Three Essays on the Impact of Automobile Production on Alabama?s
Economy
by
Sooriyakumar Krishnapillai
A dissertation submitted to the Graduate Faculty of
Auburn University
in partial fulfillment of the
requirements for the Degree of
Doctor of Philosophy
Auburn, Alabama
May 7, 2012
Keywords: Automobile Production, Regional Growth Model, Poverty, Proprietor Density,
Generalized Spatial Three-Stage Least Squares (GS3SLS) procedure
Copyright 2012 by Sooriyakumar Krishnapillai
Approved by
Henry Kinnucan, Chair, Professor of Agricultural Economics and Rural Sociology
Curtis Jolly, Professor of Agricultural Economics and Rural Sociology
James Novak, Professor of Agricultural Economics and Rural Sociology
ii
Abstract
The Alabama state government has expanded economic incentives to attract auto industry
in order to create additional employment and generate personal income for its citizens. As a
result, large automobile firms and its input suppliers have located in several Alabama counties.
This paper studies the effect of automobile production on per capita income, employment,
population, nonfarm proprietor density and poverty in Alabama?s counties, especially in the
distressed black belt counties and also examines whether ?people follow jobs? or ?jobs follow
people?.
In Chapter one, regional growth models were developed as a spatial panel simultaneous
equations model of population, per capita income and employment using county data for the
period 1970-2007 and for 1997-2005 to determine the impact of auto production on income,
population, and employment growth in the state. A partial lag adjustment was introduced into
this regional growth equilibrium model and a one-way error component model for the
disturbances was utilized. A Generalized Spatial Three-Stage Least Squares (GS3SLS)
procedure as outlined in Kelejian and Prucha (2004) was used to estimate these regional growth
models. The empirical findings suggest that the population, per capita income and employment
of the county where a plant locates increases with automobile production. The population of
neighboring counties decline but per capita income and employment increase. The percentage
increase in per capita income of distressed Black Belt Counties might be higher than the
percentage increase in per capita income of other neighboring counties. The results of this
iii
analysis show that jobs follow people and also people follow jobs. Results also indicate that the
effect of population growth on employment growth is at least 2 times as large as the effect of
employment growth on the population growth. A conclusion is that appropriate policies to lure
industrial development and improve the educational level of resident population are very
important for economic development. The existence of spatial lags indicate that population,
employment and per capita income growth are not only dependent on the characteristics of that
county, but also on those of its neighbors. These interdependences provide a need for economic
development policy coordination among the counties.
In Chapter two, regression models were estimated using county data for the period 1970-
2000 to determine the impact of auto production and local government expenditure on poverty in
Alabama, especially in the distressed Black Belt counties. The results show that automobile
production in Alabama significantly reduces the poverty rate in all counties. The impact of
automobile production on poverty reduction in distressed black belt counties is greater than in
other counties. However local government expenditures are not very effective in reducing the
poverty. The study suggests that industrial development may be more effective in reducing
poverty than government programs.
In Chapter three, a spatial panel simultaneous equations model of non-farm proprietor
densities and per capita income was developed, using county data for the period 1970-2007, to
determine the impact of auto production on the growth of non-farm proprietor densities in
Alabama?s counties. The results show that automobile production in Alabama significantly
increases the number of non-farm proprietorship in all counties. The impact of automobile
production on the growth of non-farm proprietor densities in distressed black belt counties is
iv
greater than other counties. Appropriate policies to lure industrial development become very
important to increase the self employment opportunity.
v
Acknowledgments
I would like to offer special thanks to my academic advisor, Dr. Henry Kinnucan, for his
guidance and insightful comments that have brought this work to a successful completion. I also
like to thank my committee members, Dr James Novak and Dr Curtis Jolly, Chair of Agricultural
Economics and Rural Sociology. I am also thankful to Dr. Steve Swidler as the outside reader of
this dissertation. Special thanks to Dr. Henry Thompson, and Dr. Hyeongwoo Kim,
Finally, I owe my wife and my parents a big debt of gratitude and appreciation for their
patience and support during these four years. This piece of work is especially dedicated to my
late father Kumaru Krishnapillai and my wife, Nanthinithevy Sooriyakumar.
.
vi
Table of Contents
Abstract ......................................................................................................................................... ii
Acknowledgments......................................................................................................................... v
List of Tables ............................................................................................................................... ix
List of Figures ............................................................................................................................... x
List of Abbreviations ................................................................................................................... xi
Chapter 1 The Effects of Automobile Production on Alabama?s Economic Growth and Rural
Development ................................................................................................................... 1
1. Introduction ................................................................................................................. 1
1.1 Demographic and Economic Profile of Alabama ................................................... 4
2. Literature Review .......................................................................................................... 9
2.1 Spillover Effect of Automobile Plants ................................................................. 11
3. Model .......................................................................................................................... 12
3.1 Reduced Form Estimates and Long Run Elasticity .............................................. 17
4. Data and Sources ......................................................................................................... 20
5. Estimation Issues ........................................................................................................ 21
6. Results and Discussion ............................................................................................... 23
6.1 Population Growth Equation ................................................................................ 24
6.2 Employment Growth Equation .............................................................................26
6.3 Per Capita Income Growth Equation .................................................................... 28
7. Conclusions and Policy Implications .......................................................................... 30
vii
Chapter 2 The Effects of Automobile Production and Local Government Expenditure on
Poverty in Alabama ....................................................................................................... 40
1. Introduction ............................................................................................................... 40
2. Literature Review ........................................................................................................ 42
3. Model .......................................................................................................................... 44
4. Data and Sources ......................................................................................................... 44
5. Estimation Issues ........................................................................................................ 45
6. Results and Discussion ............................................................................................... 47
6.1 Poverty Equation .................................................................................................. 47
. 7. Conclusions and Policy Implications .......................................................................... 48
Chapter 3 The Effect of Automobile Production on the Growth of Non-Farm Proprietor
Densities in Alabama?s Counties?????????????????????.5 2
1. Introduction ............................................................................................................... 52
2. Literature Review ........................................................................................................ 54
3. Model .......................................................................................................................... 57
3.1 Reduced Form Estimates and Long Run Elasticity .............................................. 61
4. Data and Sources ......................................................................................................... 64
5. Estimation Issues ........................................................................................................ 64
6. Results and Discussion ............................................................................................... 66
6.1 Proprietor Density Equation ................................................................................. 66
6.2 Per Capita Income Equation Equation ................................................................. 68
7. Conclusion and Policy Implication ............................................................................. 69
References ................................................................................................................................. 75
viii
Appendix Method of Estimation in Panel Data Spatial Simultaneous Equations Model ......... 83
ix
List of Tables
Table 1.1 Variable Description and Data sources ....................................................................... 33
Table 1.2 Descriptive Statistics for Alabama counties ............................................................... 34
Table 1.3 Structural Coefficients for the Study Period (1970-2007) .......................................... 35
Table 1.4 Structural Coefficients for the Study Period (1997-2005) .......................................... 36
Table 1.5 Reduced Coefficients, Long Run Elasticities and 10% Impacts of Automobile
Production (1970-2007) .................................................................................................... 37
Table 1.6 Reduced Coefficients, Long Run Elasticities of Exogenous Variables (1970-2007). 38
Table 1.7 Reduced Coefficients of Initial of Population, Per Capita Income and Employment
(1970-2007)....................................................................................................................... 39
Table 2.1 Variable Description and Data Sources ...................................................................... 49
Table 2.2 Descriptive Statistics for Alabama Counties .............................................................. 50
Table 2.3 The Estimation Results of Regression Models ........................................................... 51
Table 2.4 Reduced Form Coefficients and Long Run Elasticities ............................................. 51
Table 3.1 Variable Description and Data sources ....................................................................... 70
Table 3.2 Descriptive Statistics for Alabama counties ............................................................... 71
Table 3.3 Structural Coefficients ............................................................................................... 72
Table 3.4: Reduced Coefficients, Long Run Elasticities and 10% Impacts of Automobile
Production ......................................................................................................................... 73
Table 3.5: Reduced Coefficients, Long Run Elasticity of Exogenous Variables ....................... 74
x
List of Figures
Figure 1.1 Metro and Non-metro Counties in Alabama ............................................................. 4
Figure 1.2 Percentage Change in Population, 1990 ? 2000 ........................................................ 5
Figure 1.3 Percentage Change in Population, 2000 ? 2005 ........................................................ 6
Figure 1.4 African American Majority Counties, 2005 ............................................................. 6
Figure 1.5 Per Capita Income, 2004 ........................................................................................... 7
Figure 1.6 ERS County Typology: Low Employment Counties ................................................ 8
Figure 1.7 Unemployment Rate, 2005 ........................................................................................ 8
Figure 2.1 Percent of Population in Poverty, 2003 ................................................................... 41
Figure 2.2 ERS County Typology: Persistent Poverty Counties .............................................. 41
Figure 3.1 Entrepreneurship in Alabama ................................................................................... 53
xi
List of Abbreviations
AAMA Alabama Automotive Manufacturers Association
CPI Consumer Price Index
EDPA Economic Development Partnership of Alabama
ERS USDA Economic Research Service
FG2SLS Flexible Generalized Two Stage Least Square
FG3SLS Flexible Generalized Three Stage Least Square
GS3SLS Generalized Spatial Three Stage Least Square
GM Generalized Moments Estimation
L.R.E Long Run Equilibrium Elasticity
1
CHAPTER 1
The Effects of Automobile Production on Alabama?s Economic Growth and Rural
Development
1. Introduction
Strategies to improve living conditions in the rural South are receiving increased attention
(Wimberley et al, 2002). Local economic development has become a major concern of state
policy makers and local government (Isserman 1994). Since the Alabama state government has
expanded economic incentives to attract auto industry in order to create additional employment
and generate personal income, large automobile firms and input suppliers have located in several
Alabama counties. Prior to 1997, Alabama produced not a single automobile. Due to the
aggressive recruiting efforts by the state, auto production and its ancillary industries accounted
for 3.4 % of Alabama gross domestic product and for 17.5% of the state?s manufacturing gross
domestic product in 2006. The auto industry in Alabama accounted for 47,457 direct jobs and
85,700 indirect jobs through their purchases and expenditures with annual payroll of $5.2 billion
by 2007 (AAMA 2008). In addition to providing jobs to offset losses in mining, agriculture, and
textiles, these jobs are better paying. In 2004, the average weekly wage for auto manufacturing
workers in the state was $ 1,318 compared to $761 for all manufacturing and $643 for all
industries (EDPA 2006). Jobs in 40 of the state?s 67 counties now are tied directly or indirectly
to auto manufacturing (AAMA 2008).
2
Despite its growing importance, little scholarly work has been done to assess the impact
of the auto industry on the state?s economy or living standards. Gadzey et al. (2003) estimated
an econometric model, using 30 years of county level data to determine whether state assistance
to private firms increased the real value of manufacturing output. Results based on data through
1999 showed the subsidy effect to be positive, as expected, and statistically significant.
However, the measured effect was too small for the subsidies to be profitable. This finding is
important because it affirms charges of critics (Buchholz 2008) that the incentive packages given
to auto companies were excessive. Mercedes-Benz, Honda, and Hyundai each received incentive
packages worth between $100 and $300 million (Ahn 2005). More generally, it raises questions
about whether industrial policies to lure industry are a cost effective way to improve the living
standards of rural residents. A major focus on this research is to address that question.
Gadzey et al.?s findings are consistent with the substitution view of industrial subsidies,
which means replacing import with domestic production (Wren 1996). Their analysis terminates
in 1999, and thus covers only two full years of auto production. The multiplier effects of the
industrial production were not considered in this analysis. Effects of particular interest to
students of rural development include those on employment, population, and income growth
(Duffy-Deno 1998; Deller et al., 2001: Kim et al., 2005; Saint Onge et al., 2007; Hammond and
Thompson, 2008; and Wu and Gopinath, 2008). Enlarging the analysis to include income,
population, and employment effects, as proposed in this research, provides a more complete
picture of the industry?s impact on the state. Subsidies are measured as transfers of sum of local,
state and federal government funds to counties as recorded by the US Census Bureau and thus
are non-specific to the auto industry. To circumvent the problem of non-specificity, and to
provide a direct measure of impact, we propose using a simple count of auto production as the
3
causal variable. Between 1998 and 2007 car and light truck production in the state increased
from 68,800 to 739,019 units (EDPA, 2008), which provides sufficient variation to measure the
impacts reliably.
The purpose for this research is to determine the economic impact of auto production on
income, population, and employment growth in the Alabama?s counties. A major goal of this
research is to determine whether distressed counties in the state?s Black Belt benefited from the
auto boom. Of the 17 counties in the Black Belt, Governor Riley?s Black Belt Action Committee
identified 12 as ?distressed? as follows: Bullock, Choctaw, Dallas, Greene, Hale, Lowndes,
Macon, Marengo, Perry, Pickens, Sumter, and Wilcox (Morton, 2007). This research improves
on previous studies involving Alabama counties. First, a simultaneous model permits an analysis
of the feedback effects among population, employment and per capita income which has not
been seen in previous studies. A second improvements is that the initial level of employment, per
capita income and population are included in the model, which allows to test whether each
equation in the system converge with respect to the dependent variables. Third, this study
estimates the differential impact of auto production on income, population and employment
growth in the distressed black belt counties by the interaction term of auto production and these
counties. Finally, spatial components are incorporated to capture the role of population,
employment and per capita income of neighboring counties.
The research proposed herein is motivated in part by a recent study by Kinnucan et al.
(2006) on the determinants of student performance in Alabama?s county schools. Results of this
study indicated that poverty reduction and income growth were among the most potent factors
predictive of improved student scores on standardized achievement tests. This suggests industrial
policies aimed at increasing employment or family income could have important effects on rural
4
education. In Carlino and Mills? (1987) classic study, it was speculated that since ?jobs follow
people,? in slow growing or declining regions, ?public funds may be better spent on educating
the resident population than used to lure employment?. A purpose of this research is to shed light
on the validity of this hypothesis by examining the extent to which growth in the auto industry
benefited the state?s Black Belt region.
1.1 Demographic and Economic Profile of Alabama
Based on county Core Based Statistical Area classifications, in Alabama, there are 28
metropolitan counties, 15 micropolitan counties, and 24 noncore counties. Approximately
seventy one percent of Alabama?s population resides in metropolitan counties, 18.4 percent
resides in micropolitan counties, and the remaining 10.8 percent live in noncore counties
(RUPRI, 2007)
Figure 1.1:
5
From 1990 to 2000, in Alabama, 12 counties lost population, six counties experienced
population growth of over 30 percent. These six counties include five metro and one non-metro
county. Between the 2000 Census and the July 2005 estimate, Alabama?s total population grew
by 2.5 percent. But 41 counties lost population. Twenty nine of these 41 counties were non-
metro counties (RUPRI,2007)
Figure 1.2:
6
Figure 1.3:
In many Alabama counties, the African American population accounts for a significant
portion of total population. African Americans are majority population in eleven counties in
Alabama.
Figure 1.4:
7
In 2004, the per capita income in Alabama?s counties ranged from $17,976 in Bullock
County to $36,041 in Jefferson County. Seven counties had per capita income less than $20,000
in 2004, six of them non-metro. Only four metro counties had per capita income over $30,000 in
2004 (RUPRI,2007).
Figure 1.5:
The Economic Research Service of USDA classifies counties as low employment
counties if ?less than 65 percent of residents 21-64 years old were employed in 2000.? ERS
classified 21 counties in Alabama as low employment counties. 16 of these low employment
counties are non-metro counties. The unemployment rate for Alabama in 2005 was 4.0 percent,
compared to 5.1 percent for the U.S. Within Alabama, the unemployment rate ranged from 2.6
percent to 8.7 percent (RUPRI, 2007).
8
Figure 1.6:
Figure 1.7:
9
2. Literature Review
Regional Scientists adopted simple modeling techniques in very early regional economics
literature. Most of this early literature analyzed employment, population and per capita income
growth separately. Steinnes and Fisher (1974) in their study found that there is simultaneity
between employment and population growth. The classic two-equation country growth model of
Carlino and Mills (1987) is a key contribution to the literature of regional growth. Their study
has also demonstrated how combinations of exogenous variables affected regional growth. Many
studies have suggested that dual causality and stable growth characterized the interaction of
population and employment changes (Carlion and Mills, 1987, Clark and Murphy, 1996,
Mulligan et al. (1999).
Deller et al. (2001) expanded the original Carilno and Mills Model by adding income as
an endogenous variable to determine the role of income in regional growth process. They used
county data, but restricted to non-metropolitan counties. Their empirical results indicate that
counties with higher initial population will have higher employment growth. However, they
found that counties that had higher levels of population, employment, and per capita income
tended to have lower rates of overall growth. There have also been studies to model the
interdependence between employment growth and migration (Clark and Murphy 1996;
MacDonald 1992) and the interdependence among net migration, employment growth and
average per capita income (Greenwood and Hunt 1984; Greenwood et al. 1986; Lewis et al.
2002) in simultaneous-equations models. Clark and Murphy?s (1996) findings about simultaneity
between employment density and population density have been influential in regional science.
Gebremariam et al. (2010) have found a strong interdependence between employment and
10
median household income growth rates. Their study confirms the importance of spatial effects in
regional development.
Several studies have examined the effects of business location on local economy
(Gottlieb, 1994; Glasmeier and Howland, 1994), and urban and rural status on population
migration (Graves, 1980, 1983; Nord and Cromartie, 1997; McGranahan, 1999) and amenities on
local economy (Duffy-Deno, 1997; Deller and Tsai, 1999; English et al. 2000). Deller at al.
(2000) found that climate had a strong effect on population change, but it had a less effect on
employment and per capita income growth. Gottlieb (1994) found that amenity factors did not
have a strong influence on firm location decision. Nzaku and Bukenya (2005) have examined the
effects of aging population, road density, population with high school diploma, per capita tax and
the proportion of jobs in the agriculture, manufacturing and service sectors on the local economy.
A controversial issue in the estimation of Carlino and Mills-type regional development
models is the treatment of dynamics. The controversy is aptly summarized by Carruthers and
Mulligan (2007): ? One side of the argument is that the space economy exists in a state of
disequilibrium, so that people move in response to visible disparities in the cost of living, wages,
and employment opportunities. This traditional view is epitomized by the notion that
employment growth fuels population growth, but not the other way around. The other side of the
argument is that a state of equilibrium exists, so that people move in an effort to maximize their
utility which often means trading off economic comfort for natural, cultural, and/or other
environmental amenities. In this view, it is reasonable for people to relocate to areas with low
wages and even with dismal employment prospects, if they are somehow compensated by
desirable attributes of their destination. Further, in addition to the traditional growth process, it is
reasonable to expect that jobs also follow people, as a result of firms pursuing their labor pools,
11
increased demand in the service sector, and other economic forces. Although it is impossible to
know for certain which perspective is correct, the question remains a source of extensive debate
(Greenwood 1984; Evans 1990; Graves and Mueser 1993; Hunt 2006).
Duffy-Deno (1998) and MacKinnon, White, and Davidson (1983) show that a log- linear
specification is more suitable for models involving population and employment densities than a
linear specification. Following Edmiston (2004), Deller et al (2001), Henry et al (1999), Duffy
? Deno (1998), Barkley et al (1998), Boarnet, (1994), Carlino and Mills (1987), Mills and Price
(1984) who suggest that equilibrium employment , population and median household income are
likely to adjust to their equilibrium values with a substantial lag.
2.1 Spillover effect of Automobile Plants
The greatest spillover benefit of automobile plants in Alabama is the movement of input
suppliers to Alabama counties. These input suppliers cluster around automobile plants and
create additional employment and generate personal income. The multiplier effects of this
income through consumer spending generate additional employment and income. One of the
major advantages of industry clustering is the potential for labor pooling ( Krugman, 1991).
Workers usually locate near the place where there are several firms and high demand for their
skills because if they lost a job in one firm there may be another firm to hire them. Firms also
want a pool of skilled workers in order to hire easily more labor during higher demand periods
for their products. Another reason for the intra and inter industry clustering is the technological
spillovers benefits (Romer, 1986: Krugman, 1991). Financial institutions and supporting service
firms moves near these industries and this creates additional employment and personal income.
12
There can also be negative spillovers of automobile firms on other industries. Large
automobile firms increase the market demand for inputs and then increase the wages, rents and
price of other inputs. These increased input costs deter new firms moving into the counties where
large firms locate and also deter the expansion of existing firms in these counties. The congestion
of public services and infrastructure due to the large firms and population increase is another
reason for negative impact on new and existing firms. Congestion may force the local
government to raise the tax rates and this also deters the entering of new potential firms.
3. Model
The point of departure in this analysis is the regional growth model estimated by Deller et al.
(2001). This model extends the classic two-equation country growth model of Carlino and Mills
(1987) to include per capita income as an endogenous variable. It is based upon the assumption
that utility-maximizing households migrate in search of utility derived from the consumption of
market and non-market goods, and profit maximizing firms become mobile when looking for
regions that have lower production costs or higher market demand. Importantly, the extended
model retains the essential character of Carlino and Mills?s model by permitting household and
firm location choices to be interdependent. This is important because, as a by-product of the
analysis, we can address a key issue in regional growth; namely, whether ?people follow jobs? or
?jobs follow people? ( Muth 1971; Steinnes 1978; Treyz et al., 1993; and Wu and Gopinath
2008).
13
The basic specification of the model used is a simultaneous-equation system of the form:
(1)
(2)
(3)
The equilibrium levels of population, per capita income and employment are assumed to
be functions of the equilibrium values of the other endogenous variables included in right hand
side of equation and their spatial lags, automobile production, and the vectors of the additional
exogenous variables. , , and are vectors of dimension NT x 1 of the
equilibrium levels of population, per capita income and employment respectively; t denotes time.
I is an identity matrix of dimension T and, W is a row standardized N x N spatial weights matrix
with zero diagonal values. Each element of this spatial weights matrix, , represents a
measure of proximity between observation i and observation j. Based on the queen based
adjacency criteria, is equal to 1/ki, where ki is the number of nonzero elements in row i, if i
and j are adjacent, and zero otherwise. Therefore, and
stand for the equilibrium values of neighboring counties?effect. At-i is vector of
dimension NT x 1 of automobile production. BAt-i is the interaction term of the distressed black
belt county and automobile production. The matrices of additional exogenous variables that are
included in the population, per capita income and employment equations are given by ,
and respectively. These additional exogenous variables are included in the
equations to control their effects on the dependent variables. This system of equation captures
the simultaneous nature of interaction among population, employment and per capita income.
14
The nature of interaction among the endogenous variables depends on the spillover effect of
neighboring counties and the initial conditions of exogenous variables in a county.
A multiplicative functional form was used for the equations in this system. A lagged
adjustment is introduced into our model. This partial-adjustment process replaced unobservable
equilibrium allowing the model to take the general form.
(4)
(5)
(1)
(6)
Where , , , , , , for k=1,?., Kr ; r, l = 1,2, 3; and q= 1,2 are the
parameter estimates of the model and Kr is the number of exogenous variables in the respective
equations. and represent the log differences between the end and
beginning period values of population, per capita income and employment respectively
15
representing the growth of respective variables. The variable, automobile production (lnAt-i),
was constructed as ln (automobile production/ distance). The subscript t-i denotes to the variable
lagged seven years for the study period 1970-2007 or two years for the study period 1997-2005.
The coefficient for r =1,2,3 represents the speed of adjustment coefficients, the rate at which
population, per capita income and employment adjust to their respective steady state equilibrium
levels. ut,r for r=1,2,3 are NT x 1 vectors of disturbances. A Moran?s I test statistic suggested
that there is the existence of spatial autocorrelation in the errors. The test results are given in
Table 1.3. Therefore, the disturbance vector in the rth equation is generated as:
= + r =1,2.3 (7)
This specification relates the disturbance vector in the rth equation to its own spatial lag.
A one-way error component structure was utilized to allow the innovations ( to be
correlated over time, following Baltagi (1995). Therefore, the innovation in the rth equation is
given by
, r = 1,2,3. (8)
Where , , =(
IT and IN are identity matrices of dimension T and N, respectively, is a vector of ones of
dimension T, and denotes the Kronecker product. and are random vectors with zero
means and covariance matrix ( suppressing the time index):
16
(9)
Where, denotes vector of unit specific error components and contains the error
components that vary over both the cross- sectional units and time periods. Innovations are
not spatially correlated across units but they are auto-correlated over time. However, this
specification allows innovations from the same cross sectional unit to be correlated across
equations. Therefore, the vectors of disturbances are spatially correlated across units and across
equations as given in (10) the same specification was used by Kapoor, Kelejian, and Prucha
(2007); Baltagi, Song, and Koh (2003)).
= + , r = 1,2,3 (10)
The intercepts ( ) in equations (4) ? (6) represent the combined
influences of changes in the suppressed exogenous variables; the for r = 1,2,3 coefficients
are structural elasticities corresponding to the endogenous variables. A basic hypothesis to be
tested is that the coefficients are positive, which would mean an increase in automobile
production causes population, employment, and income to increase jointly, ceteris paribus. We
add the interaction terms to test whether the automobile production boom differentially affected
economic growth in the distressed Black Belt counties. Spatial components were incorporated to
capture the role of population, employment and per capita income of neighboring counties. A
Generalized Spatial Three-Stage Least squares (GS3SLS) approach as outlined by Kelejian and
Prucha (2004) into a panel data setting was used to estimate the model.
17
An important issue in regional development policy is whether ?people follow jobs? or
?jobs follow people.? For example, if people follow jobs, then policies to lure industry would be
appropriate. Conversely, if jobs follow people, public funds might be better spent educating the
resident population. The chicken or egg question can be tested by simple inspection of the t-
ratios associated the and coefficients in equations (4) and (6). For example, if =
0 and 0, then people follow jobs and the state should emphasize industrial development.
Conversely, if and 0 , then jobs follow people and the state should emphasize
educating the resident population. If > 0 and 0 migration and employment are
interrelated. In this instance, both development approaches are relevant and their relative
effectiveness would depend on the relative size of the coefficients.
3.1 Reduced Form Estimates and Long Run Elasticity
The reduced form equations are obtained by solving structural equations derived from GS3SLS
model. A spatial autoregressive model, in the context of single equation and in panel data setting,
is expressed as:
(11)
(12)
Where y is an NT x 1 vector of observations on the dependent variable. Wy is the
corresponding spatial lagged dependent variable for weights matrix W, X is NT X K matrix of
observations on the explanatory variables, u is an NT X 1 vector of error terms. is the spatial
autoregressive parameter and is a K X 1 vector of regression coefficients. is the spatial
autoregressive coefficient for the error lag and is NT x 1 vector of innovations or white
18
noise error. This single spatial autoregressive model can be extended to a system of spatially
interrelated equations. A standard G system of equations can be written as:
(13)
(14)
Y = y1, ??, y G X = x1,??.., x G U = u1,??., u G
WU = Wu1,?.., Wu G
Where yr is the NT x 1 vector of observations on the dependent variable in rth equation ,
xk is the NT x 1 vector of observations on the kth exogenous variable, ur is the NTx1 vector of
error terms in the rth equation, and B and are parameter matrices of dimension G x G and K x
G respectively. B is a diagonal matrix. W is N x N weights matrix of known constants and is
G x G matrix of parameters. Wyr and Wur are spatial lag and spatial autoregressive error term in
the rth equation respectively.
The solution for the endogenous variable can be revealed through the vector
transformation:
Letting y = vec Y , x = vec X, u= vec U and
or
19
or
After all the spatial effects and the other endogenous variables effects are controlled, the
relations between the dependent variables and the exogenous variables x can be expressed as:
The reduced form can be written as:
+u) (15)
(16)
The system of spatial structural equations was solved to obtain a system of reduced form
equations. Since the spatial weight matrix was constructed on the queen based adjacency
criteria, this system of spatial equations control the spatial spillover effect of neighboring
counties (Nzaku and Bukenya, 2005; Trendle, 2009; Gebremariam,2010). Reduced coefficients
of significant variables in the structural equations were estimated for the counties where
automobile plant locate and its neighboring counties.
20
In this system of equations, the dependent variables are the change in population,
employment and per capita income during a specific period and exogenous variables are the
initial value of those variables at the beginning of the specific period. The dependent variables
are constructed as the difference between the lnyt ? lnyt-i. One of the exogenous variables in the
right hand side of each equation is the predetermined lagged dependent variable (lnyt-i). The
long run elasticity of exogenous variable is calculated by dividing the coefficient of an
exogenous variable by the negative value of a coefficient of a predetermined lagged dependent
variable (lnyt-i).
4. Data and Sources
Data for sixty seven counties in Alabama are drawn from several sources (Table 1.1). These data
were collected for two study periods which are from 1970 to 2007 and from 1997 to 2005. The
growth of population, employment and per capital income for the study period from 1970 to
2007 were constructed using 7 years intervals between the beginning and end period, like 1970-
1977, 1980-1987, 1990-1997 and 2000-2007. These variables were constructed for the study
period from 1997 to 2005, using two years interval between the beginning and end period, like
1997-1999, 1999-2001, 2001-2003 and 2003-2005. Independent variables include demographic,
human capital, labor market, housing, amenities, automobile production, interaction term of
automobile production and distressed black belt county and policy variables. McGranahan
(1999) developed the Economic Research Service (ERS) natural amenities index, which
combines the attractiveness of mild climate, varied topography, and proximity to surface water
into one measure. This index was used for the amenity variable in this study. The initial values
of the independent variables are used as 7 year lagged values and 2 year lagged values for 1970-
21
2007 and 1997-2005 study period respectively. This formulation reduces the problem of
endogeneity. All independent variables are in log form except those that can take negative or
zero values. Automobile plants are located in only four counties of Alabama, namely
Tuscaloosa, Talladega, Madison and Montgomery. The distance between the major city of these
four counties and major city of all other counties were obtained from MapQuest. Ratios of
automobile production and distance for each county were constructed by dividing automobile
production of each plant by distance between major city of county where plant locates and the
major city of each county. Then, sum of ratios of automobile production and distance were
obtained by adding the ratios of every company. It is assumed that the effect of automobile
production decreases with the distance from automobile plant. Per capita income, property tax
and local tax were deflated by Consumer Price Index (CPI). The descriptive statistics of the
variables are given in Table 1.2.
5. Estimation Issues
Panel models for two study periods are estimated. Each panel model contains four time periods
and 67 counties. Then, 268 observations are used in the panel model for each study period. Panel
model can be used to control unobserved heterogenity and to investigate inter-temporal changes.
Since the panel data provide more information and variables, the degree of freedom and
efficiency increases and multicollinearity is less likely to occur. Following Baltagi (1995), one
way error component structure model was utilized for the panel data in this empirical study.
This system of equations has econometric issues regarding feedback simultaneity, spatial
autoregressive lag, and spatial cross-regressive lag simultaneity with spatially autoregressive
disturbances. These simultaneities create problems in estimation and identification of each
22
equation. The order condition for identification in a linear simultaneous equations model is that
the number of dependent variables on the right hand side of an equation must be less than or
equal to the number of predetermined variables in the model but not in the particular equation.
Lagged dependent variables also can be considered as predetermined variables. Kelejian and
Prucha considered that the spatially lagged dependent variables can be treated as predetermined
(Kelejian and Prucha, 2004). The order condition for each equation of the system in (4) ? (6) is
fulfilled.
The Hausman test (1983) for over identification was done to investigate whether the
additional instruments are valid in the sense that they are uncorrelated with the error term. That
is E(Q?ur) =0, Where E is the expectation operator and Q is an instrument matrix that consist
of a subset of linearly independent columns X, WX, W2X, where X is the matrix that includes
the control variables in the model. All equations are appropriately identified because the
hypothesis of orthogonality for each equation cannot be rejected even at P= 0.05 as indicated by
the test statistic in Table 1.3.
When the spatial autoregressive lag and spatial cross-regressive lag simultaneities are
present, the conventional three-stage least squares estimation to handle the feedback simultaneity
would be inappropriate. Therefore, the Method of Moments approach was used rather than
maximum likelihood because maximum likelihood would involve significant computational
complexity. Generalized Spatial Three-Stage Least squares (GS3SLS) approach outlined by
Kelejian and Prucha (2004) into a panel data setting was used to estimate the model. This new
procedure is performed in a five-step routine as given in Appendix .
23
The system of spatial structural equations was solved to obtain a system of reduced form
equations. Since spatial weight matrix was constructed on the queen based adjacency criteria,
this system of spatial equations control only spatial spillover effect of neighboring counties.
Reduced coefficient and long run elasticity of significant variables in the structural equations
were estimated to the counties where automobile plant locates and its neighboring counties.
6. Results and Discussion
The parameter estimates of the system for the two study periods were given in Table 1.3 and 1.4.
In general, the results are consistent with theoretical expectations and previous studies on
regional growth. The results show the existence of simultaneity among endogenous variables.
This indicates that there are strong interdependences among population growth, per capita
income growth and employment growth. The signs of the coefficients are consistent with
theoretical expectations.
In the model for the study period (1970-2007), the negative and significant coefficient of
lagged dependent variable in each equation indicates the conditional convergence with respect to
the respective endogenous variable of each equation. This also implies that growth of population,
per capita income and employment were higher in counties that had low initial level of
population, per capita income and employment, respectively, compared to counties with high
initial levels. These coefficients were not significant in the model for the shorter study period
(1997-2005). Reduced coefficients and long run elasticities of automobile production for study
period (1970 -2007) were given in Table 1.5. These estimates can be used to study the impact of
automobile production on Alabama?s economy. Reduced coefficients and long run elasticities of
24
other exogenous variables and reduced coefficient of initial level of population, per capita
income and employment were given in Table 1.6 and Table 1.7 respectively.
6.1 Population Growth Equation
In the equation of population growth, the coefficient of employment growth is positive as
expected and significant at the 1% level. The coefficient of employment growth (0.59) indicates
that a 1% increase in employment may result in-migrants and hence an increase in the population
by 0.59%, other things being equal. The previous studies (Carlino and Mills, 1987 and Clark and
Murphy, 1996) reported the same relationship that changes in employment are driving
population. This is interpreted as jobs follow people.
The reduced coefficients of automobile production indicate that, in the short run, growth
of population decreases with automobile production in a county where plant locates and increase
in the neighboring counties. However, the long run elasticity implies that population increases
with automobile production in a county where a plant locates and decreases in neighboring
counties. This result implies that population increases at a decreasing rate with automobile
production in the county where plant locates. In the long run, population declines in neighboring
counties with automobile production because people migrate to a county where a plant locates
for jobs. The effect of automobile production on migration decreases with distance from an
automobile plant. The long run elasticity of automobile production suggests that if automobile
production of a given plant can increase by 10%, the population of the county where the given
plant locates will increase by 0.5% but the population of neighboring counties will decline by
about 0.2%. Since the structural coefficient of the interaction term of automobile production and
25
distressed Black County is insignificant, there is no differential impact of automobile production
on population in distressed Black Belt counties.
The reduced coefficients from the study period 1970-2007 show that population growth is
positively and highly associated with the natural amenity index (Amenity) and highway density
(hway). Our results based on the natural amenity index show that counties scoring high in a
scale of these amenities are associated with high population change. Roback (1982, 1988) found
that people are willing to move to places with lower wages and fewer jobs, but that are rich in
amenities. However, long run elasticity implies that the population in a county with a low natural
amenity index is higher than a county with a high natural amenity index in long run. The county
with high natural amenity index has high land and housing values, restrictions for industrial
development and infrastructure.
The positive sign of a reduced coefficient of road density indicates that transportation is
very important for population migration and economic development. In the long run, better
infrastructure may reduce the population density. The reduced form coefficients imply that
population growth is negatively associated with the initial level of the unemployment rate
(UNEMP), and the proportions of the population above 65 years old in the short run. High
unemployment rate attracts firms into a county to take advantage of cheap labor and jobs then
workers from other counties in long run.
In this study, the long run elasticity shows that the population in a county will decline
with the increase of per capita tax and percentage of employed labor in farming. Population is
negatively associated with the initial level of average nonfarm proprietors? income. Housing and
land value may increase with the nonfarm proprietors? income and then decrease the population.
26
The structural coefficient of initial level of employment indicates that the growth of population
in a given county is positively associated with initial level of employment.
The structural coefficient of spatial autoregressive lag is positive and significant. This
indicates that population growth in neighboring counties positively influence the population
growth of a given county through immigration due to the low housing and land value. The
structural coefficient of cross-regressive lag with respect to employment growth is negative. This
may be explained as the reason that people are moving to neighboring countries for jobs. These
results show that the growth of population and employment in neighboring counties has spillover
effect on the growth of population in a given county. Global Moran?s I statistic and indicate
there is a spatial spillover effect with respect to the error terms. This indicates that random
shocks to the system affect not only the county where the shock originates and its neighbors, but
also create shockwaves across the study area, because of the structure of the autoregressive
errors.
6.2 Employment Growth Equation
In both study periods, the employment growth in a given county is positively and highly
associated with population growth and per capita income growth. In the study period 1970 -2007,
the coefficient of population growth (1.42) indicates that a 1% increase in the population is
associated with around a 1.42 % increase in the employment. This supports the hypothesis that
people follow jobs. The coefficient of per capita income growth (0.58) shows that there is
almost 0.58% increase in employment for 1% increase in per capita income. Carlino and Mills
(1987) found that population is driving employment growth and also the increase in income led
to employment growth.
27
The reduced form coefficient and long run elasticity of the automobile production are
positive. Long run elasticity of automobile production suggests that if automobile production of a
given plant can increase by 10%, the employment of a county where the given plant locates will
increase by 0.1 % in long run and the employment of neighboring counties will increase by
0.2%. The counties where plant locates have major cities of Alabama and the initial level of
employed people is very high, compare to neighboring counties. Since the structural coefficient
of the interaction term of automobile production and distressed Black Belt County is
insignificant, there is no differential impact of automobile production on employment in
distressed counties.
The results from the employment growth equation indicate that the growth of
employment in a given county is positively associated with initial level of unemployment rate
and average nonfarm proprietor income. A county with high unemployment and nonfarm
proprietor income has a higher potential for economic growth and employment growth than a
county which has a low unemployment rate and nonfarm proprietor income. This is consistent
with previous research findings that high regional unemployment rate indicates a plentiful supply
of cheap labor, which attracts firms into the region (Marston, 1985). The results show that the
initial level of population, per capita income and nonfarm proprietor income positively influence
employment growth. A county with high initial level of population and per capita income has a
greater demand for goods and services. An increase in the demand for goods and services can
create more employment. The reduced coefficient and long run elasticity implies that road
density is also very important for employment and economic growth.
28
In the employment growth equation, structural coefficients of spatial auto regressive lag
effect are positive and significant. This implies that employment growth in a given counties
depends on the averages of employment growth of neighboring counties. This positive
autoregressive lag effect implies that the spillover effect of employment growth in neighboring
counties positively affects the employment growth in a given county. New jobs may be created
due to the positive spillover effect of industrial clustering and availability of supporting services.
Employment growth in neighboring counties attracts job seekers to commute from a given
county.
The coefficient of cross regressive lag with respect to the population growth is negative.
This means that population growth in neighboring counties may attract more firms from a given
county. These results indicate that the population and employment growth in neighboring
counties have spillover effect on the employment growth of a given county. The existence of
spatial dependencies in the error terms imply that random shocks to the system affect not only a
county where the shock originates and its neighbors, but also the entire study area.
6.3 Per capita Income Growth Equation
The per capita income growth in a given county is positively and highly associated with the
growth of employment. This result is consistent with theoretical expectations. Nzaku and
Bukenya (2005) found that employment has a strong positive effect on per capita income. In the
study period 1970 -2007, the coefficient of employment growth (0.38) implies that a 1% increase
in employment is associated with around a 0.38 % increase in per capita income.
The reduced coefficient and long run elasticity suggests that automobile production of a
plant positively influences the per capita income of a county where the plant locates and of its
29
neighboring counties. The long run elasticity of automobile production indicates that if
automobile production of a given plant can increase by 10%, the per capita income of a county
where the given plant locates will increase by 0.3 % and the per capita income of neighboring
counties will increase by the range of 0.05% - 0.09%. Since the structural coefficient of the
interaction term of automobile production and distressed Black County is significant at the 10%
level, there might be differential impacts of automobile production on per capita income in
distressed Black Belt counties. Long run elasticity estimates of the interaction term suggests that
per capita income of distressed Black Belt County may rise by about 0.08 % for a 10% increase
in automobile production.
Although the results show that the per capita income in a given county is positively
associated with the initial level of the employment, the reduced coefficient and long run
elasticity of percentage of employed labor in farming suggests that the dependence of a large part
of employment on farming negatively influences the per capita income of a county. Per capita
income in a given county is shown to be negatively associated with the initial level of
unemployment rate and population. These results are consistent with previous research (Nzaku
and Bukenya, 2005). The negative reduced coefficient and long run elasticity of average nonfarm
proprietor income indicate the conditional convergence with respect to the per capita income.
This also implies that growth of per capita income were higher in counties that had low initial
level of nonfarm proprietor income, compared to counties with high initial level of nonfarm
proprietor income. The positive sign of road density indicates that transportation is very
important for economic development.
The structural coefficient of spatial auto regressive lag effect is positive and significant.
The per capita income growths in neighboring counties have positive spillover effect on the per
30
capita income growth of a given county. The structural coefficient of cross regressive lag effect
with respect to employment growth is negative. The higher employment growth in neighboring
counties makes neighboring counties more attractive to new and existing firms. These results
imply that the per capita income growth of a particular county is depend on the average of
employment growth and per capita income growth of neighboring counties. This is important
from a policy perspective because per capita income depends not only on the characteristics of
that county, but also on the characteristics of its neighbors. The disturbances indicate the
existence of spatial dependencies in the error terms. This means that random shocks to the
system affect not only a county where the shock originates and its neighbors, but also the entire
study area.
7. Conclusions and Policy Implications
The empirical findings suggest that if automobile production in a given plant increases by 10%
the population, per capita income and employment of a county where the plant locates will
increase by 0.5%, 0.3% and 0.1% respectively. But the population of neighboring counties will
decline by about 0.2%. However, the employment of neighboring counties will increase by 0.2%.
The per capita income of neighboring counties will increase by the range of 0.05% - 0.09%.
Since the structural coefficient of the interaction term of automobile production and distressed
Black County is significant at 10% level, there might be differential impact of automobile
production on per capita income in distressed Black Belt counties. The per capita income of
distressed Black County may rise by about 0.08 %. There is no differential impact of automobile
production on population, and employment in distressed Black Belt counties.
31
Automobile plants are located in major cities and jobs in these plants are better paying
jobs. Because of large automobile plants, input suppliers have located in neighboring counties
and created employment opportunities. The prevailing high unemployment rate in neighboring
counties may be the reason for the increase in employment in these counties even though people
move from there to major cities for better paying jobs.
There are significant feedback simultaneities among the growth of population, per capita
income and employment in Alabama counties during the study period. Employment growth in
the population growth equation and population growth in the employment growth equation have
positive strong effect. The conclusion can only be that jobs follow people and also people follow
jobs. The results show that a 1% increase in population growth is associated with a 1.42 %
increase in employment growth but a 1% increase in employment growth is associated with only
0.6 % increase in population growth. This means that the effect of population growth on
employment growth is at least 2 times as large as the effect of employment growth on population
growth. Appropriate policies to lure industrial development and improve the educational level of
resident population become very important for economic development.
The results of this study also show the existence of positive spatial autoregressive lag
with respect to the growth of population, employment and per capita income. The population
growth equations show the existence of negative spatial cross regressive lag effect with respect
to growth of employment in neighboring counties. The growth of per capita income equation
shows the existence of negative spatial cross regressive lag effect with respect to growth of
employment in neighboring counties. The growth of employment equation shows the existence
of negative spatial cross regressive lag effect with respect to the growth of population in
neighboring counties. These findings are important from an economic perspective because the
32
existence of these spatial lag effects indicates that growth of population, employment and per
capita income are not only dependent on the characteristics of that county, but also on those of its
neighbors. These interdependences provide the need for economic development policy
coordination among the counties. This finding has economic policy implications. It indicates that
sector specific policies should be integrated in order to achieve desired outcomes.
33
Table 1.1: Variable Description and Data Sources
----------------------------------------------------------------------------------------------------------------------------------------
Variable variable Description unit Source
----------------------------------------------------------------------------------------------------------------------------------------
POPG Population Growth % A, B
PCIG Per capita income Growth % A, B
EMPG Employment Growth % A, B
pop population number B
pci per capita income $/person B
emp employment number B
auto No. of automobile/distance Number/mile A, J, K
autoblack Interaction of auto and Black Belt county
unemp unemployment rate % E
17years % of population below 17years % C, D
65years % of population above 65years % C, D
hsch % of high school degree or above % C, D
bach % of bachelor degree or above % C, D
pov poverty rate % D
protax per capita property tax $/person D
tax per capita local tax $/person D
owner owner occupied housing in percent % D
farm % employed in farming % B
manu %employed in manufacturing % B
serv %employed in other sectors % B
amenity Natural Amenities Index ERS index H
anfpin average nonfarm proprietor?s income $ B
hway road density mile/square mile I
dista distance from metro area mile J
metro dummy variable for metro area dummy value
Spatial lag of POPG % A, B
Spatial lag of PCIG % A, B
Spatial lag of EMPG % A, B
-----------------------------------------------------------------------------------------------------------------------------------------
A- Computed, B- US Department of Commerce, Bureau of Economic Analysis (REIS database), C- County &
City Data Book, D- U.S Census Bureau, E- Bureau of Labor Statistics, F- American Medical Association, G-
Federal Bureau of Investigation, H- Economic Research Service, USDA, I ? US Bureau of Transportation
Statistics, J- Map Quest, K - Mercedes-Benz U.S. International, Tuscaloosa, AL, Honda Manufacturing of
Alabama, Lincoln, AL, Hyundai Motor Manufacturing Alabama, Montgomery, AL, Toyota Motor Manufacturing
Alabama, Huntsville, AL, Automotive News Market Data Book
34
Table 1.2: Descriptive Statistics for Alabama Counties
------------------------------------------------------------------------------------------------------------------------------------------
1970-2007 1997-2005
--------------------------- ------------------------
Variable variable Description Mean Std Dev Mean Std Dev
------------------------------------------------------------------------------------------------------------------------------------------
POPG Population Growth, t 1.05 0.09
1.00 0.02
PCIG Per capita income Growth, t 1.14 0.1
1.04 0.03
EMPG Employment Growth, t 1.1 0.14
1.01 0.06
pop population, t-i 59149.84 93442.84
66241.42 98368.96
pci per capita income, t-i 18225.76 4978.4
23675.26 3662.83
emp employment, t-i 28441.48 55516.72
35295.81 66389.74
auto No. of automobile/distance, t-i years 494.7 4892.61
3284.24 15224.87
autoblack Interaction of auto and Black Belt county 55.24 283.85
246.25 741.71
unemp unemployment rate, t-i years 8.8 4.93
6.62 2.15
17years % of population below 17years, t-i 30.03 4.99
25.49 4.18
65years % of population above 65years, t-i 12.74 2.53
13.91 2.01
hsch % of high school degree or above, ,t-i 53.37 14.97
69.65 6.73
bach % of bachelor degree or above, t-i 9.95 5.61
13.49 6.32
pov poverty rate, t-i 23.09 10.33
18.05 5.35
protax per capita property tax, t-i 81.21 80.5
201.82 100.38
tax per capita local tax, t-i 208.02 181.56
482.31 217.88
owner owner occupied housing in percent, t-i 72.97 7.1
farm % employed in farming, t-i 9.03 6.69
5.58 3.69
manu %employed in manufacturing, t-i 25.4 10.42
20.52 9.21
serv %employed in other sectors, t-i 16.98 5.87
54.38 9.63
amenity Natural Amenities Index, t-i 1.87 1.79
1.87 1.79
anfpin average nonfarm proprietor?s income, t-i 11312.53 4988.92
18172.37 6034.65
hway road density, t-i 0.13 0.03
0.13 0.03
dista distance from metro area 34.72 25.18
34.72 25.18
metro dummy variable for metro area 1.31 0.66
1.31 0.66
(I?W)POPG Spatial lag of POPG, t 1.05 1.06
1 1.01
(I ?W)PCIG Spatial lag of PCIG, t 1.13 1.07
1.03 1.02
(I?W)EMPG Spatial lag of EMPG, t 1.09 1.07
1.01 1.03
---------------------------------------------------------------------------------------------------------------------
i is 7 for the period 1970 -2007 and 2 for the period 1997- 2005.
35
Table 1.3: Structural Coefficients for the Study Period (1970-2007)
------------------------------------------------------------------------------------------------------------------------------------------------
POPG Equation EMPG Equation PCIG Equation
----------------------------- ----------------------------- --------------------------------------
Variable Coeff. z-stat Coeff. z-test Coeff. z-test
------------------------------------------------------------------------------------------------------------------------------------------------
POPG
1.424 17.38
-0.148 -1.42
PCIG -0.173 -2.02
0.587 5.13
EMPG 0.594 16.01
0.377 6.33
pop -0.069 -2.68
0.133 3.83
-0.077 -2.38
pci -0.020 -0.46
0.139 2.24
-0.289 -5.68
emp 0.052 2.34
-0.102 -3.39
0.056 2
auto -0.007 -4.18
0.006 2.39
0.007 3.1
autoblack -0.002 -1.05
0.001 0.29
0.004 1.62
unemp -0.049 -5.91
0.084 6.89
-0.054 -5.22
17years 0.001 0.02
-0.030 -0.54
0.063 1.23
65years -0.060 -2.55
0.064 1.81
0.035 1.14
hsch -0.028 -0.81
-0.001 -0.02
0.082 1.8
bach -0.008 -0.62
0.022 1.14
-0.016 -1.01
pov 0.004 0.29
-0.009 -0.44
0.012 0.67
protax 0.005 0.55
-0.003 -0.25
-0.019 -1.77
tax 0.033 2.44
-0.060 -3
0.055 3.31
owner 0.011 0.4
farm 0.008 1.35
-0.003 -0.35
-0.017 -2.41
manu -0.005 -0.59
0.010 0.86
-0.009 -0.92
serv -0.008 -0.6
0.020 1
-0.022 -1.24
amenity 0.013 3.89
-0.018 -3.7
0.001 0.34
anfpin -0.029 -2.07
0.058 2.88
-0.052 -3.24
hway 0.025 2.25
-0.042 -2.59
0.032 2.28
dista -0.008 -1.36
0.010 1.16
-0.002 -0.2
metro -0.032 -1.44
0.033 1.01
0.019 0.68
(I ? W) POPG 0.469 4.02
-0.812 -4.86
0.131 0.85
(I ? W) PCIG 0.017 0.25
-0.147 -1.49
0.267 2.95
(I ? W) EMPG -0.376 -4.13
0.713 5.67
-0.283 -2.66
0.967 2.15
-2.529 -4.11
3.121 6.1 Constant
RHO(?) -0.321 -8.33b
-0.538 -3.88b
-0.121 -1.18b
SIG V 0.001 29.1b
0.003 7.3b
0.003 12.24b
SIG 1 0.002 16.59b
0.004 3.52b
0.002 6.82b
NR2 - ?2 (39,41,40) 31.770 0.7877c
31.283 0.8364c
44.135 0.3405c
Moran I 0.149 0.022
0.065 0.244
0.144 0.027
N 268
268
268
---------------------------------------------------------------------------------------------------------------------------------------------------
b: t-static value, c: p-value
36
Table 1.4: Structural Coefficients for the Study Period (1997-2005)
------------------------------------------------------------------------------------------------------------------------------------------------
POPG Equation EMPG Equation PCIG Equation
----------------------------- ----------------------------- -----------------------------
Variable Coeff. z-stat Coeff. z-test Coeff. z-test
------------------------------------------------------------------------------------------------------------------------------------------------
POPG
1.5919 5.99
-0.7680 -3.5
PCIG -0.1742 -3.3
1.0307 8.12
EMPG 0.2201 6.23
0.5545 8.35
pop 0.0006 0.08
0.0312 1.48
-0.0205 -1.37
pci 0.0062 0.48
0.0456 1.37
-0.0172 -0.72
emp -0.0004 -0.05
-0.0237 -1.32
0.0136 1.09
auto -0.0006 -1.78
-0.0002 -0.24
-0.0004 -0.71
autoblack -0.0005 -1.32
-0.0015 -1.34
0.0018 2.41
unemp -0.0290 -6.62
0.0381 2.81
-0.0183 -1.86
17years -0.0025 -0.31
0.0103 0.5
-0.0167 -1.22
65years -0.0274 -3.59
0.0021 0.1
-0.0004 -0.03
hsch 0.0070 1.74
-0.0096 -0.91
0.0073 1
bach -0.0018 -0.42
0.0076 0.71
-0.0015 -0.2
farm 0.0036 1.8
0.0065 1.32
-0.0050 -1.57
manu -0.0012 -0.4
0.0116 1.55
-0.0023 -0.46
serv -0.0018 -0.14
0.0563 1.88
-0.0141 -0.7
amenity 0.0022 1.99
-0.0023 -0.79
0.0009 0.45
anfpin -0.0024 -0.69
0.0033 0.37
-0.0063 -1.04
hway 0.0040 1.14
-0.0085 -0.92
0.0036 0.57
dista -0.0025 -1.26
-0.0032 -0.62
0.0033 0.99
metro -0.0120 -1.64
-0.0133 -0.69
0.0117 0.88
(I ? W) POPG 0.1031 0.93
-0.8776 -3.42
0.4821 2.75
(I ? W) PCIG -0.0486 -0.6
-0.9182 -4.88
0.9099 7.41
(I ? W) EMPG -0.0740 -1.38
0.7485 6.98
-0.4111 -4.6
0.1093 0.77
-0.9666 -2.96
0.4564 0.31 Constant
RHO(?) 0.0089 0.05b
-0.6998 -6.52b
-0.8148 -6.51b
SIG V 0.0002 6.05b
0.0009 7.08b
0.0004 5.31b
SIG 1 0.0002 3.52b
0.0009 3.49b
0.0042 1.94b
NR2- 2(37,37,37) 28.6 0.838c
27.7 0.867c
35.2 0.554c
Moran I 0.0310 0.498
0.1490 0.025
0.0200 0.592
N 268
268
268
---------------------------------------------------------------------------------------------------------------------------------------------------
b: t-static value, c: p-value
37
Table 1.5: Reduced Coefficients, Long Run Elasticities and 10% Impacts of Automobile Production (1970-2007)
-------------------------------------------------------------------------------------------------------------------------------------------------------
Reduced Coefficient Long Run Elasticity 10% impact
----------------------------------------------- --------------------------------------- ----------------------------
County POPG PCIG EMPG pop pci emp pop pci emp
-------------------------------------------------------------------------------------------------------------------------------------------------------
Toyota
Jackson 0.005 0.003 0.015
-0.029 0.015 0.032
-0.183 0.092 0.199
Limestone 0.005 0.003 0.014
-0.028 0.013 0.030
-0.193 0.085 0.206
Madison -0.008 0.008 0.005
0.048 0.034 0.010
0.483 0.337 0.099
Marshall 0.004 0.002 0.012
-0.024 0.009 0.025
-0.155 0.059 0.162
Morgan 0.005 0.003 0.013
-0.027 0.011 0.029
-0.173 0.072 0.184
Hyundai
Autauga 0.004 0.002 0.013
-0.026 0.011 0.028
-0.191 0.079 0.202
Bullock 0.005 0.003 0.015
-0.031 0.012 0.032
-0.191 0.074 0.200
Crenshaw 0.005 0.002 0.013
-0.027 0.010 0.029
-0.164 0.062 0.171
Elmore 0.004 0.002 0.012
-0.025 0.010 0.027
-0.177 0.069 0.185
Lowndes 0.004 0.002 0.012
-0.026 0.010 0.028
-0.177 0.068 0.190
Macon 0.005 0.002 0.013
-0.028 0.011 0.029
-0.176 0.066 0.184
Montgomery -0.008 0.008 0.004
0.050 0.033 0.009
0.495 0.333 0.094
Pike 0.005 0.002 0.014
-0.029 0.011 0.031
-0.178 0.065 0.185
Honda
Calhoun 0.005 0.003 0.014
-0.029 0.011 0.030
-0.195 0.077 0.205
Clay 0.005 0.003 0.013
-0.028 0.011 0.030
-0.194 0.077 0.204
Cleburne 0.005 0.003 0.014
-0.030 0.012 0.031
-0.191 0.074 0.200
Coosa 0.004 0.002 0.012
-0.025 0.010 0.026
-0.156 0.060 0.164
St. Clair 0.004 0.002 0.012
-0.025 0.010 0.026
-0.161 0.062 0.168
Shelby 0.004 0.002 0.011
-0.024 0.009 0.025
-0.151 0.059 0.159
Talladega -0.008 0.008 0.004
0.052 0.033 0.008
0.524 0.330 0.083
Mercedes -Benz
Bibb 0.004 0.002 0.011
-0.024 0.009 0.025
-0.154 0.060 0.162
Fayette 0.005 0.003 0.014
-0.030 0.011 0.031
-0.185 0.071 0.194
Greene 0.004 0.002 0.013
-0.027 0.011 0.029
-0.173 0.070 0.183
Hale 0.004 0.002 0.012
-0.026 0.011 0.027
-0.162 0.067 0.172
Jefferson 0.004 0.002 0.011
-0.024 0.009 0.025
-0.142 0.055 0.149
Pickens 0.005 0.003 0.014
-0.030 0.012 0.031
-0.194 0.075 0.203
Tuscaloosa -0.009 0.007 0.003
0.054 0.033 0.008
0.540 0.328 0.077
Walker 0.004 0.002 0.012
-0.025 0.009 0.026
-0.151 0.054 0.156
--------------------------------------------------------------------------------------------------------------------------------------------------------
38
Table 1.6: Reduced Coefficients, Long Run Elasticities of Exogenous Variables (1970-2007)
-------------------------------------------------------------------------------------------------------------------------------------------------------
Reduced Coefficient Long Run Elasticity
----------------------------------------------- ------------------------------------
Variable County POPG PCIG EMPG pop pci emp
-------------------------------------------------------------------------------------------------------------------------------------------------------
Madison
0.018 -0.036 0.097
-0.107 -0.157 0.207
anfpin Montgomery 0.018 -0.036 0.095
-0.104 -0.158 0.206
Talladega
0.016 -0.036 0.091
-0.099 -0.16 0.203
Tuscaloosa 0.015 -0.036 0.089
-0.097 -0.16 0.201
Madison
-0.026 -0.049 0.061
0.156 -0.217 0.13
unemp Montgomery -0.027 -0.05 0.058
0.164 -0.219 0.126
Talladega
-0.029 -0.05 0.053
0.179 -0.22 0.119
Tuscaloosa -0.03 -0.05 0.051
0.188 -0.221 0.115
Madison
-0.101 0.022 -0.028
0.599 0.095 -0.06
65years Montgomery -0.101 0.021 -0.03
0.612 0.092 -0.065
Talladega
-0.103 0.02 -0.034
0.64 0.09 -0.076
Tuscaloosa -0.104 0.02 -0.037
0.656 0.089 -0.082
Madison
0.058 0.073 0.052
-0.347 0.321 0.111
tax Montgomery 0.059 0.073 0.052
-0.355 0.323 0.114
Talladega
0.059 0.073 0.054
-0.369 0.322 0.12
Tuscaloosa 0.06 0.073 0.055
-0.377 0.322 0.124
Madison
0.027 -0.011 0.027
-0.158 -0.048 0.057
farm Montgomery 0.027 -0.011 0.027
-0.161 -0.048 0.058
Talladega
0.027 -0.011 0.027
-0.166 -0.047 0.06
Tuscaloosa 0.027 -0.011 0.027
-0.169 -0.047 0.061
Madison
0.003 -0.003 -0.029
-0.019 -0.015 -0.061
amenity Montgomery 0.003 -0.003 -0.028
-0.02 -0.014 -0.06
Talladega
0.004 -0.003 -0.026
-0.024 -0.013 -0.059
Tuscaloosa 0.004 -0.003 -0.026
-0.026 -0.013 -0.058
Madison
0.051 0.048 0.049
-0.3 0.21 0.105
hway Montgomery 0.051 0.048 0.049
-0.307 0.211 0.107
Talladega
0.051 0.048 0.05
-0.318 0.211 0.112
Tuscaloosa 0.051 0.048 0.051
-0.324 0.211 0.115
Bullock
---- 0.0004 ----
---- 0.0018 ----
autoblack Greene ---- 0.0004 ----
---- 0.0019 ----
Hale
---- 0.0004 ----
---- 0.0019 ----
Lowndes ---- 0.0003 ----
---- 0.0015 ----
Macon
---- 0.0003 ----
---- 0.0015 ----
Pickens
---- 0.0004 ----
---- 0.0019 ----
-------------------------------------------------------------------------------------------------------------------------------------------------------
39
Table 1.7 Reduced Coefficients of Initial of Population, Per Capita Income and Employment (1970-2007)
------------------------------------------------------------------------------------------------------------------------------------------
Variable county POPG PCIG EMPG
------------------------------------------------------------------------------------------------------------------------------------------
Madison
0.168
0.191
-0.165
Montgomery 0.165
0.188
-0.163
pop Talladega
0.161
0.186
-0.159
Tuscaloosa 0.158
0.185
-0.157
Madison
0.027
-0.227
-0.041
pci Montgomery 0.027
-0.227
-0.040
Talladega
0.025
-0.227
-0.039
Tuscaloosa 0.024
-0.228
-0.039
Madison
0.513
0.335
-0.467
emp Montgomery 0.505
0.331
-0.460
Talladega
0.492
0.325
-0.449
Tuscaloosa 0.485
0.322
-0.443
--------------------------------------------------------------------------------------------------------------------------------------------
40
CHAPTER 2
The Effects of Automobile Production and Local Government Expenditure on Poverty in
Alabama
1. Introduction
Poverty reduction is one of the major concerns for policy makers and local governments in most
of the countries. In the United States, Poverty is unevenly distributed across counties. Poverty
rates remains high in the most isolated rural counties, particularly in counties far from
metropolitan areas (Glasmeier and Farrigan, 2003; Swaminathan and Findeis, 2004; Partridge
and Rickman, 2006, Ch. 2). Poverty rate in the United States increased from 11.3% in 2000 to
12.3% in 2006 (DeNavas-Walt, Bernadette, and Smith 2007). In the United States, the
Appalachian Region has been the center of attention for poverty reform because most of the
counties are isolated rural counties and far behind in the social and economic development from
the rest of the nation (Pollard, 2003). National and local policy programs to alleviate poverty in
this region have shown a substantial improvement in economic conditions over the past several
decades.
The poverty rate in Alabama was 15.3 percent in 2003. In Alabama counties, the poverty
rate ranged from 6.8 percent in Shelby County to 28.7 percent in Perry County. Among the
counties, fourteen had poverty rates of 20 percent or higher. The Economic Research Service,
USDA, classifies counties as persistent poverty counties if they have had poverty rates of 20
percent or higher in each decennial census from 1970 through 2000. In Alabama, 22 counties are
41
classified as persistent poverty counties. 17 counties of these 22 counties are non-metro counties
(RUPRI, 2007).
Figure 2.1:
Figure 2.2:
42
Alabama is the tenth poorest state in the nation and one of 20 states that have established
a commission on poverty. The Alabama state legislature has formed the State Commission to
study state-supported programs, policies and services and make recommendations on proposed
legislation concerning poverty. Since the state government has expanded economic incentives to
attract auto industry to create additional employment and generate personal income, large auto
mobile firms and its input suppliers have located in several Alabama counties. The auto industry
in Alabama accounted for 47,457 direct jobs and 85,700 indirect jobs through their purchases
and expenditures with annual payroll of $5.2 billion by 2007 (AAMA 2008). Jobs in 40 of the
state?s 67 counties now are tied directly or indirectly to auto manufacturing (AAMA 2008).
Private investments and government expenditures are sources of both employment and income.
In addition to the socio, economic, demographic and other factors, these two sources
substantially contribute to the family poverty rate on county level.
2. Literature Review
There are several studies on the determinants of poverty in urban and rural areas. Most of these
studies model poverty rates or changes in poverty rates as functions of demographic
characteristics and local economic conditions, using county-level data. Gibbs (1994) and Davis
and Weber (2002) argue that rural labor markets are thinner with poorer employer-employee
matches than their urban counterparts, Fisher (2005; 2007) shows that while part of the higher
rate of poverty in rural areas is attributable to poor economic opportunities in rural areas and
self-selection of poor people locating into rural areas. Some studies have examined spatial
externalities in poverty research. Rupasingha and Goetz (2000) have developed a spatial
43
econometric model and found that changes in poverty are affected by the poverty of neighboring
counties.
Several researchers have investigated the effects of changes in economic, social, political,
and demographic conditions on the poverty rate. Levernier, Partridge, and Rickman (2000) in
their study found that economic development targeting African-American communities and non-
MSA counties would be most effective in reducing poverty. Triest (1997) concluded that
increased employment and educational opportunity of the low-income population would narrow
the interregional gap in poverty. Rupasingha and Goetz (2007) suggested that public investment
in social capital can reduce poverty rates by easing transaction costs paid by local associations.
Fan, Linxiu, and Xiaobo (2002) concluded that government expenditures on rural
education and infrastructure reduced the rural poverty rate. Jung and Thorbecke (2003) found
that increased expenditure on education can contribute to economic growth and poverty
alleviation by supplying more educated and skilled labor. Education is another key for reducing
poverty rates for the counties with minorities (Swail, Redd, and Perna 2003). But, Gomanee et
al. (2005) found that public spending on social services was ineffective in reducing poverty and
suggested that new techniques should be developed to improve the efficiency of public spending.
Industry composition also can affect the poverty rate. Levernier, Partridge, and Rickman (2000)
found that counties with above-average shares of employment in agriculture, trade, and services
have higher poverty rates.
The purpose of the research is to determine the effects of the auto industry and local
government expenditure on the poverty of Alabama?s counties. This research improves on
existing research in many ways. First, we include the initial level of poverty rate, which allow us
to test whether the equation converges with the respective to dependent variable. Second, we are
44
able to estimate the differential impact of auto production on proportional change in the poverty
rate in the distressed black belt counties by introducing an interaction term of auto production
and these counties. Third, we incorporate spatial components to capture the role of poverty rate
of neighboring counties. Finally, we include the initial level of employment, per capita income,
population, and other socio, economic, demographic and policy variables to control their effect
on the dependent variable. Since these variables are 10 years lagged, the endogenous problem
from these variables can be avoided. The analysis will be based on county data for the study
period 1970 ? 2000. A major goal is to determine whether poverty rate in the distressed counties
in the state?s Black Belt are reduced from the auto boom.
3. Model
This model developed using the idea that private investments are important sources for
generating employment and income. In addition to the socio, economic, demographic and other
factors, private investments can substantially influence the poverty rate on a county level.
Poverty rate, in county level are influenced by the socio, economic, demographic and policy
variables and spatial components of poverty rate of neighboring counties.
4. Data and Sources
Data for sixty seven counties in Alabama are drawn from several sources (Table 2.1). These data
were collected for the study period for the years 1970 to 2000. The growth of poverty rate was
constructed using 10 years interval between the beginning and end period, like 1970-1980, 1980-
1990 and 1990-2000. Independent variables include demographic, human capital, labor market,
automobile production, interaction term of automobile production and distressed Black Belt
45
county and policy variables. The initial values of the independent variables are lagged 10 years.
But automobile production variable is lagged 2 years in this equation. This formulation reduces
the problem of endogeneity. The variable, automobile production (lnAt-i), was constructed as ln
(automobile production/ distance). All independent variables are in log form except those that
can take negative or zero values. Per capita income, per capita local government expenditure
and per capita local tax were deflated using consumer price index (CPI). The descriptive
statistics of the variables are given in Table 2.2.
5. Estimation Issues
A panel model is estimated using 201 observations. This panel model contains three time periods
for 67 Alabama counties. This panel model was used to control unobserved heterogenity and to
investigate inter-temporal changes. Since the panel data provide more information and variables,
the degree of freedom and efficiency increases and multicollinearity is less likely to occur.
Many studies suggest that geographical location and location parameters significantly
affect productivity, inequality and growth (Quah 1996, Redding and Venables 2002, Rupasingha
et al. 2002, Rupasingha and Goetz 2007). The presence of spatial dependence can result in
misleading results from employing models using OLS (LeSage 1999). Poverty in a given county
may have spillover effects to the neighboring county. Then, the errors are dependent. In this
study, three alternative spatial specification models and a model without spatial component were
estimated. The three alternative spatial specification models are Spatial Lag model, Spatial Error
Model (SEM) and Spatial Autoregressive model (SAR). Spatial Lag model was estimated by
Maximum Likelihood Estimation method. Spatial Error Model and Spatial Autoregressive
Model were estimated by a Method of Moments Approach. The Spatial Lag model accounts for
46
the spatial dependence in the dependent variable and SEM incorporates spatial dependence in the
error term. The SAR model accounts for both spatial dependence in the dependent variable and
error term.
The Spatial Lag model takes the following form:
(1)
Y is the dependent variable and X is a vector containing all the independent variables and
is a normally distributed error term. is autoregressive coefficient and W is the weighting
matrix that was constructed on the queen based adjacency criteria. This weight matrix controls
only spatial spillover effect of neighboring counties.
Spatial dependence could also arise if a shock to an omitted variable in the model affects
the dependent variable. The SEM takes the following:
(2)
(3)
Where is the scalar spatial error coefficient. The spatial Autoregressive Model
incorporates spatial dependence in both the dependent variable and shocks to omitted variables
in the model. It takes the following form:
(4)
(5)
47
6. Results and Discussion
The parameter estimates of the four regression models and long run elasticity were given in
Table 2.3 and Table 2.4 respectively. In general, the results are consistent with theoretical
expectations and previous studies. The results of Moran I statistics and spatial dependence
models indicate that there is no spatial dependence in dependent variable and in error terms. The
significant coefficient of initial level of poverty rate (0.169) implies that there is conditional
convergence with respect to the poverty rate. It also indicate that, other thing being equal, a
county which had higher initial level of poverty rate will have higher poverty rate than a county
which had lower initial poverty rate.
6.1 Poverty Rate Equation
In the equation of poverty rate, the coefficient of automobile production (-.0326), and the
interaction term of automobile production and distressed black belt county (-0.026) are negative
and significant at the 5% level. The long run elasticity of automobile production (-.039), and the
interaction term of automobile production and distressed Black Belt county (0.031) suggest that
if automobile production in a given plant can increase by 10%, the poverty rate of a county
where the plant locates will decrease by 0.39% but if a county is a distressed Black Belt County,
the poverty rate will decrease by 0.7%. The poverty rate of other counties decreases but this
decrease in poverty declines with distance from a county where the plant locates. This result
shows that automobile production in Alabama significantly reduced the poverty of the distressed
Black Belt counties, compare to other counties.
48
The coefficient of female household head (0.29) suggests that the poverty rate in a given
county is positively associated with the percentage of female headed households. The long run
elasticity of female household head (0.35) indicates that a 10% increase in the percentage of
female household heads in a given country is associated with 3.5% increase in the poverty rate in
the given county. This positive sign is consistent with previous studies. Poverty rates are also
higher for female-headed families, among most minority groups and among families with larger
numbers of children (Farmer et al., 1989, Levernier et al., 2000). The results show that
unemployment is positively related to the poverty rate. The long run elasticity of unemployment
rate (0.116) suggests that a10 % increase in unemployment rate of a given county will raise the
poverty rate of the county by 1.6%.
In this study period, the coefficient of per capita local government expenditure is
insignificant. It indicates that the local government expenditure is ineffective in reducing the
poverty rate of the given county. Gomanee et al. (2005) also found that public spending on social
services was not effective in reducing poverty and highlighted the need for new techniques to
improve the efficiency of public spending. The results show that the poverty rate is negatively
associated with the initial level of per capita income. The long run elasticity of the initial level of
per capita income (-1.28) implies that a 10 % increase in the initial level of per capita of income
of a given county will reduce the poverty rate of the county by 12.8%.
7. Conclusions and Policy Implications
The empirical findings suggest that automobile production in Alabama significantly reduced the
poverty rate in all counties. The impact of automobile production on poverty reduction in
distressed Black Belt counties is greater than in other counties. Local government expenditures
49
aimed at reducing poverty was found to be effective. This result suggests that industrial
development may be more effective in reducing poverty than government programs.
Table 2.1: Variable Description and Data Sources
----------------------------------------------------------------------------------------------------------------------------------------
Variable variable Description unit Source
----------------------------------------------------------------------------------------------------------------------------------------
POV Poverty Rate, t % A, D
Auto No. of automobile/distance, t- 2 years Number/mile A, G, H
Autoblack Interaction of auto and Black Belt county
Lpov Poverty Rate, t- i % D
Lgexp Per Capita Local Government Expenditure, t-i $/person D
Lpop population, t-i number B
Lpcip per capita income, t-i $/person B
Ltem employment, t-i number B
Unemp Unemployment Rate, t-i year % E
D17years % of population below 17years, t-i % C, D
D65years % of population above 65years, t-i % C, D
Hsch % of high school degree or above,t-i % C, D
Fhh % of Female household Head family, t-i % C, D
Tax per capita local tax $/person D
Hway road density, t-i mile/square mile F
Metro dummy variable for metro area Dummy value
Spatial Lag of Growth Rate of Poverty, t % A, D
-----------------------------------------------------------------------------------------------------------------------------------------
A- Computed, B- US Department of Commerce, Bureau of Economic Analysis (REIS database), C-
County & City Data Book, D- U.S Census Bureau, E- Bureau of Labor Statistics, F ? US Bureau of
Transportation Statistics, G- Map Quest, H - Mercedes-Benz U.S. International, Tuscaloosa, AL, Honda
Manufacturing of Alabama, Lincoln, AL, Hyundai Motor Manufacturing Alabama, Montgomery, AL, Toyota Motor
Manufacturing Alabama, Huntsville, AL, Automotive News Market Data Book
50
Table 2.2: Descriptive Statistics for Alabama Counties
----------------------------------------------------------------------------------------------------------------------------------------
Variable variable Description Mean Std Dev
----------------------------------------------------------------------------------------------------------------------------------------
POV Poverty Rate, t 19.93 10.697
Auto No. of automobile/distance, t-2 years 566.4 4853
Autoblack Interaction of auto and Black Belt county 62.25 280.5
Lpov Poverty Rate, t- i 25.12 12.05
Lgexp Per Capita Local Government Expenditure, t-i 1614 784.6
Lpop population, t-i 56717 91682
Lpcip per capita income, t-i 16476 4009
Ltem employment, t-i 25899 50298
Unemp Unemployment Rate, t-i year 9.6 5.22
D17years % of population below 17years, t-i 31.52 4.78
D65years % of population above 65years, t-i 12.36 2.59
Hsch % of high school degree or above,t-i 47.93 12.92
Fhh % of Female household Head family, t-i 18.18 7.23
Tax per capita local tax 292.61 131.09
Hway road density, t-i 0.126 .031
Metro dummy variable for metro area .179 .384
Spatial Lag of Growth Rate of Poverty, t - 13.9 11.1
-----------------------------------------------------------------------------------------------------------------------------------------
i is 10 years
51
Table 2.3: The Estimation Results of Regression Models
--------------------------------------------------------------------------------------------------------------------------------------------
OLS Spatial Lag SEM Spatial Autoregressive
--------------------- --------------------- ------------------- -----------------
Variable coeff. t coeff. t coeff. t coeff. t
--------------------------------------------------------------------------------------------------------------------------------------------
const 10.16 4.54
10.29 4.47
9.59 4.29
9.71 4.22
auto -0.0326 -2.79
-0.0322 -2.73
-0.0327 -2.85
-0.0324 -2.79
autoblack -0.0259 -2.26
-0.0259 -2.25
-0.0267 -2.32
-0.0266 -2.31
unemp 0.0966 2.37
0.0975 2.38
0.1000 2.49
0.1006 2.48
d17years 0.0880 0.37
0.0827 0.35
0.1317 0.56
0.1256 0.53
d65years 0.1567 1.32
0.1545 1.29
0.1653 1.39
0.1634 1.36
hsch 0.1292 0.76
0.1253 0.73
0.1164 0.69
0.1136 0.67
fhh 0.2903 3.61
0.2907 3.6
0.2832 3.54
0.2837 3.54
tax -0.0376 -0.52
-0.0360 -0.49
-0.0428 -0.6
-0.0414 -0.57
lpop -0.0298 -0.3
-0.0298 -0.3
-0.0276 -0.28
-0.0277 -0.28
lpci -1.0738 -4.77
-1.0793 -4.76
-1.0297 -4.61
-1.0354 -4.6
ltem 0.0623 0.68
0.0628 0.69
0.0624 0.69
0.0628 0.69
lgex 0.0505 0.81
0.0502 0.81
0.0561 0.92
0.0556 0.9
lpov 0.1642 2.01
0.1638 2
0.1677 2.07
0.1674 2.06
hway -0.0231 -0.35
-0.0245 -0.37
-0.0179 -0.28
-0.0191 -0.29
metro -0.0121 -0.23
-0.0118 -0.22
-0.0169 -0.32
-0.0165 -0.31
(I?W)povr
-0.0191 -0.25
-0.0144 -0.19
Rho
0.1165 0.62
0.1385 0.74
sigv
0.0390 4.83
0.0391 4.8
sig1
0.0436 3.25
0.0434 3.25
Adj R-
squared 0.7521
0.7508
0.7565
0.7551
N 201
201
201
201
--------------------------------------------------------------------------------------------------------------------------------------------------------
Table 2.4: Long Run Elasticities of Exogenous Variables
--------------------------------------------------------------
auto
-0.039
autoblack
-0.031
unemp
0.116
fhh
0.347
lpci
-1.285
-------------------------------------------------------
52
CHAPTER 3
The Effect of Automobile Production on the Growth of Non-Farm Proprietor Densities in
Alabama?s Counties
1. Introduction
Entrepreneurship is a key catalyst for economic growth and regional development. State and
local policymakers are allocating considerable resources to promote entrepreneurship. In the
United States, the number of full and part time non-farm self employed, or proprietors, grew by
around 300% or from 9.6 million in 1969 to 29.2 million in 2004. In comparison, the number of
full and part time wage and salary workers grew by only 77% or from 78.8 million in 1969 to
138.8 million workers in 2004. The ratio of self to wage and salary employment nearly doubled,
from 0.12 to 0.21, over this period (Goetz and Rupasingha, 2009).
In 2006, nonfarm proprietor employment accounted for 18.8 percent of total nonfarm
employment in United States. In Alabama, this percent was 17.8, and ranged from 10.4 percent
to 43.3 percent. Microenterprise employment represented 17.7 percent of U.S nonfarm
employment and 16.7 percent of Alabama nonfarm employment. Within Alabama, this ranged
from 12.6 percent to 30.5 percent (RUPRI, 2007). Over the past two decades the focus of
economic development policy has shifted more heavily toward entrepreneurship. This increased
interest in the entrepreneur?s role in the economy has led to a growing body of research
attempting to identify the factors that promote entrepreneurship. Most applied economic research
53
on entrepreneurship uses the number of nonfarm self-employed individuals as a share of the
labor force as a measure of entrepreneurship.
Figure 3.1:
The greatest spillover benefit of automobile plants in Alabama is the movement of input
suppliers and supporting services to Alabama counties. These firms cluster around automobile
plants. Clusters are characterized by a focus on one particular industrial activity and the fact that
many small firms specialize in different phases of the production process (OECD, 1996).
Clusters enhance the competitiveness of established small businesses and thereby influencing the
survival rate of these businesses. Clustering thus can have an impact on the level of
entrepreneurship through both entry and exit. Automobile production in Alabama helped spur
the formation of new businesses and increased the growth of existing firms.
54
This paper studies the impact of automobile production on the ratio of non farm
proprietorships to all full and part time workers. This study also examines how county level
economic, social variables and county level spillover effects influence rates of non-farm
proprietorships density. In this study, non-farm proprietorships density and per capita income
are considered to be interdependent.
2. Literature Review
Entrepreneurship is important because the competitive behavior of entrepreneurs drives the
market process and leads to economic progress (Kirzner 1973). From society?s perspective, the
profits earned by entrepreneurs represent gains to society as a whole. Entrepreneurs deal with
uncertainty about the future, not with risk. Probabilities can be estimated for risky activities and
thus are insurable. Since entrepreneurs are dealing with uncertainty about the profitability of
their new combinations of resources, entrepreneurs cannot insure against the probability that new
goods and services will not be liked. Entrepreneurs bear the burden of the uncertainty associated
with the market process (Cantillon et al.1921). Berkowitz and DeJong (2005) find a strong
relationship between economic growth and the rate of entrepreneurial activity within a country
over the years. Kreft and Sobel (2005) find the same relationship across U.S. states. Henderson
(2002) finds it to hold at the local level within the United States.
Most studies of entrepreneurship examine the factors that influence an individual?s
choice between wage employment or self employment. One factor that influences an
individual?s decision to become an entrepreneur is the availability of funding. Homeownership
and housing values significantly improve prospective entrepreneur?s ability to borrow capital to
initiate new business because homes can be used a major source of loan collateral (Robson
55
1998a,b). The amount of dollars deposited per capita in local bank can be used as a proxy for
availability of capital even though proprietors have access to national credit markets to borrow
capital, (Malecki ,1994).
Countries that experience rapid population and work force growth have a growing share
of self-employed people in the work force, whereas countries experiencing low population
growth have a diminishing share of entrepreneurs in the labor force (ILO, 1990). Population
growth may lower wages through increasing the labor supply. However, population growth will
also create a future increase in the demand for goods and services. Expectations of potential
entrepreneurs of future entrepreneurial opportunities are likely to stimulate start-ups (Reynolds,
Hay and Camp, 1999).
High population density in urban areas may be an important reason for the existence of
small businesses in urban areas and the startup of new businesses (Reynolds et al., 1994 and
Storey, 1994). The age structure of the population may have direct and indirect impact on the
level of entrepreneurship. Evans and Leighton (1989a) found that many entrepreneurs start a
business in their mid-thirties and that the average age of an entrepreneur is over 40 years.
Goetz and Freshwater (2001) in their study conclude that Individuals with more
education are more likely to become entrepreneurs. Bates (1993) found that educated and skilled
potential entrepreneurs are highly sensitive to the opportunity costs of self employment because
they need to sacrifice high wage positions as employees. Self employment rates increase with
age, because of greater experience levels and potential age discrimination in the labor market
(Evans and Leghton 1989b).
56
Several studies have examined the relationship between ethnic diversity and economic
development. Alesina and La Ferrara (2000) find that involvement in associational activities is
significantly lower in ethnically fragmented localities. Rupasingha et al. (2006) in their study on
social capital found that ethnically fragmented societies have less social capital. Social
interaction among local entrepreneurs is important for sustaining and enhancing local
entrepreneurship. Greater diversity may lead to diversified consumer demand patterns leading to
specialization among firms and niche markets. Females are less likely than males to be self
employed. Parker (1996) found that the proportion of time allocated by an individual to self-
employment is inversely related to the riskiness of returns to self-employment and the degree of
risk aversion.
Per capita income also reflects aggregate demand in an economy (Robson 1998b). Large
aggregate demand in a given county attracts big firms to migrate in or gives incentives to expand
the existing firms. This may also work to deter small business firms from expanding and new
small entrepreneur from starting. The impact of economic growth on the level of
entrepreneurship is however ambiguous. It appears that economic growth can either have a
positive or a negative impact on the level of entrepreneurship, depending on the stage of
economic development.
Various studies found that economic development is associated with a decrease in the
self-employment rate (Kuznetz, 1966; Schultz, 1990; Bregger, 1996). Several arguments have
been given to support the theory of negative impact of economic growth on the level of self-
employment (Carree, Van Stel, Thurik and Wennekers, 2002). Economic development is
accompanied by an increase in wage levels. Higher real wages increase the opportunity costs of
self-employment and this makes wage employment more attractive (EIM/ENSR, 1996).
57
Marginal entrepreneurs may be induced to become employees (Lucas, 1978). At the macro level
a high rate of unemployment can negatively impact the level of entrepreneurship because of the
decline in the availability of business opportunities induced by a depressed economy. Moreover,
the failure rate of established businesses rises because of low revenues (EIM/ENSR, 1996).
The impact of taxes on the level of entrepreneurship is complex and inconsistent. In
OECD (1998) it is argued that high tax rates reduce the returns on entrepreneurship and can deter
the start-up of new firms and expansion of established firms. On the other hand, it has been
hypothesized that self-employment offers better opportunities to avoid tax liabilities than wage-
employment (Parker, 1996). Amenities and rural/urban status of a county may also affect the
density of proprietorship in a given county. Employment shares by industry influence
proprietorship growth in a given county (Malecki 1994; Armington and Acs 2002).
3. Model
In this study, non-farm proprietorships density and per capita income are considered to be
simultaneously related to each other. Non-farm proprietor densities in a given county are
influenced by returns from self-employment and wage employment and self-employment risk,
socio, economic, demographic, regional, and government policy variables and spatial
components of non-farm proprietor densities and per capita income of neighboring counties.
County-level aggregates are used as a proxy for the characteristics of the pool of individuals
from which entrepreneurs potentially emerge, and the local market conditions facing the self-
employed. The basic specification of the model is a simultaneous-equation system of the form:
58
(1)
(2)
The equilibrium levels of proprietorship density and per capita income are assumed to be
functions of the equilibrium values of the endogenous variable included in right hand side of
equation and their spatial lags, automobile production and the vectors of the additional
exogenous variables. Where, and are vectors of dimension NT x 1 of the
equilibrium levels of proprietorship density and per capita income respectively; t denotes time. I
is an identity matrix of dimension T and, W is a row standardized N x N spatial weights matrix
with zero diagonal values. Each element of this spatial weights matrix, , represents a
measure of proximity between observation i and observation j. Based on the queen based
adjacency criteria, is equal to 1/ki, where ki is the numbers of nonzero elements in row i, if i
and j are adjacent, and zero otherwise. Therefore, and stands for
the equilibrium values of neighboring counties? effect. At-i is vector of dimension NT x 1 of
automobile production. BAt-i is the interaction term of the distressed black belt county and
automobile production. The matrices of additional exogenous variables that are included in the
proprietorship density and per capita income equations are given by and respectively.
Where i is 7 years in both equations. These additional exogenous variables are included in the
equations to control their effects on the dependent variables. This controlling makes estimates
on the relationship between the variables we are interested in more precise. A multiplicative
functional form was used for the equations in this system. A lagged adjustment is introduced into
our model. This partial-adjustment process replaced unobservable equilibrium which allowed the
model to take the general form as follows:
59
(3)
+
(4)
Where , , , , , , for k=1,?., Kr ; r, l = 1,2; and q= 1,2 are the
parameter estimates of the model and Kr is the number of exogenous variables in the respective
equations. and represent the log differences between the end and beginning
period values of proprietorship density and per capita income respectively. Then, they represent
the growth rates of the respective variables. The variable, automobile production (lnAt-i), was
constructed as ln (automobile production/ distance). The subscript t-i denotes to the variable
lagged 7 years for study period 1970-2007 and for r =1,2 are the speed of adjustment
coefficients, the rate at which proprietorship density and per capita income adjust to their
respective steady state equilibrium levels. ut,r for r=1,2 are NT x 1 vectors of disturbances. A
Moran?s I test statistic suggested that there is the existence of spatial autocorrelation in the
errors. The test results are given in Table 3. Therefore, the disturbance vector in the rth equation
is generated as:
= + r =1,2 (5)
60
This specification relates the disturbance vector in the rth equation to its own spatial lag.
A one-way error component structure was utilized to allow the innovations ( to be
correlated over time, following Baltagi (1995). Therefore, the innovation in the rth equation is
given by
, r = 1,2 (6)
Where , , = (
IT and IN are identity matrices of dimension T and N, respectively, is a vector of ones of
dimension T, and denotes the Kronecker product. and are random vectors with zero
means and covariance matrix ( suppressing the time index):
(7)
Where, denotes the vector of unit specific error components and contains the
error components that vary over both the cross- sectional units and time periods. The innovations
are not spatially correlated across units but they are auto-correlated over time. However,
this specification allows innovations from the same cross sectional unit to be correlated across
equations. Therefore, the vectors of disturbances are spatially correlated across units and across
equations as given in (8) as was used by Kapoor, Kelejian, and Prucha (2007); Baltagi, Song, and
Koh (2003)).
= + , r = 1,2 (8)
61
The intercepts ( ) in equations (9) ?(10) represent the combined
influences of changes in the suppressed exogenous variables; the for r = 1,2 coefficients are
structural elasticities corresponding to the endogenous variables; and the for r = 1 coefficients
are structural elasticities corresponding to automobile production. We add the interaction terms
to test whether the automobile production boom differentially affected the growth rate of
proprietorship density in the distressed Black Belt counties. We incorporate spatial components
to capture the role of proprietorship density and per capita income of neighboring counties
3.1 Reduced Form Estimates and Long run Elasticity
The reduced form equations are obtained by solving structural equations derived from
Generalized Spatial Three Stage Least Square (GS3SLS) model. A spatial autoregressive model,
in the context of single equation and in panel data setting, is expressed as:
(9)
(10)
Where y is an NT x 1 vector of observations on the dependent variable. Wy is the
corresponding spatial lagged dependent variable for weights matrix W, X is NT X K matrix of
observations on the explanatory variables, u is an NT X 1 vector of error terms. is the spatial
autoregressive parameter and is a K X 1 vector of regression coefficients. is the spatial
autoregressive coefficient for the error lag and is NT x 1 vector of innovations or white
noise error. This single spatial autoregressive model can be extended to a system of spatially
interrelated equations. A standard G system of equations can be written as:
62
(11)
(12)
Y = y1, ??, y G X = x1,??.., x G U = u1,??., u G
WU = Wu1,?.., Wu G
Where yr is the NT x 1 vector of observations on the dependent variable in rth equation ,
xk is the NT x 1 vector of observations on the kth exogenous variable, ur is the NTx1 vector of
error terms in the rth equation, and B and are parameter matrices of dimension G x G and K x
G respectively. B is a diagonal matrix. W is N x N weights matrix of known constants and is
G x G matrix of parameters. Wyr and Wur are spatial lag and spatial autoregressive error term in
the rth equation respectively. The solution for the endogenous variable can be revealed through
the vector transformation:
Letting y = vec Y , x = vec X, u= vec U and
or
or
63
After all the spatial effects and the other endogenous variables effects are controlled, the
relations between the dependent variables and the exogenous variables x can be expressed as:
The reduced form can be written as:
+u) (13)
(14)
The system of spatial structural equations was solved to obtain a system of reduced form
equations. Since spatial weight matrix was constructed on the queen based adjacency criteria,
this system of spatial equations control spatial spillover effects of neighboring counties (Nzaku
and Bukenya, 2005; Trendle, 2009; Gebremariam,2010). Reduced coefficients of significant
variables in the structural equations were estimated for the counties where automobile plants
locate and for its neighboring counties. The long run elasticity of automobile production and
other exogenous variables in the per capita income and nonfarm proprietor density of these
counties was calculated from these reduced form coefficients.
In this system of equations, the dependent variables are the change in per capita income
and nonfarm proprietor density during the specific period. Exogenous variables are the initial
value of those variables at the beginning of specific period. The dependent variables are
constructed as the difference between the lnyt ? lnyt-i. One of the exogenous variables in the right
hand side of each equation is the predetermined lagged dependent variable (lnyt-i). The long run
64
elasticity of exogenous variable is calculated by dividing the coefficient of exogenous variable
by negative value of the coefficient of the predetermined lagged dependent variable (lnyt-i).
4. Data and Sources
Data for sixty seven counties in Alabama are drawn from several sources (Table 3.1). These data
were collected for study periods which are from 1970 to 2007. In this study, the non-farm
proprietorship density is constructed as the ratio of non farm proprietorship to total employment.
The growth of non-farm proprietor density and per capita income are constructed using 7 years
interval between the beginning and end period, like 1970-1977, 1980-1987, 1990-1997 and
2000-2007. Independent variables include demographic, human capital, labor market,
automobile production, interaction term of automobile production and distressed black belt
county and policy variables. The initial values of the independent variables are used as 7 year
lagged values. This formulation reduces the problem of endogeneity. All independent variables
are in log form except those that can take negative or zero values. The initial non-farm
proprietorship density and per capita income are included in this model to control for the relative
size of the existing proprietor base and per capita income in the county and to test for conditional
convergence with their respective endogenous variable. The descriptive statistics of the
variables are given in Table 3.2.
5. Estimation Issues
Panel models can be used to control unobserved heterogenity and to investigate inter-temporal
changes. Since panel data provides more information and variables, the degree of freedom and
efficiency increases and multicollinearity is less likely to occur. For this study, a panel model
65
was estimated containing three time periods for 67 counties. A total of 268 observations are
used in the panel model. Following Baltagi (1995), one way error component structure model
was utilized for the panel data in this study.
This system of equations has econometric issues regarding feedback simultaneity, spatial
autoregressive lag, and spatial cross-regressive lag simultaneity with spatially autoregressive
disturbances. These simultaneities create problems in estimation and identification of each
equation. The order condition for identification in a linear simultaneous equations model is that
the number of dependent variables on the right hand side of an equation must be less than or
equal to the number of predetermined variables in the model but not in the particular equation.
Lagged dependent variables also can be considered as predetermined variables. Kelejian and
Prucha considered that the spatially lagged dependent variables can be treated as predetermined
(Kelejian and Prucha, 2004). The order condition for each equation of the system in (3) ? (4) is
fulfilled.
A Hausman test (1983) for over identification was done to investigate whether the
additional instruments are valid in the sense that they are uncorrelated with the error term. That
is E(Q?ur) =0, Where E is the expectation operator and Q is an instrument matrix that consist of
a subset of linearly independent columns X, WX, W2X, where X is the matrix that includes the
control variables in the model. All equations are appropriately identified because the hypothesis
of orthogonality for each equation cannot be rejected even at P= 0.05 as indicated by the
test statistic in Table 3.3.
When the spatial autoregressive lag and spatial cross-regressive lag simultaneities are
present, the conventional three-stage least squares estimation to handle the feedback simultaneity
would be inappropriate. Therefore, the Method of Moments approach was used rather than
66
maximum likelihood because maximum likelihood would involve significant computational
complexity. Generalized Spatial Three-Stage Least squares (GS3SLS) approach outlined by
Kelejian and Prucha (2004) into a panel data setting was used to estimate the model. This new
procedure is performed in a five-step routine as given in Appendix .
6. Results and Discussions
The parameter estimates of the system are given in Table 3.3. In general, the results are
consistent with theoretical expectations and previous studies. In the model, the negative and
significant coefficient of the lagged dependent variable in each equation indicates the conditional
convergence with respect to the respective endogenous variable of each equation. The results
show the existence of simultaneities between growth of proprietorship density and per capita
income growth. This indicates that there is strong interdependence between growth of
proprietorship density and per capita income growth. The signs of the coefficients are consistent
with theoretical expectations. The reduced coefficient and long run elasticities of significant
variables in the structural equations of the system were calculated and given in Table 3.4.
6.1 Proprietorship Density Growth Equation
In the equation for growth of proprietorship density, the per capita income growth is negatively
and highly associated with the growth of proprietorship density. Several studies have found the
negative impact of economic growth on the level of self-employment (Carree, Van Stel, Thurik
and Wennekers, 2002). The structural coefficient of the variable automobile production is
positive and significant at 5% level. The reduced coefficient and long run elasticity suggests that
automobile production of a plant positively influences the proprietorship density of a county
67
where a plant locates and on its neighboring counties. Long run elasticity of automobile
production indicates that if automobile production of a given plant can increase by 10%, the
proprietorship density of a county where the given plant locates will increase by 0.6 % and the
proprietorship density of neighboring counties will increase by the range of 0.011% - 0.021%.
The structural coefficient of the interaction term of automobile production and distressed Black
Belt County is significant at 10% level there might be differential impact of automobile
production on proprietorship density between distressed black counties and other counties. Long
run elasticity of interaction term suggests that proprietorship density of distressed Black Belt
Counties may rise by about 0.04 % for a 10% increase in automobile production.
The negative structural and reduced coefficients of unemployment rate equations indicate
that the proprietorship density in a given county is negatively associated with unemployment
rate. The long run elasticity of unemployment rate (-0.15) suggests that a 10 % increase in
unemployment rate will decrease the proprietorship density by 0.15% in long run. This result is
consistent with many research studies related to proprietorship density. The proprietorship
density in a given county is positively associated with the percentage of high school degree and
higher education. The long run elasticity of the percentage of high school degrees and above
(0.924) indicates that a 10% increase in the percentage of the percentage of high school degrees
and higher education in a given country is associated with 0.92% increase in the proprietorship
density in the given county. But the long run elasticity of bachelor degrees of above (-0.437)
implies that the proprietorship density in a given county is negatively associated with the
percentage of bachelor degrees and above.
The coefficient of the spatial lag of endogenous variables is significant. This indicates
the presence of spatial autoregressive lag effect in this study period. This means that the growth
68
of proprietorship density in neighboring counties has positive spillover effects on the growth of
proprietorship density in a given county. Global Moran?s I statistic and indicate there is a
spatial spillover effect with respect to the error terms in this study period. This indicates that
random shocks originated in a given county will affect its neighbors.
6.2 Per Capita Income Growth Equation
The reduced coefficient and long run elasticity suggests that automobile production of a plant
positively influences the per capita income of the county where a plant locates and of its
neighboring counties. The long run elasticity of automobile production indicates that if
automobile production of a given plant can increase by 10%, the per capita income of a county
where the given plant locates will increase by 0.12 % and the per capita income of neighboring
counties will increase by the range of 0.018% - 0.037%. The structural coefficient of the
interaction term of automobile production and distressed Black Belt County is significant at 5%
level. There is a differential impact of automobile production on per capita income in distressed
Black Belt counties. The long run elasticity of a interaction term suggests that per capita income
of a distressed Black Belt County may rise by about 0.03 % for a 10% increase in automobile
production.
The reduced coefficient and long run elasticity of the percentage of employed labor in
farming suggest that a large dependence on employment in farming negatively influences the per
capita income of a county. The per capita income in a given county is negatively associated with
the initial level of unemployment rate and per capita property tax and positively associated with
initial level of percentage of high school degree or higher education. These results are consistent
with previous research (Nzaku and Bukenya, 2005).
69
The results show the existence of spatial autoregressive lag effects and spatial cross-
regressive lag effects with respect to endogenous variables. These results imply that the per
capita income growth of a particular county is depend on the average growth of proprietorship
density and per capita income of neighboring counties. This is important from policy
perspectives because the per capita income depend not only on the characteristics of that county,
but also on the characteristics of its neighbors. The disturbances from the equation indicate the
existence of spatial dependencies in the error terms. This means that random shocks to the
system affect not only a county where the shock originates and its neighbors, but also the entire
study area.
7. Conclusions and Policy Implications
The empirical findings suggest that automobile production in Alabama significantly increases
nonfarm proprietorship in all counties. A appropriate policies to lure industrial development are
thus very important to increase self employment opportunities. There is significant spatial lag
effects and spatial error effect between non-farm proprietor densities and per capita income. This
interdependence provides the need of economic development policy coordination among the
counties.
70
Table 3.1: Variable Description and Data Sources
-------------------------------------------------------------------------------------------------------------------------------------------
Variable variable Description unit Source
-------------------------------------------------------------------------------------------------------------------------------------------
DPRO Growth rate of non-farm proprietor density %
A,B
DP Growth rate of per capita income, t %
A,B
lpci per capita income, t-7
$/person
B
lpro non-farm proprietor density,t-7 nonfarm proprietor/ employment A,B
lpop population, t-7
number
B
lemp employment, t-7
number
B
unemp unemployment rate, t-7 years
%
E
auto No. of automobile/distance, t-7 years number/mile
A,J,K
autoblack Interaction of auto and Black Belt county
d17years % of population below 17years, t-7 %
C,D
d16years % of population above 65years, t-7 %
C,D
hsch % of high school degree or above,t-7 %
C,D
bach % of bachelor degree or above,t-7 %
C,D
farm % employed in farming, t-7
%
B
manu %employed in manufacturing, t-7 %
B
serv %employed in services, t-7
%
B
tax per capita local tax, t-7
$/person
D
protax per capita property tax, t-7
$/person
D
anfpin average non-farm proprietor?s income,t-7 $/person
B
awas average wage and salary,t-7
$/person
B
popden population density, t-7
number/square mile A,B
nonwhite % of nonwhite, t-7 years
%
D
owner owner occupied housing in percent, t-7 %
D
dista distance from metro area
mile
J
amenity Natural Amenities Index, t-7
ERS index
H
hway road density, t-7
mile/square mile
I
metro dummy variable for metro area
cv coefficient of variation of anfpin, t-7
A,B
female female labor participation,t-7
%
D
bdep Bank deposits,t-7
$
D
mvh median housing value,t-7
$
D
lpov poverty rate, t-7
%
D
(I?W)DPRO spatial lag of DPRO
%
A,B
(I?W)DPCI spatial lag of DPCI
%
A,B
-------------------------------------------------------------------------------------------------------------------------------------------
A- Computed, B- US Department of Commerce, Bureau of Economic Analysis (REIS database), C- County &
City Data Book, D- U.S Census Bureau, E- Bureau of Labor Statistics, F- American Medical Association,
G-Federal Bureau of Investigation, H- Economic Research Service, USDA, I ? US Bureau of Transportation
Statistics, J- Map Quest, K - Mercedes-Benz U.S. International, Tuscaloosa, AL, Honda Manufacturing of
Alabama, Lincoln, AL, Hyundai Motor Manufacturing Alabama, Montgomery, AL, Toyota Motor Manufacturing
Alabama, Huntsville, AL, Automotive News Market Data Book
71
Table 3.2: Descriptive Statistics for Alabama Counties, 1970-2007
------------------------------------------------------------------------------------------------------------------------------
Variable Variable Description Mean Stdev
-------------------------------------------------------------------------------------------------------------------------------
DPRO
Growth rate of non-farm proprietor density
1.07 0.16
DP
Growth rate of per capita income, t
1.14 0.10
lpci
per capita income, t-7
18225.76 4978.40
lpro
non-farm proprietor density,t-7
0.20 0.06
lpop
population, t-7
59149.84 93442.84
lemp
employment, t-7
28441.48 55516.72
unemp
unemployment rate, t-7 years
8.80 4.93
auto
No. of automobile/distance, t-7 years
494.70 4892.61
autoblack
Interaction of auto and Black Belt county
55.24 283.85
d17years
% of population below 17years, t-7
30.03 4.99
d16years
% of population above 65years, t-7
12.74 2.53
hsch
% of high school degree or above,t-7
53.37 14.97
bach
% of bachelor degree or above,t-7
9.95 5.61
farm
% employed in farming, t-7
9.03 6.69
manu
%employed in manufacturing, t-7
25.40 10.42
serv
%employed in services, t-7
16.98 5.87
tax
per capita local tax, t-7
208.02 181.56
protax
per capita property tax, t-7
81.21 80.50
anfpin
average non-farm proprietor?s income,t-7
11312.53 4988.92
awas
average wage and salary,t-7
14314.31 8104.24
popden
population density, t-7
71.93 86.59
nonwhite
% of nonwhite, t-7 years
29.14 20.97
owner
owner occupied housing in percent, t-7
72.97 7.10
dista
distance from metro area
34.72 25.18
amenity
Natural Amenities Index, t-7
1.87 1.79
hway
road density, t-7
0.13 0.03
metro
dummy variable for metro area
1.31 0.66
cv
coefficient of variation
0.16 0.09
female
female labor participation,t-7
42.30 4.38
bdep
Bank deposits,t-7
594538.77 1382885.90
mvh
median housing value,t-7
66720.39 20393.22
pov
poverty rate, t-7
23.09 10.33
(I?W)DPRO spatial lag of DPRO
1.05 0.07
(I?W)DPCI spatial lag of DPCI
1.12 0.04
-------------------------------------------------------------------------------------------------------------
72
Table 3.3. Structural Coefficients
-----------------------------------------------------------------------------------------------------------------------------------
DPCI Equation DPRO Equation
------------------------------------ ------------------------------------
Variable Coeff. z-stat Coeff. z-test
------------------------------------------------------------------------------------------------------------------------------------
DPCI
-0.698 -4.02
DPRO -0.221 -4.7
lpci -0.247 -4.42
-0.142 -1.26
lpop 0.027 0.74
0.075 0.95
ltem -0.024 -0.82
-0.070 -1.03
lpro
-0.174 -3.53
auto 0.006 2.46
0.014 3.04
autoblack 0.008 2.56
0.009 1.6
unemp -0.062 -5.83
-0.069 -3.17
hsch 0.121 2.25
0.244 2.1
bach -0.019 -0.93
-0.089 -2.22
farm -0.025 -2.42
-0.013 -0.6
manu 0.000 0.01
-0.029 -1.35
service -0.019 -0.89
-0.025 -0.64
anfpin -0.026 -1.34
0.034 0.98
popden -0.018 -0.89
-0.021 -0.54
nonwhite -0.003 -0.43
-0.018 -1.14
tax 0.067 3.6
0.027 0.73
protax -0.037 -2.93
-0.033 -1.28
d17years 0.041 0.62
0.151 1.21
d65years 0.005 0.14
-0.004 -0.05
dista -0.009 -0.86
-0.027 -1.47
amenity -0.004 -0.84
0.001 0.14
hway 0.020 1.15
0.030 0.94
metro -0.015 -0.41
-0.083 -1.3
pov 0.004 0.22
awas
8.780 1.63
owner
-0.078 -0.62
cv
-0.004 -0.23
female
0.056 0.43
bdep
-0.021 -0.88
mvh
-0.041 -0.54
(I ? W) DPCI 0.336 3.51
0.254 1.38
(I ? W) DPRO 0.147 2.41
0.493 4.54
const 2.252 4.03
0.292 0.22
Rho -0.249 -3.26b
-0.338 -2.89b
sigv 0.003 17.58b
0.010 9.68b
sig1 0.003 10.16b
0.010 5.66b
NR2 ?
X2(50,45) 47.2 0.59c
40.3 0.67c
Moran I 0.144 0.03
0.150 0.02
N 268
268
----------------------------------------------------------------------------------------------------------------------------------------
b: t- static value, c: p - value
73
Table 3.4: Reduced Coefficients, Long Run Elasticities and 10% Impact of Automobile Production
-------------------------------------------------------------------------------------------------------------------------------------------------------
Reduced Coefficient Long Run Elasticity 10% impact
---------------------------------------- ------------------------------------ ----------------------------
County ?PCI ?PRO PCI PRO PCI PRO
-------------------------------------------------------------------------------------------------------------------------------------------------------
Toyota
Jackson
0.0015
0.0007
0.0058
0.0034
0.0365
0.0213
Limestone 0.0012
0.0006
0.0044
0.0027
0.0302
0.0185
Madison
0.0032
0.0126
0.0121
0.0593
0.1208
0.5928
Marshall
0.0007
0.0004
0.0027
0.0021
0.0175
0.0133
Morgan
0.0010
0.0005
0.0037
0.0026
0.0235
0.0164
Hyundai
Autauga
0.0010
0.0005
0.0036
0.0024
0.0265
0.0177
Bullock
0.0015
0.0013
0.0055
0.0063
0.0345
0.0391
Crenshaw 0.0008
0.0004
0.0030
0.0020
0.0182
0.0123
Elmore
0.0008
0.0004
0.0031
0.0021
0.0214
0.0146
Lowndes
0.0012
0.0011
0.0046
0.0053
0.0317
0.0359
Macon
0.0012
0.0012
0.0047
0.0053
0.0293
0.0332
Montgomery 0.0031
0.0125
0.0120
0.0592
0.1200
0.5923
Pike
0.0008
0.0004
0.0031
0.0021
0.0187
0.0129
Honda
Calhoun
0.0010
0.0005
0.0037
0.0025
0.0251
0.0173
Clay
0.0010
0.0005
0.0037
0.0025
0.0251
0.0173
Cleburne 0.0010
0.0006
0.0037
0.0026
0.0240
0.0168
Coosa
0.0008
0.0004
0.0031
0.0021
0.0192
0.0132
St. Clair
0.0008
0.0004
0.0031
0.0021
0.0195
0.0134
Shelby
0.0008
0.0004
0.0030
0.0020
0.0192
0.0129
Talladega 0.0032
0.0125
0.0120
0.0592
0.1203
0.5925
Mercedes -
Benz
Bibb
0.0008
0.0004
0.0031
0.0021
0.0198
0.0136
Fayette
0.0010
0.0005
0.0037
0.0025
0.0229
0.0156
Greene
0.0015
0.0013
0.0056
0.0064
0.0358
0.0411
Hale
0.0015
0.0013
0.0055
0.0063
0.0352
0.0399
Jefferson 0.0008
0.0004
0.0030
0.0020
0.0179
0.0119
Pickens
0.0015
0.0014
0.0057
0.0066
0.0370
0.0426
Tuscaloosa 0.0032
0.0125
0.0120
0.0592
0.1203
0.5925
Walker
0.0007
0.0004
0.0026
0.0018
0.0159
0.0110
------------------------------------------------------------------------------------------------------------------------------------------------------
74
Table 3.5: Reduced Coefficients and Long Run Elasticities of Exogenous Variables
------------------------------------------------------------------------------------------------------------------------------------------------
Reduced Coefficient Long Run Elasticity
---------------------------------------------- -----------------------------------
Variable county ?DPCI ?DPRO PCI PRO
-------------------------------------------------------------------------------------------------------------------------------------------------
Bullock
0.0005
0.0008
0.0018
0.0038
autoblack
Greene
0.0005
0.0008
0.0019
0.0038
Hale
0.0005
0.0008
0.0018
0.0038
Lowndes
0.0004
0.0007
0.0015
0.0032
Macon
0.0004
0.0007
0.0016
0.0032
Pickens
0.0005
0.0008
0.0019
0.0039
Madison
-0.0579
-0.0324
-0.2186
-0.1521
unemp
Montgomery -0.0573
-0.0320
-0.2184
-0.1514
Talladega
-0.0574
-0.0321
-0.2185
-0.1517
Tuscaloosa -0.0574
-0.0321
-0.2185
-0.1516
Madison
-0.0266
0.0043
-0.1005
0.0203
farm
Montgomery -0.0264
0.0044
-0.1005
0.0206
Talladega
-0.0264
0.0043
-0.1005
0.0205
Tuscaloosa -0.0264
0.0043
-0.1005
0.0205
Madison
0.0746
-0.0229
0.2817
-0.1073
tax
Montgomery 0.0739
-0.0228
0.2817
-0.1079
Talladega
0.0740
-0.0228
0.2817
-0.1077
Tuscaloosa 0.0740
-0.0228
0.2817
-0.1077
Madison
-0.0361
-0.0095
-0.1364
-0.0445
protax
Montgomery -0.0358
-0.0093
-0.1364
-0.0441
Talladega
-0.0358
-0.0094
-0.1364
-0.0443
Tuscaloosa -0.0358
-0.0094
-0.1364
-0.0442
Madison
0.0844
0.1974
0.3189
0.9251
hsch
Montgomery 0.0834
0.1954
0.3178
0.9239
Talladega
0.0836
0.1958
0.3182
0.9243
Tuscaloosa 0.0836
0.1957
0.3181
0.9243
Madison
0.0004
-0.0932
0.0015
-0.4368
bach
Montgomery 0.0005
-0.0924
0.0021
-0.4367
Talladega
0.0005
-0.0925
0.0019
-0.4367
Tuscaloosa 0.0005
-0.0925
0.0019
-0.4367
-----------------------------------------------------------------------------------------------------------------------------------------------------
75
References
Ahn, YoungKyoo. 2005. ? Alabama?s Automotive Industry and Direct Foreign Investment?.
http://www.umich.edu/~econdev/alabama auto/ index.html (accessed 20 June 2008).
Alabama Automotive Manufacturers Association. 2008. ? Alabama?s Automotive Manufacturing
Industry Continues to Grow?. Press Release. February 1.
Alesina, A., & La Ferrara, E. (2000). Participation in heterogeneous communities. Quarterly
Journal of Economics, 115, 847?904.
Armington, C., & Acs, Z. (2002). The determinants of regional variation in new firm formation.
Regional Studies, 36, 33?45.
Baltagi, Badi. H, 1995, Econometric Analysis of Panel Data, New York: John Wiley and Sons.
Baltagi, Badi H., Seuck Heun Song, and Won Koh, 2003, ?Testing Panel Data Regression
Models with Spatial Error Correlation,? Journal of Econometrics, 117: 123-150.
Barkley DL, Henry MS, Bao S (1998) The Role of Local School quality and Rural Employment
and Population Growth. Rev Reg Stud 28(1): 81-102.
Bates, T. (1993). Theories of entrepreneurship. In R. D. Bingham & R. Mier (Eds.), Theories of
local economic development. Newbury Park: Sage Press.
Berkowitz, Daniel, and David N. DeJong. 2005. ?Entrepreneurship and Post-socialist Growth.?
Oxford Bulletin of Economics and Statistics, 67(1): 25-46.
Boarnet MG (1994) An Empirical Model of Intra-Metropolitan Population and Employment
Growth. Pap Reg Sci 73(2): 135-153.
Bregger, J.E., 1996, Measuring self-employment in the United States, Monthly Labor Review,
January/February, 3-9.
Buchholz, David E. 2008 ? Business Incentive Reform Case Study: Mercedes ? Benz: The Deal
of the Century.? Corporation for Enterprise Development: Washington, D.C.
http://www.cfed.org/focus.m?parentid=34siteid=1629&id=1653 (accessed 20 June 2008).
Cantillon, R. 1931. Essai sur la nature du commerce. Edited by H. Higgs. London: Macmillan,
for The Royal Economic Society.
Carlino, Gerald A., and Edwin S. Mills. 1987. ? The Determinants of County Growth,? Journal
of Regional Science, 27, 39-54.
Carree, Martin, Andr? van Stel, Roy Thurik and Sander Wennekers (2002), ?Economic
development and business ownership: an analysis using data of 23 OECD countries in the period
76
1976-1996,? Small Business Economics 19 (3), 271-290.
Carruthers, John I. and Gordon F. Mulligan. 2007. ? Land Absorption in U.S. Metropolitan
Areas: Estimates and Projections from Regional Adjustment Models.? Geographical Analysis.
39,pp. 78-104.
Clark, D. E. and C. A. Murphy. 1996. ?Countywide Employment and Population Growth: An
Analysis of the 1980s,? Journal of Regional Science, 36, 235?256.
Davis, Elizabeth E. and Bruce A. Weber, 2002. ?How Much Does Local Job Growth Improve
Employment Outcomes of the Rural Working Poor?? The Review of Regional Studies 32(2), 255-
274.
Deller, Steven C., Tsung-Hsiu Tsai, David W.Marcouiller, and Donald B.K. English. 2001. ?
The Role of Amenities and Quality of Life in Rural Economic Growth,? American Journal of
Agricultural Economics, 83, 352-365.
DeNavas-Walt, C., D.P. Bernadette., and J. Smith. 2007. Income, Poverty, and Health Insurance
Coverage in the United States: 2006 U.S. Census Bureau, Current Population Reports.
Washington DC: U.S. Government Printing Office, pp. 60?233.
Duffy-Deno, Kevin T.1998. ?The Effect of Federal Wilderness on County Growth in the
Intermountain Western United States,? Journal of Regional Science, 38, 109-136.
Economic Development Partnership of Alabama. ? Alabama Automotive Industry Profile.?
http://www.edpa.org/docs/automotive-industry-profile.pdf (accessed 28 June 2008).
Edmiston KD (2004) The Net Effect of Large Plant Locations and expansions on county
employment, J Reg Sci 44(2):289-319.
EIM/ENSR, 1996, The European Observatory for SMEs; fourth annual report, EIM Business
and Policy Research, Zoetermeer.
English, D.B.K., Marcouiller, D.W. and Cordell, H.K. 2000. Linking Local Amenities with Rural
Tourism Incidence: Estimates and Effects. Society and Natural Resource 13, 185-202.
Evans, David S. and Linda S. Leighton. 1989. ?Some Empirical Aspects of Entrepreneurship.?
American Economic Review, 79(3): 519-35.
Evans, David S., and Boyan Jovanovic. 1989. ?An Estimated Model of Entrepreneurial Choice
under Liquidity Constraints.? Journal of Political Economy, 97(4): 808-27.
Evans, A.W.1990. ?The Assumption of Equilibrium in the Analysis of Migration and
Interregional Differences: A Review of Some Recent Research.? Journal of Regional Science 30,
515-31.
77
Fan, S., L. Zhang, and X. Zhang. 2002. ?Growth and Poverty in Rural China: The Role of Public
Investment.? International Food Policy Research Institute Policy Report 125.
Farmer, F.L., Ilvento, T.W., Luloff, A.E., 1989. Rural community poverty: a LISREL
measurement model. Rural Sociology 54, 491?508.
Fisher, Monica, 2005. ?On the Empirical Finding of a Higher Risk of Poverty in Rural Areas: Is
Rural Residence Endogenous to Poverty? Journal of Agricultural and Resource Economics
30(2), 185-99.
Fisher, Monica, 2007. Why Is U.S. Poverty Higher in Nonmetropolitan than in Metropolitan
Areas? Growth and Change 38(1), 56-76.
Gadzey, Anthony, Henry Thompson, and Osie Agyeman Yeboah. Undated. ? Industrial
Subsidies in Alabama: Economic Impact Across Counties?, Southern Economics & Business
Journal, 2003. http://www.aum.edu/uploadedfiles/Upload_Vol26_%203_4.pdf (accessed 20
June 2008)
Gebremariam, G., Gebremedhin G., and Schaeffer, P. 2010. ?Analysis of county employment
and income growth in Appalachia?, a spatial simultaneous-equations approach, Empirical
Economics 38: 23-45.
Gibbs, Robert M., 1994. ?The Information Effects of Origin on Migrants? Job Search Behavior,?
Journal of Regional Science 34, 163-178.
Glasmeier, A. and Howland, M.1994. Service-led Rural Development: Definitions, Theories, and
Empirical Evidence. International Regional Science Review 15, 197-229.
Glasmeier, A.K and T.L. Farrigan. 2003. ?Poverty Sustainability and the Culture of Despair: Can
Sustainable Development Strategies Support Poverty Alleviation in America?s Most
Environmentally Challenged Communities.? The Annals of the American Academy of Political
and Social Science 590, 131-149.
Goetz, S. J., & Freshwater, D. (2001). State-level determinants of entrepreneurship and a
preliminary measure of entrepreneurial climate. Economic Development Quarterly, 15, 58?70.
Goetz. S.J & Rupasingha,A. 2009. "Determinants of growth in non-farm proprietor densities in
the US, 1990?2000," Small Business Economics, Springer, vol. 32(4), pages 425-438
Gomanee, K., S. Girma, O. Morrissey, and A. Verschoor. 2005. ?Aid, Government Expenditure
and Aggregate Welfare.? World Development 33(3): 355?370.
Gottlieb, P.D. 1994. Amenities as an Economic Development Tool: Is There Enough Evidence?
Economic Development Quarterly 8, 270-285.
Graves, P. E. 1980. Migration and Climate. Journal of Regional Science 20, 227-237.
78
Graves, P. E. 1983. Migration with a Composite Amenity: The Role of Rents. Journal of
Regional Science 23, 541-546.
Graves, P. E., and P.R. Mueser. (1993). ? The Role of Equilibrium and Disequilibrium in
Modeling Regional Growth and Decline: A Critical Reassessment.? Journal of Regional Science
33, 69-84.
Greenwood, Michael J., and Gary L. Hunt. 1984. ?Econometrically Accounting for Identities and
Restrictions in Models of Interregional Migration,? Regional Science and Urban Economics, 14,
113-128.
Greenwood MJ, Hunt GL, McDowel JM (1986) Migration and employment change: empirical
evidence on spatial and temporal dimensions of the linkage. J Reg Sci 26(2): 223-234
Hammond, George W. and Eric C. Thompson 2008. ? Determinants of Income Growth in
Metropolitan and Nonmetropolitan Labor Markets.? American Journal of Agricultural
Economics. 90. 783-793.
Hausman, J.A., 1983. SpeciGcation and estimation of simultaneous equations systems. In:
Griliches, Z., Intrilligator, M.D. (Eds.), Handbook of Econometrics, Vol. 1. Elsevier Science,
Amsterdam.
Henderson, Jason. 2002. ?Building the Rural Economy with High-Growth Entrepreneurs.?
Federal Reserve Bank of Kansas City Economic Review, 87(3): 45- 70.
Henry MS, Schmitt B, Kristensen K, Barkley DL, Bao S (1999) Extending Carlino-Mills
Models to Examine Urban Size and Growth Impacts on Proximate Rural Areas. Growth Change
30(4):526-548.
Hunt, Gary L. 2006. ?Population Employment Models: Stationary, Cointegration, and Dynamic
Adjustment.? Journal of Regional Science, 46, pp.205-244.
International Labor Organisation (ILO), 1990, The Promotion of Self-Employment, ILO, Geneva.
Isserman, A.M. 1994. State economic development policy and practice in the United States: A
survey article. International Regional Science Review 16,1&2: 49-100.
Jung, H., and E. Thorbecke. 2003. ?The Impact of Public Education Expenditure on Human
Capital, Growth, and Poverty in Tanzania and Zambia: a General Equilibrium Approach.?
Journal of Policy Modeling 25: 701?725.
Kapoor, M., Kelejian, H.H. and I. R. Prucha (2007). Panel data models with spatially correlated
error components, Journal of Econometrics, 140, 97-130.
Kelejian HH, Prucha IR (2004) Estimation of Simultaneous Systems of Spatially Interrelated
Cross Sectional Equations. J Econ 118(1):27-50.
79
Kim, Kwang-Koo, David W. Marcouiller, and Steven C.Deller. 2005. ? Natural Amenities and
Rural Development: Understanding Spatial and Distributional Attributes.? Growth and Change,
36,273-297.
Kinnucan, H.W., Y.Zheng, and G.Brehmer. 2006. State Aid and Student Performance: A
Supply-Demand Analysis. Education Economics, 13, 486-509.
Kirzner, Israel.(1973) Competition and Entrepreneurship. Chicago: University of Chicago Press.
Kreft, Steven F., and Russell S. Sobel. 2005. ?Public Policy, Entrepreneurship, and Economic
Growth.? Cato Journal, 25(3): 595-616.
Krugman, Paul 1991. Geography and Trade. Cambridge, MA: MIT Press.
Kuznetz, S., 1966, Modern Economic Growth: Rate, Structure and Spread, Yale University
Press, New Haven.
LeSage, James, P. (1999). The Theory and Practice of Spatial Econometrics. Department of
Economics, University of Toledo.
Lewis DJ. Hunt GL, Plantinga AJ (2002) Does public land policy affect local wage growth .
Growth Change 34(1): 64-86
Levernier, W., M.D. Partridge, and D.S. Rickman, 2000. ?The Causes of Regional Variations in
U.S. Poverty: A Cross-County Analysis,? Journal of Regional Science 40, 473-98.
Lucas, Robert E. Jr. (1978), ?On the size distribution of business firms,? Bell Journal of
Economics 9, 508-523.
MacKinnon, J. G., White, H., & Davidson, R. (1983). Tests for model specification in the
presence of alternative hypotheses: Some further results. Journal of Econometrics, 21, 53?70.
MacDonald JF (1992) Assessing the development status of metropolitan areas. In: Mills ES,
MacDonald JF (eds) Sources of metropolitan growth. Cent. For Urban Policy Res, New
Brunswick
Malecki, E. J. (1994). Entrepreneurship in regional and local development. International
Regional Science Review, 16, 119?153.
Marston, S. (1985). Two views of the geographic distribution of unemployment, The Quarterly
Journal of Economics. 100: 57-79.
McGranahan, D.A. 1999. Natural Amenities Drive Rural Population Change. Food and Rural
Economics Division, Economic Research Service, US Department of Agriculture. [AER-781],
1?24.
80
Mills, E., Price, R., 1984. Metropolitan suburbanization and central city problems. Journal of
Urban Economics 15, 1?17.
Morton, Jason.2007. ?Riley: Black Belt-like Initiative Will HelpRural Areas.? Tuscaloosa News,
August 10. http://www.tuscaloosanews.com/article/20070810/NEWS/708100349/-1/NEWS03
(accessed 17 June 2008).
Mulligan, G.F., A.C. Vias, and S.M.Glavac (1999). Initial diagnostics of a regional adjustment
model, Environment and Planning A. 31: 855-898.
Muth, Richard F.1971. ? Migration: Chicken or Egg, ?Southern Economic Journal, 37,295-306.
Nzaku, Kilungu and James O. Bukenya. 2005. ?Examining the Relationship between Quality of
Life Amenities and Economic Development.? Review of Urban and Regional Development
Studies, 17, 89-103.
Nord, M. and Cromartie, J.B. 1997. Migration: The Increasing Importance of Rural Natural
Amenities. Choices 3, 22-23.
OECD (1996), Technology, Productivity and Job Creation, Paris.
OECD, 1998, Fostering Entrepreneurship, the OECD jobs strategy, OECD, Paris.
Parker, S. C. (1996). A time series model of self-employment under uncertainty. Economica, 63,
459?75.
Partridge and Rickman,. 2006. The Geography of American Poverty: Is There a Role for Place-
Based Policies? The W.E. Upjohn Institute for Employment Research: Kalamazoo, MI.
Pollard, K.M. 2003. Appalachia at the Millennium: An Overview of Results from the Census
2000. Washington, D.C: Population Reference Bureau and Appalachian Regional Commission.
Quah, D.T. 1996. ?Regional Convergence Clusters Across Europe?, European Economic Review
40, 951-958.
Redding.S and Venables A.J, 2002. "Explaining Cross-Country Export Performance:
International Linkages and Internal Geography," CEP Discussion Papers dp0549, Centre for
Economic Performance, LSE.
Reynolds, P. (1994). Autonomous firm dynamics and economic growth in the United States,
1986?1990. Regional Studies, 28, 429?42.
Reynolds, Paul D., Michael Hay, and S. Michael Camp. 1999. Global Entrepreneurship Monitor.
Kansas City: Kauffman Center for Entrepreneurial Leadership.
81
Roback, J. 1982. Wages, Rents, and the Quality of Life. Journal of Political Economy 90, 1257?
1277.
Roback, J. 1988. Wages, Rents, and Amenities: Differences among Workers and Regions.
Economic Inquiry 26, 23?41.
Robson, M. T. (1998a). Self-employment in the UK regions. Applied Economics, 30, 313?22.
Robson, M. T. (1998b). The rise in self-employment amongst UK males. Small Business
Economics, 10, 199?212.
Romer, Paul. 1986. ?Increasing Returns and Long Run Growth, ?Journal of Political Economy,
94, 1002- 1037.
Rupasingha, A., Goetz, S.J., Freshwater, D., 2000. Social capital and economic growth: a
county-level analysis. Journal of Agricultural and Applied Economics 32, 565?572.
Rupasingha.A, Goetz.S.J, and Freshwater.D, 2002. "Social and institutional factors as
determinants of economic growth: Evidence from the United States counties," Papers in
Regional Science, Springer, vol. 81(2), pages 139-155
Rupasingha A., Goetz, S. J., & Freshwater, D. (2006). The production of social capital in U.S.
counties. Journal of Socio-Economics, 35, 83?101.
Rupasingha.A, and Goetz.S.J, 2007. "Social and political forces as determinants of poverty: A
spatial analysis," The Journal of Socio-Economics, Elsevier, vol. 36(4), pages 650-671, August.
RUPRI, 2007. Rural Policy Research Institute, http:// www.rupri.org (accessed February 2011).
Saint Onge, Jarron M., Lori M. Hunter and Jason D. Boardman.2007. Population Growth in
High-Amenity Rural Areas: Does it Bring Socioeconomic Benefits to Long-Term Residents??
Social Science Quarterly. 88, pp.366-381.
Schultz, T.P. (1990), Women?s changing participation in the labor force: a world perspective,
Economic Development and Cultural Change 38, 457-488.
Steinnes, D.and Fisher, W. 1974. An Econometric Model of Intra-Urban Location. Journal of
Regional Science 14, 65-80.
Steinnes, Donald N. 1978. ?Causality and Migration: A Statistical Resolution of the ?Chicken or
Egg Fowl-Up,? Southern Economic Journal, 45, 218-226.
Storey, D.J., 1994, Understanding the Small Business Sector, Routledge, London/New York.
Swail, W.S., K.E. Redd, and L.W. Perna. 2003. ?Retaining Minority Students in Higher
Education: A Framework for Success.? ASHE-ERIC Higher Education Report No. 2.
82
Washington, DC: The George Washington University, School of Education and Human
Development.
Swaminathan, Hema and Jill L. Findeis, 2004. ?Policy Intervention and Poverty in Rural
America,? American Journal of Agricultural Economics 86(5), 1289-1296.
Trendle, Bernard, 2009. ? The Determinants of Population and Employment Growth in Small
Queensland Regions?. Economic Analysis & Policy, 39, 295-307.
Treyz, George I., Dan S. Rickman, Gary L. Hunt, and Michael J. Greenwood. 1993. ? The
Dynamics of U.S. Internal Migration,? The Review of Economics and Statistics, 75,209-214.
Triest, R.K., 1997. Regional differences in family poverty. New England Economic Review
(January/February), 3?17.
Wimberley, Ronald C. and Libby V.Morris.2002. ? The Regionalization of Poverty: Assistance
for the Black Belt South?? Southern Rural Sociology. 18, 294-306.
Wren, Colin.1996. ?Fund Substitution and the Incentive Effect of Public Investment Subsidies.?
Scottish Journal of Political Economy. 43, 534-548.
Wu, Junjie and Munisamy Gopinath.2008. What Causes Spatial Variations in Economic
Development in the United States?? American Journal of Agriccultural Economics. 90,pp.392-
408.
83
Appendix: Method of Estimation in Panel Data Spatial Simultaneous Equations Model
This estimation procedure has five-step. In the first step, Generalized Two-Stage Least Squares
(G2SLS) is used to estimate the parameter vector consisting of [ , using an
instrument Matrix Q that consists of a subset of X, (I W)X, (I 2X, where X represents a
matrix that includes all control variables in the model, I is the identity matrix of dimension T,
is the Kronecker product, and W is a row standardized queen-based contiguity spatial weights
matrix. Using estimates for [ , from G2SLS, the disturbances for each
equation are computed.
In the second step, The computed disturbances are used to estimate the spatial
autoregressive parameter ? and the variance components, and , using the generalized
moment procedure suggested by Kapoor, Kelejian and Prucha?s (2004). For this generalized
moment procedure, two orthogonal and symmetric idempotent matrices, P and H, are defined.
Where P is a matrix that averages the observations across time for each individual and H is a
matrix which obtains the deviations from the individual means. These P and H matrices are used
to define the generalized moment estimators of ?, and in terms of six moment conditions.
This second step has two parts. In the first part, un-weighted initial generalized moment
estimators of ?, and are computed. In the second part, weighted GM estimators of ?,
and are computed.
In the third step, the weighted GM estimators of the spatial autoregressive parameter ?
are used to transform the data, using Cochran ?Orcutt-type transformation. Then, the
transformed data are further transformed using the variance components and by their
weighted GM estimators.
84
In the fourth step, these transformed data were used to estimate the Feasible Generalized
Spatial Two Stage Least Squares (FGS2SLS) estimates for [ , , using a
subset of the linearly independent columns of [X, (I W)X, (I 2X] as the instrument matrix.
Even though this GS2SLS takes the spatial correlation into account, it does not take into account
the potential cross equation correlation in the innovation vectors , r = 1,2,3.
In the fifth step, the full system information is utilized by stacking the transformed
equations in order to jointly estimate them. The FGS3SLS estimators of [ ,
are obtained by estimating this stacked model.