Clark University

Clark University The Demand and Supply of Labor and Interstate Relative Wages: An Empirical Analysis Author(s): Gordon L. Clark and Kenneth P. Ballard Source: Economic Geography, Vol. 57, No. 2 (Apr., 1981), pp. 95-112 Published by: Clark University Stable URL: http://www.jstor.org/stable/144135 Accessed: 27-03-2015 11:58 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Clark University is collaborating with JSTOR to digitize, preserve and extend access to Economic Geography.

http://www.jstor.org

This content downloaded from 163.1.39.173 on Fri, 27 Mar 2015 11:58:44 UTC All use subject to JSTOR Terms and Conditions

VOL.

57

APRIL, 1981

No. 2

THE DEMAND AND SUPPLY OF LABOR AND INTERSTATE RELATIVE WAGES: AN EMPIRICAL ANALYSIS' GORDON L. CLARK Harvard University KENNETH P. BALLARD

U.S. Departmentof Commerce The impact of labor on interstate wage relativities are analyzed in this paper. Migration, unemployment, and employment growth related variables are integrated into a short- and long-run time-series model of relative wage determination. It is found that in and out gross migration may increase or decrease relative wages as may employment growth. As expected, increasing unemployment tends to induce decreases in average relative wages. The model is analyzed over the period 1958-1975 using data derived from the Continuous Work History Sample of the Social Security Administration.

Interstate wage variations have been discussed and analyzed in the United States for many years [4; 17; 37]. When linked to the parallel issue of income inequalities, interstate wage inequality remains a lively and important topic for both academics and policy makers [13; 18]. At least three different sets of reasons could be advanced to help explain interest in this topic. First, many individuals and governments have an ethical and equity related interest in the level of regional inequality [11]. Second, academics have been particularly interested in the causes of interstate wage inequalities. This has often involved the testing of theories, empirical methodologies, and conceptual approaches for understanding why wage disparities * Thanks to the National Science Foundation (Grant SES 7909370) for research support at Harvard. The views expressed are those of the authors and do not reflect those of the Department of Commerce or the Bureau of Economic Analysis.

occur. Third, governments have also been interested in wage disparities since many policies and programs depend upon wage standards set on the basis of local and national conditions. Academics and planners have been involved with two aspects of the wage issue: regional wage levels per se, and interstate relative wages. The level of local wages is of immediate relevance for policy makers, particularly those concerned with local welfare, living standards, and minimum income support schemes.' However, interstate wage relativities are also important since the I For example, CETA (Comprehensive Employment and Training Act of 1978) uses a wage index system, based on national and local conditions, for determining average and maximum wages that can be paid to local Public Service Employees (PSE). Wage differences between areas and the national average are standardized in a general index system so as to maintain "fairness" (however defined) between different CETA areas and employees (see Mangum et al. [33] for further details).


96

ECONOMIC GEOGRAPHY

federal government often uses identified relativities for setting wage standards, hours of work, and the like for government procurement contracts.2 Differences in wage levels between states determine their relative attractiveness for capital investment, production, and economic growth. Thus, when being concerned with patterns of regional growth and the locational determinants of capital, wage relativities have been of central concern. Neoclassical theory predicts that over the long-run, interstate wage differences will tend to converge to the national norm. Studies of interstate income and wage convergence abound [2; 29; 42], and are often based upon Borts and Stein's [5] notions of long-run equilibrium and factor adjustment (price and quantity). The long-run in their models is usually conceived of in terms of decades or even centuries. However, the long-run depends upon sequences of adjustment which occur over much smaller time frames in response to changing economic conditions. As Kalecki [26, p. 165] noted .. . the long-run trend is but a slowly changing component of a chain of short-period situations." Whilst studies of convergence often deal with the cross-sectional properties of interstate wage relativities, a neglected but as important task is to analyze the temporal adjustment path. This paper is concerned with the shortrun (year to year) adjustment of state wages relative to the national norm, the role of labor market characteristics in determining relative wages, and the various empirical approaches to this problem. Analysis is based upon a yearly time-series model which links labor market phenomena such as gross migration flows, employment growth, and unemployment to relative, average wage

Models of interstate wage disparities are typically framed in terms of the longrun. Wage levels are linked to the process of regional economic growth and decline. State income and wages are derivative of the structure of economic growth models which deal with issues such as resource endowment, the availability of capital, and the productivity of labor. Two competing theories are used to explain interregional growth and welfare disparities: the neoclassical model which predicts wage convergence and factor equilibrium, and the cummulative disequilibrium model (associated principally with Myrdal and Kaldor) which predicts inequality and divergence of interregional wage differentials. Other theories of regional growth have also been developed, however, the focus, in terms of academic popularity, has been reserved for the neoclassical and cummulative desequilibrium models (see Miernyk [35] and Clark [8] on some recent competing theories). Borts and Stein's [5] explanation of interregional economic growth was based upon a neoclassical model and five specific assumptions: pure competition in each region; each region produces the same output using the same production function; labor tends to flow from low wage to high wage regions;

2 The Davis-Bacon Act of 1931, for example, sets the wage and hour standards for federally funded construction work. Wage levels are set according to local standards and with reference to federal minimum wages determined through the Walsh-Healey Public Contracts Act of 1936 [30].

3 It could be questioned theoretically whether or not states are the appropriate unit for analyzing labor markets in a functional sense. In reality, however, many government programs, like unemployment insurance [22], assume that states are labor markets. This administrative practice is followed in this paper.

changes. The model is estimated over the period of 1958 to 1975. In contrast to many previous studies, gross migration estimates are used directly within the model's structure to determine aggregate interstate wage relativities. The approach is aggregate in that it deals with the demand and supply of labor at the state level for all industries.3 ALTERNATIVE PERSPECTIVES ON INTERSTATE WAGE DISPARITIES


INTERSTATE RELATIVE WAGES

and, the growth of capital is a function of the marginal product of capital (MPC) in each region. Assuming pure competition, there is a negative relationship between the real wage (the marginal product of labor) and the MPC, consequently two implications were suggested: first, low wage regions will experience relatively rapid economic growth as capital flows in and labor flows out; and, second, low wage regions will also experience relatively rapid growth in wages because of the growth in the marginal product of labor and the capital-labor ratio. Whether or not these implications are borne out in reality depends upon the supply characteristics of labor in the low wage and high wage regions. For example, if labor is not responsive to interregional wage differentials, then economic growth could still occur in low wage regions without inducing spatial factor price equilibrium. Labor would have to be both responsive to interregional wage differentials and actually move to the correct destinations (high wage regions) for the model to work. Of course, individuals may improve their wages by migrating without necessarily moving to the high wage region. In common with many models of economic activity, the neoclassical model of regional growth does not directly integrate labor supply within its structure. Of the wage studies that have dealt with labor supply and migration, almost all have been framed in terms of the long-run. Further, Lowry's [32] model, which has dominated much of interregional migration analysis, has encouraged the conclusion that out-migrants from low wage areas are in fact unresponsive to origin characteristics. Essentially, out-migration is thought to occur because of life-cycle or psychological characteristics [6]. Recent work by Fields [16] has brought into question the Lowry migration method and has shown that there need be no symmetry between the choice of origins and des-

97

tinations implicit in the Lowry methodology. These issues have been extended by Clark and Ballard [10] and Ballard and Clark [1] in a series of analyses of the patterns of short-run adjustment of gross migration to economic determinants. Three conclusions derived from these studies which bear upon the neoclassical model of interstate wage disparities deserve attention. First, it was shown that in the short-run, workers of depressed regions are responsive to relative interstate wage differences. Second, workers by migrating do improve their individual well-being, although,, third, individual optimality may not bring about aggregate efficiency. That is, workers may not necessarily come from and go to the correct regions (low wage to high wage regions). Consequently, the neoclassical model may not hold in reality because of the macro-temporal and spatial misallocation of migrants. While the neoclassical model predicts long-run equilibrium, Kaldor's [25] model predicts that initial regional economic disparities are often reinforced and that divergence rather than convergence is the rule. Kaldor's [25] argument was extended by Dixon and Thirlwall [15] and depends upon what has been termed the Verdoorn "effect," which is interpreted in terms of cummulative causation where regions maintain their position through increasing returns to scale [38]. Consequently, wage increases can be offset through increased productivity. The model is also export and demand orientated, implying that leading regions come to dominate other low wage regions by virtue of their product cycles and return on capital invested [40]. There is a very important role for labor in this process. For regional growth to be maintained, export prices must be competitive within an integrated interregional market system. The price, availability, and productivity of labor are important components of export pricing. Like the neoclassical model, the demand and supply of labor are important elements of the


98

ECONOMIC GEOGRAPHY

Kaldorian model. However, given increasing returns to scale in the high wage regions, migration of labor from low to high wage regions in the Kaldorian model would simply reinforce the dominance of the spatio-economic core. Both models place heavy emphasis on the role of labor in determining interregional wage relativities. Further, both assume that labor prices are a central element of the regional growth process itself. Consequently, the supply of labor and its demand determined price (wage) relative to other areas has an important role to play in promoting the economic growth or decline of different units of the spatial system. Notice, however, that supply is relatively fixed in both instances and that demand is deemed to be the crucial variable. Perhaps a reason for the lack of concern with the supply adjustment issue has been the presumption that capital and especially labor take a long time to adjust to economic opportunities. Yet recent evidence has suggested that labor is quite responsive to short-run changes in economic activity and to the landscape of economic opportunities [1; 9]. Moreover, the separation of the long-run from the short-run may, in fact, be quite artificial. Modeling the path of regional economic adjustment and supply of labor requires an integration of both elements as well as explicit concern for the demand and supply of labor. Similarly, models of interstate wage relativities also need to be framed in terms of different time paths of adjustment and both the demand and supply of labor. GEOGRAPHICAL PATTERNS OF WAGES AND EMPLOYMENT

In this section a set of labor related variables are analyzed as being determinants of interstate wage relativities. The selection and emphasis on these variables is determined with reference to both the neoclassical and Kaldorian cummulative disparities models with the

empirical modeling strategy being developed in the next section. No replacement theory is proposed, rather an empirical strategy is suggested which integrates the short-run and long-run with labor demand and supply determinants of regional wages. A number of writers have recently noted evidence of interregional convergence tendencies of per capita nominal incomes in the United States [2]. Interstate wage variations have shown a similar convergence, although there are variations between industry sectors. Figure 1 shows for a variety of employment sectors the decline in interstate wage variations since 1958.4 Wages are defined in nominal terms as total compensation derived from employment divided by the number of workers employed in each state and industry for each year. The advantage of using this wage definition is that differences in labor participation and the number of hours worked are accounted for implicitly in the measurement procedure. Moreover, transfer payments and government welfare expenditures are excluded from the analysis so that labor market phenomena are directly considered. Emphasis is placed on what workers actually receive, on average, rather than an hourly standard [20]. Although there is evidence of convergence in average wages over the period 1958-1978, it is apparent (Figure 1) that the trend is not overwhelming. In fact, the coefficient of variation of state average wage levels has increased 4

V

The coefficient of variations (V) is defined as: 1=->il

(Xi-X)/n X

* 100

In this case, V was determined over each sector i for each year t over the fifty states. Washington, D.C. was not included as were unusual outliers in a number of years which would have distorted the overall trends. Use of V implies equal weighting to each state relative to the nation. The U.S. mean (X) was used independently of the sum of state means and is consequentially the population mean rather than the sample mean.


99


24 All Government o

23

22

>

o C

o

~~~~~~~Construction 21

20 19

/\.

o's.*

Services

\ 18

16

:~~~~/\,

y\

>

WiManufacturing

(

/

'\

--

15 All Sectors 14 13. 0

-

58

60

62

64

66

68

70

72

74

76

78 time

Fig. 1. Interstate Average Wage Variations.

for all sectors, construction, services, and manufacturing from the early 1970s. Only in the government sector was the decline in interstate wage differentials pronounced and maintained over the total period. Notice that the path of wage differentials is not stable despite the smoothing effect of having observations at two-year intervals. Rather, short-run variations and increases and decreases are typical of most series. What is also apparent is the trend towards wage convergence in the 1960s and the reversal

of this trend in the 1970s. In the first period, an unparalleled national economic boom prevailed, in the second period, stagflation. The ratio of state to national, nominal, average wages is shown in Figure 2 for a selected group of states over the period 1958-1975. While some states such as Alabama (AL), Alaska (ALR), and North Carolina (NC) clearly improved their positions relative to the national average, other states, including Texas (TX) and Indiana (IND), remained relatively


100

ECONoMic GEOGRAPHY ( %)

i

ALR

X

140 130 120

NYD

110

iCT __

~~~~~NJ

=

100

IND H Y

M~~~N

90 -T

70 60 59

61_3-

59

61

5

67

69

71

73

7

65

67

69

71

73

75

50

63

time

Fig. 2. Ratio of State to National Average Wages.

stable vis-A-vis the national average. States such as New York (NY) and New Jersey (NJ) show slight declines in their relative positions. Some states had little fluctuation in average relative wages while other states, such as Alaska (ALR), had wider variations. The aim here is to empirically analyze year to year changes in the ratio of state (Wr) to national wages (W.) in terms of labor demand and supply. The level of demand for labor at any point in time is reflected in unfilled vacancies and the level-of unemployment Changes in the level of demand are likely to reflect the path of adjustment to variations in economic growth and labor

supply. Over the long-run, the demand for labor is a function of capital expansion. Thus, the effect of employment growth on wages is likely to be distributed over a number of years. Consequently, the long-run can be conceptualized as being a composite of year to year changes linked consecutively to one another, perhaps analogous to the polynomial distributed lag models of Almon and Koyck [36]. In the short-run, the level of demand for labor is unlikely to be determined by changes in capital or innovation. Rather, short-run variations in output and the level of inventories are important. Essentially, labor is assumed to


INTERSTATE

RELATIVE WAGES

work longer hours, and/or more peripheral workers are employed to take up the short-fall in available labor. The shortterm level of demand could also affect the volume of in-migration, depending upon the speed of migration response. In a relatively tight local labor market, firms could increase their labor force 'by importing employees from outside the labor market. Otherwise, firms might have to bid for other firms' labor. Migrants in this instance would appear as new hires to the firm and may command relatively higher wages. Thus, in-migration could be interpreted as being a component of short-run labor turnover [21]. Clark [9] has shown that local turnover can be important in determining local wages, more so in fact than local unemployment. Consequently, in-migrants could initiate increases in average state wages if the level of demand for labor is strong enough and if union or wage bargains involve the maintenance of wage relativities within and between firms [34]. There are two sources of labor supply in the short-run: those already unemployed and looking for work, and inmigrants from outside the region. The relationship between unemployment and wages has often been noted and summarized via the Phillips curve [7; 14]. In recent years, however, this relationship has been subject to severe criticism. Many have noted the phenomena of "sticky" wages, expectations, and stagflation. Moreover, policy has been shown to be severely compromised if based upon this supposed trade-off [3]. Moreover, Kraft et al. [28] noted that over the long-run there is apparently little relationship between the level of wages and the level of unemployment for different regions of the United States. Figure 3 summarizes the pattern of state unemployment rates (Ur) relative to the nation since 1958. For a selected group of states, there are wide variations and fluctuations in their unemployment relative to the national average. Indiana (IND) and California (CA), for example,

101

are virtually inverses of one another. Unemployment in California first increases rapidly, relative to the national average, then holds its level, and in the late 1970s begins to decline relative to the nation. Indiana on the other hand, first declines, holds steady, and then increases its level of unemployment relative to the nation. Massachusetts (MA), however, shows a steady and dramatic increase, while New Hampshire (NH) and Oklahoma (OK) exhibit steady declines in their levels of unemployment relative to the nation. The patterns reflect the general path of the United States economy over the period 1958-1975 and the differential impacts of national cycles on regional economies. Migration (Mr) may have a variety of effects on local average wages. For example, in-migration as new hires to local firms could increase wages if the level of local labor demand was very high. Wages may be increased to hire and attract new workers from outside the labor market. Such increases could have further spill-over effects as local revenue is increased and firms attempt to hire even more employees. On the other hand, in-migration, if not directly linked to the level of demand, could induce a decline in average local wages as labor supply increases relative to demand. Out-migration could also have a variety of effects: reducing labor supply and thus increasing wages, or decreasing average wages as high wage earners leave and are not immediately replaced. The latter case may occur if more senior workers quit their jobs for opportunities elsewhere, and are replaced by younger or less skilled workers who do not command the same average wage. Outmigration could also be viewed as quits to the firm and local labor market. If the volume of out-migration is large enough, wages in the local area may have to rise so as to hold on to the labor force (especially if labor demand is already high) [39]. Figure 4 summarizes the increasing importance that migration plays in state


102

ECONoMIc GEOGRAPHY

(%) 140 MA

120 59

61

63

65

67

69

71CA VT

GA

100

NC 90 IND 80 OK NH

70

~~~~~~~~~tm

11-~

NBS SD

0

59

61

63

65

67

69

71

73

75 time

Fig. 3.

Ratio of State to National Unemployment Rates.

labor markets. The total volume of migration (in-migration plus out-migration without regard to sign) as a proportion of the total, non-moving work force (Lr) has increased in most states over the period 1958-1975. In recent years, many states including Michigan (MI) and Alabama (AL), have seen rapid increases in the significance of migration in terms of its contribution to the total work force. A number of states, including Nevada (NV) and Alaska (ALR), exhibit variations or fluctuations than others on this variable. The rapid shifts of Alaskan migration might, however, be due to the data sources rather than an inherent process, although in most cases the trend is stable and increasing.

EMPIRICAL FRAMEWORK

Relative interstate average wages were explored in a short-term time series model which incorporated both labor demand and supply characteristics. Following the discussion above concerning alternative wage and regional growth theories and the role of labor, the empirical relationship analyzed can be specified as: (Wr (Wn)

Ur Mr) f(Er = ( En ' Un ' Lr )

1

Equation (1) links state wages (Wr) relative to the nation (Wn) at a particular point in time (t) to the levels of relative


103


%) 100 90

80 ALK

70

NV

50

40 AZ

3030

NN AL

20

MN ml

10

0

t 59

61

63

65

67

69

71

73

75 time

Fig. 4.

Total In- and Out-Migration as a Percentage of State Work Force.

employment (Er/En), unemployment (Ur/Un), and migration relative to the local labor force (Mr/Wn), also at time (t). The dependent variable (Wr/Wn) is a relative measure for two reasons: first, to enable cross-state comparisons, which is made possible because state average wages are normalized according to the national average; and, second, to enable analysis of the process of state average wage adjustment vis-A-vis the national average and thus implicitly the degree of convergence or divergence. Essentially, the model enables empirical analysis of the role of different labor variables in determining changes in the ratio of state to national average wages.5 As previously noted, average state wages (Wr) were defined as the level of total

employee compensation divided by the total number of employees for each state and year. The sources of data were the Bureau of Economic Analysis and the Bureau of Labor Statistics (BLS). The employment variable Er/En measured the aggregate number of persons employed in both the state, r, and the nation, n, for each year. State and national employment and data unemploy5 The model proposed and tested here is not designed to be a complete specification of all possible determinants of relative wages. Rather, the aim is to investigate the empirical relationship between a selected set of variables, emphasizing labor supply and demand components. Following Hicks [23], there is an implied, but weak, separable causality in that the specified variables are deemed to be causes of interstate wage relativities but not exclusive causes.


104

ECONOMIC

ment were obtained from the BLS. The measure of unemployment is quite specific, since it includes only those individuals currently available for work and actively looking for work. Those who looked for work in the previous four weeks were defined as active. Discouraged workers are not included. Consequently, although the numbers of employed and unemployed could be thought of conceptually as being the inverse of one another, in reality the counting system is so restrictive that they are very different aspects of the local labor market. The measure is biased towards the frictional aspects of the labor market rather than long-term unemployment. Gross migration sources are much harder to find than the more traditional employment measures. In fact, to adequately test for the impact of migration in this context, migration needs to be disaggregated by employment status, age, gross flow (in- and out-migration), and year. The Continuous Work History Sample (CWHS) of the Social Security Administration (SSA) and the Bureau of Economic Analysis (BEA) is perhaps the best source for such types of data. Actually, this is probably the only source for such data in the United States, although researchers have developed data series on a limited basis from the Internal Revenue Service. The CWHS data file used in this study included those individuals employed and between the ages of 24 and 64, inclusive. Excluded were the unemployed, retired, and student age groups. To enable comparison between states of the migration variables, both in-migration to state, r, and outmigration from state, r, were made relative to each state's non-moving work force. A supplementary variable was also tested, the ratio of in-migration to out-migration for each state r and time t. The relationship tested for each state of the United States over the period 1958-1975 linked Wr/Wn to a variety of employment, unemployment, and mi-

GEOGRAPHY

gration variables. A time-series testing strategy was developed utilizing either the Cochrane-Orcutt [12] method of controlling for first-order, autoregressive, serial autocorrelation or the Koyck [27] arithmetic distributed lag procedure. Since the Cochrane-Orcutt method involves first differences, the actual relationship tested (a variant of equation (1)) was in fact an economic adjustment model. Similarly, the Koyck procedure is also a time dependent formulation (also implicitly an adjustment model). Both the employment and migration variables were modified, the former by invoking a Koyck distributed-lag term and the latter by specifying different in- and out-migration variables. The choice between different specifications was based upon the overall fit of each combination of variables and equations. The model combined labor demand and supply determinants in an adjustment framework emphasizing both the long-run effects of employment growth and the short-run effects of migration and unemployment. The supply aspect was a central feature of the model in contrast to many studies which have either ignored the issue or have been unable to adequately specify short-run gross migration flows. RESULTS OF ANALYSIS

Tables 1 and 2 summarize the findings derived from estimating the general relationship embodied in equation (1) for the fifty states of the United States over the period 1958-1975. Only in seven states did the adjusted R2 fall below 0.59.6 Those states were Idaho, North Dakota, South Dakota, Wyoming, Indiana, Iowa, and Maine. The first four states are clearly spatially and economically interrelated and are relatively isolated from the national economy. The results for Indiana, 6 The coefficient of determination (R2) was adjusted for the degrees of freedom making more meaningful comparisons between different state models possible (see Intriligator [24, p. 127] for more details). The average percent error was used to discriminate between different variables (see Glickman [19, p. 142] for further discussion).


105


cn-

(

=;~~0

LOn

00

iC

iC

cn

]-

C

0 C0

0~~~~~~~C C 0

O0

CC 0

O~

V)q>_

-

_

C] 0q

CD

C

t

00

LO

i

C] C]

iC 0

-t

C

C

]

Q

-Qn

LO

0

0

0

-

0t_

~

n X 0

CD

It

C] 0

>n

::;

~ n~~~~~~~o

_ m

0 n

g

OL s

CC t

CD 0 n

N

ql \e 1.

CD)

CD

0

m

O C t

t

m N

0 o

Cli o

o

6-~~~~~~ c~~~~-

c cqC C'i'~~~~~~~~ '~~~~C]~

co

-

00

co0 c Io

_c

~~~~~

>

~ ~ ~ ~ ~~ H I

c

_C

01t-cq

~ ~ ~~~~~ ^

0

~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

^

c

O

CO

-0]I

?

c-

~~~~~~~~~~~~~-~~c

0

cq

0o

0C cq

~

.

oC

I

I

i

U

-

C

t

l

E~~~~~~~~~~~~~~0 &~~~ ;

H

rr>~~~~ SC;

,z H6

I'

6

_6

c6 LO

06

00

I co C"

4I6 c6 6Cs

C6c

C

CL0

c

r-

C'

I _

I

o

6

-

c'i6

X=

i _

O 6C 6L

O

me

n3 X

C

Q

rCC

_

c

-r

6

-

C]

tz On tOn -t~~~~~~~~~~~~~~~~C. bt

t0

U z~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ _

fg

,~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~-

b c l_>

W5

-.

W

cll


-_

106

ECONOMIC

0>

~Z Xa

0>

O

6 LQ 10i

0

-~

10 O

GEOGRAPHY

0 c

] m0

c

6

6 -ACI

6

o 0

CO CO

0>

-A

A

t

_~

CO 0o

-o 10

X

0

_> _0

U:t

-~~~~ ~ ~ ~~~~~~~~~~~~~~~~~I

t_

-

-

O

1

]

O

C]

C]

0 CO

a>

0

10 10

C]

1 '0

C] 10

0

0

10

10

O~

a> C]

C CO~~~~~7~

10

Nt s o il cc i OO C tCO m

tr-A

0

ci:

C-

H

H 0

CO

1 0~-

-

O-

CO

tO

-~ i-

C]

C

CO 0CIO 0=q

0

COC

0

a>

CC)~ H

f-

C]

_

m; N

to

O > *

x

cq

_0-

H

OC

C

f

t

10D-

'c

a>

r--O

mt-

a>-

_

0

a;

U) cq

10

10O-

-A

C-

10Cc

H

>

I

c

-t

>

0-

00

30

z

Z

t>

10~O3

O

C0

C

om

-

-01

)

ci

CO

C-

xC

11-

_

ci

6

~

0 ci

C"

_

-

10

-1--

"

M

MM

1

U

r.

OO~

-;

C-C

ana

0-C

c-f

C]

100

C-

~~~~~~~~~~~~~~~~~~~~~-4

U

a)_

m

U H2 0 >

~

ca>

C10-

-Co

C---C

C-C

0

OC-

100]

0]CO

/]

>-

F

I C>

C.

00> 0-O

N

COC 0-_ 10X

_

I

~

00 C-

I

COC _ -

0>

0]]

e~~~~~~~~~~~m

>

t

"C

so

X

0D-

0>0

1

0>-] m

.

,_

~

z

U 1) -0;

0 z

Z

z

z


N C3i

in

-

'C-i

CD

C7) .

[rD .

.

, cq

[e

I--i [n

C

.

.

_

.

t

.

I

C>

C>O

Co

-0

~ ~

~

t

c co

ic-

c

t

.C

t-

_

z'-

.

-

.0

.

Cq

3

C-

CQ

.

.

CC

.

C.

.

cs

c-l-t

_C

CC

I-

__

f-

0)

C

0 I'

0)IO

I-CC 0)-

o

CC X)C~

U

-

.

-

~Ct i _

i cU _

i-~~~~~~~ c-l~~~

mc-l0

-"

I-

CCO CC-"

00

~~.

q

CCCs

_

cvi

iDI)

C

_

t-i

CD cvi

-

-

;

1oIr

*;

0 Cl)

.. I o cv

t6ic

t-

-I)

rt

_

CO

~-

.

I

.'i .

ar . t

Q

-e nm l

zC

> co C~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~c S~~~~~~~~~IS

0>

_T 0

.

sI

0

X

CCD

..i .

.

_6

C,,C.zat

,,CC>

_

ID

CO

CDc-

00

-

.~

.. C.I_____

-~~~~~~~ q

CO

0

~ ~~ ~ ~ ~~~~f

;C

t-

Iv _

C~~~~c-i

O-

I

CQ

0

-

~~ ~ ~

z _,

.

0 C>

-_

c--

CCQ

cu

-I t C6

>

1-

I.-~~~ *e

CC

_C

4c-i ;0 r)-

t*

t-

0

~~~~~~ ~~~~~~~

_)

C~~c-l~~~0c-lCC~t

CQ

0 Iq

C C

0

CO

~~~ ~~

'4~

t-0

1-

O

CD

I

CO~

X.)

Ci

I

)

C-I

C

c's

II_

0)

C6

.

co ;x N

cZ

C

-

cI C

I

_i

e

C

0

C 1-

0

CO

m

co

x C 0

in I1-

C,

3 ; 0-

C Ci

~~ ~~~~~~~~~~co ~ ~~ Ci~~~~~~~~~~~~~~~~~~0C cs t.

cl

? n

0~

O

OC

-, C,

C,, )

'C

'x O

,-

Cl Cq

_C

CO

_

m -

0f

-

C,

x t

3m

C,, C 0q

0

I 1-

1-

0n

m

o t

-

,C> US)

cl

~

C,, 0 0-I

3m 0r 0r

107

RELATIVE WAGES

INTERSTATE

-

Min

-O.

;0

t

a H

=

Cu H

C

x-

C CQtC DCs- s zl -C-Cr

D

-

COF

e

C0

lo

; C--


108

ECONoMic GEOGRAPHY TABLE2 RESULTS OF ANALYSIS CROSS-CLASSIFIED BY STATE AND SIGNIFICANT VARIABLES

Emp)loyment

Unemployment

Variables

Current

Additional Variables

D~ynamic

Illinois, Maine, South Carolina

Georgia, Ohio, Pennsylvania, South D~akota

Alabamna,Kentucky, Massachusetts, Nevada, Washington

Iowa, New Jersey, North Carolina, Oklahoma

Unemployment,

Inmigration

Arizona

Unemployment,

Inmigration,

Alaska, Connecticut,

Louisiana,

Mississippi

New Mexico, Texas, West Virginia,

Outmigration

Wisconsin

Unemployment, In/outmigration

Rhode Island, Utah

Arkansas, Indiana

Inmigration

Missouri

Maryland

Idaho, Delaware

Outmigration Inmigration, Outmigration

Minnesota, New lIampshire

In/outmigration

Montana, New York, Virginia

California, Florida, Ilawaii, Nebraska Vermont

Kansas, North Dakota, Tennessee, VXyoming

Unemployment, In/outmigration alone Unemployment,

Inmigration

alone

Michigan, Oregon Colorado

Iowa, and Maine were more surprising since all three have strong links with the northern United States manufacturing belt. First-order, autoregressive, serial autocorrelation (measured by the Durbin-Watson statistic) was not of significance, at the 95 percent level, in all states but Alabama. Virtually all parameter estimates reported in Tables 1 and 2 were significant at the 95 percent level and most were significant at the 99 percent level. Student "t"statistics are included in Table 1 immediately under the estimate of the variable parameter. Employment growth is an important factor in promoting local wage rate increases for both the neoclassical and Kaldorian cummulative in-quality models. In all states except Michigan, Oregon, and Colorado, either the short-run current (differenced one period) or long-run dynamic (Koyck lag) employment variable was significant in determining the ratio of state to national average wages. Seven states, in fact, had only the employment parameter as a significant determinant of relative wages. The signs of the employment parameters are, how-

ever, mixed. Some are negative, implying that as employment increases in state r relative to the nation, average wages tend to fall relative to the nation; and some are positive, implying that relative employment growth causes relative wage increases. The signs on the shortrun employment change parameters are more often negative rather than positive as predicted by neoclassical theory. On the other hand, states with significant long-run dynamic employment parameters (Koyck lagged structures) typically had positive signs, although California, for one, had a negative sign on its dynamic employment parameter. It should be remembered at this stage that each equation was estimated so as to establish the "best fit" relationships. Thus, the variations in signs and the association of negative parameters with the current employment variable reflects empirical testing. It is apparent that these results are not accidental. Perhaps one answer could be found in the path of adjustment. Presumably, it is sustained growth over a number of years that pressures local labor supply and thus leads to


INTERSTATE

109

RELATIVE WAGES

increases in wages. It is conceivable that short-run growth (one period to the next) simply draws upon lower paid marginal workers, which may reduce average wages although the number of workers actually increases. Even over the long-run some states, including Mississippi and Nebraska, experienced rapid employment growth relative to the nation but still experienced reduced relative wages. Here it could be argued that the skill and occupational mix of particular industries may hold the key in understanding the wage effect. Clearly if low skill jobs are expanded and if women are induced to participate more in the labor market, average relative wages could decline. Consequently, employment growth in either the shortrun or long-run is not necessarily likely to bring increased relative average wages. Casual observation of the results in Table 1 does suggest that southern states (for example, Oklahoma) are more likely to have negative parameters than northern states. This is in contrast to the common perception that southern areas are growing rapidly in both employment and average wages. However, it should be emphasized that there are exceptions, including Indiana and Ohio. Results concerning the significance and signs of the unemployment variable's parameter had fewer surprises. A little over half the states had significant unemployment parameters and all the signs were negative: as unemployment in state r rises relative to national unemployment, average wages fall. This result is as expected. The actual parameter values were quite small, indicating that, relative to other variables such as employment or migration, unemployment was not as important as perhaps one might assume. In three states (Michigan, Oregon, and Colorado), unemployment and a migration variable were the sole determinants of wages. However, it is also apparent that in just less than twenty states, unemployment was not significant at all. All areas of the country were represented in this group as were fast growing states

(Florida, for example) and slower growing or declining states (for example, Minnesota) . The migration variables also had different effects than prediced by the neoclassical or cummulative causation models. Some states had negative signs on their migration parameters and others had positive signs. Typically, the size of these parameters were larger than the unemployment parameters and thus ranked second in terms of relative importance in determining relative state wages. A negative sign on the in-migration parameter implies that as in-migration increases as a proportion of the state's work force, then the state's relative average falls. For example, in Alaska in-migration is likely to induce a decline in that state's relative average wages. CONCLUSIONS

The variety of results indicated in Tables 1 and 2 concerning the significance of different parameters, their signs and values, was unexpected. No ready classification of the results is practical whether by geographical region (the South versus the North; cf. Scully [41]) or by the pace of economic growth and decline. There is one slight exception: California, Florida, and Hawaii (all rapidly growing states) are grouped together visa-vis their significant variables and almost all signs are consistent and positive (with the exception of the employment parameter for California which is negative). Further, in some states the significance of the dynamic employment variable emphasizes the long-run nature of adjustment, while in many other cases labor adjustment and wage response take place over relatively shorter time frames (year to year). In summary, the results support a contention made in this paper that wage adjustment and the demand and supply of labor can be usefully analyzed over both the short-run and the long-run. Migration disaggregated by gross flows, in particular, was shown to have significant effects


110

ECONOMIC GEOGRAPHY

over the short-run. In fact, short-run migration was an important determinant in almost all cases, and the values of the parameters were consistently larger than those derived for the unemployment variable. Similarly, the employment variable was shown to be important over either the short-run and the long-run, although it is not clear why some states are dominated by short-run change, others by long-run change. The results do not completely reject or force the acceptance of either the neoclassical or cummulative causation models of regional growth and wage determination. In some states employment growth may increase average relative wages while in others average relative wages may decline. Further, in-migration may not decrease average relative wages as out-migration may not lead to an increase in average relative wages. In fact, in-migration may increase relative wages as out-migration may decrease average wages. It is apparent that the aggregate effects of labor market determinants on interstate wage relativities may be quite mixed. Consequently, any policy that presumes that economic growth will automatically bring about increases in average relative wages would be seriously misplaced. Further, any policy that ignored the substantial shortrun impact of migration on relative wages would also be seriously compromised. Employment growth and migration are more important determinants of relative average wages than, for example, the relative level of unemployment. These results suggest that much more work is needed before any one model of regional growth and wage disparities can be accepted unconditionally. Two avenues of research deserve more attention. First, it is clear that the skill, occupation, and sex aspects of new employment and migration may hold the key for understanding whether or not changes in these variables will bring about increases or decreases in relative average state wages. This paper demonstrated, in the aggregate, what Lonsdale and Seyler [31]

found through case study: employment growth may not lead to increased average regional wages if much of the growth is concentrated in unskilled and low-paying occupations. Second, we emphasized the path of short-run adjustment and the possibility of analyzing migration, in particular, as a short-run phenomenon. It is apparent that migration may be more volatile than previously thought, as may its effects on the local labor market. Again, more attention should be placed on this aspect of labor market adjustment rather than assuming labor supply to be fixed in the short-run. LITERATURE CITED 1. Ballard, K. P. and G. L. Clark. "The Short-Run Dynamics of Inter-State Migration: A SpaceTime Economic Adjustment Model of InMigration to Fast Growing States," Regional Studies, 15 (forthcoming, 1981). 2. Berry, B. J. L. "Inner City Futures: An American Dilemma Revisited," Transactions, Institute of British Geographers, 5 (1979), pp. 1-28. 3. Blanchard, 0. J. "The Monetary Mechanisms in the Light of Rational Expectations," Rational Expectations and Economic Policy. Edited by S. Fischer. Chicago: Chicago Unifersity Press, 1980. 4. Block, J. W. "Regional Wage Differentials, 1970-1946," Monthly Labor Review, 66 (1948), pp. 371-77. 5. Borts, G. H. and J. L. Stein. Economic Growth in a Free Market. New York:Columbia University Press, 1964. 6. Bradfield, M. "Necessary and Sufficient Conditions to Explain Equilibrium Regional Wage Differentials," Journal of Regional Science, 16 (1976), pp. 247-55. 7. Brunner, K. and A. M. Meltzer. The Phillips Curve and Labor Markets. New York: North Holland, 1976. 8. Clark, G. L. "Capitalism and Regional Disparities," Annals, Association of American Geographers, 70 (1980), pp. 226-37. 9. Clark, G. L. "A Hicksian Model of Labor Turnover and Local Wage Determination," Environment and Planning (forthcoming, 1981). 10. Clark, G. L. and K. P. Ballard. "Modeling-Out Migration from Depressed Regions: The Significance of Origin and Destination Character-



111

istics," Environment ana(Planning A, 12 (1980), pp. 799-812.

25. Kaldor, N. "The Case for Regional Policies," Scottish Journal of Political Economy, 1970.

11. Clark, T. A. "Regional Income Inequality in the United States: The Relevance of Neoclassical and Myrdal-Kaldor (Cummulative Causation) Models," paper presented at the annual meeting of the Association of American Geographers, Louisville, 1980.

26. Kalecki, M. Selected Essays on the Dynamics of the Capitalist Economy. Cambridge: Cambridge University Press, 1971.

12. Cochrane, 1). and G. M. Orcutt. "Application of Least Squares Regression to Relationship Containing Autocorrelated Error Terms," Journal, American Statistical Association, 44 (1949), PP. 32-61.

28. Kraft, G., A. R. Willens, J. B. Kaler, and J. R. Meyer. "On the Definition of a Depressed Area," Essays in Regional Economics. Edited by J. Kain and J. Meyer. Cambridge: Harvard University Press, 1971.

13. Coelen, S. P. "Regional Income Convergence/ Divergence Again," Journal of Regional Service, 18 (1978), pp. 447-57.

29. Lande, P. S. and P. GCordon."Regional Growth in the United States: A Re-examination of the Neoclassical Model," Journal of Regional Science, 17 (1977), pp. 61-9.

14. Desai, M. "The Phillips Curve: A Revisionist Interl)retation," Economica, 42 (1975), pp. 119. 15. l)ixon, R. and A. P. Thirwall. "A Model of Regional Growth Rate Differences on Kaldorian Lines," Oxford Economic Papers, 27 (1975), PP. 201-14.

27. Koyck, L. M. Distributed Lags and Investment Analysis. Amsterdam: North Holland.

30. Levitan, S. A. and R. S. Belous. More than Subsistence: Minimum Wages for the Working Poor. Baltimore: Johns Hopkins University lPress, 1979. 31. Lonsdale, R. E. and LI. L. Seyler (eds.). Nonmetropolitan Industrialization. New York: Winston Press, 1979.

16. Fields, G. S. "Labor Force Migration, Unemployment and JobTlurnover," Review of Economics and Statistics, 58 (1976), pp. 407-15.

32. Lowry, I. Migration and Metropolitan Growth. San Francisco: Chandler, 1966.

17. Fuchs, V. R. and R. Perlman. "Recent Trends in Southern Wage Differentials," Review of Economics and Statistics, 42 (1960), pp. 292-300.

33. Mangum, C. S., J. Morlock, M. W. Pines, and 1). Snedecker. Job Market Futurity: Planning and Managing Local Manpower Programs. Salt Lake City: Olympus Press, 1979.

18. Garnick, 1). H1."A Reappraisal of the Outlook for Northern States and Cities in the Context of U.S. Economic History." Washington, D.C.: Bureau of Economic Analysis, 1978.

34. Marshall, A. "Labor Markets and Wage Growth: The Case of Argentina," Cambridge Journal of Economics, 4 (1980), pp. 27-60.

19. Glickman, N. Econometric Analysis of Regional Systems. New York: Academic Press, 1977.

35. Miernyk, W. H. "A Note on Recent Regional Growth Theories," Journal of Regional Science, 19 (1979), pp. 303-08.

20. Goldfarb, R. S. and A. M. J. Yezer. "Evaluating Alternative Theories of Intercity and Intern regional Wage Differentials," Journal of Regional Science, 16 (1976), pp. 345-63.

36. Moriarty, B. M. "The Distributed Lag Between Metropolitan Area Employment and Population Growth," Journal of Regional Science, 16 (1976).

21. Greenwood, M. J. "Research on Internal Migration in the United States: A Survey," Journal of Economic Literature, 14 (1976), pp. 397-433.

37. Moriarty, B. M. "A Note on Unexpected Residuals in North-South Wage Differential Models," Journal of Regional Science, 18 (1978), pp. 105-08.

22. Grossman, II. I. "Rational Expectations, Business Cycles, and Government Behavior," Rational Expectations and Economic Policy. Edited by S. Fischer. Chicago: Chicago University Press, 1980. 23. Hicks, J. R. Causality in Economics. Oxford: Basil Blackwell, 1979. 24. Intriligator, M. V). Econometric Models, Techniques and Applications. Englewood Cliffs: Prentice-Hall Inc., 1978.

38. Myrdal, G. Rich Lands and Poor Lands. New York: Harper and Row, 1957. 39. Raimon, R. I. "Interstate Migrations and Wage Theory," Review of Economics and Statistics, 44 (1962), pp. 428-38. 40. Rees, J. "Regional Industrial Shifts in the U.S. and the Internal Generation of Manufacturing in Growth Centers of the Southwest," Interregional Movements and Regional Growth:


112

ECONOMIC GEOGRAPHY

COUPE Papers on Public Economics. Edited by W. C. Wheaton. Washington, D.C.: Urban Institute, 1979.

42. Smith, D. M. "Neoclassical Growth Models and Regional Growth in the U.S.," Journal of Regional Science, 15 (1979), pp. 165-81.

41. Scully, G. W. "Interstate Wage Differentials: A Cross-Section Analysis," American Economic Review, 69 (1969), pp. 757-73.
