GENERAL MOTORS LORDSTOWN: A SIMULATION IMPACT STUDY ON THE
YOUNGSTOWN-WARREN ECONOMY
by
Joseph Bartholomew
Submitted in Partial Fulfillment ofthe Requirements
for the Degree of
Master of Arts
in the
Economics
Program
YOUNGSTOWN STATE UNIVERSITY
June, 1999
GENERAL MOTORS LORDSTOWN: A SIMULATION IMPACT STUDY ON THE
YOUNGSTOWN-WARREN ECONOMY
Joseph Bartholomew
I hereby release this thesis to the public. I understand this thesis will be housed at the
Circulation Desk ofthe University library and will be available for public access. I also
authorize the University ofother individuals to make copies ofthis thesis as needed for
scholarly research.
Signature: h~
Joseph Bartholomew, Student
~ Ji9 Jet'!
~ IDate
Approvals:
Date
Peter 1. Kas msky, Dean ofGraduate S
f1----'i/~--H-'==---=--------;----_+_''__+_-C{ 7
Dennis etrus ,Ph.D, Committee Member / '
~/~~
Abstract
General Motors' "acceptance" of its recent decline in market shares has prompted
the company to make drastic changes in its production methods. These improvements
will begin with a change from on-site assembly to modular assembly of certain car
components. Along with a new production technique, General Motor's will also build
new plants to accommodate modular assembly. These announced plans threaten existing
plants, many of which desperately need capital improvements. Ultimately, those
locations that are not selected for the new plants will probably witness the closing of their
assembly plants. The General Motors plant in Lordstown, in particular, falls in this
category. Built in 1966, the plant houses General Motors' small car operations, and, at
that time, was the largest and most automated assembly plant. The plant has also become
an important manufacturing firm for Mahoning, Trumbull and Columbiana counties.
General Motors has not yet confirmed any plans to reinvest in Lordstown; therefore, if a
shutdown occurs, the region may suffer serious economic consequences. This paper
studies the possible ramifications ofboth a decision to build and a decision to shut down
the plant on the Youngstown-Warren economy. Specifically, a regional econometric
model is used in four different scenarios that reflect the two possible decisions. Scenario
1 and scenario 2 focus on the selection ofLordstown for the new plant while scenario 3
and scenario 4 take the alternative view. Regional multipliers and time series graphs then
are used to study these effects and draw conclusions for the scenarios.
111
ACKNOWLEDGEMENTS
I would like to thank Dr. Liu and my committee for their assistance and support in
completing this thesis. I want to especially thank Dr. Stocks for volunteering his time in
explaining TROLL and helping with the data. I would also like to express my thanks to
Dena for her support and making data entry interesting, and my parents, for their
encouragement.
Table OfContents
Chapter 1 Introduction 1-2
Chapter 2 History 2-9
Chapter 3 The "Battle" Between Lordstown and Lansing 9-14
Chapter 4 Literature Review........................................... 15-33
4.1 Econometric Techniques 15-24
4.2 Non-Econometric Techniques 24-33
Chapter 5 Model 33-36
Chapter 6 Data 36-40
Chapter 7 Methodology 40-44
Chapter 8 Results....................................................... 45-57
Chapter 9 Conclusions 57-58
Chapter 10 Limitations , 59-60
Appendix 61-75
AppendixA 61-64
Appendix B 65
Appendix C 66-75
References 76-78
Chapter 1 Introduction
General Motors' market share ofthe domestic auto industry has been slowly
shrinking in the last decade, and several elements can explain this fall from supremacy.
First, General Motors is far behind Ford and Chrysler in producing and marketing mini 
vans and sport utility vehicles, which is now one of the fastest growing segments of the
market. Second, their cars lack the engineering quality and designs of their competitors,
resulting in reduced sales. Finally, the continued use oftraditional assembly methods in
aging assembly plants has increased costs and reduced efficiency. In an effort to reverse
this trend, General Motors is gearing up for a massive reformation project.
One of the key features ofthis project rests with the commitment to abandon old
production methods and, instead, build cars using modules. With modular assembly.. less
laborers will be needed to build the cars, and outsourcing to suppliers will result in lower
prices for car parts. Since first announcing these changes, General Motors has nearly
selected all the locations for the new plants, They have however left the last remaining
site unmentioned, responding by only saying that the choice has been narrowed to two
possible locations: Lansing and Lordstown. Without hesitation, both cities began local
campaigns to help convince General Motors to reinvest in their respective region.
The possibility ofLordstown shutting down has spawned many questions about
its impact on the Youngstown-Warren economy in the event ofthis decision. This paper
examines these questions and attempts to put into "prospective" the possible economic
consequences ofits closure. In the following chapters ahead, four scenarios are
introduced relating to the two possible outcomes of a decision. A regional econometric
model analyzes these different scenarios and delivers many important conclusions.
Before drawing any conclusions about Lordstown's future, the paper begins with a
detailed discussion ofits thirty-three year old history.
Chapter 2 History
This section follows the history ofthe Lordstown's assembly and Fisher Body
plant from its rocky beginnings and prosperous years to the present.
The Beginnings
The post World War II automobile industry experienced a steady increase in the
demand for its products, and General Motors especially felt this surge. For General
Motors, this phenomenon lead to a nationwide investment ofnew plants, equipping them
with the latest automobile technology and assembly line methods (Vindicator, 1966, p.
E10). One ofthese sites, Lordstown, Ohio, brought an innovative plant design of a dual
Fisher Body and Chevrolet Assembly Plant to advance the process ofbuilding
automobiles. An official indication ofthis planned investment came in February, 1956
when General Motors bought 775 acres ofland, and, later, in May when they purchased
an additional 300 acres (Reiss, 1956, p.1). This location, a mile south ofLordstown at
the intersection of Young-Hallack and Bailey-Ellsworth roads, was located in the middle
ofa quiet farming community of2,000 residents. Initially, the plant had an estimated
employment capacity of 8,000 to 20,000 workers, total area of 2,250,000 feet and
2
estimated costs of$75 million. Preliminary figures indicated that the plant would be the
world's largest automobile assembly plant (Vindicator, 1956, p. 13).
After several delays in the start ofits construction, groundbreaking ceremonies
finally commenced on September 30, 1964. With construction finally underway,
expectations of the new plant's economical impact on the surrounding area remained
high. Former Governor Rhodes, for instance, said, "This is a giant step in the economic
growth ofthe Mahoning Valley" (1agnow, 1964, p.1). Estimations of its impact concur
with his statements as forecasted purchases from suppliers around Mahoning County
ranged form $9 million to $11 million and payroll and other estimated purchases reached
nearly $45 million (1agnow, 1964, p.1). These figures then indicated that important
secondary effects would appear because ofthe new plant, including the development of
new businesses and expansions ofexisting businesses.
Construction was expeditiously completed, and the plant produced its first car on
March 4, 1966--one year and seven months after initial ground breaking. At this time,
only 1,500 workers were employed at the plant, and production rate was only five cars an
hour; however, with an expected increase in production, General Motors Lordstown was
hiring hundreds of workers a week (Vindicator, 1966, p. E5). While General Motors
intended on hiring more workers, the list ofapplicants greatly exceeded the number of
jobs available. Astonishingly, personal managers ofboth the Fisher Body and Chevrolet
Assembly Plants received over 35,000 job applications, some as far away as Florida and
California but most within the regional area. (Vindicator, 1966, p. E5). The expected job
opportunities made available by the new manufacturing plant rendered it the title ofthe
area's most important and promising new business ofthe 1960's.
3
The 1970's: The Turbulent Years
For the Mahoning Valley, the end ofthe 1960's were a period ofboth anticipation
and relief as the "world's largest" automobile plant was finally built and producing cars.
The 1970's however brought a new set ofchallenges for the automobile plant. With
pollution control issues and rising fuel costs making headlines, the demand for more fuel 
efficient cars was rising. Small car imports made deep inroads into the domestic
automobile market, reducing the demand for domestic products and snatching profits
away from American automakers. General Motors, vowing to reclaim the market for
domestic cars, introduced a new small car that had, as the company described, "maximum
combination of low pollutant emissions, fuel economy and performance" (Vindicator,
1970, p.1). This changing market would soon alter the role of Lordstown in producing
cars.
General Motor's' optimism about a resurgence in demand for domestic cars was
attributed to its plan to build its new line ofcompacts--known as the Vega 2300--at the
company's newest and most automated plant in Lordstown. In preparation, General
Motors added a new section to its complex: a 2.2 million square foot metal stamping
plant. In addition, the company redesigned the assembly line and Fisher Body plant to
accommodate the production of the new small car. General Motors wasted no time in
equipping the plant for production of the new Vega (Vindicator, 1970, p.1).
Within a few months, all necessary preparations for the Vega's arrival were
complete; however, production was halted by a two-month strike--the first major work
4
stoppage at the plant. The strike, resulting from a contract dispute, was however settled
within two months, and, with production resumed, the first Vega was produced in
November, 1970 (Vindicator, 1970, p.1). Concurrent with General Motor's pursuit of
building small cars, they also saw potential profits in another area of the auto industry:
the van market.
To gain a foothold in the market, General Motors once again selected Lordstown
as the site to build a new product line. As Reiss (1970) reported, this plan called for the
building ofa 630,000-foot assembly plant that would ultimately cost $25 million and
employ 1,500 workers (p.1). For Lordstown, 1970 brought much promise and success;
however, as the years rolled on and production ofsmall cars increased to record levels,
problems began to occur.
In 1972, General Motors Lordstown was distinguished with the fastest assembly
line (100 cars/hr), and, ironically, this achievement led to the most significant strike in
the plant's history. Problems first began to surface when the company hired General
Motor's Assembly Division, a new management team, to take over the management of
the plant. Specifically, disputes occurred after the new team attempted to increase
efficiency and reduce costs by cutting back on jobs; moreover, the remaining workers
were responsible to pick up the slack (Automotive Industries, 1972, p. 18). Workers, in
opposition, retaliated and caused production setbacks through increased absenteeism and
number of defects, low productivity and outbreaks of worker sabotage. In fact, workers
even publicly admitted to committing sabotaging acts such as failing to do their job,
incorrectly installing parts and purposely causing damage to cars (Rothschild, 1972, p18;
Given, 1972, p.50). Management, enraged at these actions, developed more strict
5
policies, and eventually the workers went on strike. In response to the strike, Emma
Rotheschild (1972) said, "conditions there are bound to be important for years" (p. 18).
This statement suggests that even though the strike was brief (three weeks) and did not
severely handicap production, the strike's underlying issues would continue to be a source
ofproblems for Lordstown.
Redemption in the 1980's
By the end of the 1970's, General Motors Lordstown's role in producing
automobiles was secured. Lordstown had been producing the Vega since 1970, and, just
as the car's luster began to disappear, General Motors awarded the plant with a new
contract for building compacts. This new group consisted of the Chevrolet Monza,
Pontiac Sunbird, Oldsmobile Starfire and the Buick Skyhawk, and this responsibility set
company records for the number of models produced at one plant (Reiss, 1977, p.1).
The beginning ofthe 1980's brought even greater prominence to Lordstown as it
again won the bid to produce the next line of compacts, the J cars (Mahoney, 1981, p.3).
Once recognized as incorrigible, young and fearless, this now tamer group ofemployees
was again given the task ofrevitalizing GM's small car market. Also, at this time,
General Motors Lordstown's operations were thriving as employment reached its highest
levels and demand for its products were strong.
While GM's valley operations remained strong, other businesses were
experiencing different economical conditions. For example, Youngstown's industrial
foundation, Steel, began its decline in the late 1970's, and the full ramifications did not
6
appear until the early 1980's. As unemployment soared and thousands ofjobs were lost,
General Motors arguably served as the crutch to a fallen manufacturing sector. It can
also be hypothesized that Lordstown's existence aided in an economic rebound years
later.
With the decline ofthe steel industry, Lordstown, now assumed a new identity.
As the area's leading manufacturing company, it was imperative that Lordstown cleanup
its reputation and avert labor-management problems. Fortunately, the 1980's served as a
non-strike era, and some officials gave credit for this accomplishment to the presence and
services ofCharles Abernathy. A former manager of the assembly plant, Abernathy's
friendliness and insistence ofopen communication among workers and managers won
votes among the employees (Vindicator, 1976, p. B17).
Although General Motors Lordstown's operations ran smoothly in the 1980's, the
plant was struck with an unforeseeable disappointment. The dissatisfying news came in
the late 1980's when General Motor's announced that it would shut down the van
assembly plant and eliminate 2,000 jobs (Jensen and Livingston, 1987, p. A2).
The van plant however did not officially closed its doors until March 1992, and,
importantly, the event jumpstarted questions about the effects ofa complete shutdown
(Williams, 1992, p. C2). Recently this question has again sparked a lot ofinterest as GM
seeks a new solution to gain profits and regain supremacy in the industry.
These new changes however entail much more severe and complex changes than
past attempts. Most importantly, these decisions will create one ofthe greatest
challenges for General Motors Lordstown.
7
The Fate of Lordstown
In an effort to downsize and reduce costs, General Motor's is restructuring its
assembly methods and consequently its family ofassembly plants. This new idea, called
Project Yellowstone, involves modular assembly or the delivering of completed sections
of the car, dashboards and engines for example, to its assembly plants for final
production. Therefore, the car will be built by linking orjoining together different
modules, bypassing traditional assembly techniques. With sections ofthe car delivered
intact, less laborers will be needed and total assembly time per vehicle will be cut; thus
labor, material and factory costs will be reduced. The responsibility for building the
modular parts will be outsourced, and those suppliers will also be involved with its
design. GM vice president Mark Hogan believes that this process will not only greatly
reduce costs but also improve quality and efficiency. In fact, he predicts that total factory
costs will decrease by 20%, and, most importantly, he believes that this new production
method will finally allow GM to earn profits on small cars (Kubik, 1999, p.9).
Yellowstone plants will differ from traditional assembly plants by the size (1.2
million square feet versus three million square feet) and employment (average of 2,100
workers compared to 3,600). Total production does not differ substantially between the
two types ofplants (200,000 automobiles for traditional plants and 215,000 autos for
Yellowstone plants per year) (Kubik, p. 9).
For employees of Yellowstone plants, they will endure 250 hours oftraining, and
an additional 75 hours of training per year. Employees will also have more responsibility
in their jobs and efforts will focus on, according to Mark Hogan, "ergonomics, worker
8
safety, teamwork, and seek[ing] to reduce strain and repetitive motion injuries by rotating
workers every two hours" (Kubik, p.9). This change will relieve some of the boredom
that past assembly line workers often felt with their jobs (Salpukas, 1972, p. E10).
Monetarily, the total cost of a Yellowstone plant is $300 million and construction will not
begin until 2002
Lordstown is especially challenged by this new ideal because General Motors has
declared Lordstown as one of two possible sites for the last Yellowstone plant. However,
for General Motors Lordstown to win the bid for the last plant, the community must
compete with Lansing and convince GM to build here instead.
Chapter 3 The "Battle" Between Lansing and Lordstown
As General Motors begins to finalize the plans for Project Yellowstone, they must
decide soon on the location on the final assembly plant. Although they continue to
postpone this decision, both Lordstown and Lansing remain optimistic about their
chances in being selected. Their hopefulness is ushered by their ongoing campaigns and
programs to build support for a new assembly plant. The objectives ofthese campaigns
include offering tax incentives, improving education and amassing local support groups
to oversee activities concerning this cause. In the meantime, both communities are
keeping the company aware of these efforts and on any improvements made in these
areas.
9
The concerns and issues listed above are a few of the important factors in a firm's
decision to locate in a certain region. The research regarding these elements and the
progress made by each community in these areas are discussed below.
There is widespread belief that the business climate ofa particular state is
instrumental in attracting new businesses. Consequently, an unhealthy business climate,
such as high state and local taxes, absence of right-to-work legislation, low quality oflife,
inadequate spending on public goods and governmental apathy to expansion ofbusiness
opportunities can harm the possibility ofbringing in new industry (Patch, 1995, p.35).
On the other hand, some ofthe major components of a positive business climate
include the existence of opportunities for financing projects through state and local
revenue bonds, a state's willingness to loan funds for initial construction costs, and
issuance of tax exemptions on land, capital and equipment. In examining this variable's
position in the location process, Patch examines a 1979 study by Carlton. This study
shows that the above characteristics of a business climate to be significant in attracting
new firms in fabricated plastic products, communication transmitting equipment and
electronic component industries. These results however may not be a true reflection of
all businesses since the study only considered three industries (p.36).
One ofthe most important aspects ofthe business climate mentioned above
involves taxes. According to Walthall (1996), at least twenty-five states maintain various
"economic development credits and incentives" to lure new businesses into their region
(p.43). Many experts believe that taxes playa major role in determining the best
location of new plants; however, conflicting evidence of this belief exists. Early studies,
pre 1960, find that taxes play no significant role in a firm's decision to locate in a certain
10
area; instead variables such as proximity to markets, labor accessibility and power and
fuels are determined to be more significant. According to Patch, these early studies
support the widespread opinion that taxes are only a small part oftotal costs; hence small
differences in taxes have little impact on decision-making (p.38). Additionally, Walthall
states that these types ofpackages do little to bring in new business, and instead use up
valuable funds that could have been spent elsewhere (pA3).
Studies following 1960 bring both similar and different conclusions than the
earlier studies. For instance, some deem taxes as important in the decision-making
process. Specifically, Patch cites two different studies, both using the corporate tax rate,
to reveal this detail. First, Bartik and Newmann study employment growth by industry
and conclude that, for capital concentrated industries, the corporate tax rate significantly
accounts for an increase in new industry (pAO). Similarly, Newmann's study also finds
that the corporate tax has a positive role in attracting new businesses, especially in the
durable goods industries.
Differently, in a study by Walthall (1996), he finds the opposite case in the
automobile industry, which is also one ofthe largest durable goods industries. According
to Walthall, transportation issues played the biggest role in the decision of General
Motors to build a Saturn plant in Tennessee while taxes played no role. Interestingly, tax
breaks were completely irrelevant since Tennessee legislation forbids giving any kind of
tax incentives to firms. Walthall does however agree that taxes are still a factor in
location decision making, and he believes that tax issues become more important as
businesses begin the final phase in the selection of sites. In fact, he notes that this issue
was paramount in the attraction ofa Mercedes-Benz plant to Alabama.
11
Therefore, the conflicting results from the studies above hinder any rational
conclusions about the importance oftaxes in the location decision equation. However
enough evidence exists to consider this variable to be important in the decision-making
process (pp. 43-47).
In the present "battle" for the new assembly plant, Nequetto and Shilling (1998)
report that both Lansing and Lordstown are offering General Motors tax incentive
packages for re-establishing operations. For Lordstown, the effort is lead by
Congressman James Traficant and includes many prominent politicians and businessman
(p. AI). The package consists of various tax and other local, state and federal incentives
and mimics the ones offered by Toledo when they won the bid for a Jeep assembly plant
(Welker, 1998, p. AI). Besides tax incentives, both communities are pursuing efforts to
improve their respective region's quality oflife.
Quality of life refers to the alluring qualities of a region that make it a favorable
area to live and run a business (Patch, pAl). According to Carn and Rabianski (1991),
the quality of life factors should include differences in the cost of living, quality of
housing as well as other elements that will provide a healthy living and business
environment. (p.320). In obtaining this type ofenvironment, Patch believes that a better
quality oflife can be reached as government spending on public goods increases,
especially on education (pAl).
Numerous studies have been conducted researching the amount of money spent
on this area and its influence on the location of new plants. In her book, Patch especially
talks about the importance ofeducation as a quality oflife factor, and she mentions
several studies contributing to this importance. For example, Wasylenko and McGuire
12
and Plaut and Pluta conclude that the percentage of income spent on education is
positively related to new employment within a state (pA2). The conclusions above can
however be disputed by examining other studies on the subject. For example, studies by
Bartik find no significant connection ofeducation spending and employment growth
(Patch, pA2). Patch however believes that inconsistent findings do not discredit the
positive role education plays in improving the quality oflife of an area.
Quality of life involves much more than the amount of money spent on education,
and recent research has encountered several other factors. For instance, Patch notes that
several studies indicate a positive statistical relationship between expenditures on public
goods and new business, including the importance ofproximity to rail lines and interstate
highways (Patch, p.43). These results indicate that government spending on
infrastructure will also aid in attracting new business.
Many government policies can be implemented to increase the quality of life of a
region; thus as governments appropriate more funds toward community improvements,
the quality oflife increases for its residents. Believing in this cause, both Lordstown and
Lansing have worked to improve their quality of life.
The majority ofthe work undergone in this area by the communities, according to
Shilling (1998), concerns improvements in education. For Lansing, they adopted
education improvement programs that focused on decreasing drop out rates, increasing
efficiency test scores and cooperative education programs. Widespread community
support for these programs through volunteer tutoring by adults and college students and
hiring senior citizens to patrol for delinquent children have also helped with this cause.
Thus, within the last two years, these programs have proven successful as test scores rose
13
and drop out rates decreased (p. A3). Although Shilling did not report the efforts ofthe
communities surrounding Lordstown toward this cause in his article, news reports have
indicated that similar programs are being adopted to improve education.
Besides education, Lansing and Lordstown's campaigns are both heavily
concerned with gathering community support for their causes. These projects include
both advertising (newspaper, radio, television and special news coverage) and holding
informational meetings about the plants importance and meaning to their respective areas
(Shilling p. AI; Welker p. AI). Evidently, both communities are working hard to
convince GM to continue operations in their areas.
14
Chapter 4 Literature Review
The research regarding impact studies ofplant openings and closings deals
primarily with Ex Post studies ofexiting plants and both Ex Post and Ex Ante studies of
new plants to a region. The models used for these studies can be divided into one oftwo
main types: econometric and non-econometric. Much of the available research
concerning the impact ofplant closings involve the non-econometric technique of input 
output analysis while studies on plant openings consist ofboth types.
This section will present some ofthe important studies surrounding this topic, and
its content will focus primarily on the technique used to perform the impact study.
Furthermore, because limited examples ofplant impact studies exist for econometric
techniques, several important models will instead be discussed, and the impact studies
conducted with these models will also be mentioned.
Chapter 4.1 Econometric Techniques
One ofthe common techniques in analyzing and explaining a particular regional
economy involves forecasting and impact studies. In order to apply these techniques,
first, an economic model, summarizing the various industries and business
interrelationships, ofthe region is built. Although many different modeling techniques
exist, one ofthe most common choices involves econometric modeling. Bolton (1985)
describes a regional econometric model as "a set ofequations, perhaps highly
15
simultaneous, describing the economic structure of a regional economy, usually a state or
province or metropolitan area" (pA95).
Most of the early research, the 1950's and 1960's, and models use and apply the
theory from national scale models; however, today, with the complexity of markets,
regional econometric models obtained their own identity (Bolton, pA95). Nevertheless,
most econometric models still retain a unique relationship with national models.
According to Bolton, it is necessary to join a regional model and national model
because typically the national economy will influence the regional economy (pA98).
These type ofmodels are known as "top down" models which are different from "bottom
up" models. Bottom up models describe the situation where the regional economy affects
the national economy while top down models characterize the national influence on the
local economy. This difference can especially be seen in the specification ofeach model.
For top down models, the national variables remain exogenous to the model whereas, for
bottom up models, the national variables are defined as endogenous. When examining
existing econometric models, top down models tend to be the most popular (Bolton,
pA99).
Econometric regional model building however has one very important
shortcoming: its costs. Significant costs are incurred in both building a model and
continued updating ofthe model for future forecasts. Additionally, these costs are
compounded as model size and intricacy increases. This fact has lead to smaller scale
econometric models thereby reducing the project's total cost (Fiske, Lamb and Moress,
1991, pA9).
16
A structured econometric model contains a list ofequations where each dependent
variable is explained by a number of independent variables. Two important types of
SEM's include the simultaneous system and the recursive system. For recursive systems,
Cleary and Levenbach (1986) note that the equations must be logically ordered in order
to account for multiple effects among the independent variables--independent variables
being influenced by previous values of other independent variables (p.191). Because of
this characteristic, these types of models are also known as recursive models. One of the
many problems that can occur with these models include specification errors and, for
simultaneous models, the "simultaneity" problem. Specification errors occur through the
misstating of the econometric equation while the simultaneity problem relates to the
correlation among independent variables.
One ofthe early examples ofa SEM model follows the regional economy ofthe
vast metropolitan area of Los Angeles. Uniquely, Hall and Lecari's (1974) regional
model of the city uses SMSA data for the historical values of the endogenous variables.
The structure ofthe model is governed by a product-total production relationship where
economic activity is measured by summing production from each economic sector.
Specifically, the model separates Los Angeles into these four sectors of manufacturing:
wholesale trade, regional trade, other (including finance services and construction) and
government (federal, state, county and city) (p.338).
The variables used in this model contain numerous endogenous and exogenous
variables; endogenous variables are variables determined within the model while
exogenous variables are determined outside the model.
17
The complete model ofLos Angeles contains 29 equations (nineteen stochastic
and ten identities). To estimate the stochastic equations, Hall and Lecari perform both an
OLS (ordinary least Squares) regression and two stage least square (TLS). Although not
shown here, the two regression procedures produce satisfactory results as nearly every
coefficient had significant t scores and the coefficient's of determination (R2 ) remain
close to one. Additionally, the calculated DW (Durbin Watson) statistic, nearly two for
all equations, indicates no significant presence ofautocorrelation (Hall and Lacari,
pp.340-341).
In addition, the model's accuracy is determined by computing the MAPE or mean
absolute percent error of the endogenous variables. The equation for MAPE is shown
below where Y*j (t) is the historical value for the ith variable at time t and N is the sample
size.
(1)
Many statisticians believe that MAPE values at 5% or less is sufficient for
designating a model as being accurate. In response to their model's accuracy, Hall and
Licari note that the average MAPE for the endogenous variables hovers around 2%
(p.343).
Besides being a reliable forecast model, Hall and Lucari explain that the model
can also calculate "impact multipliers." Multipliers help explain, empirically, the impact
18
ofexternal or internal changes in the economy. Generally, these changes occur with the
endogenous variables and are calculated by taking the differences in the endogenous
variables resulting from a change in an exogenous variable. From the results of
multipliers, appropriate business decisions and policy changes can be made and harmful
decisions can be avoided (p. 344).
In their study, Hall and Lecari used the model to conduct various impact studies.
For instance, in one study, they measured the impact of a one billion-dollar increase in
GNP in the export sector. Ultimately the impact generated a $26.9 million increase in
Gross Revenue Product or a multiplier of 26.9 and positive multipliers in several other
sectors (p. 347). Table I below shows the results mentioned above, and it also includes
other multipliers as responses to increases in other exogenous variables.
Table I: Impact Multipliers for Exogenous and Endogenous Variables
USGNP
Exogenous
TAXR
Variables
POP GREVA
C 10.125 173.33 -.047 .744
EGOV .253 22.333 -.001 .094
GEXP 2.539 223 -.012 .948
GREV 2.679 235.333 -.012 1
GRP 26.901 465.333 -.074 1.926
LF 1.321 29.667 .104 .127
NWY 6.513 100.667 -.018 .466
P -.007 .333 0 .001
TEMP 1.524 34.333 -.003 .146
Source: Journal ofRegional Science. "Building Small Region Econometric Models: Extension of
Glickman's Structure To Los Angeles", p. 347.
Specifically, for the average tax rate (TAXR), table 3 shows the results of a I% increase
in the variable on the selected endogenous variables in the far-left column. Similar
19
impact studies are recorded for one thousand increases in the regional population (POP)
and autonomous revenues (GREVA) (p. 347).
The model has also been used to conduct employment and income multipliers.
These multipliers in particular measure the change in aggregate employment or income
from a change in employment or income in a particular sector. Furthermore, the
econometric and the input-output method remain the two most popular methods for their
calculation.
Because an input-output model also existed for Los Angeles, the authors were
able to make comparisons between the multipliers by initiating the same changes for each
model (pp. 348-349). Table 4 summarizes these results, and interestingly nearly identical
solutions exist for the two models.
Table 2: Comparison ofRegional Models for Employment and Income
Multipliers
Employment Multiplier Income Multiplier
Sector Multiplier Model Sector Multiplier Model
Private exports 2.76 Econometric Private exports 2.43 Econometric
Private exports 2.13 Input-Output Private exports N/A Input-Output
Local Gov't 1.54 Econometric Local Gov't 2.03 Econometric
Local Gov't 1.71 Input-Output Local Gov't N/A Input-Output
Source: Journal ofRegIOnal SCience. "BUIldmg Small RegIOn Econometnc Models: ExtenSIOn of
Glickman's Structure To Los Angeles", p. 350.
Similar regional models have been built using the idea ofthe SEM model with
SMSA data, and one such important model concerns the model ofChicago.
20
Chicago Model
Another famous use ofSMSA data and econometric modeling is the Duobinis
(1981) Chicago model (pp.293-316). The general specification ofthe model uses the
"Transcendental logarithmic production function" to describe productions of goods in
each sector. Eventually the model uses each sector's figures to calculate Gross
Metropolitan Product (GMP) which is equivalent to its national scale counterpart Gross
Domestic Product. According to Duobinis, fluctuations in the values ofthis variable will
determine the amount ofexpansion taking place in the Chicago statistical area (pp. 293-
296).
The model's business sectors are broken down into four main divisions:
manufacturing, government, and farming and mining. Fourteen industries comprise the
manufacturing sector, the largest of the divisions. Interestingly, Duobinis reveals that
most of the demand for this sector's goods lies outside the region, but local linkages also
exist. In order to express this local demand relationship, he constructs input-output tables
using "estimation location quotients. (p.296). The local linkage and external demand for
each manufacturing industry is determined by the following equation.
(2) InVdi = lnal = Bl + B2 *In LAL + B3 *In NAL
where: VDj = amount of output
di =national price of the product for each industry
LALj =local level of activity
NA~ =National activity level in each manufacturing industry
21
In addition, a "transcendental logarithmic function" is added to each of these equations to
describe the entire effect (Duobinis, p.296).
The model estimates similar equations for the remaining sectors and represents
them with numerous local and national variables. The complete model contains 170
equations with the majority being identity equations (Duobinis, pp. 306-312). Because of
the numerous linkages, the flow diagram is extremely complex and therefore will not be
shown. In addition the size of the model will also prevent the presentation of the model's
equations. However, as noted before, most ofthe equations follow the structure in
equation 1. Also, specifically, the model is a "block recursive system" where each block
is solved separately, and the information solved in one block can be used to solve the
next set ofblocks.
The model's accuracy was tested using the Mean Absolute Percent Error (MAPE)
or a procedure Duobinis refers to as "Ex Post forecasting." His summary ofMAPE
results shows that the model performed favorably with over half of all variables having a
MAPE less than 1% (p.313).
One ofthe main functions of this model, according to Duobinis, is to perform
simulations. Specifically, Duobinis performs two simulations by first altering a regional
variable and then changing a national variable. In the first simulation, the local tax rate is
increased by 1% for each year ofthe simulation. Thus, Duobinis goal is to determine the
local impact ofa change in a regional variable. The second simulation study reports the
results of a percentage decrease in federal taxes (pp. 315-316). Evidently, Duobinis'
model and experiments show that local impacts can be determined from regional models.
22
Impact Study with Automobiles
In the research on automobile plants, only one impact study could be found, and
this example concerned the analysis of a new plant. Additionally, this study is the sole
representative on econometric techniques for plant openings and closings. Specifically,
Campbell (1989) performed an impact study of Diamond Star Motors, a new business
entrant in the Kentucky Bluegrass region. From the plants initial construction to
completion, Campbell, examined the effects ofthree different stages including a building
stage, producing stage, and the multiregional effects. To estimate this impact, Campbell,
uses the conjoined input-output model (a model that combines an input-output and
econometric model). Uniquely, the model is driven by employment in Motor Vehicles
upon which data was readily available. Like many new entrants, employment levels
slowly increased as the plant increased its production levels (p.31).
In analyzing the impact ofthe actual plant's construction stage, the effect was
solely concentrated in one specific region: the plants actual location. The multiplier
effects of this impact revealed however that some new employment was created. The
construction stage however lasted for only a year and more dramatic effects were
witnessed when the plant began operations (Campbell, p.33).
The impact after the plant started operations showed a much larger multiplier
effect than the construction phase. Some sectors such as Wholesale trade and Finance,
Insurance and Real Estate increased employment twofold while others reported more
moderate gains. In addition, demand for certain occupations (secretaries and engineers,
23
for example) also increased. Thus the initial employment of the plant's 2,900 workers
created many more jobs (Campbell, 1989, p.33).
The last stage ofthe analysis examined an interregional effect. Immediate
impacts occurred in the region ofthe plant's actual location and some adjacent ones while
less of an impact occurred in the remote areas. Statistically, Campbell finds that 30% of
the jobs created were formed from increases in "consumer spending" and increases in
several durable and nondurable good sectors (p.33).
Chapter 4.2 Non-Econometric Models
Amid the ongoing expansion in regional model building and interest in impact
analysis, the input-output has become one ofthe most popular design choices for new
models. An input-output model describes the interdependency and unique relationships
among a region's economic sectors by aggregately organizing each sector's source of
inputs and resulting outputs in a single table. Individually, each sector's economic
activity is represented by a linear equation, and, together, these equations describe the
workings of region's economy. Traditionally, the entire regional economy and its
interindustry relationships are transformed into matrices, and these relationships are
structurally labeled in a coefficient matrix (Leontief 1986 p. 19).
Table 3 below shows a typical input-output table for a simplified economy
composed of three sectors.
24
Table 3 Input-Output Table for a Simple Three-Sector Economy
Outputs From: Outputs To:
I 2 3 Final Demand Total Output
1 X ll X l2 X l3 YI Xl
2 X 21 X 22 X 23 Y2 X 2
3 X 31 X 32 X 33 Y3 X 3
Value Added VI V2 V3 V
Source: Input-Output and Regwnal Economics, p. 27
Using row two as an example, the total output equation is calculated as:
where Xij is the purchases ofindustry j from industry i
If a constant relation among the amount ofgoods bought by the industries from
industry 2 and their output exist, equation 3 is formed (Richardson 1972 p.26).
The aij'S are commonly known as "direct input coefficients" and represent the amount of
inputs needed by industry j from industry i per unit of output from industry j (Richardson
pp. 26-27). When assuming the constant relation above, a mathematical relationship then
exists between final demand and total output. Using this information, any economy,
whether simple or complex, can then be represented by the equation below.
25
(5) X =AX + Y
where: X is a (n X I) vector matrix oftotal output for n sectors
A is the coefficient matrix
Y is the vector matrix for final demand
Then by manipulating the matrix, equations 6 and 7 result
(6) XCI - X) =Y
(7) X =(1 - A) -I Y
where 1is the identity matrix
Next, by letting B =(1 -A) -I, equation 8 can be formed.
(8) X =BY
Each bij represent the "indirect and direct purchases from industry i by industry j
in satisfying one additional unit offinal demand" (Schaffer p.157).
Using equation 3, Harris (1993) then illustrates a technique to measure the
economical impact of a new plant and new sector on the economy of Nevada (pp. 100-
106). He first partitions matrix A in the following manner (p. 102).
26
A = (All Al2)A21 An
where: All is a (n-p) X (n-p) coefficient matrix
A1Z is a (n-p) X p coefficient matrix
AZ1 is a p x (n-p) coefficient matrix
Azz is a p x p coefficient matrix
P =new sectors
Given the partition ofmatrix A above, we can then rewrite equation 5 as:
(10) (Xl)=(I-AII -AI2)-I(Yl)X
2 - A21 I - An Y2
The new sector's impact is measured by total output Xz and final demand Yz.
Next, Harris revises equation 8 by considering only the new industry's impact on
existing sectors and this new equation becomes:
where the Y's are sectors that are exogenous to the model.
Harris then multiplies the A matrix in equation 9 by equation 11 below to obtain
an inverse (p.102).
27
where: I =(n-p) X (n-p) identity matrix
01 =(n-p) X p null matrix
O2 =P X (n-p) null matrix
03 =P X P null matrix
The inverse then becomes
Finally, from equations 6 and 7 the impact of the new industry p becomes:
where: D =n x n matrix ofcoefficients that shows the interindustry relationships with
the new sector.
Harris then used the model to study the impact of a new potato processing plant
on two counties in Nevada. Also, in studying the impact, Harris includes the household
sector as endogenous to the model to measure the "induced" impact ofeach sector's sales
to final demand (p.l03).
The analysis begins with a breakdown ofequation 10. In the equation, the (I-A) -1
matrix measures the impact of the regional economy with the potato processing plant
28
included while the coefficients of the (I - JpAJp) -I measures the impact without the new
sector. Then, by comparing the differences between the two matrices, the impact ofthe
new industry is measured (p.l04).
Harris also reveals that his model can perform an impact for an exiting plant.
However, the procedures can become more complex, Harris contends, when an entire
sector is removed rather than just one plant. With this scenario, he argues that the
analysis must consider many important details and will become multi-faceted. For
instance, one must now explore the impact on the sectors that buy imports from the
exiting sector, and also the ramifications on those sectors that sell inputs to the departing
sector. According to Harris, all these circumstances will then make the analysis more
complicated (p. 105).
The technique used above to measure the impact of a new industry is just one way
to perform an impact study using the input-output method. Differently, the next example
shows an alternative way to perform an impact study using the input-output method
Schaffer (1976) also constructs an input-output model to describe the impact of a
new plant in Georgia. Similar to Harris, he includes households as endogenous to the
model by including it among the interindustry matrix (p.IS7).
The study focuses on the impact of a new plant in the already flourishing veneer
and plywood industry. The Georgia input-output tables used to summarize this impact
emulate the U.S. input-output tables. For example, the model contains 300 ofthe 367
industries of the U.S. model; thus 67 of the industries lie outside the region. The model
also includes information about the factor ofproduction payments; federal, state and local
29
taxes and other expenses by the industries. These figures then become the foundation for
various income multipliers (pp. 160-161)
To analyze the impact of the new plant, the standard input-output is adjusted to
include the veneer and plywood industry as a separate industry. Then by taking the
inverse, the standard input-output table changes and organized to conduct an impact
analysis. A reduced version ofthe original inverted table is shown below.
Table 4. Reduced Version of Table Indicating the Changes in the Georgia Economy
From A New Plant
Industry Changes in:
Output Number of Personal City and State Revenue
Value Percent Employees Income County
Revenue
Agriculture .2 0 0 .1 0 0
Mining 0 0 0 0 0 0
Lumber and .7 .2 33 .2 0 0
wood products
Veneer and 5.3 12.2 318 2 0 0
plywood
Households 4.2 0 0 0 .1 .1
Source: On the Use ofInput-Output Models for Regional Planning, pp. 162-163.
To find the impact, the column in the table is multiplied by unity (representing output)
and one of the following: a ratio ofemployee to output and an income-output ratio for
one of the levels of government. Finally, each column is multiplied by a change in
industry output resulting from the new plant (p.161). This procedure for each applied to
each column and industry is described below
30
(15) bij * Vpj * ~ej
where: j represents the new industry
~ej =expected change in industry output
bij = direct, indirect and induced sales of industry i to industry j (inverse of A)
Vpj =employee-output or income -output ratio
By summing each columns bij *Vpj or Ibij *Vpj, a multiplier is found and then the
sum ofthe columns will equal the multiplier times the expected change (p. 161). Table 2
shows the results ofthe procedures listed above in calculating the various multipliers for
selected industries. The multipliers indicated in the table are calculated for a regional
economy encompassing an entire state while the previous example considered a regional
economy made up oftwo counties. The next example considers an economy even bigger
then the previous studies: a country. However, a regional study still exists because the
example calculates an impact for the various regions within the country.
In Sweden, the possible abolishment ofNuclear power and hence the closedown
of all nuclear power plants became an issue in the late 1970's. To study this impact, the
Commission on Consequences was created, and this group reported the economical
effects on employment, regional planning, the environment and household consumption
in Sweden (Albegov, Andersson and Snickaris 1982 p.361). The commission
concentrated heavily on assessing the regional impacts using various types of models.
For determining the economic impacts, the input-output model described below was
used. The impact addressed the question of the consequences of a decrease in production
and plant closings in energy dependent sectors which is the direct result ofeliminating
the power plants (Albegov et al. p. 371).
31
The model begins by letting ait' = the purchased products ofsector j region i to
produce an additional unit of output from sector I region k. Also let fik = final demand of
sales of sector i, region k; Xik = operating level in sector i, region k and 8t = labor-output
ratio in sector i, region k.
Then, if A is a n X n matrix ofait1and F and X are vectors ofthe elements above,
the following equation can be written (p. 372).
(16) X=AX+For
X =(I - A) -I F
Now, letting (I-A) -I = B
(17) X = BF
Using the equations above, a certain decrease in final demand in sector j results in an
identical decrease in output. Therefore 8/ represents the average reduction in
employment in sectorj. region 1(p. 372).
Using equation 16, the interindustry effects are determined. Now, letting L1s/ be
the reduction in employment in sector j region 1, the multiple employment loss is equal
the L1l below.
32
The results of the study were then used to determine the appropriate policy
regarding the elimination of nuclear power in Sweden.
Chapter 5 The Model
The Youngstown-Warren Econometric Model is a "labor oriented" model
consisting of72 equations, 44 stochastic and 28 identity and can be considered both a
forecasting and an impact study tool. According to its developers, Yih-Wu Liu and
Anthony Stocks, the forecasts can be used by "local development agencies in evaluating
the impacts of new business entrants or the expansion or contraction ofexisting local
firms" (p.l)
Table 4 below shows the regions diverse business sectors and the disaggregated
variables representing each one. In addition, the model's structure indicates that it can
predict changes in four main areas: employment, weekly wages, wage bill and man
hours. According to the authors, these four divisions are chosen because of the
availability ofthe data and the wide applicability of the results (Liu and Stocks, p.l).
33
Table 5 The Model's Endogenous Variables
Sectors Employment, Weekly Wages, Man Hours, Wage
Bill
Construction SECC, SWWCC, SMHCC, SWBCC
Stone, Clay and Glass SEMFD32, SWWMFD32, SMHMFD32, SWB32
Primary metals SEMFD33, SWWMFD33, SMHMFD33, SWB33
Fabricated metals SEMFD34, SWWMFD34, SMHMFD34, SWB34
Non-electric Machinery SEMFD35, SWWMFD35, SMHMFD35, SWB35
Electrical Machinery SEMFD36, SWWMFD36, SMHMFD36, SWB36
Transportation Equipment SEMFD37, SWWMFD37, SMHMFD37, SWB37
Manufacturing Nondurable Goods SEMFN, SWWFN, SMHMFN, SWBMFN
Transportation and Public Utilities SETU, SWWTU, SMHTU, SWBTU
Finance, Insurance and Real Estate SEFIR, SWWFIR, SMHFIR, SWBFIR
Services SESER, SWWSER, SMHSER, SWBSER
Government SEGV, SWWGV, SMHGV, SWBGV
Retail Trade SERT, SWWRT, SMHRT, SWBRT
Wholesale Trade SEWT, SWWWT, SMHWT, SWBWT
The flow chart in appendix B describes the ties between the sectors ofthe local
economy and the top down approach ofthis relationship. Therefore it pictorially depicts
the notion that the national economy drives the local economy with no feedback from the
local economy. It explicitly shows this specification as national exogenous variables of
man-hours, weekly wages and others driving several local sectors. In tum, these local
sectors influence other industry sectors until the loop is complete. The diagram also
shows the relationships among the four areas listed in the table above.
For example, the flow diagram shows that each business segments' employment
levels are derived from man-hours, and, then employment becomes an explanatory
variable for weekly wages. The connection continues as employment and weekly wages
combine to determine the wage bill. Finally, the Youngstown-Warren wage bill, found
by adding together the wage bill of the manufacturing and non-manufacturing sectors, is
used as an independent variable in certain nondurable sectors.
34
This relationship can also be observed by referring to the entire model in
appendix A. A close examination of the model will reveal that for each of employment,
man-hours, and weekly wages, an imprecise equation can be adopted. For example,
employment equations can be represented as:
(19) In(E) = Ao + Al *In(E(-1)) + A3 *In(MH(-1)) + A4 *T1 + As *SD1 + A6 *SD2
+ A7 *SD3+ As* DSHIFT+?I
where: MH =quarterly manhours
E =Employment
T1 =trend variable
DSHIFf =dummy variable for MSA update
SD1, SD2, SD3 =seasonal dummies
E =disturbance term
Similarly, man-hours can be reduced to a general equation, but unlike employment, most
of these equations contain a proxy variable. In this case, the proxy variable represents
real output and originates from national variables. This format is shown below.
(20) In(MH) = Bo+ BI *In(MH(-1)) + B3 *In(PV) + B4 *T1 + Bs *SD1 + B6 *SD2
+ B7 *SD3 + Bs *DSHIFT + ?2
where MH =man hour
PV= proxy variable
E 2 = stochastic term
Finally, wages in manufacturing and nonmanufacturing can be generalized by the
following equation:
35
* SD2 + Cs * SD3 + C6 *Tl + C7 * DSHIFT + E 3
where WWMF =regional average weekly wages in manufacturing
WWus =national average weekly wages in each manufacturing sector
E 3 =stochastic term
Chapter 6 Data
The model's data is broken down into fourteen business sectors and four areas of
interest per sector: employment, weekly hours, weekly earnings and the wage bill. From
its inception in the early 1980's, the model has consistently generated forecasts for the
Youngstown - Warren regional economy; therefore data for the variables were already
collected up to the last forecast (1996). Preliminary steps were then taken to update all
exogenous and endogenous variables from the 1996 period to date.
For the endogenous variables, this process began by locating and gathering two
important sources of information: The Labor Market Review and Covered Employment
and Payroll. The Labor Market Review is published by The Ohio Bureau of Employment
Services and records monthly employment, average weekly wages, and average weekly
hours for designated Metropolitan Statistical Areas (MSA's) and the state of Ohio. The
Ohio Bureau of Employment Services also publishes the second source, Covered
Employment and Payroll. This particular publication supplies income and employment
data for the various business sectors of Ohio's MSA's. Using these two publications, data
1 For the nonmanufacturing equation, this variable will become In(WWMF).
36
was recorded form 1996 through second quarter 1998, overlapping some of the existing
data. This process allowed for adjustments on any figures that were previously
estimated. After obtaining each sector's employment, average weekly hours and average
weekly wages, the information was then used to calculate the same sector's wage bill and
man-hours. The following equations below describe the process.
(22) WBi=WWi*E*13
where: WBj =the Wage Bill in the ith sector
WWj =the average weekly wage ofthe ith sector
Ej =the employment in the ith sector
(23) MHi =WHi *E *13
where: MHi = the man hours in the ith sector
WHj = the average weekly hours in the ith sector
Special proxies and conversions were also used for those sectors that lacked
available data; these particular variables and their origins are shown below.
37
Table 6 Conversions For Endogenous Variables
Employment
Variable Method ofCalculation
SEMFD36 (Electrical Machinery) 10% of Durable Goods
SEMFD32 (Stone, Clay and Glass) 6% of Durable Goods
Weekh Wages
Variable Method of Calculation
SWWMFD32 (Stone Clay and Glass) Ohio as ProXy
SWWMFD36 (Electrical Machinery) Ohio as Proxy
SWWFIR (Finance, Insurance and Real Estate) SWWFIR =SWBFIR / (SESER*13)
SWWSER (Services) SWWSER =SWBSER / (SESER * 13)
SWWGV (Governent) SWWGV =SWBGV / (SEGV *13)
SWWCC (Construction) Ohio as ProXy
SWWTU (Transportation and Public Utilities) SWWTU =SWBTU / (SETU *13)
Weeki Hours
Variable Method ofCalculation
SWHSER (Services) Average of Ohio data including Business, Health
and Engineering
SWHMFD36 (Electrical Machinery) Ohio as Proxy
SWHFIR (Finance, Insurance and Real Estate) National Data as Proxy
SWHCC (Construction) Ohio as Proxy
SWHTU (Transportation and Public Utilities) National Data as Proxy
SWHMFD32 (Stone, Clay and Glass Ohio as Proxy --
After finishing updating all endogenous variables, attention then shifted towards
the exogenous variables. This process entailed searching through the various sources
listed on the next page.
38
Table 7 Sources For Endogenous Variables
Exogenous Variables Source
SMSACPI (Consumer Price Index for Youngstown Monthly Labor Review (Average of Pittsburgh and
- Warren) Cleveland)
UIIPR (US Production Index in Primary Metals) Federal Reserve Bulletin
UIPIMVP(US Production Index for Motor Vehicles Federal Reserve Bulletin
and Parts)
UMHMFD36 (US Man Hours in Manufacturing Federal Reserve Bulletin
Center in Electrical Machinery)
UMHMFD35 (US Man Hours In Non-Electric Federal Reserve Bulletin
Machinery)
UMHMFD33 (US Man Hours in Manufacturing Federal Reserve Bulletin
Center in Primary Metals)
USCMI (US Conventional Mortgage Rates on New Federal Reserve Bulletin
Homes Including Finance Charges)
UWWMFD37 (US Average Weekly Wage in Internet (BLS website)
Transportation Equipment)
UWWMFD35 (US Average Weekly Wage in Non- Internet (BLS website)
Electric Machinery)
UWWMFD33 (US Average Weekly Wages in Internet (BLS website)
primary Metals)
UWWMFD32 (US Average Weekly Wages in Internet (BLS website)
Stone, Clay and Glass)
In examining Table 7, the consumer price index was found by taking the average
price index of Pittsburgh and Cleveland. This particular method was chosen because the
Monthly Labor Review alternates the recording of the two cities' indices every other
quarter. In addition, the data gives a good approximation to Youngstown's index due to
proximity.
The table also reveals that several other variables were gathered in unique ways.
For instance, US Weekly wages in Transportation Equipment; Non-electric machinery;
Primary metals; and Stone, Clay and Glass were found using the Bureau of Labor and
Statistics website. Using this information, the final block ofexogenous variables was
39
calculated using formula 1 or updated using dummy values. The fortunate existence and
accessibility of the historical tables for the information above simplified the data 
gathering step and decreased the total time expended on this step. On the otherhand, the
most difficult task of the project concerned the development of the models in order to do
the various analyses.
Chapter 7 Methodology
General Motors has been a vital part of the economies of Mahoning and its
surrounding counties for over thirty years. During this period, the Lordstown plant, as
highlighted in the history chapter, adapted to many changes and conquered even more
challenges. Since the closing of the van plant in 1992, many questions have been raised
concerning Lordstown's permanence. In recognition ofthese questions, the project will
consider four separate scenarios, each pertaining to a separate situation.
Each situation and their corresponding question will be investigated and
conclusions will be reached by analyzing time series graphs ofpredIcted values and by
calculating multipliers. In obtaining these summary statistics, each model is first
simulated in Troll, a program devoted to forecasting and simulation, and the data is
entered in SPSS for constructing graphs. After declaring each scenario, the original
model, shown in appendix A, was altered to fit the particular situation. These changes as
well as each scenario's description is reported below.
40
Scenario 1
The first situation addresses the thoughts ofthose who believed in a possible
shutdown ofLordstown at the beginning ofthe 1990's. In examining this particular case,
several introductory steps are taken to properly set up the model. First, all exogenous and
endogenous variables ofthe Transportation and Equipment sector are removed; thereby
reducing the model to sixty-eight equations. This action is taken because employment at
the Lordstown facility occupies nearly all ofthe sector's employment totals. The
remaining employees mayor may not be associated with General Motors; however; their
inclusion or deletion will hardly affect the results. Finally, after eliminating the sector,
aggregate values are recalculated and renamed to express the entire shock. This step
concludes the preliminary work needed to prepare the model for simulation, and, to
separate it from other models, it is renamed SMSA2 .
With the original model altered, in Troll, an OLS (Ordinary Least Squares)
procedure is performed from the first quarter of1981 to the fourth quarter of1997 to
estimate the equations; these dates are chosen to account for the revised statistical area2.
This procedure is consistent with this type ofmodel as shown in the literature review.
Finally, a historical simulation procedure is performed from the first quarter of1990 to
the fourth quarter of 1997, and these values will serve as comparative figures to establish
conclusions. Historical simulation uses previous values ofendogenous variables to
perform the simulation; however, the technique can only be used when data for these
variables exist. Historical simulation will therefore outperform the alternative technique
ofdynamic simulation.
41
Scenario 2
Using model SMSA2, the second scenario builds upon the results obtained in the
first one. This particular situation studies the impact of the economy where scenario 1
left off (from the first quarter of 1998) and extends this forecast until the fourth quarter of
2002. In order to run this simulation, exogenous data through the simulation period
are estimated, and two different techniques are implemented to obtain these statistical
projections. First, in SPSS, using historical values for each exogenous variable, a curve
estimation procedure is run to obtain a best-fit line. After testing several different types
of curves, the linear best-fit line produced the largest R2 values; consequently, those
values are used for the initial guesses. Similarly, in Troll, a NAINTERP command also
serves to find initial guesses for the endogenous variables. NAINTERP replaces any
missing data, represented as NA, with values calculated by "linear interpolation". The
combination of these two methods gives the necessary starting points for projecting
values. In addition, time series graphs are plotted to view the historical path of the data
and to gain further prospective on its future course. Blending the two procedures' results,
using the time series graphs and with a little intuition, values are selected for each
variable.
Since this scenario uses the same model as scenario I and an OLS procedure was
previously performed, the coefficient values of the equations are already saved.
2 Beginning in 1981, the Youngstown-Warren MSA data includes Columbiana County
42
Therefore, to conclude the analysis, the simulation task is performed and the
results are saved. The simulation task in this scenario however differs from the
simulation technique in the first scenario.
In this situation, an historical simulation does not apply--no historical data exist
for the future--so it relies on dynamic simulation. This simulation technique uses the
simulated values as the starting values ofthe endogenous variables. The technique,
although not as accurate as historical simulation, produces good results.
The next two sections bring an added dimension to this project by making an
important assumption. It assumes that Lordstown loses the bid for the Yellowstone
project and General Motor's decided to build the plant in 1990. The key feature relates to
the changing ofthe employment level in Transportation Equipment sectors from their
historical levels to no employment.
Scenario 3
Scenario 3 studies the impact ofthe Youngstown - Warren economy from the first
quarter of1990 to the fourth quarter of1997 assuming General Motors had built the new
plant in 1990. Similar to scenario 1, the model is refined to estimate this impact. First,
employment levels in the relevant sector are converted from the original levels to a
constant 2,100. Next, related variables are calculated and renamed to separate them from
their original values. Two variables however involving the transportation and equipment
sector remain the same. These include weekly wages and weekly hours variables, and it
43
is determined that a new plant probably will not have any affect on them. The altered
model is then named SMSA3 and is shown in the appendix B.
Similar to scenario 1, an OLS procedure is run to solve for the coefficients in the
various stochastic equations, and, after, a simulation command is performed. The results
from both these procedures are saved for future forecasts and analysis.
Scenario 4
The last scenario extends the research in the third scenario and simulates from the
first quarter 1998 to the fourth quarter 2002. Using the same predicted exogenous
variables as described in scenario 2 and the same coefficient values as estimated in
scenario 3, a simulation task is run and the results are saved.
The results will ultimately declare the intensity ofthe future impact on the market
with a new plant. Both scenario 2 and this case relate to the present situation of General
Motors, Lordstown; therefore their results will be more important to this study.
44
Chapter 8 Results
The results for the scenarios described earlier are shown below. Each of these
situations presents a separate case involving an alteration of General Motors Lordstown's
current operations. Therefore, each circumstance carries its own specifications and
conclusions, and this chapter is divided into four sections to handle each case separately.
The following terms will explain the language contained in both the explanations and the
sources used for conclusions.
Definitions:
Original Model (SMSA) =the model assuming the current operations of
Lordstown
Excluded Model (SMSA2) =the model without Lordsown
Reduced Model (SMSA3) =the model with a new assembly plant
In addition to these definitions above, the reader should be aware that all results
presented here have been annualized from quarterly values to make the data more
meaningful and manageable. Moreover, some of these results are embedded in the text
while others are located in Appendix C. Also, as noted in the last section, the model has
been producing forecasts for over fifteen years; therefore, from the beginning, a
reliability assumption is made about the model. In respect to this decision, no regression
results nor mean percentage errors are reported.
45
Scenario 1
In brief, SMSA2 describes the impact on the Youngstown - Warren economy
during the 1990's, assuming that General Motors Lordstown closed operations at the
beginning of the decade. The graphs of aggregate variables below give a visual
representation ofthis presumption.
Figure 1 Total Employment
SMSA Employment (in Thous) for Various
Models, 1990 - 2002
260
+-'c 250
~ >.
o 240a.
Ew
ro 230
:::lc
~ 220
! --------::--
.'-- "+-----:::--"
...- - -::-=. -.- _........... ----..... --::::--'" I
..... "
,/, .. - I./,' I
./"
......... ":... ' I-------,,- I
----------------
SESMSA
SESMSA2
210 SESMSA3
1990 1992 1994 1996 1998 2000 2002
1991 1993 1995 1997 1999 2001
YEAR
46
Figure 2 Total Wage Bill
SWBSMSA
SWBSMSA2
SWBSMSA3
2002
2001
1996 1998 2000
1997 1999
1994
1993 1995
SMSA Wage Bill (in Thous) for Various
Models, 1990 - 2002
I
II
11 ()()()()()
1000000i---,r----r--y--r-----.---r--y-----J1-----.----r---r-----w
1990 1992
1991
1800000
17()()()()()
i:ii 1600000
~ 15()()()()()ctS:s:
14()()()()()
ctS 13()()()()():::J
C.a:
12()()()()()
YEAR
The time series plots above follow the movement of the economy's total
employment and total wage bill through the year 2002. The reference line at the year
1998 indicates where the break occurs between simulated values and historical values.
As shown in the legend, the solid curves refer to the original model while the dotted and
dashed curves represent the excluded and reduced models respectively, all of which were
described earlier. Each point then represents annualized data ofa specific model over the
relevant time period
In analyzing these two graphs for scenario 1, a few interesting patterns emerge.
For the graph oftotal employment, the original model's curve (SESMSA) and the
excluded model's curve (SESMSA2) seem nearly parallel. This picture then suggests that
47
the curves have similar slopes at various tangent lines along the curves. Since the slope
ofa line is equivalent to equation 24 below, and the slopes of the graphs appear the same,
the L1E would be almost equal. From this analysis, it may then be hypothesized that the
economy, during this period, shows no signs of recovering the initial jobs lost from the
plant closing down.
(24) tillL1T
where: E = Employment
T=Time
This type of conclusion may be a realistic outcome for several reasons. For one,
the majority of workers with seniority at General Motor plants have clauses in their
contracts that guarantee them jobs in the event of a complete shutdown. At Lordstown,
many workers at the plant fall in this category; consequently, they can escape the
hardships and frustrations offinding new jobs. Therefore, the existence of a guaranteed
work clause may lead to a departure from the area and relocation at another plant. With
this situation, it may not be necessary for these people to search for local work; therefore,
the community will not be responsible for recovering these lost jobs. This occurrence
may also culminate in an increase in residential homes for sale and, importantly, a
decrease in overall consumer spending in the area. This last reason especially needs
addressed since the average pay for a General Motor employee is generally higher than
most other manufacturing firms. In sum, a comparison of the curves says a lot about the
ability ofthe community to recoup any losses from the economic shock.
48
Instead ofcomparing graphs, by examining only the shape ofthe excluded
model's curve, some positive remarks can be made. For instance, the graph shows an
upward trend for this period, indicating that, aside from the initial shock, the economy
has rebounded quickly. This fact is evident between the years 1993 and 1996 when a
significant increase occurs and afterwards when a more moderate increase occurs.
Besides time series plots, another way to conduct this analysis is to examine the
actual data. The tables in appendix C shows the data plotted in both figures for the
original and excluded models. In addition, the changes in absolute and percent changes
are listed to describe the differences. Given the data, an employment multiplier can be
calculated that indicates the strength ofthe impact on employment. In this case, the
initial reduction in employment in transportation and equipment (8,175) in 1990 caused
an 8,400 decrease in employment or a multiplier of1.03 by the end ofthat same year.
Intuitively, this number means that for a 1,000 change in employment, total employment
would change by 1,030. The weak multiplier effect shows that little linkage exists among
the various sectors ofthe economy. A further analysis ofmultipliers can also take the
form ofexamining the various sectors to determine where the multipliers rest. In this
scenario, by examination ofthe manufacturing and non-manufacturing sector tables in
appendix C, the impact is shown to take place exclusively on the manufacturing sector.
A further analysis, within the manufacturing sector, comparing durable goods and non 
durable goods shows that the change in employment had a heavier impact on the former.
Derived from employment, the wage bill also has very powerful explanatory
capabilities. For example, the variable is used to approximate the amount oftotal income
earned in the economy. From these figures, two important conclusions can be drawn.
49
First, income or its proxy variable the wage bill determines the amount oftaxable
income generated in the economy. Second, with taxes withdrawn, disposable income
determines how much money is available for consumption on goods and services.
Because car manufacturing jobs pay considerably more than other manufacturing jobs, it
is expected that total emploYment loss in the sector to have adverse effects on the wage
bill. Indeed, by simultaneously examining figures 1 and figures 2 and the tables in
appendix C, this contention is confirmed.
From the tables in appendix C, it can be found verified that the wage bill dropped
significantly during 1990, decreasing nearly $78,500,500 or 7%. From these figures, an
income multiplier can be calculated indicating the degree ofthis change. Thus the loss of
General Motors Lordstown during this period resulted in an initial change in the wage bill
of$78,487,900 and comparing this figure to the tota110ss in the wage bill results in a
multiplier near one. This figure indicates that the changes in total wage bill and the wage
bill oftransportation and equipment were nearly equal. A further analysis ofthis drop in
the wage bill can also be obtained by examining figure 2.
When comparing the original (SWBSMSA) and excluded model (SWBSMSA2)
curves, a similar pattern as in emploYment emerges. The curves remain almost parallel,
again indicating that the market does not seem to be making up for the initia110sses in
total income.
While the economy does not seem to recoup the initia110ss in income from the
plant closing down, the inference that tax money will be significantly reduced and never
return to previous levels can also be made. The initial drop in income can now be used to
draw conclusions on its effects on the economy.
50
As stated before, a drop in income decreases the amount of taxes received by
local, state and federal governments. In this case, the impact will affect the local
governments, especially the village of Lordstown. Depending on the particular tax rate,
the local governments combined can lose millions of dollars. A series of consequences
can arise including a decrease in spending on public goods or governments cutting
funding in programs to make up for the losses.
Another significant effect occurs in the drop in consumer spending. This
particular occurrence can have a large long-term multiplier effect as businesses fold due
to the decrease in business activity. In total, the twofold impact has adverse effects on all
aspects ofthe economy.
The graphs and thus the economy however could have demonstrated different
patterns. For instance, a divergent series, as shown in figure 4, could have developed. In
this type ofsituation, employment in the excluded mode! declines while the employment
level in the original model continues to rise. With this phenomenon, the loss of
Lordstown would result in much bigger employment losses (higher multipliers) and no
signs ofrecovery.
Figure 3 Convergent Series
EMP
SMSA
/o~"
'---------T
51
Figure 4 Divergent Series
EMP SMSA
OTHER
'---------T
A second type ofpossibility would be a strong convergent series as shown in
figure 3. In this case, the economy has rebounded quickly and any unemployed workers
are able to find jobs elsewhere. In comparison ofthe two graphs, this situation represents
the ideal situation for any region experiencing business loss.
Scenario 2
Scenario 2 extends the notion ofa complete shutdown by examining the impact
through the year 2002. The results ofthe simulation appear in figure 3 and figure 4, and
the dashed line boundary at the year 1998 separate it from the first scenario. Notice that
in extending this simulation well beyond the present year, most ofthe conclusions
reached in the first scenario continue to hold. In reference to these figures, any future
projections beyond one period should be handled very carefully since the probability of
miscalculating numbers increase with each successive period forecasted.
As mentioned in scenario 1, the graphs ofthe excluded and the original model
remain nearly parallel throughout the 1990's; however, beyond this period, some changes
start to occur. The graphs begin to show signs ofconverging, especially toward the end
ofthe simulation period. Therefore, while convergence did not occur for the majority of
the time series, in a longer time span, the economy is beginning to show signs of
regaining those lost jobs.
An examination ofthe excluded model's curve indicates that employment
continues to rise throughout this period; therefore, the effects ofGeneral Motors,
according to the forecasts, did not significantly hamper business growth. In fact, this
52
curve is nearly a positive straight line, indicating a relatively constant growth rate through
the 1990's.
In order to gain perspective on the impact ofthe closing during the simulation
period, an employment multiplier can be calculated. In 1997, a decrease in employment
in transportation and equipment of8,270 caused a reduction in employment in 1998 of
8,380 or a multiplier of1.01. Therefore a 1,000 decrease in total employment would
cause employment to decrease by 1,001.
Looking at appendix C, this impact can be compared among several sectors. For
instance, the impact between employment in non-manufacturing and manufacturing and
employment in manufacturing non-durable goods and manufacturing durable goods
follow scenario 1's findings. Therefore, the brunt ofthe impact was concentrated in
manufacturing and specifically manufacturing durable goods. Additionally, the results
from comparing non-manufacturing to manufacturing say that linkages among these two
sectors are weak.
Similar to the findings on employment, the graphs ofthe wage bill remain
parallel. One ofthe major differences however is that there are no signs ofconverging;
thus the wage bill continues to rise and at the same rate as the original model. This
parallel course indicates that the initial income lost from the shutdown is not being made
up in the future.
Along with this analysis, a multiplier can be calculated to gain further information
ofthis impact. The wage bill multiplier in this case is calculated by taking the change in
the wage bill oftransportation equipment in 1997 (-$115,871,500) and describing its
53
changes on the total wage bill in 1998. The multiplier then is .93 or less than the
multiplier obtained in scenario 1.
Scenario 3
This particular section takes a more optimistic view ofthe fate ofLordstown. In
this and the last scenario, it is assumed that Lordstown has won the bid for a new plant,
and employment in the relevant sector remains at 2,100 (the average number of
employees at Yellowstone plants). Different from scenario 4, this situation analyzes the
impact ofthis type ofplant on the economy had it been built during the early 1990's.
Again, referring to figure 1 oftotal employment, the dashed curve SESMSA3
describes this scenario's course through the year 2002. Similar to scenario 1, the
excluded model, it remains parallel to the original model's curve, and the curve, as
expected, lies above the curve plotted without Lordstown. In relation to the difference in
employment between the original model and the reduced model, in 1990, Lordstown had
an average employment of8,200 workers, and with a new plant, employment is reduced
to 2,100.
Since parallel curves exist, similar conclusions to those listed in scenario 1 are
considered. These include, among others, that the economy does not appear to recover the
losses in employment from the shutdown--at least in this time period. With a smaller
employment loss than in scenario 1, however, it is assumed that a lower employment
multiplier will occur. In applying this multiplier, the initial decrease in 1990 in
employment in transportation and equipment amounts to 8,200, and the total employment
54
reduction in that same year is 5,930. From this data, a multiplier of.72 is achieved and
concurs with the expectations. In comparing this impact among the manufacturing and
non-manufacturing and the durable goods and non-durable goods sectors, as witnessed
before, the manufacturing and, within manufacturing, the durable goods sectors
experience more ofthe impact.
In examining figure 2 ofthe wage bill, it again gives much ofthe same results as
seen in the first scenario. For example, the curves remain parallel to the original model's
curve with Lordstown, indicating that the initial reduction in income is not being
recouped in later years. However, because the reduced model's curve has a steady
incline, there is no indication ofany additional adverse effects on the economy.
Therefore, even though the new plant reduces total income, the effects seem to pass
quickly.
In order to determine the strength ofthe impact, as before, an income multiplier
is calculated. The initial change in income in 1990 in transportation and equipment
(-$78,487,900) is then compared to the change in the total wage bill in 1990
(-$76,183,250). Using this data, a multiplier of.97 is obtained, indicating a less then
equal change in total income from a change in the individual's sector income.
Additionally, from the tables in appendix C, it is shown that the majority ofthe
impact is felt by the manufacturing sector and specifically for durable goods.
55
Scenario 4
With a relatively steady rise in total employment and total wage bill for the
reduced model from 1990 to 1997, it is safe to assume similar predictions for the future.
Looking at figure 2 and figure 3 and beyond the reference line, the conclusions are
corroborated. For total employment, simulated future values show the extension ofthe
general trend in the 1990's; however the curve is beginning to show signs ofconvergence
with the original model's curve. In response to the reduced model's total employment
curve, the trend again indicates that business growth occurs throughout this period, and
the economy continues to be healthy. Next, because the curve is beginning to converge
with SESMSA, similar reasons discussed in scenario 1 apply. Thus the economy is
beginning to make up the loss in employment from the building ofa smaller plant. Since
this situation applies to the present situation ofLordstown, the impact multiplier is
especially relevant to this project
In determining the multiplier, the initial impact within the transportation and
equipment will occur in 1997 while the total employment change will be calculated for
1998. For transportation and equipment, the change in employment in the above year is
-8,100, and this response on total employment results in a decrease of6,400. The
employment multiplier then becomes .79.
In analyzing the impact ofthis model on the wage bill, similar strategies are used
as those used on employment. Notice, from figure 2, the trend continues from its 1990
course with relatively parallel lines existing between all three models. The paths
however are extremely linear and therefore have constant growth throughout the period.
56
In addition, the impact or the job loss continues to have no additional ramifications on the
economy. Again, to empirically describe this impact, an income multiplier is calculated.
Using the same years as for employment, the reduction in transportation and equipment is
$115,871,500, and the decrease in the total wage bill in $91,943,300 or a multiplier of
.79. The impact as seen in the appendix is again concentrated in manufacturing and
specifically the manufacturing durable goods.
Chapter 9 Conclusions
The idea for this project originated in response to General Motor's announcement
ofits plans to restructure its assembly line methods. Since the late 1980's, questions
about General Motor's permanence at Lordstown have consistently been addressed. The
goal ofthis project then was to provide some statistical answers to these questions,
expressed within four distinct scenarios. Using SMSA data for the Youngstown - Warren
region and a previously defined model ofthis area, simulations were run for each
possible shutdown case and the results organized to provide the conclusions. Then, from
this data, several conclusions were given concerning the overall effects on employment
and the wage bill.
First, for all four scenarios, the impact concentrated strictly on the aggregate
sector ofdurable goods with little leakage to the other sectors ofthe economy. As noted
before, this result reveals that the other aggregated sectors (non-durable goods and non 
manufacturing for example) are not very dependent on Lordstown for business.
57
Second, for both employment and income (wage bill), the graphs ofthe original
model and the changed models are nearly parallel to each other and for most ofthe time
period. Once again, this observation leads to the beliefthat the economy is struggling to
make up for the initial losses.
Finally, the low values calculated for the multipliers indicate that the initial
impact ofeach situation dominates the total impact with little repercussions elsewhere.
This outcome affected all four scenarios in the study. General Motor's influence on other
businesses may therefore be less than what was expected.
When examining all the scenarios as a whole, the conclusion that no matter what
situation Lordstown faces, the economy will respond with continued growth can be
reached. However, within the scope ofthe model, the economy will struggle to recover
the initial losses. Although General Motors is important to the study's economy, the
above analysis indicates that the economy will wholly survive a shutdown.
58
Chapter 10 Limitations
The limitations surrounding this impact study--and all impact studies--are varied.
One ofthe most important ofthese constraints rests with the dilemma ofobtaining
inaccurate forecasts, especially with those residing in the future. In this project,
simulations were extended four years past the present; therefore, those predictions three
or four years in the future will probably not produce accurate predictions. These errors
originate from the estimation that occurs for the exogenous variables, which are part of
the preliminary steps to running a dynamic simulation procedure. Thus, without
historical values, the forecasters must rely on computer-generated approximations and
intuition to provide these values. Additionally, as mentioned in the methodology, certain
variables are used as proxies for the regional variables. These replacements may then
misrepresent the true values ofthe variables and correspondingly produce inaccurate
predictions for these variables.
A specific limitation on the data and the predictions occurs when considering the
impact ofthe new plant. Unequivocally, the Yellowstone project will provide a
completely new plant and drop employment in the assembly plant to 2,100; however,
unknown is the degree in which this plant stimulates additional growth. For example, the
new plant may bring branch plants (paint spraying, for example) that assist in the
assembly methods. Additionally, the plant may attract new businesses that establish
operations in close proximity and provide the modules for assembly. The model takes a
"worst case scenario" stand on this issue by only considering the influence ofa new
59
assembly plant. The absence of any additional items occurs because of the unavailability
ofthe data and extremity ofassumptions.
Related, no model can precisely describe the workings of the economy and
completely predict all changes in it. It can then be assumed that errors are inherently
present in modeling. Keeping this idea in mind, errors then will naturally emanate from
the model and circulate to the predicted values. However, most models will generate
predictions that are relatively close to the actual values, measured by the MAPE. This
model falls in this category. When forecasting in the future, an "accurate" model will
provide a person with good approximations to the actual data; however, nothing can be
more reliable than an ex post analysis. This project may be a good candidate for this type
of study since one of these scenarios will become reality.
60
Appendix A: The Model
Model SMSAI
ENDOGENOUS
SECC SEFIR SEGV SEM SEMFD SEMFD32 SEMFD33 SEMFD34 SEMFD35 SEMFD36 SEMFD37
SEMFN SENM SERT SESER SESMSA SETU SEWT SMHCC SMHFIR SEMHGV SMHM SMHMFD
SMHMFD32 SMHMFD33 SMHMFD34 SMHMFD35 SMHMFD36 SMHMFD37 SMHMFN SMHNM
SMHRT SMHSER SMHSMSA SMHTU SMHWT SWBCC SWBFIR SWBGV SWBM SWBMFD
SWBMFN SWBNM SWBRT SWBSER SWBSMSA SWBTU SWBWT SWB32 SWB33 SWB34 SWB35
SWB36 SWB37 SWWCC SWWFIR SWWGV SWWM SWWMFD SWWMFD32 SWWMFD33
SWWMFD34 SWWMFD35 SWWMFD36 SWWMFD37 SWWMFN SWWNM SWWSER SWWSMSA
SWWTUSWWWT
EXOGENOUS
DSHIFT SOl SD2 SD3 SMSACPI Tl UIIPR UIPIMVP UMHMFD33 UMHMFD35 UMHMFD36
USCMI UWWMFD32 UWWMFD35 UWWMFD37
EQUATIONS
1: LOG(SMHMFD33) = Al + A3 * LOG(SMHMFD33(-I) + A4 * T1 + A5 * SOl + A6 * SD2 + A7 *
SD3 + A9 * UIIPR + AlO * DSHIFT
2: LOG(SMHMFD34) = G1 + G2 * LOG(UMHMFD33(-I)) + G3* LOG(SMHMFD34(-I)) + G6 * SOl +
G7 * SD2 + G8 * SD3 + G9 * Tl+ GlO * DSHIFT
3: LOG(SMHMFD35) =BI + B2 * LOG(UMHMFD35(-I)) + B3* LOG(SMHMFD35(-l)) + B4 * SOl +
B5 * SD2 + B6 * SD3 + B7 * TI+ B8 * DSHIFT
4: LOG(SMHMFD36) =01 + D2 * LOG(UMHMFD36(-I)) + D3* LOG(SMHMFD36(-I)) + D6 * SDI +
D7 * SD2 + D8 * SD3 + D9 * TI+ DIO * DSHIFT
5: LOG(SMHMFD37) =Cl =C2 * LOG(SMHMFD37(-I)) + C3 * SOl + C4 * SD2 + C5 * SD3 * SD3 +
C8 * LOG(UIPIMVP) + C6 * DSHIFT + C7 * 1'1
6: LOG(SMHMFD32) = FI + F3 * LOG (SMHMFD32(-I)) + F4 * Tl + F5 * SOl + F6 * SD2 + F7 *
SD3 + F2 * DSHIFT
7: LOG(SMHMFN): EI + E3 * LOG(SMHMFN(-I)) + E5 * SDI + E6 * SD2 + E7 * SD3+ E4 *
DSHIFT + E8 * Tl
8: LOG(SMHCC) = KI + K2 * LOG(USCMI)-2)) + K3 * LOG(SMHCC(-I)) = K4 * SDI + K5 * SD2 +
K6 * SD3+ K7 * DSHIFT + K8 * T1
9: LOG(SMHWT) =KlO + K12 * LOG(SMHWT(-I)) + K13 * T1 + KI4 * SOl + K15* SD2 + K16 *
SD3 + Kll * DSHIFT
10: LOG(SMHRT) =Ml + M2 * LOG(SMHRT(-I)) + M3 * LOG(SWBSMSAlSMSACPI) + M4 * T1 +
M5 * SOl + M6 * SD2 + M7 * SD3 + M8 * DSHIFT
11: LOG(SMHTU) =NI + N2 * LOG(SMHTU(-I)) + N3 * LOG(SMHMFD) + N4 * T1 + N5 * SOl +
N6 * SD2 + N7 * SD3 + N8 * DSHIFT
12: LOG(SMHSER) =Ll + L2 * LOG(SMHSER(-I)) + L3 * T1 + L4 * SOl + L5 * SD2 + L6 * SD3 +
L7 * DSHIFT
61
13: LOG(SMHFIR) =LIO + LII *LOG(SMHFIR(-I)) + LI2 *LOG(SWBSMSA/SMSACPI) + LI3 *T1
+ LI4 *SDI + LI5 * SD2 +LI6 * SD3 + LI7 *DSHIFT
14: LOG(SMHGV) =PI + P2 *LOG(SMHGV(-I)) + P3 *Tl + P4 *SDI + P5 *SD2 + P6 *SD3 + P7 *
DSHIFT
15: SMHMFD =SMHMFD32 + SMHMFD33 + SMHMFD34 + SMHMFD35 + SMHMFD36 +
SMHMFD37
16: SMHM =SMHMFD + SMHMFN
17: SMHNM =SMHCC+ SMHWT + SMHRT+ SMHTU + SMHSER + SMHGV + SMHFIR
18: SMHSMSA =SEM + SENM
19: LOG(SEMFD34) =G21 + G22 *LOG(SEMFD34(-I)) + G23 *LOG(SMHMFD34) + G24 *Tl +
G26 *LOG(SMHMFD34(-I)) + G27 * SDI + G28 *SD2 + G29 * SD3 + G30 *DSHIFT
20: LOG(SEMFD33) =A21 + A22 *LOG(SEMFD33(-I)) + A23 *LOG(SMHMFD33) + A24 *T1 +
A25 *LOG(SMHMFD33(-I)) + A26 * SDl + G27 * SD2 + G28 * SD3 + G29 *DSHIFT
21: LOG(SEMFD32) = F21 + F22 *LOG(SEMFD32(-I)) + F23 *LOG(SMHMFD32) + F24 *TI + F26*
LOG(SMHMFD33(-I)) + F27 * SDl + F28 * SD2 + F29 *SD3 + F30 *DSHIFT
22: LOG(SEMFD37) = C21 + C22 * LOG(SEMFD37(-I)) + C23 *LOG(SMHMFD37) + C24 *T1 +
C26 *LOG(SMHMFD37(-I)) + C27 * SDI + C28 * SD2 + C29 *SD3 + C30 *DSHIFT
23: LOG(SEMFN) =E21 =E22 *LOG(SEMFN(-I)) + E23 *LOG(SMHMFN) =E24 *T1 + E25 *SDl
+ E26 * SD2 + E27 *SD3 + E28 *DSHIFT
24: LOG(SECC) =K21 + K22 *LOG(SECC(-I)) + K23 *LOG(SMHCC) + K24 *LOG(SMHCC(-1)) +
K25 *T1 + K26 *SDl + K27 *SD2 + K28 * SD3 + K29 *DSHIFT
25: LOG(SEWT) ==K31 + K32 *LOG(SEWT(-1)) + K33 *LOG(SMHWT) + K34 *LOG(SMHWT( 
1)) + K35 *T1 + K36 * SDl + K37 * SD2 + K38 * SD3 + K39 *DSHIFT
26: LOG(SERT) ==M21 + M22 *LOG(SERT(-1)) + M23 *LOG(SMHRT) + M24 *LOG(SMHRT(-I))
+ M25 *T1 + M26 * SDI + M27 * SD2 + M28 *SD3 + M29 *DSHIFT
27: LOG(SETU) ==N21 + N22 *LOG(SETU(-I)) + N23 *LOG(SMHTU) + N24 *LOG(SMHTU(-I))
+ N25 *Tl + N26 *SDI + N27 * SD2 + N28 * SD3 + N29 *DSHIFT
28: LOG(SEMFD35) = B21 + B22 * LOG(SEMFD35(-I)) + B23 *LOG(SMHMFD35) + B24 *
LOG(SMHMFD35(-I)) + B25 *SDI + B26 *SD2 + B27 *SD3 + B29 *DSHIFT
29: LOG(SEMFD36) =D21 + D22 *LOG(SEMFD36(-I)) + D23 *LOG(SMHMFD36) + D24 *
LOG(SMHMFD36(-I)) + D25 * SDl + D26 *SD2 + D27 * SD3 + D29 *DSHIFT
30: LOG(SESER) =L21 + L22 *LOG(SESER(-I)) + L23 *LOG(SMHSER) + L24 *T1 + L25 *SDl +
L26 *SD2 + L27 * SD3 + L28 *DSHIFT
31: LOG(SEFIR) = L31 + L32 *LOG(SEFIR(-I)) = L33 *LOG(SMHFIR) + L34 *LOG(SMHFIR(-I)) +
L35 *T1 + L36 * SDI + L37 * SD2 + L38 *SD3 + L39 *DSHIFT
62
32: LOG(SEGV) =P21 + P22 * LOG(SEGV(-I? + P25 * SOl + P26 * SD2 + P27 * SD3 + P23 *
LOG(SMHGV) + P24 * Tl + P28 * DSHIFT
33: SEMFD =SEMFD32 + SEMFD33 + SEMFD34 + SEMFD35 + SEMFD36 + SEMFD37
34: SEM = SEMFD + SEMFN
35: SENM = SECC + SEFIR + SEWT + SERT + SETU + SESER + SEGV
36: SESMSA = SEM + SENM
37: LOG(SWWMFD32) = F31 + F32 * LOG(SWWMFD32(-I? + F33 * LOG(UWWMFD32) + F34 *
SDI + F35 * SD2 + F36 * SD3
38: LOG(SWWMFD33) = A31 + A32 * LOG(UWWMFD33) + A33 * SOl + A34 * SD2 + A35 * SD3
39: LOG(SWWMFD34) = G31 + G32 * LOG(SWWMFD34(-1)) + G33 * Tl + G34 * SOl + G35 * SD2
+ G36 * SD3 + G38 * LOG(l/SEMFD34)
40: LOG(SWWMFD35) = B31 + B32 * LOG(SWWMFD35(-I? + B33 * LOG(UWWMFD35) + B34 *
SOl + B35 * SD2 + B36 * SD3 + B37 * LOG(l/SEMFD35)
41: LOG(SWWMFD36) =D31 + D32 * LOG(SWWMFD36H? + D34 * Tl + D35 * SDI + D36 * SD2
+ D37 * SD3
42: LOG(SWWMFD37) = C31 + C32 * LOG(SWWMFD37(-I)) + C33 * LOG(UWWMFD37) + C34 *
SD 1 + C35 * SD2 + C36 * SD3 + C37 * LOG( IISEMFD37)
43: SWWMFD = (SWWMFD32 * SMHMFD32 + SWWMFD33 * SMHMFD33 + SWWMFD34 *
SMHMFD34 + SWWMFD35 * SMHMFD35 + SWWMFD36 * SMHMFD36 + SWWMFD37 *
SMHMFD37)1 SMHMFD
44: LOG(SWWCC) = K41 + K42 * LOG(SWwcq-1)) + K43 * Tl + K44 * SOl + K45 * SD2 + K46 *
SD3 + K47 * LOG(IISECC)
45: SWWWT= K50 + K51 * SWWWT(-I) + K52 * Tl + K53 * SOl + K54 * SD2 + K55 * SD3
46: LOG(SWWRT) = M31 + M32 * LOG(SWWRT(-1)) + M33 * SD1 + M34 * SD2 + M35 * SD3 +
M36 * LOG(SWWMFD)
47: LOG(SWWTU) =N31 + N32 * LOG(SWWTU(-I? + N33 * Tl + N34 * SOl + N35 * SD2 + N36 *
SD3 + N37 * LOG(SWWMFD)
48: LOG(SWWFIR) = L41 + L42 * LOG(SWWFIR(-I? + L43 * SDI + L44 * SD2 + L45 * SD3 + L46 *
LOG(SMSACPI)
49: LOG(SWWSER) = L51 + L52 * LOG(SWWSER(-I)) + L53* Tl + L54 * SOl + L55 * SD2 + L56 *
SD3 + L57 * LOG(l/SESER)
50: LOG(SWWGV) =P31 + P32 * LOG(SWWGV(-I? + P33* SOl + P34 * SD2 + P35 * SD3 + P36 *
LOG(l/SEGV)
51: SWWMFN = E40 + E41 * SWWMFN(-I) + E42 * SWWMFD + E43 * Tl + E44 * SOl + E45 * SD2
+ E46 * SD3
52: SWWM = (SWWMFD * SMHMFD + SWWMFN * SMHMFN) I SMNM
63
53: SWWNM = (SWWCC *SMHCC + SWWRT *SMHRT + SWWTU *SMHTU + SWWFIR *
SMHFIR + SWWSER *SMHSER + SWWWT *SMHWT + SWWGV *SMHGV) / SMHNM
54: SWWSMSA = (SWWM *SMHM + SWWNM *SMHNM) / SMHSMSA
55: SWB32 =SEMFD32 * SWWMFD32 * 13
56: SWB33 =SEMFD33 *SWWMFD33 * 13
57: SWB34 = SEMFD34 *SWWMFD34 * 13
58: SWB35 = SEMFD35 *SWWMFD35 * 13
59: SWB36 =SEMFD36 *SWWMFD36 * 13
60: SWB37 = SEMFD37 * SWWMFD37 * 13
61: SWBCC =SECC * SWWCC * 13
62: SWBMFN =SEMFN *SWWMFN * 13
63: SWBTU =SETU * SWWTU * 13
64: SWBFIR =SEFIR *SWWFIR * 13
65: SWBSER = SESER *SWWSER * 13
66: SWBGV = SEGV *SWWGV * 13
67: SWBRT =SERT *SWWRT * 13
68: SWBWT = SEWT * SWWWT 8 13
69: SWBMFD =SWB32 + SWB33 + SWB34 + SWB35 + SWB36 + SWB37
70: SWBM = SWBMFD + SWBMFN
71: SWBNM = SWBCC + SWBTU + SWBFIR + SWBSER + SWBRT + SWBWT + SWBGV
72: SWBSMSA =SWBM + SWBNM
64
>"'d
"'d(D
~0-.....
><
to~
tlanUfactUrfnl
IialeBillI
~ nurableGoodsBill
Local Economic Activity In,
Durable~Goods
WeeklyI
rlwage
National !nputs
Illlrnble Goods 1>1Irabie
tlnnhours Goods
Industrial Pro- .~t.::la::n::h:.:o:.:u:.:r,-=s:..... ..J
duct Ion Index
Hanufacturfng nurables
Ileek Iy Wage ) ,
Fi/:III'., I: A Simplified n",,,, Chart of the Youngstown-liarren Econometric ~I(\del
Transportation Transportation Transportation
, Public Util- , Public Util- , Public Util-
ities Hanhours ities Wage ities Wage Bill
J.
~
'"rj........
o
~
(1=r
~o
I-+).....
=r(D
~o
~
~(JQ
\/:J.....o
~
~
~
~(D
~m
(')o
~o
~
H YOUniit~~Warren SHSAWage Bill
HNonmanuf8cturlngl/ageBill
~~ NondurableGoodsWage Bill
J
I IConstruction
?-+JWage Bill
;onstruction
leekly Wage
Wholes81e , Re- Wholesale -,-Re-
tail Trade , Serv tail Trade , Serv-
ices Weekly Wage ices Wage Bill
Government
Wage Bill
Finance, Insur- Finance, Insur-
ance, , Real Es- once, , Real Es-
tate Weekly Wa e tate Wage Bill
Hondurable.oods WeeklylageNOridurableGoodsEmployment
Construction Construction
Hanhours Employment '
IGovernment
HFinance, Insur ance, , Real
rlWholesale , Re tail Trade ,
Nondurable Goods JNondurable
tl'lIlhours Goods
Hanhours
Conventional t1ort 
Rage Rate
0\Ul
Appendix C Tables ofExogenous Variables
Table CI Total Employment W/O GM (in thousands)
YEAR** ACTUAL/FITTED* TOTAL CHANGE %CHANGE
TOTAL EMPLOYMENT
EMPLOYMENT W/OGM
(SESMSA2)
1990 226.19 218.15 -8.04 -3.6%
1991 224.05 217.48 -6.57 -2.9%
1992 223.78 217.51 -6.27 -2.8%
1993 224.9 217.88 -7.02 -3.1 % --
1994 230.6 222.33 -8.27 -3.6%
1995 239.45 228.87 -10.58 -4.4%
1996 244.2 233.7 -10.5 -4.3%
1997 243.81 235.79 -8.01 -3.3%
1998 245.97* 237.59 -8.38 -3.4%
1999 247.07* 238.82 -8.25 -3.3%
2000 247.62* 240.07 -7.55 -3.1 %
2001 249.62* 242.64 -6.98 -2.8%
2002 252.67* 246.24 -6.43 -2.6% --
** Note: All data IS annualized from quarterly values
Table C2 Total Wage Bill W/O GM (in billions)
YEAR** ACTUAL/FITTED* TOTAL WAGE CHANGE %CHANGE
TOTAL WAGE BILL W/O GM
BILL (SWBSMSA2)
1990 1.200842 1.107981 -.09286050 -7.7%
1991 1.211626 1.128167 -.0834590 -6.9%
1992 1.265525 1.169124 -.0964008 -7.6%
1993 1.326941 1.217547 -.109394 -8.2%
1994 1.406273 1.294729 -.11l545 -7.9%
1995 1.469244 1.336544 -.132699 -9.0%
1996 1.498718 1.373702 -.125015 -8.3%
1997 1.512067 1.423197 -.0888698 -59%
1998 1.572638* 1.464601 -.108037 -6.9%
1999 1.607754* 1.498462 -.109293 -6.8%
2000 1.643705* 1.535865 -.107840 -6.6%
2001 1.690705* 1.586630 -.104076 -6.2%
2002 1.747261 * 1.647607 -.0996535 -5.7%
** Note: All data IS annualized from quarterly values
66
Table C3 Total Employment With New Plant (in thousands)
YEAR** ACTUAL/FITTED* TOTAL CHANGE %CHANGE
TOTAL EMPLOYMENT
EMPLOYMENT WITH NEW
PLANT
(SESMSA3)
1990 226.19 220.26 -5.93 -2.6%
1991 224.05 219.58 -4.48 -2.0%
1992 223.78 219.60 -4.19 -1.9%
1993 224.90 220.00 -4.90 -2.2%
1994 230.60 224.43 -6.17 -2.7%
1995 239.45 230.93 -8.52 -3.6%
1996 244.20 235.79 -8.41 -3.4%
1997 243.81 237.86 -5.94 -2.4%
1998 245.97 239.94 -6.04 -2.5%
1999 247.07 241.19 -5.88 -2.4%
2000 247.62 242.50 -5.12 -2.1%
2001 249.62 245.13 -4.49 -1.8%
2002 252.67 248.77 -3.90 -1.5%
** Note: All data IS annualized from quarterly values
Table C4 Total Wage Bill With New Plant (in billions)
YEAR** ACTUAL/FITTED* TOTAL WAGE CHANGE %CHANGE ~
TOTAL WAGE BILL WITH
BILL NEW PLANT
(SWBSMSA3)
1990 1.200842 1.124659 -.07618325 -6.3%
1991 1.211626 1.144908 -.0667180 -5.5%
1992 1.265525 1.186657 -.0788680 -6.2%
1993 1.326941 1.236348 -.0905925 -6.8%
1994 1.406273 1.315041 -.0912320 -6.5%
1995 1.469244 1.354804 -.114440 -7.8%
1996 1.498718 1.391816 -.106902 -7.1 %
1997 1.512067 1.441566 -.0705013 -4.7%
1998 1.572638 1.480695 -.0919433 -5.9%
1999 1.607754 1.516653 -.0911018 -5.7%
2000 1.643705 1.555427 -.0882783 -5.3%
2001 1.690705 1.607230 -.0834753 -4.9%
2002 1.747261 1.669015 -.0782463 -4.5%
** Note: All data IS annualized from quarterly values
67
Table C5 Employment in Non-Manufacturing WIO GM (in thousands)
YEAR** ACTUAL/FITTED* EMPLOYMENT IN CHANGE %CHANGE
EMPLOYMENT IN NON-
NON- MANUFACTURING
MANUFACTURING WIO GM (SENM2)
1990 171.65 171.33 -.32 -.2%
1991 171.35 172.80 1.45 .9%
1992 172.25 173.68 1.43 .8%
1993 172.53 174.18 1.66 1.0%
1994 177.38 177.12 -.25 -.1%
1995 182.08 180.73 -1.35 -.7%
1996 184.28 183.66 -.63 -.3%
1997 183.92 184.30 .38 .2%
1998 187.42 187.26 -.16 -.1%
1999 189.65 189.13 -.52 -.2%
2000 191.36 190.79 -.57 -.3%
2001 193.22 192.68 -.54 -.3%
2002 195.32 194.83 -.48 -.3%
** Note: All data IS annualIzed from quarterly values
Table C6 Employment in Manufacturing WIO GM (in thousands)
YEAR** ACTUAL/FITTED* EMPLOYMENT IN CHANGE %CHANGE
EMPLOYMENT IN MANUFACTURING
MANUFACTURING W10 GM (SEM2)
1990 54.54 46.82 -7.72 -14.2%
1991 52.70 44.68 -8.02 -15.2%
1992 51.53 43.83 -7.69 -15.0%
1993 52.38 43.71 -8.67 -16.6%
1994 53.23 45.21 -8.02 -15.1 %
1995 57.38 48.14 -9.23 -16.1%
1996 59.92 50.04 -9.88 -16.5%
1997 59.88 51.49 -8.39 -14.0%
1998 58.56 50.33 -8.23 -14.1%
1999 57.17 49.69 -7.48 -13.1%
2000 56.26 49.29 -6.98 -12.4%
2001 56.41 49.96 -6.45 -11.4%
2002 57.35 51.41 -5.94 -10.4%
** Note: All data IS annualIzed from quarterly values
68
Table C7 Employment in Non-Durable Goods WIO GM (in thousands)
YEAR** ACTUALIFITTED* EMPLOYMENT IN CHANGE %CHANGE
EMPLOYMENT IN NON-DURABLE
NON-DURABLE GOODS GOODS WIOGM
(SEMFN2)
1990 6.80 6.92 .12 1.8%
1991 6.73 6.85 .12 1.8%
1992 7.08 7.02 -.05 -.7%
1993 6.90 7.06 .15 2.3%
1994 7.08 7.14 .07 .1%
1995 7.30 7.29 -.01 -.2%
1996 7.35 7.28 -.07 -.9%
1997 7.78 7.42 -.36 -4.6%
1998 7.42 7.30 -.12 -1.7%
1999 7.34 7.33 -.01 -.2%
2000 7.37 7.37 .00 0%
2001 7.40 7.40 .00 0%
2002 7.43 7.43 .00 0%
** Note: All data IS annualized from quarterly values
Table C8 Employment in Durable Goods WIO GM (in thousands)
YEAR** ACTUALIFITTED* EMPLOYMENT IN CHANGE %CHANGE
EMPLOYMENT IN DURABLE GOODS WIO
DURABLE GOODS GM (SEMFD2)
1990 47.74 39.90 -7.85 -16.4% --
1991 45.98 37.83 -8.15 -17.7%
1992 44.45 36.81 -7.64 -17.2%
1993 45.48 36.65 --
1994 46.15 38.07 -8.08 -17.5%
1995 50.08 40.85 -9.22 -18.4%
1996 52.57 42.76 -9.81 -18.7%
1997 52.10 44.07 -8.02 -15.4%
1998 51.13 43.03 -8.10 -15.9%
1999 49.83 42.36 -7.47 -15.0%
2000 48.90 41.92 -6.98 -14.3%
2001 49.01 42.56 -6.45 -13.2%
2002 49.92 43.97 -5.95 -11.9%
** Note: All data IS annualized from quarterly values
69
Table C9 Employment in Manufacturing With New Plant (in thousands)
YEAR** ACTUALIFITTED* EMPLOYMENT IN CHANGE %CHANGE
EMPLOYMENT IN MANUFACTURING
MANUFACTURING WITH NEW PLANT
(SEM3)
1990 54.54 48.94 -5.61 -10.3%
1991 52.70 46.78 -5.93 -11.2%
1992 51.53 45.92 -5.60 -10.9%
1993 52.38 45.82 -6.56 -12.5%
1994 53.23 47.31 -5.92 -11.1%
1995 57.38 50.22 -7.16 -12.5%
1996 59.92 52.13 -7.78 -13.0%
1997 59.88 53.58 -6.31 -10.5%
1998 58.56 53.04 -5.52 -9.4%
1999 57.17 52.27 -4.90 -8.6%
2000 56.26 51.85 -4.41 -7.8%
2001 56.41 52.53 -3.88 -6.9%
2002 57.35 53.98 -3.37 -5.9%
** Note: All data IS annualIzed from quarterly values
Table ClO Employment in Non-Manufacturing With New Plant (in thousands)
YEAR** ACTUALIFITTED* EMPLOYMENT IN CHANGE %CHANGE
EMPLOYMENT IN NON-
NON- MANUFACTURING
MANUFACTURING WITH NEW PLANT
(SENM3)
1990 171.65 171.33 -.32 -.2%
1991 171.35 172.80 1.45 .9%
1992 172.25 173.68 1.43 .8%
1993 172.53 174.18 1.66 1%
1994 177.38 177.12 -.25 -.1%
1995 182.08 180.73 -1.35 -.7%
1996 184.28 183.66 -.63 -.3%
1997 183.92 184.30 .38 .2% --
1998 187.42 186.90 -.52 -.3%
1999 189.65 188.92 -.73 -.4%
2000 191.36 190.65 -.71 -.4%
2001 193.22 192.60 -.62 -.3%
2002 195.32 194.79 -.53 -.3%
** Note: All data IS annualIzed from quarterly values
70
Table C11 Employment in Non-Durable Goods With New Plant (in thousands)
YEAR** ACTUAL/FITTED* EMPLOYMENT IN CHANGE %CHANGE
EMPLOYMENT IN NON-DURABLE
NON-DURABLE GOODS GOODS WITH NEW
PLANT (SEMFN3)
1990 6.80 6.92 .12 1.8%
1991 6.73 6.85 .12 1.8%
1992 7.08 7.02 -.05 -.7%
1993 6.90 7.06 .15 2.3%
1994 7.08 7.14 .07 1.0%
1995 7.30 7.29 -.01 -.2%
1996 7.35 7.28 -.07 -.9%
1997 7.78 7.42 -.36 -4.6%
1998 7.42 7.42 .00 0%
1999 7.34 7.34 .00 0%
2000 7.37 7.37 .00 0%
2001 7.40 7.40 .00 0%
2002 7.43 7.43 .00 0%
** Note: All data IS annualIzed from quarterly values
Table C12 Employment in Durable Goods With New Plant (in thousands)
YEAR** ACTUAL/FITTED* EMPLOYMENT IN CHANGE %CHAi
EMPLOYMENT IN DURABLE GOODS
DURABLE GOODS WITH NEW PLANT
(SEMFD3) --
1990 47.74 42.01 -5.73 -12.0%._--
1991 45.98 39.93 -6.05 -13.2%
1992 44.45 38.90 -5.55 -12.5%
1993 45.48 38.77 -6.71 -14.8%
1994 46.15 40.17 -5.99 -13.0%
1995 50.08 42.93 -7.15 -14.3%
1996 52.57 44.85 -7.71 -14.7%
1997 52.10 46.15 -5.94 -11.4%
1998 51.13 45.62 -5.52 -10.8%
1999 49.83 44.93 -4.90 -9.8%
2000 48.90 44.48 -4.41 -9.0%
2001 49.01 45.13 -3.88 -7.9%
2002 49.92 46.55 -3.38 -6.8%
** Note: All data IS annualIzed from quarterly values
71
Table C13 Wage Bill in Durable Goods W/O GM (in billions)
YEAR** ACTUALIFITTED* WAGE BILL IN CHANGE %CHANGE
WAGE BILL IN DURABLE GOODS W10
DURABLE GOODS GM (SWBMFD2)
1990 .3977115 .3142445 -.0785 -16.4%
1991 .4010920 .3065133 -.0815 -17.7%
1992 .3967485 .3160595 -.0764 -17.2%
1993 .4390025 .3326780 -.0883 -19.4%
1994 .4754123 .3689850 ,.0808 -17.5%
1995 .4845610 .3664618 -.0922 -18.4%
1996 .4984555 .3779593 -.0981 -18.7%
1997 .5138063 .4019018 -.0802 -15.4%
1998 .5202435 .4111583 -.0810 -15.9%
1999 .5176918 .4110938 -.0747 -15.0%
2000 .5189518 .4150653 -.0698 -14.3%
2001 .5293975 .4294113 -.0645 -13.2%
2002 .5474445 .4515133 -.0595 -11.9%
** Note: All data IS annualIzed from quarterly values
Table C14 Total Wage Bill in Non-Durable Goons W/O GM (in billions)
YEAR** ACTUALIFITTED* WAGE BILL IN NON- CHANGE %CHANGE
WAGE BILL IN DURABLE GOODS W/O
NON-DURABLE GOODS GM (SWBMFN2)
1990 .03903832 .03928915 .00025083 .6%
1991 .03866238 .03956288 .00090050 2.3%
1992 .04150398 .04006288 -.00144110 -3.5%
1993 .04110250 .04155125 .00044875 1.1%
1994 .04170490 .04258180 .00087690 2.1%
1995 .04184960 .04267403 .00082443 2.0%
1996 .03922023 .03943860 .00021838 .6%
1997 .04281328 .04084523 -.00196805 -4.6%
1998 .04268883 .04306350 .00037468 .9%
1999 .04321218 .04372460 .000512.42 1.2%
2000 .04411898 .04439465 .00027568 .6%
2001 .04498083 .04507243 .0009160 .2%
2002 .04580220 .04575630 -.0004590 -.1%
** Note: All data IS annualIzed from quarterly values
72
Table C15 Wage Bill in Manufacturing (in billions)
YEAR** ACTUALIFITTED* WAGE BILL IN CHANGE %CHANGE
WAGE BILL IN MANUFACTURING
MANUFACTURING W/O GM (SWBM2)
1990 .4367500 .3535335 -.08321650 -19.1%
1991 .4397540 .3460763 -.0936778 -21.3%
1992 .4382525 .3561223 -.0821303 -18.7%
1993 .4801048 .3742290 -.105876 -22.1%
1994 .5171170 .4115668 -.105550 -20.4%
1995 .5264105 .4091358 -.117275 -22.3%
1996 .5376755 .4173978 -.120278 -22.4%
1997 .5566198 .4427473 -.113873 -20.5%
1998 .5629323 .4542220 -.108710 -19.3%
1999 .5609038 .4548185 -.106085 -18.9%
2000 .5630708 .4594595 -.103611 -18.4%
2001 .5743785 .4744838 -.0998948 -17.4%
2002 .5932468 .4972700 -.0959768 -16.2%
** Note: All data IS annualIzed form quarterly values
Table C16 Wage Bill in Non-Manufacturing W/O GM (in billions)
YEAR** ACTUALIFITTED* WAGE BILL IN NON? CHANGE %CHANGE
WAGE BILL IN MANUFACTURING
NON- W10 GM (SWBNM2)
MANUFACTURING
-1990 .7640928 .7544483 -.00964450 -1.3%
1991 .7718718 .7820913 .01021950 1.3%
1992 .8272725 .8130023 -.0142703 -1.7%
1993 .8468363 .8433185 -.00351775 -.4%
1994 .8891565 .8831625 -.00599400 -.7%
1995 .9428335 .9274095 -.0154240 -1.6%
1996 .9610423 .9563053 -.00473700 -.5%
1997 .9554475 .9804508 .02500325 2.6%
1998 1.009706 1.010380 .00067400 .1%
1999 1.046851 1.043644 -.00320725 -.3%
2000 1.080634 1.076405 -.00422925 -.4%
2001 1.116323 1.112146 -.00417625 -.4% --
2002 1.154015 1.150338 -.00367625 -.3%
** Note: All data IS annualIzed from quarterly values
73
Table C17 Wage Bill in Durable Goods With New Plant (in billions)
YEAR** ACTUALIFITTED* WAGE BILL IN CHANGE %CHANGE
WAGE BILL IN DURABLE GOODS
DURABLE GOODS WITH NEW PLANT
(SWBMFD3)
1990 .3977115 .3311148 -.06659675 -16.7%
1991 .4010920 .3234535 -.0776385 -19.4%
1992 .3967485 .3337860 -.0629625 -15.9%
1993 .4390025 .3516153 -.0873873 -19.9%
1994 .4754123 .3894205 -.0859918 -18.1 %
1995 .4845610 .3851948 -.0993663 -20.5%
1996 .4984555 .3965843 -.101871 -20.4%
1997 .5138063 .4208388 -.0929675 -18.1 %
1998 .5202435 .4329900 -.0872535 -16.8%
1999 .5176918 .4343543 -.0833375 -16.1%
2000 .5189518 .4393948 -.0795570 -15.3%
2001 .5293975 .4546013 -.0747963 -14.1 %
2002 .5474445 .4774728 -.0699718 -12.8%
** Note: All data IS annualIzed from quarterly values
Table C18 Wage Bill in Non-Durable Goods With New Plant (in billions)
YEAR** ACTUALIFITTED* WAGE BILL IN NON- CHANGE I%CHANGl
WAGE BILL IN DURABLE GOODS
NON- WITH NEW PLANT
DURABLE GOODS (SWBMFN3)
1990 .03903832 .03931265 .00027433 .7%
1991 .03866238 .03958575 .00092338 2.4%
1992 .04150398 .04006795 -.0143603 -3.5%
1993 .04110250 .04154380 .00044130 1.07%
1994 .04170490 .04255165 .00084675 2.0%
1995 .04184960 .04268940 .00083980 2.0%
1996 .03922023 .03945825 .00023803 .6%
1997 .04281328 .04086143 -.0195185 -4.6%
1998 .04268883 .04268590 -.00000292 0%
1999 .04321218 .04319650 -.00001568 0%
2000 .04411898 .04409830 -.00002067 -.1%
2001 .04498083 .04494385 -.00003697 -.1%
2002 .04580220 .04573300 -.00006920 -.2%
** Note: All data IS annualIzed from quarterly values
74
Table C19 Wage Bill in Manufacturing With New Plant (in billions)
YEAR** ACTUALIFITTED* WAGE BILL IN CHANGE %CHANGE
WAGE BILL IN MANUFACTURING
MANUFACTURING WITH NEW PLANT
(SWBM3)
1990 .4367500 .3704278 -.06632225 -15.2%
1991 .4397540 .3630393 -.0767148 -17.4%
1992 .4382525 .3738540 -.0643985 -14.7%
1993 .4801048 .3931593 -.0869455 -18.1 %
1994 .5171170 .4319723 -.0851448 -16.5%
1995 .5264105 .4278840 -.0985265 -18.7%
1996 .5376755 .4360425 -.101633 -18.9%
1997 .5566198 .4617000 -.0949198 -17.1%
1998 .5629323 .4756755 -.0872568 -15.5%
1999 .5609038 .4775508 -.0833530 -14.9%
2000 .5630708 .4834933 -.0795775 -14.1%
2001 .5743785 .4995448 -.0748338 -13.0%
2002 .5932468 .5232055 -.0700413 -11.8%
** Note: All data IS annualized from quarterly values
Table C20 Wage Bill in Non-Manufacturing With New Plant (in billions)
-----YEAR** ACTUALIFITTED* WAGE BILL IN NON- CHANGE %CHANGE
WAGE BILL IN MANUFACTURING
NON- WITH NEW PLANT
MANUFACTURING (SWBNM3)
1990 .7640928 .7542318 -.0986100 -1.3% ..-
1991 .7718718 .7818690 .0999725 1.3%
1992 .8272725 .8128040 -.144685 -1.8%
1993 .8468363 .8431898 -.0364650 -.4%
1994 .8891565 .8830695 -.0608700 -.7%
1995 .9428335 .9269200 -.159135 -1.7%
1996 .9610423 .9557740 -.0526825 -.6%
1997 .9554475 .9798665 .02441900 2.6%
1998 1.009706 1.005019 -.0468650 -.5%
1999 1.046851 1.039102 -.0774875 -.7%
2000 1.080634 1.071934 -.0870025 -.8%
2001 1.116323 1.107686 -.0863700 -.8%
2002 1.154015 1.145810 -.0820500 -.7%
** Note: All data IS annualized from quarterly values
75
References
Albegov, M., AKE, A. E., & Snickers, F. (1982). Regional Development Modeling:
Theory and Practice. New York: North-Holland Publishing Company.
Bolton, R (1985). "Regional econometric models." Journal ofRegional Science, 25
(4),495-317.
Campbell, H. S. (1989, Summer). "State and regional economic impact ofDiamond 
Star Motors." Economic Development Review, 7(3), 31-34.
Cam, N. G. & Rabianski, J. (1991, August). "Selecting industrial locations, sites."
National Real Estate Investor, 33, 28-29.
Cleary, C. P. & Levenbach, H. (1986). The Professional Forecaster: The Forecasting
Process Through Data Analysis. California: Lifetime Learning Publications.
Duobinis, S. F. (1981). "An econometric model of the Chicago Standard Metropolitan
Statistical Area." Journal ofRegional Science, 21 (3), 293-316.
"Filter 35,000 applications to fill 2,500 plant jobs." (1966, October 16). Youngstown
Vindicator. p. E5.
Fiske, J. R, Lamb, J. C. & Morss, M. F. (1991, July). "Practical economic forecasting
for small regions." Business Economics, 26,49-54.
"Get 775 acres for GM plant." (1956, February 21). Youngstown Vindicator. p.l3
Given, K. (1972, May). "The unmaking of the Vega." Motor Trend. p.50.
"GM traces lineage to pioneers, Durant, Chevrolet, A P. Sloan." (1966, October 16).
Youngstown Vindicator. p. ElO.
"GM's new mini car is named 'Vega 2300'." (1966, October 16). Youngstown
Vindicator. p.1.
Hall, O. P. & Lecari, J. A. (1974) "Building small region econometric models:
Extension of Glickman's structure to Los Angeles." Journal ofRegional Science, 14
(3),221-251.
76
Harris, T. R. (1983). "Using Input-Output Analysis for Estimation of Distributional
Impacts from Plant Openings and Closings." In D. M. Otto and T.G. Johnson (Eds).
Microcomputer-Based Input-Output Modeling (pp. 100-106). Boulder: Westview
Press.
Jagnow P. (1964, September 30). "Six blasts signal start on $40,000,000 industry."
Youngstown Vindicator. p.l.
Jensen, C. & Livingston S. (1987, October 13). "GM's Lordstown van plant to shut."
Youngstown Vindicator. p.A2.
Kubik, M. (1999, January). "Project Yellowstone promises profits on GM's small cars."
Business Journal. p.9.
Leontief, W. 1986. Input-Output Economics:. New York: Oxford University Press.
Liu, Y. & Stocks, A. H. (1983). "A Labor-Oriented Quarterly Econometric Forecasting
Model OfThe Youngstown-Warren SMSA." Regional Science and Urban Economics,
13,317-340.
"Lordstown turns out Vegas." (1970, November 24). Youngstown Vindicator. p.l.
"Magazine Features GM Manager - Lordstown auto complex purrs under Abernathy's
direction." (1976, July 26). Youngstown Vindicator. p. B17.
Mahoney, M. (1981, April 1). "20 J-cars per hour is GM goal for today at Lordstown
Plant." Youngstown Vindicator. p.B17.
"The 'new breed' surfaces at Lordstown, Ohio." (1972, March 1). Automotive Industries.
p.18.
Niquette, M. & Shilling D. (1998, November 13). "Efforts on GM factory diverge."
Youngstown Vindicator:. p A3.
Patch, E. P. (1995). Plant Closings and Employment Loss in Manufacturing. New York:
Garland Publishing.
Reiss, G. R. (1956, May 9). "GM plant is delayed six months." Youngstown Vindicator.
p.l+.
---. (1970, March 17). "Lordstown turns out vans." Youngstown Vindicator.
p.l.
---. (1977, May 21). "Recall set of 1,600 workers." Youngstown Vindicator. p. 1+.
Richardson, H. W. 1972. Input-Output and Regional Economics. New York: John
77
Wiley & Sons.
Rothschild, E. (1972, January 23). "GM in more trouble." N.Y. Review. pp. 18-24.
Salpukas, A. (1972, February 6). "Tum the nut, tum the nut, tum the ...." New York
Times. p. EI0.
Schaffer, W. A. 1976. On the Use ofInput-Output Model for Regional Planning.
Leiden: Martinus Nijhoff Social Sciences Division.
Shilling, D. (1998, November 22). "Lansing's bid for GM began before valley's."
Youngstown Vindicator. p. AI.
Walthall, R. C. (1996, June). "State and local tax incentives." International Tax Review.
pp.43-47.
Welker, G. (1998, November 22). "In Toledo, Jeep effort succeeded." Youngstown
Vindicator. pAl.
Williams, L. (1992, August 28). "Dealers wait for repercussions to hit sales."
Youngstown Vindicator. p. C2.
78