population size, 1950, for the 186 SMSAs is
.42. Using Sakoda's logic, this amount of relationship is definitely
too low to justify using a net migration rate as the dependent variable.
The absolute amount of net migration is therefore retained as the

dependent variable.

Table 3. Simple Correlation Matrix and Basic Statistics for Part II

 

Variable Standard
Number Vapiaple _l_vag _3 4 5__ 6 Mean Deviation

 

1 Net Migration 1.00 - - - - 43.4 133.6

132562.3 299334.5

2 Pryroll Shift .74 1.00 - - -

3 Retirees .59 .881.00 - - - 14765.7 32016.0
4 Administration .71 .71 .741.00 - - 2511.1 4522.4
5 Military .51 .33 .24 .511.00 - 1924.1 5395.2
6 Population .42 .86 .96 .68 .18 1.00 441760.9 942206.2

 

As can also be seen in Thble 3, each of the four export base
variables is highly related to net migration, the correlations ranging
from .51 to .74: payroll shift,has the highest correlation with migration.
Three of the four independent variables, payroll shift, retirees, and
administration, are highly interrelated, however, with intercorrelations
ranging from .71 to .78, so that each one can be eXpected to have only
a small correlation with the migration variable independent of the other
two. Likewise, each of the three is rather strongly related to population
size (r = .68 to .96).

47

Retirees has such a strong correlation with population size (.96)
that it may be regarded as nothing more than an alternative indicator of
the emme variable. Possibly, most of the differential increase in persons
over 65 during the decade was due to population aging rather than to net
migration.

The high intercorrelations among the independent variables suggest
that they may be an interdependent complex rather than a set of distinguish-
able factors, or that each variable is actually measuring the same general
phenomenon, perhaps metropolitanization. Statistically they are likely
to be an ineffective combination of variables to use in a multiple
correlation analysis, because of their high intercorrelations.

Proceeding nevertheless with the multiple correlation analysis,
it can be seen, in Thble 4, that the independent variables account for
a substantial portion of the net migration variance (32:.68). As
expected, removing any one of the variables does not cause much of a
reduction in the amount of variance explained. Removing military causes
the multiple correlation coefficiaxt of determinatim to drop from
.68 only to .66, while removing retirees or administration likewise
produces only a slight drop. The effect of payroll shift is somewhat
greater. Hypothesis 9 must be rejected, since this combination of
variables is ineffective in jointly explaining the net migration
statistical variance. Table 5 summarizes the evaluation of hypotheses.

The correlation between payroll shift and migration is an impressive
.74. Previous investigations, using changes in the number of manufacturing
employees or other measures to reflect manufacturing increases, have not
found such a high correlation. It is possible that previous studies
failed to reflect a multiplier effect, since such other'measures usually

involve numbers of people rather than dollar increases in local income.

48

'Bable 4. Multiple Correlation of Export Base Variables with Net Migration

 

Coefficient Partial

 

 

Regression Standard Beta Correlation R2
Variable Coefficient Error wgght gpeffipient Delete‘
Intercept ~12.52929 6.42016
Payroll Shift .00036 .00004 .81 .55 .55
Retirees -.00190 .00041 -.45 -.32 .65
Administration . 01140 . 00212 . 39 . 37 . 63
Military .00369 .00126 .15 .21 .67

R=.83 R2=.68

 

‘Final column represents the value that R? would assume if the
variable listed in the corresponding row were deleted from.the analysis.
The addition of each variable except military is statistically signi-
ficant to at least the .01 level, according to an F test of significance.
(Draper and Smith, 1966:67-69).

Table 5. Sum of Results for Hypotheses 5 to 9

 

 

 

Simple Correlation Hypothesis
Hyppthesis Variable Coefficient Supported
5 Payroll Shift .74 Yes
6 Retirees .59 Yes
7 Administration .71 Yes
8 Military .51 Yes
9 Combined Relationship No

 

'ma—i—mu. Atm-

49
In summary, Hypotheses 5 through 8, concerning the relationship
of each of the four independent variables to the dependent variable,
are all individually confirmed. Hypothesis 9, concerning their
effectiveness as joint predictors, is rejected. The practical and
theoretical implications are less clear-cut, as will be discussed in

the following section.
DISCUSSION

The high correlation between migration and the manufacturing
variable suggests that the latter'moy be a good measure of the export
base. It is reasonable to argue that.a causal link exists between
these two variables. Given the capital intensive nature of most types
of manufacturing, and the resulting inability to respond to population
shifts in a short period of time, it is probable that manufacturing
growth acts predominantly as a cause, rather than an effect, of migration
in any causal relationship.

Based.upon the findings of the present chapter, the manufacturing
variable is the best single predictor of net.migration. Not only does
it have the highest correlation, but the manufacturing industry employs
many more persons in SMSAs than either public administration or the
military establishment, so that manufacturing shifts are likely to have
a stronger impact upon the local economy, and ultimately upon migration
patterns. The impact of retirees and tourists is also likely to be less
than that of manufacturing, except in special cases such as Saint
Petersburg.

From the high correlations among all of the independent variables,

except the one reflecting changes in the military establishment, with

 

50
population size, it is evident that the largest increases in manufacturing,
the number of elderly persons, and of governmental employees in public
administration, all occurred in large population centers during the decade
of the 1950s. Net migration was generally greatest in large cities as
well, and it is possible that the relationship between the independent
variables and the dependent variable is to an extent just an artifact of
population size. Manufacturing payroll, for example, may have increased
in large cities primarily because of unionized wage increases and
inflationary pressures, which are usually greatest in large cities, while
the net-inflow of migrants was due to the appeal of large cities or
other factors. In any event, a thorough analysis of the ecological
factors associated with manufacturing changes and with differential
net migration may help clarify these questions. This will be done in
the next chapter.

CHAPTER V

ECOLOGICAL FACTORS ASSOCIATED WITH MflNUFACTURING SHIFTS

The fact that the manufacturing variable correlates highly with
net migration lends some support to a "pull” theory of migration based
upon differential growth of economic opportunities. Such opportunities
presumably arise in response to the potential for economic development,
which in turn is produced by a strong export base. Expansion of industry
in a given area creates new jobs in manufacturing; the income derived
from such jobs that is spent in the local economy creates further
economic growth and additional Jobs in the service sector.

It is hypothesized that in the United States in recent decades
industry has expanded primarily through a national decentralization
of alreadybexisting industries. Decentralization of productive
facilities may minimize distribution costs for those companies selling
in a national market, since there is presumably a point beyond which a
division of labor in one location has sharply diminishing returns.

The ecological characteristics of a region which.make it
attractive for industrial expansion are probably many and varied. Seven
specific ecological factors, derived from the theoretical framework of
chapter I, will be evaluated in the present chapter in terms of their
statistical relationship to manufacturing growth or decline. They are:
population size, distance to nearest larger SMSA, population potential,
amount of employment in service industries, climate, population density,
and median income. The logic behind the selection of these variables
involves the general hypothesis that these factors indirectly promote a

51

 

52
high degree of net in-migration, by attracting industry. Job openings in
'manufacturing and in an expanding service sector in turn attract migrants.
The same sample of 186 SMSAs used in chapter IV are again used in the

present analysis, with the same adjustments made to the 19h? payroll data.
OPERATIONAL HYPOTHESES

Seven remaining hypotheses will be tested in this chapter, accord-
ing to the logic of the industrial mobility scheme presented in chapter I.

fiypothesis 10: Manufacturing payroll growth is directly pr0portional
to the distance to nearest larger SMSA.

Industry should be attracted to areas isolated from other centers
of industrial production. Distance is measured in tonne of airline miles
between SMSAs (United States Department of Comerce, 1961).

fizpothesis 11: manufacturing payroll growth is directly propor-
tional to population size.

Larger cities have a greater market for the products of industry.

fixpothggis 12: Manufacturing payroll growth is inversely propor-
tional to population potentia . Industry is presumably decentralizing
away from the high population potential areas of the northeastern regions.

Population potential is a measure of the distribution of population
over territory in relation to a given point. As Duncan (1957:367)
interprets the measure: "If potential at La figures out at 350,000, this
means that, 'on the average,’ a person at LG is within a mile of 350,000
other persons." For the United States in 1950, the measure is essentially
an indicator of regional location, with low values in the south and west.

Eypothesis 1}: Manufacturing payroll growth is inversely propor-

tional to median income. The relationship is expected to be an inverse

53

one because of the argument that industrial mobility is toward areas of
low labor costs, so these are the areas with employment opportunities.

Median income is for families in 191.9, as given in the United
States Ezreau of the Census (1953).

flypothesis ;u. Payroll growth is directly proportional to climate.

The climate index will be an adaptation of a measure used by
Karp and Kelly (1971:2h-26): mean January temperature divided by mean
July temperature. Presumably people (and industries) seek moderate
temperatures, in the form of cool summers and warm winters.

gzpothesis l5: manufacturing payroll growth is directly propor-
tional to percentage employment in ggzgigg,industries. Industries may
be expected to move toward areas where large manufacturing concentrations
are not already in existence.

This variable is measured as a residual category which includes
all activities other than resource-extracting activities and manufacturing.

fizpothesis 16: Manufacturing payroll growth is inversely propor-
tional to population density. Areas with ample room for large new

physical plants for industry should attract more industry.
RESULTS

The zero-order correlation matrix is contained in Table 6. Only
population size and distance show a high correlation with manufacturing
payroll shifts (.86 and .73, respectively). 0f the remaining five inde-
pendent variables, just two are significantly related to payroll shift, and
these, population density and median income, have correlations too low (.26

and .29) to have practical significance. Moreover, they are in the Opposite

5h
direction to the one hypothesized. Statistically, four of the seven
Impotheses in this chapter are thus confirmed (Table 7), but most of the
explained variance in payroll shift is accounted for by population size
and distance, as will be seen below. Moreover, the latter two concepts
have a rather high intercorrelation (.82), so they are likely to be an
ineffective combination to use as independent variables in a multiple
correlation equation. Their high intercorrelation is to be expected,
since a large population size implies that one mst generally travel
a greater distance to find an even larger one.

Table 6. Simple Correlation Matrix for Ecological Variables and Net
Migration

 

 

 

 

Variable V e be .
N332: Variable J. 2 L a 5 L A _9

1 Population 1.00

2 Distance 0 .82 1.00

3 Potential .20 .09 1.00

1} Services -.03 .18 -.62 1.00

5 Climate .00 .12 «‘43 .lb2 1.00

6 Density .3“ .20 .31 -.10 -.06 1.00

7 Median Incas .26 .12 .19 -.21 -.’+7 .18 1.00

8 Payroll Shift .86 .73 .13 -.07 .06 .26 .29 1.00

9 Net Migration .142 .51 -.21 .16 .35 .00 .15 .74 1.00

 

The final row of the Table 6 matrix indicates that the relation-
ship of each variable to net migration cannot be predicted by knowledge
of its relationship to payroll shift shown in the row imediately above.

Three of the seven ecological variables (population potential, services,

55

Table 7. Basic Statistics and Summary of Results for Hypotheses 10
to 16, Concerning Relationships of Ecological Variables to Manufacturing
Payroll Shifts.

 

Simple

Ecological Standard Correlation Hypothesis
Vapipble Mean Deviation gypothesis Coefficient Supported
Distance 138.8 270.6 10 .73 Yes
Population ##1760.9 9h2206.2 11 .86 Yes
Potential 312.2 121.” 12 .13 No
Income 3198.1 511.9 13 .29 No
Climate 0.“? 0.1“ 14 .06 No
Services 68.5 12.0 15 -.07 No

Density 423.8 1132.0 16 .26 No

 

56
and climate) have a stronger relationship to net migration than to payroll
shift. Climate in particular shows a fairly substantial (.35) correlation
with net migration and a random (.06) linear correlation with payroll
shift. This pattern of relationships clashes with the theoretical frame-
work of the study, which hypothesizes that each of the seven ecological
variables influences net migration primarily through its effect upon.
manufacturing growth: in such a case, the correlation of each of the
variables with payroll shift should be higher than its correlation with
net migration. The relatively high correlation of climate with net migra-
tion lends support rather to a non-manufacturing growth theory. The
relation of population size and distance to net migration is as expected:
the correlations are high (.42 and .51), but lower than their correlations
with payroll shift. In the case of population and payroll shift, there
appears to be a multicollinearity problem. The latter two variables do
have a clearly distinguishable relation to net migration, however.

The multiple correlation analysis in Table 8 shows that the
seven ecological variables jointly account for a substantial 77% of the
payroll shift variance, but as Table 9 shows, these same variables account
for only 43% of the net migration variance, unless payroll shift is
included as an independent variable. The higher correlation of the
ecological variables with payroll shift than with net migration is as
expected. However, the partial correlation between the latter two variables
is a very high .86, implying that there is a substantial relation between

these two variables that is left unexplained by the seven ecological variables.

Additional Multiple Correlation Analyses
As indicated above, several of the ecological variables do not

have a significant relationship to payroll shift, and inclusion of them

57

Table 8. Multiple Correlation Analysis of Seven Ecological Variables
with Payroll Shift.

 

Coefficient Partial 2
Regression Standard Beta Correlation R

Variable Coefficient Error Weight Coefficient Delete
Intercept -97805.99 157439-40

 

 

Climate 305188.95 99721.67 .14 .22 .76
Potential -ll7.45 139.19 -.05 -.07 .77
Distance 138.95 77.87 .13 .13 .77
Medan Income 75 a 12 25 e ’43 a 13 o 21 a 76
Population 0.24 0.02 .74 .60 .64
DenSity '8008 10.55 -003 -006 077
Seflicas -3m’6068 1114'9072 "e 1“ 'e22 076
= .88 R2 = .77

 

Table 9. Multiple Correlation Analysis of Seven Ecological Variables
and Payroll Shift with Net Migration

 

 

Coefficient Partial

Regression Standard Beta Correlation R2
Variable Coefficient Egror, weight Coefficient Delete
Intercept -291.511 56.988
Climate 246.554 36.993 .26 .45 .81
Patenthl -0 e 013 O 0 01‘8 " a 01 - a 02 e 85
Distance 0.119 0.028 .24 .30 .84
Median Income 0.037 0.009 .14 .28 .84
Population 0.00013 0.00001 -.95 -.69 .72
Density -0.008 0.004 -.07 -.16 .85
Services 1.046 0.426 .10 .18 .85
Payroll Shift 0.00060 0.00002 l..35 .86 .43

R = .92 a? = .85

 

 

58
in the mltiple correlation equation probably distorts the correlations
of the ranaining variables, or at least adds nothing to the equation.
The analysis is therefore redone, using only those variables which are
both substantively and statistically significant. This includes just
population size and distance, since these are the only two ecological
variables whose relationship to manufacturing shifts is both statistically
significant and sufficiently high to warrant further examination.

These two variables explaining 74% of the variance in
payroll Shift (Table 10) , but it is population size which contributes
substantially more to the explained variance of payroll shift. The
partial between the latter two variables is .66, while the partial
between distance and payroll shift is only .11. It may therefore be
concluded that the combination of ecological variables chosen for this
investigation does not support the theoretical stmcture presented;
population size alone statistically explains about as much of the
variance in payroll shift as any combination of the variables. This,
of course, provides no basis for drawing conclusions about the causal
relationships among the variables, since the particular choice of
variables to be included in the analysis strongly influences the pattern
of observed partial correlations. Moreover, the ecological variables
are rather highly intercorrelated, and as Gordon (1968) shows, the sizes
and patterns of correlations among the independent variables affect
their relationship to the independent variable.

As shown in Table 11, population size and distance are
much less effective in explaining the variance of net migration than in
explaining the variance of payroll shift (26% versus 74%): in addition,
distance has the higher correlation (.32) to net migration. With distance,

held constant, population has a perfectly randan relation to net migration.

 

 

I!
III
It}!

59

Table 10. ‘Multiple Correlation Analysis of Population Size and Distance
with Manufacturing Shifts

 

 

Coefficient Partial
Regression Standard Beta Correlation R
vs e oe i t Error W" t C eff c t e te
Intercept 3018.88" 12596.82
Distance 10u.97‘ 71.56 .09 .11 .7h
POPUl‘tion 0.29 0002 078 066 .5“
R = 086 R2 = a 7'4

 

.‘Coefficient not statistically significant.

Table 11. IMultiple Correlation Analysis of Population Size and Distance
with.Net.Migration

 

Coefficient Partial
Regression Standard Beta Correlation R2
Variable C e ficient we t C efficient Delete
Intercept 8-558 9.57
Distance 0.25 0. 05 . 50 . 32 . 17
Population 0. 0000012‘ 0. 000015 . 01 . 01 . 26
R = .51 R2 = .26

 

“Coefficient not statistically significant.

 

60

The fact that population size and distance explain less of the
variance in net migration than in payroll shift is to be expected in
terms of the present theoretical framework, which hypothesizes that
these variables exert their influence on migration through their ability
to attract industry. The size of the difference, however, suggests that
the model is incomplete or incorrect. The results imply that other
ecological factors, not included in the analysis, are important in pro-
moting manufacturing growth, as it relates to migration. This is
suggested by the figures in Table 12, which contains the multiple
correlation analysis of population size, distance, and payroll shift
with net,migration. The inclusion of distance and population in the
calculations should sharply reduce the first order correlation of
payroll shift with net migration. Instead, the correlation actually
rises, to .833 with payroll shift held constant, population size shows
a strong negative relationship to net»migration. Apart from their
function as centers of economic opportunity, as reflected in manufactur-
ing payroll growth, larger cities apparently tended to repel migrants rather
than attract them. This may be due to an amenities factor, combined
with societal affluence: cities in the 19503 were well recognized as
centers of congestion, crime, and pollution. At the same time, many
persons were taking increasing advantage of social security and retirement
plans to settle away from the larger cities.

The final column of Table 12 implies that distance contributes
only a modest amount to the multiple correlation of these variables with
net migration. With distance removed, population size and payroll shift
still have an R? of .73 with net migration, while with population size

removed, R2 drops to .55. and with payroll shift removed, it plummets to .26.

61

The latter two variables alone account for a remarkably large proportion
of the overall net migration variance (73%), compared to previous
studies. Watson's (1959) four independent variables (economic structure,
SMSA maturity, ecological location, and amount of economic opportunity)
could jointly explain only l+9% of the net migration variance, and other
studies have seldom accounted for more net migration variance. Net
migration is apparently a complex variable that has traditionally not
correlated highly with gross social and economic variables. The
multiple correlation results of payroll shift and population with
net migration is given in Table 13.

As previously revealed by Table 7, climate and net migration
have a substantial intercorrelation, tut climate does not correlate
highly with payroll shift. Climate is therefore included in a final
multiple correlation analysis, along with population size and nanufactur-
ing. As shown in Table 1’4, these variables statistically account for
a full 80% of the net migration variance. Like population size, there-
fore, clinate makes a substantial contribution to the variance of

net migration, independent of its relation to payroll shift.

Effects of Wage Inflation
The last finding of the investigation concerns the relationship

of a large initial payroll to subsequent manufacturing payroll increases
and to net migration. In using aggregate payroll in dollars as an
indicator of industrial mobility, there is the danger that inflation

and unionized wage increases may produce substantial increases in payroll
that are not actually due to manufacturing mobility or growth. Those
SMSAs with a large initial payroll would show the largest such

inflationary increases, and this factor could conceivably distort the

62

Table 12. Multiple Correlation Analysis of Distance, Population Size,
and Payroll Shift with Net Migration

 

Coefficient Partial

Regression Standard Beta Correlation R2

c t 'w t f c t eteb
Intercept 3.45638‘ 5.319966 '7“
Disunce O a 18069 0 c 03036“ a 37 a 1+0 0 73 .
Population -0.00016 0.000012 -1.10 -.71 . 55 "
Payroll Shift 0.00063 0.000031 1.u2 .83 .26

R = e 88 f = a 77

 

“Coefficient not statistically significant.

crements to 32 significant for each variable at .01 level or
greater (See Table 1!, Footnote a).

Table 13. Multiple Correlation Analysis of Population Size and
Payroll Shift with Net Migration

 

Coefficient Partial
Regression Standard Beta Correlation R2
v e c e c w n c f t De e b
Intercept 9 . 60764 " 5 . 688116
Poplflfition -0 c 00012 O a 00001.1 - 0 8a - a 63 a 51"
Payroll Shift 0.00065 0.000031; 1.116 .82 .17
R = e 85 R2 = 0 73

 

“Coefficient no statistically significant.
bIncrements to R significant for both variables at .001 level or
greater (See Table 1*, Footnote a).

63
analysis. This possibility is enhanced by the finding that payroll in
1947 correlates .95 with population size, 1950, and .81 with change in
labor payroll, 1947-58. However, payroll in 191+? correlates only .29
with net migration, in contrast to the correlation of . 71+ behveen
payroll shift and net migration. This suggests that the strong statis-
tical relationship between payroll shift and net migration may be due
to real shifts in manufacturing, rather than simply to inflation and
similar factors. Indeed, with payroll shift, the other independent vari-
able, held constant, payroll in 19+? has a negative relationship (-.78)
to net migration, as shown in Table 15, while payroll shift retains its
high partial of . 90.

DISCUSSIW

Though the export base approach to migration analysis undertaken
in this investigation appears promising, the present attempt to find
comunity ecological correlates of the export base must be regarded as
largely unsuccessful. Either mobility decisions are made independently
of ecological characteristics at the destination, or else the important
ecological characteristics are not included in the present study. In
any event, the results do not support the theory of industrial mobility
presented in chapter I. Only three ecological variables, population
size, distance, and climate, have substantial correlations with net
migration, and these correlations are largely independent of the
relationship between manufacturing and net migration. If a cause and
effect relationship among climate, distance, population size, and net migra-
tion exists, it is apparently independent of any local attraction for
industry. It may be that a warm climate contributes in unexamined

ways to an area's export base, such as by attracting tourists, who can

Table 14. Multiple Correlation Analysis of Climate, Population, and
Manufacturing Payroll Shifts with Net Migration

 

Coefficient Partial

Regression Standard Beta Correlation R2
Variable Coefficient Error Weight Coefficient Deletea

 

Intercept -109.52489 15,8630310

Climate 25h.73440 32.245003“ .2? .51 .73

Population -0.000ll 0.0000093 -.79 -.67 .60

Payroll Shift 0.00063 0.0000294 1.00 .84 .29
R=.89 122:.80

 

aIncrements to R2 significant for each variable at .001 level or
greater (See Table a, Footnote a).

Table 15. Multiple Correlation Analysis of Manufacturing Payroll 19h?
and Manufacturing Payroll Shifts with Net.Migration

 

 

Coefficient Partial 2
Regression Standard Beta Correlation R
Variable Coefficient i_§rror 'Weight Coeffic ent De ete
Intercept 0.83619 4.58635
Payroll ’47 - a 00032 a 00001 - e 89 - e 78 a 5""
Pay Shift .0006“ .00002 l.h5 .90 .08

R = .91 R2 = .82

 

 

 

Ill] Ill-‘1'! 'l I .tllv. ll ’1'. I

65
expand an area's service inchistry, or by attracting ”footloose“ service
industries like research and development firms.

The effect of population size, which is negative with paYPOII shift
held constant, my be due to an amenities factor in which migrants avoid
the congestion and inconvenience of the larger cities when possible.
Distance has a pattern of correlations similar to population size, but
weaker. Distance is a mixed measure of population size and metropolitan
dominance, and these mo aspects of the distance measure may exert
opposing tendencies. Long distance migrants usually have a main
population center as their destination, while the amenities factor
referred to in the previous paragraph may encourage migrants more
generally to avoid large cities.

Median income, population density, and population potential may
have mixed effects in attracting industry. An area with low median
income, for example, might appeal to industry in terms of low potential
production costs, but be undesirable from the standpoint of distribution
costs, since areas with low median incane are genenlly ranote from the
largest population centers. A similar argument would apply to the
lack of significant findings for populatim potential. The population
density variable depends to a large extent on the amount of outlying
rural territory included in a given SMSA, and this may make it a
completely arbitrary and unreliable variable: some SMSAs, especially
in western states, include enormous stretches of rural territory, while
others extend only a short distance beyond the urbanized area.

The low correlation of percent employment in services with net
migration and manufacturing shifts is more difficult to explain, since

high service employment is a characteristic of both large cities and

 

 

 

llll‘l‘

66

southern and western states, all areas of expected attractiveness for
industry. Industry is supposedly decentralizing away from the large
cities of the northeast, so this variable too probably has mixed effects.

It must be concluded that the variables used in the present
investigation are excessively crude indicators for the most part, and
they may blur any important relationships that do exist: furthermore,
some are susceptible to multiple interpretations. This is a common
problem with census data used for research. In any event, the theoretical
structure of Part III of the dissertation is not supported.

0n the positive side, three ecological variables are remarkably
effective in predicting net migration. These variables, which are
largely independent of the migration process, are capable, in the present
instance at least, of predicting much more variance in net migration than
are ecological measures used in previous studies. They have a correlation
'with net migration in combination that is not a great deal lower than
the one found by Lowry for ”change in employment level," a much more
causally ambiguous and hard-to-interpret variable. The Table 1“
multiple regression equation thus predicts 80% of the net migration
variance among SMSAs, 1950-60. The same equation, expressed in different
terms, implies that an increase of .01 in the climate index is associated
with an increase of 2500 in net migration; an increase of ten thousand
dollars in manufacturing payroll is associated with an increase of six
in net migration; and a decrease of ten in population size is associated
with an increase of one in net migration. SMSAs most likely to attract
large numbers of migrants during the decade were ones which were small

in size, able to attract new industry, and located in a warm climate.

 

CHAPTER VI
SUMMARX AND EVALUATION

In a discussion of multiple correlation techniques, Guttman
(19u1:292) once pointed out that: ”There is nothing in the statistical
theory of prediction as such that warrants any interpretations of the
effect of one or more variates on one or more other variates in any
type of regression, be it.multiple, partial, or other.” It is well to 4L
bear this comment in mind in drawing theoretical conclusions from the
present study. No theory has been either proven or disproven; the
results are merely consistent with certain theoretical interpretations
and not with others.

Another point that it is well to remuber is that the investigation
analyzed census data for a period during which there were general
increases in most of the indicators used. National population size and
urban pepulation size were steadily increasing, inflation was occurring,
more persons were being employed in service activities, and so forth.

These influences alone were enough to insure positive correlations among
many of the variables, so the mere fact that a relationship is statistically
significant does not mean that it has practical significance. The question
of practical significance is a matter of individual judgment by the
researcher and the reader. For this reason, and even more importantly,

for the reason cited in the ensuing paragraph, no significance tests
were performed in connection with the present investigation.

67

68

Statistical significance tests were not appropriate in the present
investigation, because, with. minor exceptions, the complete universe of
SMSAs was used in the analysis, rather than a sample. Such tests might
have been included for the benefit of those who feel that worthwhile
conclusionscan be drawn fran them, such as in the case of their being
a sample from an infinite hypothetical universe. The author agrees with
most of the criticisms voiced in the volume by Morrison and Henkel (1970),

however, and felt that their inclusion was inappropriate.
CONCLUSIONS

Remembering these caveats and problems of inference, the major
conclusions of the investigation may be summarised as follows:

1. For most purposes, present export base methodology is deficient.
The index of total surplus workers, like the minimum requirements technique,
is unsatisfactory as a reflection of the overall commity export base.

2. An alternative approach to eocport base measurement, based
upon the measurement of differential manufacturing growth, is very
promising.

3. Net migration analysis based upon the use of a “mltiplier
effect, " in conjunction with an export base theoretical framework, is
likewise promising. This is implied by' the superiority of changes in
labor payroll, compared to measures of the number of manufac‘lmring
employees used in previous studies, in predicting net migration.

1}. The theory of industrial mobility presented in the present
investigation is not supported. Community ecological characteristics

statistically explaining the high correlation between change in labor
payroll and net migration could not be found. Mobility choices are

5.

 

., Ill... ‘lllll'll‘
‘lulu '

69
either: (a) unrelated to ecological characteristics of the proposed
destination: (b) governed by ecological considerations not included in
the present study; or (c) too complex or variable to be detected by the
rather»gross indicators used herein.

5. In spite of the absence of theoretical support for the
proposed theory of industrial mobility, ecological factors theoretically
and operationally independent of the migration process were found
capable of statistically explaining a large proportion of netwmigration
variance. The three factors of climate, population size, and change
in labor>payroll for’manufacturing had a multiple correlation of .89
with net.migration.

Generally then, an evaluation of the results of the study leads
one to conclude that while present export base methodology is inadequate,
the alternate scheme proposed herein for measuring the export base and
linking it to net migration is in some manner deficient as well. The
hypothesised high correlation was found between change in labor payroll
and net.migration, strongly supporting an export base theoretical
framework, but the combination of community ecological variables could
not be unearthed which would statistically explain the majority of
variance in both net.migration and change in labor payroll. Some possible
explanations for this are mentioned above. It is possible that a different
choice of variables and operational definitions would produce a better
reflection of the export base and.provide clearer insight into the
nature of the economic growth process.

Since change in labor payroll was highly related to netymigration,
it appears that manufacturing growth may indeed act as an initiator of
economic growth, and that an analysis of other local or regional character-

istics may lead to a better understanding of the factors which attract

 

Jill

 

7O
industry and lead to a self-sustaining growth cycle.

Of course this process, even if better defined and understood,
could be restricted to a particular stage of economic development. The
same technological breakthroughs which have led to an ever-increasing
percentage of the labor force into tertiary sector employment may
eventually reduce the significance of the secondary sector to such an
extent that an entirely new pattern of growth dynamics will emerge.

On the whole, the findings of the study indicate that an export
base approach to analysis may be fruitful in future research in accounting

If h

for patterns of population growth due to net migration. It seems likely 4“
that manufacturing growth is intimtely linked to migration; further
analysis of ecological factors potentially leading to industrial

differential. growth is definitely warranted.
FUTURE RESEARCH POSSIBILITIES

Further constructive research could take any of several directions.
Since it is not presently clear what ecological variables contribute to
industrial mobility, a case stuck of one or several SMSAs with differing
migration patterns would undoubtedly help to indicate what factors are
important in attracting industry and producing a large amount of migration.
For example, a detailed comparative analysis of a half dozen SMSAs with
high, medium, and low amounts of migration, focusing upon ecological
structure and the degree and type of industrial expansion, would be
beneficial. Of course there is the unavoidable problem in such studies
that case studies are of questionable generality, but they can help one

to get insights and a better graSp of underlying processes.

71

A second avenue of future research would be to obtain infonmation
from industrial executives on why they relocate in one community nlther
than another. This type of study could provide direct clues concerning
the ecological variables considered important by those making the actual
mobility decisions. Unfortunately, such studies are fraught with
methodological problems. In particular, there is an apparent tendency
for executives to give ”socially desirable" responses on questionnaires
inquiring about such.moves, such as lower’production costs, when the
executives may hove other private reasons for choosing a particular
area, such as a good school system or other amenities. Power relation-
ships may be at least as relevant as ecological ones.

A third possibly helpful type of research would be an analysis of
local dollar flows, such as the economists' input-output matrix method,
to get a clearer notion of what types of investment and what types of
economic structure lead consistently to economic growth and high rates
of net migration. This kind of analysis can be very informative, though
it tends to be extremely expensive. The only thorough study of this type
known to the writer was done by Fortune (1936) magazine for the city of
Oskaloosa, Iowa.

Other comparative studies similar’to the present investigation,
testing a wider arroy of ecological variables for their association with
differential manufacturing growth and net migrdion, would be constructive.
There is still a relative dearth of such investigations, as there was
thirteen years ago when Whtson (1959319) commented:

Thus, it is clear that both case studies and comparative

studies are required, the fonmer to provide depth and the latter
broad scope. To date there have been many excellent case studies

but practically none of the other variety, and the urgent need
for comparative studies of metropolitan migration remains.

72

A particularly useful type of comparative study would be one which
breaks migration into its components of in-migration and mt-migration,
according to Duncan's (1966) suggestions. This could not be done in
the present study for lack of data, ‘mt as mentioned previously, the
factors influencing in-migration my differ considerably from those
influencing out-migration. It may therefore be an improvement in
methodology to analyze migration in terms of its component parts. This
may soon be possible as a result of continuing census improvements;
census data for SMSAs in particular is being gathered in greater detail.

A final research strategy which would perhaps be the most
effective of all would be to overcome census data limitations by gathering
data specifically on the variables desired through a separate survey.
One example of such an improved strategy is that of Taeuber (1967) in
his”residential histories" approach to migration measurement, which links
an ecological analysis to actual movements of individuals, rather than

merely inferring individual movment patterns from gross census totals.

 

l.' '7 II I I! i! Illillll
lull! l
l

 

BIBLIOGRAPHY

BIBLIOGRAPHY

Anderson, Theodore R. 1953. Characteristics of Metropolitan Subregions
Associated with Intermetropolitan Migration, 1935-40. Ph.D. Disserta-
tion. University of Wisconsin.

Andrews, Richard B. 1953. Mechanics of the Urban Economic Base: Histori-
cal Development of the Base Concept. Land Economics 29 (May): 161-67.

Bacon, LLoyd. 1971. Poverty Among Interregional Rural-to-Urban Migrants.
Rural Sociology 36 (June): 125-40.

Beale, Calvin L. 1967. The Relation of Gross Outmigration Rates to Net
Migration. Paper presented at annual Population Association meetings.

Blalock, Hubert M; 1963. Correlated Independent Variables: the
Problem of Multicollinearity. Social Forces XLII (December): 233-37.

1964. Causal Inferences in Nonexperimental Research. Chapel
Hill: University of North Carolina Press.

Blanca, Cicely. 1962. The Determinants of Regional Factor Mebility.
Ph.D. Dissertation. Nederlandsche Economische Hoogeschool te Rotterdam.

1964. Prospective Unemployment and Interstate Population
Movements. Review of Economics and Statistics 46 (Mhy): 221-22.

Blumenfeld, Hans. 1955. The Economic Base of the Metropolis. Journal
of the American Institute of Planners 21 (Fall): 114-32.

Bogue, Donald J. and Dorothy L. Harris. 1954. Comparative Population
and Urban Research via Multiple Regression and Covariance Analysis.
Studies in Population Distribution No. 8, Scripps Foundation. Oxford:
Miama University and Population Research and Training Center.

. Shryock, Henry S. and A. Hoermann Siegfried. 1957. Subregional
Migration in the United States, 1935-40. Volume I. Streams of Migration
Between Subregions. Oxford, Ohio: Scripps Foundation, No. 5.

Braschler, Curtis H. 1971. Importance of Mhnufacturing in Area Economic
Growth. Land Economics 47 (February): 109-11.

Cohen, Jacob. 1968. Multiple Regression as a General Data-Analytic
System. Psychological Bulletin 70: 426-43.

Darlington, R. B. 1968. Multiple Regression in Psychological Research
and Practice. Psychological Bulletin 69: 161-82.

73

7n

Draper, Norman R. and H. Smith. 1966. Applied Regression Analysis.
New York: John Wiley and Sons.

Duncan, Otis D. 1957. Pepu'lation Distribution and Community Structure.
Cold Springs Harbor Symposia on Quantitative Biology XXII: 357-71.

1966. Path Analysis: Sociological.Ekamp1es. The American
Journal of Sociology 72 (July): 1-16.

 

and Albert J. Reiss. 1956. Social Characteristics of Urban
Communities, 1950. New York: John Wiley and Sons.

 

Fennessey, James. 1968. The General Linear Model: A New Perspective
on some Familiar Topics. The American Journal of Sociology 74: 1-27.

Fortune Magazine Research Staff. 1936. Oskaloosa Vs. the United States.
Fortune 13: 82-105.

Fuchs, Victor. 1962. Changes in the Location of Manufacturing in the
United States since 1929. New Haven: Yale University Press.

Goodrich, Carter. 1936. Migration and Economic Opportunity. Philadel-
phia: University of Pennsylvania Press.

Gordon, Margaret S. 1954. EmP1oyment EXpansion and Population Growth:
the California Experience, 1900-1950. Berkely and Los Angeles: The
University of California Press.

Gordon, Robert A. 1968. Issues in Multiple Regression. The American
Journal of Sociology 73 (March): 592-616.

Guttman, Louis. 1941. Mathematical and Tabulation Techniques. The
Prediction of Personal Adjustment. New York: Social Science Research
Council: 251-364.

Haenszel, William. 1967. Concept Measurement and Data on Migration
Analysis. Demography 4: 253-61.

Hawley, Amos H. 1949. The Population of Michigan 1840 to 1960: An
Analysis of Growth, Distribution and Composition. Michigan Governmental
Studies. No. 19. Ann Arbor: University of Michigan Press.

1950. Human Ecology. New York: Ronald Press Co.

 

Kariel, Herbert G. 1963. Selected Factors Areally Associated with
Population Growth Due to Net Migration. American Association of
Geographer Annals 53: 210-23.

Karp, Herbert H. 1971. Migration and Functional Expansion: An
Ecological Analysis of Pepulation Redistribution. Land Economics
XLVII (November): 365-72.

and K. Dennis Kelly. 1971. Toward an Ecological Analysis
of Intermetropolitan Migration. Chicago: Markham Company.

 

75

Lampard, Eric E. 1955. The History of Cities in the Economically
Advanced Areas. Economic Development and Cultural Change III (January):
81-136 .

Lansing, John B. and Eva Mueller. 1967. The Geographic Mobility of
Labor. Ann Arbor: Institute for Social Research.

Lee, Everett S. 1965. A Theory of Migration. In Jackson, J. A. 1969.
Migration. London: Cambridge University Press.

Lowry, Ira S. 1966. Migration and Metropolitan Growth: Two Analytical
Models. San Francisco: Chandler Publishing Company.-

Makower, H., J. Marschak, and H. W. Robinson. 1939. Studies in the
Mobility of Labour: Analysis for Great Britain. Part I. Oxford
Economic Papers 2 (May): 70-97.

Mattila, John M. and Wilbur R. Thompson. 1955. The Measurement of the
Economic Base of the Metropolitan Area. Land Economics XXXI (August):
215 -28 0

Miller, Ann R. 1967. The Migration of Employed Persons to and from
Metropolitan Areas of the United States. Journal of American
Statistical Association 62 (December): 1418-32.

Morrison, Denton E. and Ramon E. Henkel (eds.). 1970. The Significance
Test Controversy. Chicago: Aldine Publishing Company.

North, Douglas C. 1961. The Economic Growth of the United States:
1790-1860. Englewood Cliffs: Prentice-Hall, Inc.

Ogburn, William F. 1937. Social Characteristics of Cities. Chicago:
The International City Managers' Association.

Pfouts, Ralph W. (ed.). 1960. The Techniques of Urban Economic Analysis.
West Trenton, New Jersey: Chandler-Davis Publishing Company.

Pratt, Richard T. 1968. An Appraisal of the Minimum Requirements
Technique. Economic Geography 44 (April): 117-24.

Price, Daniel 0. 1951. Some Socio-Economic Factors in Internal Migration.
Social Forces 29: 409-15.

Ravenstein, E. G. 1885. The Laws of Migration. Journal of the Royal
Statistical Society 48 (June): 167-235: 52 (June, 1889): 241-305.

Rutman, Gilbert L. 1970. Migration and Economic Opportunities in west
Virginia: A Statistical Analysis. Rural Sociology 35 (June): 206-17.

Sakoda, James M. 1966. The Construction of Indices. Providence:
Department of Sociology, Brown University. (Mimeographed.)

Shryock, Henry S. 1964. Population Mobility within the United States.
Chicago: Community and Family Study Center, University of Chicago.

 

76

Stouffer, Samuel A. 1940. Intervening Opportunities: A Theory Relating
Mobility and Distance. American Sociological Review 5 (December): 845-67.

Taeuber, Karl. 1967. The Residential Redistribution of Farm-Born
Cohorts. Rural Sociology 32: 20-36.

Thomas, Edwin N. 1960. Areal Associations Between Population Growth
and Selected Factors in the Chicago Urbanized Area. Economic
Geography 36 (April): 158-70.

Thompson, Wilbur. 1965. A Preface to Urban Economics. Baltimore:
The John Hopkins Press.

Ullman, Edward L. 1954. Amenities as a Factor in Regional Growth.
The Geographical Review XLIV: 119-32.

1968. Minimum Requirements after a Decade: A Critique and
an Appraisal. Economic Geography 44v(October): 364-69.

and Michael F. Dacey. 1960. The Minimum Requirements Approach
to the Urban Economic Base. Papers and Proceedings of the Regional
Science Association 6: 175-94.

United States Bureau of the Budget. 1961. Standard Metropolitan
Statistical Areas. Wishington, D. C.: United States Government
Printing Office.

United States Bureau of the Census. 1947. Postwar Migration and Its
Causes in the United States: August 1945 to October 1946. Current
Population Reports. Population Characteristics. Series P-ZO, No. 4.
United States Government Printing Office.

1952. United States Census of the Population: 1950.
Volume II. Characteristics of the Population. Washington, D. C.:
United States Government Printing Office.

 

1953a. United States Census of Population: 1950. Volume II.
Characteristics of the Population. Part 1. United States Summary.
thhington, D. C.: United States Government Printing Office.

 

1953b. City and County Data Book, 1952. Wishington, D. C.:
United States Government Printing Office.

 

1957. City and County Data Book, 1956. Washington, D.C.:
United States Government Printing Office.

 

1962. Components of Population Change, 1950 to 1960, for
Counties, Standard Metropolitan Statistical Areas, State Economic Areas,
and Economic Subregions. Current Population Reports. Series P-23, No. 7.
washington, D.C.: United States Government Printing Office.

 

1963. United States Census of Population 1960. Subject
Reports. Mobility for States and State Economic Areas. Final Report
PC(2)-2B. ‘thhington, D. C.: United States Government Printing
Office. Table 31.

 

 

 

 

llll {I III‘ '1 1'1 I 11'. III. I!!! ‘3
:

III"

.

77

United States Department of Commerce. 1961. Air-Line Distances between
Cities in the United States. Special Publication No. 238. United States
Coast and Geodetic Slrvey. Washington, D. C.: United States Government
Printing Office.

1965-66. Office of Business Economics. Growth Patterns in
Employment by County 1940-50 and 1950-60. Lowell D. Ashby. fishington,
D. C.: United States Government Printing Office. 8 Volumes.

 

Vance, Rupert B. 1952. Is Theory for Demographers? Social Forces 31:
9-13.

Watson, Walter B. 1959. Metropolitan Migration in the United States,
1949-50. Ph.D. Dissertation. University of Wisconsin.

Weimer, Arthur M. and Homer Hoyt. 1939. Economic Base Analysis. In Pfouts.
Wendt, Paul and Hal Varian. 1970. A Note on Hoyt's 'Importance of
Manufacturing in Basic Finployment. ' Land Economics 47 (February):
109-11.

Zuiches, James J. 1970. In-Migration and Growth of Non-Metropolitan
Urban Places. Rural Sociology 35 (September): 410-20.