Local preferences for economic development outcomes: analytical hierarchy procedure

Growth and Change Vol. 31 (Summer 2000), pp. 341-366

Local Preferences for Economic Development Outcomes: Analytical Hierarchy Procedure ANNA M. COX, JEFFREY ALWANG, AND THOMAS G. JOHNSON ABSTRACT Governments frequently formulate policies designed to stimulate regional economic development. Rarely, however, are efforts made to measure local preferences for economic development outcomes. While the political process should eventually sort out how well local governments are meeting the needs of their constituents, the irreversible nature of many development outcomes makes it preferable to incorporate local preferences directly into the decision making process. This paper presents a straightforward means of measuring preference trade-offs. The analytical hierarchy procedure is applied to local economic development outcomes in three Virginia counties and is shown to improve the targeting of industries by incorporating local preferences in the targeting process. The method has wide applicability for different development decisions.

Background tate and local governments in the U.S. have tried to stimulate economic development in rural areas for over 100 years. Tools used for industrial development include shell buildings, infrastructure investments, local marketing boards, and incentive programs such as tax abatements and credits, and others. Governments, mostly in the Southern U.S., began to use direct financial incentives in the 1930s, and by the 1980s almost every state had an economic development policy that included different measures to induce firm location (Isserman 1994). Currently, economic incentive programs are funded at historical highs (Venable 1994). There is a lively debate in the economic development literature concerning the efficacy of such programs. Surprisingly, considering the magnitude of the dollars being spent on economic development programs, there is little research examining community preferences for economic development outcomes. Knowledge of these preferences

S

Anna M. Cox is a research associate at the Community Policy Analysis Center and Jeffrey Alwang is an associate professor of agricultural and applied economics, at Virginia Tech, Blacksburg;and Thomas G. Johnson is vice provost for extension, agricultural economics at the University of Missouri-Columbia. Submitted July 1997, revised Jan., Oct. 1999. © 2000 Gatton College of Business and Economics, University of Kentucky. Published by Blackwell Publishers, 350 Main Street, Malden MA 02148 US, and 108 Cowley Road, Oxford, OX4 1JF, UK.

342 GROWTH AND CHANGE, SUMMER 2000

can be used to target industries for certain localities as a part of a proactive economic development program. This knowledge can also be used to structure the development package itself. Development packages, whether tax abatements, shell buildings, or community marketing, can affect the development outcome. Inconsistencies can exist between the form of the package and the stated goal. For example, politicians often claim to be interested in increasing employment, yet the majority of state and local economic development resources are used to subsidize capital rather than labor (Courant 1994). Community preferences should matter. Economic development outcomes, regardless of the program or policy in question, are felt locally in a number of ways. For example, a firm startup creates employment and generates incomes. These economic impacts depend on the industry and local economic structure. Other impacts include increased demands for public services, congestion, changes in property values, etc. Evaluating the desirability of these outcomes represents a form of multi-attribute decision making. The critical component of such decision making is determining the weights to attach to different attributes. This paper describes research that tested a means of incorporating local preferences in industry targeting strategies. The approach involves three steps. Local decision makers are interviewed in the first step, and the analytical hierarchy process (AHP) is used to create cardinal weights for different local impacts of economic development outcomes. The weights from this process are consistently derived and theoretically sound. In the second step, critical development impacts of industry locations are identified and quantified for each of several “industries” that survive an initial screening. The final step involves applying the weights to the measured attributes of each industry to derive a community-specific measure or score for each industry.

Weighting Community Preferences When development outcomes are multi-dimensional, a means of weighting these dimensions is needed to rank the events that cause the outcomes. If community preferences were known over the K impacts of an event (industry recruitment in this case), then scoring each of the j=1...J industries would be straightforward: S L = I L* • w*L , (1) where SL is a J G 1 vector of scores the J industries (ordered arbitrarily) receive from the L community, I L* is a J GK matrix of impacts of the J industries on th

the L community , and w*L is a KG1 vector of weights of the K impacts by the th L community. To date, little work has been done to measure these weights. th

1

LOCAL PREFERENCES 343

Obviously, the I L* matrix will depend on the economic structure of the community (see, for example, Siegel et al. 1995 a & b). Weighting systems have been hampered by two major shortcomings. First, most known attempts focus on placing weights on the direct impact of the industry or firm—the number of jobs, wages, etc. Few, if any, have tried to weight the final impact of the firm—changes in population, property values, environmental outcomes, in addition to employment and income effects. For example, Shaffer (1989) recommends use of a screening system where a community assigns weights to the set of firm characteristics. Johnson et al. (1994) had community members assign weights to a list of screening criteria that were used to identify industries for targeting. The Shaffer and Johnson et. al. procedures focus on firm attributes, and neither examines the final impact of a development outcome. The second major shortcoming is the reliance on ad hoc scoring methods that do not adhere to basic principles. Less-rigorous techniques include ordinal ranking or fixed point scoring. These methods are simple to use, but may lead to hasty or ill-considered judgments. Alston et al. (1995) criticize scoring methods (in the context of agricultural research) showing that they frequently lack rigor, have no theoretical basis, and are fraught with inconsistencies in assigned weights. The bottom line is “weights on objectives should reflect clients’ value judgments about trade-offs among objectives (p. 467).” Since weights measure the relative contribution of each criterion to a decision-maker’s overall objective, the representation of the decision-maker’s weights will determine the validity of the multi-criteria decision-making model. Methods other than pairwise comparison as used in the AHP are often used to obtain preferences from decision-makers. As mentioned, ordinal rankings and fixed point scoring are two of the less rigorous methods available. Other preference elicitation methods include, but are not limited to, the Delphi technique, the multi-attribute utility procedure, and the Clarke-Groves voting procedure (Cohon 1978, Harker 1989, Romero and Rehman 1987, Tideman and 2 Tullock 1976). The Delphi technique obtains preferences (weights) from decision-makers through anonymous questionnaires, controlled feedback, and statistical analysis of the results (Dalkey et al. 1972). Each person ranks the criteria using a scale of importance and explicitly states her underlying assumptions. The ranking is accomplished using an ordinal scale (i.e. 1 = very important, 2 = moderately important, etc.). The assessments and assumptions of the group members are analyzed, medians and quartiles are calculated, and the results distributed to the group. Each member then has the opportunity to revise her earlier assessment, based on the results of the group. This process is iterated until a consensus has been reached (Hampton et al. 1973).

344 GROWTH AND CHANGE, SUMMER 2000

The multiple-attribute utility procedure requires the decision-maker to answer questions dealing with probabilities, usually in a lottery framework (Roberts 1979). Decision-makers are asked to predict the probability of a particular consequence (criterion). As the probabilities being elicited are usually subjective in nature, they “represent the ‘degree of certainty’ or ‘degree of conviction’ that the expert has that an event will occur” (Roberts 1979, p. 372). The outcomes derived from the probabilities represent the utility (weight) of each criterion. Another process, which has not been used explicitly as part of multiplecriteria decision modeling, is the demand-revealing voting process. Also called the Vickrey-Groves-Clarke mechanism, the demand-revealing process elicits preferences in a manner that rewards truthful representation of the intensity of the decision-maker’s preferences (Auerbach and Feldstein 1987). In this procedure, each person is asked which of two or more options she prefers, and how much she would be willing to pay to have her preferred option rather than the others. The outcome is reached by summing the dollar amounts for each option. The highest dollar amount is the most preferred. The mechanism ensures each person accurately revealed her preferences by levying a “tax” on each person based on her impact on the outcome (Tideman and Tullock 1976).

The Analytical Hierarchy Procedure (AHP) The AHP, developed by Saaty, is a means of weighting or prioritizing impacts through a systematic representation of a problem. Through pairwise comparisons, the relative importance, or weights, of different factors can be measured; tradeoffs between objectives are explicitly considered in these pairwise comparisons. The pairwise comparison process imposes rigor that is missing when directly assigning weights to a number of impacts, because possible inconsistencies (intransitivities and inconsistent weights) in the judgments can be calculated and reexamined. Even with subjective criteria, the weights obtained through the AHP are “ratio scale numbers and correspond to so-called hard numbers” (Saaty and Kearns 1985, p. 19). Thus, the derived priority weights are cardinal. One foundation of the AHP is the observation that human decision making is not always consistent. Consistency suffers when the criteria being compared are subjective in nature. The AHP provides a standard by which the degree of consistency can be measured. If inconsistency exceeds an established threshold then participants can reexamine their judgments. AHP has been applied to different problems, including electric utility planning, portfolio management, conflict management, advertising, and resource allocation (for examples, see Hämäläinen and Seppäläinen 1986; Lauro and Vepsalainen 1986; Saaty and Alexander 1989; Dyer et al. 1992; Saaty 1980). No known application exists of AHP to community preferences for development outcomes.

346 GROWTH AND CHANGE, SUMMER 2000

If the actual utility weights of the criteria being compared were known, the pij would represent the ratio of the weights: P = W,

(2) th

where wij = wi/wj, and wi is the utility weight of the i criterion (see equation 1). When the weights (w1, w2,..., wK) are unknown, then the pairwise comparison is performed using subjective judgments estimated numerically from a scale of numbers. The scale recommended by Saaty (1977) has been validated for effectiveness in different applications (Table 1). TABLE 1. SCALE OF RELATIVE IMPORTANCE Intensity of Relative Importance

Definition

Explanation

1

Equal importance

Two activities contribute equally to the objective

3

Moderate importance of one over another

Experience and judgement slightly favor one activity over another

5

Essential or strong

Experience or judgment strongly favors one actiivty over another

7

Demonstrated importance

An activity is strongly favored and its dominance is demonstrated in practice

9

Extreme importance

The evidence favoring one activity over another is of the highest possible order of affirmation.

2,4,6,8 Intermediate values between the two adjacent judgments

When compromise is needed

Source: Saaty and Kearns 1985

Solving for preference weights. Obtaining the relative weights would be simple if there were no errors or inconsistencies in comparisons. In such a case, w could be solved as an eigenvector of P corresponding to an eigenvalue equal to K, which, in the perfectly consistent case, is also the matrix rank (Saaty, 1980): PW = KW, (3)

LOCAL PREFERENCES 347

where P is the comparison matrix, w is the eigenvector corresponding to K, and K is the number of rows and columns of P, or, in other words, the number of impacts being assigned weights. In general, the P matrix will contain errors or inconsistencies. The inconsistencies can be attributed to the limitations put on the comparison by the scale being used. That is, each pij is based on subjective estimate, not on exact measurement. The judgment matrix is not really a ratio of wI /wj, but only those elements of the scale. By using integers and their inverses, the rounding error can reach 50 percent (Fichtner 1986). Human error can also cause inconsistencies. Prior rankings of each pair of elements are difficult for people to consider as they compare other criteria. This limitation is the primary justification for keeping the number of elements to be compared at each level of the hierarchy below nine (see Saaty 1980). Given this 3 error, the maximum eigenvalue is used in place of K to solve for w:

PW = λ max w,

(4)

where λmax is the maximum eigenvalue of P, and w is the eigenvector corresponding to λmax. Weights are calculated using the eigenvector corresponding to the largest eigenvalue, λmax. The elements of w are normalized using:

wi* =

wi å wi

(5)

These results ( wi* ) are the cardinal, or relative, values (weights) of the criteria. The vector of these weights for the l community wi* is used as in equation 1 (for detailed information see Saaty 1980). A strength of the AHP is its ability to estimate preference weights even in the advent of intransitivity among criteria and inconsistency in the intensity with which judgments are expressed. The AHP provides a way for inconsistencies to be measured. It is desirable for judgments to fall under a consistency threshold (see Appendix for the consistency index). th

Structure of the Problem for Evaluation of Economic Development Outcomes In Virginia, economic development directors or their counterparts develop local target industry strategies. These directors typically have a commission, board, or authority, or an informal network with which they confer. Different communities conceivably weight criteria differently but all have, as their goal, positive impacts on the community (Bailey 1996). The AHP is used in the current problem to measure preferences by the development directors and other

348 GROWTH AND CHANGE, SUMMER 2000

local leaders for alternative development outcomes. Here, the procedure is not intended to measure the preferences of the general population or, for example, a median voter. Depending on the goals of the preference measuring process, more (or different) respondents could be brought into the process. While there are many well-known problems associated with eliciting weights for a social welfare function, the AHP can estimate the cardinal weights of a small group of decision makers for the outcomes in question. It thus adds substantial information to the industry-targeting process. The goal is to target the industry with the most favorable total impact on the locality. The second level then represents the criteria by which the top level is judged. The criteria, in the present case, consist of the various types of impacts that a firm can have on a locality. Following a review of findings from other studies, and a survey of economic development directors in the state of Virginia 4 (Bailey 1996), the following impacts were identified: number of jobs created, average wage or salary, average level of capital investment, average level of utility requirements, environmental impacts, effect on population growth, and impact on property values. Respondents were asked to weight these impacts in the AHP interview process. These categories of impacts were chosen because of their demonstrated importance to communities. The number of newly created jobs is often highlighted as an indication of success in economic development. Localities are interested in job quality and the average wage or salary is an indicator of quality. Capital investment is important because it shows commitment to the community, and it increases the stream of property tax revenues. Utility requirements were included as a measure of the importance of local utility capacity constraints. Economic development professionals indicated that water and sewer demands were a concern or at least a consideration when granting incentives to locating industries (Bailey 1996). Property values are important to the local government primarily because increased property values increase the tax base. The perceived cleanliness of industry was included to determine the importance of environmental considerations to decision-makers in these communities. The impact of population growth is a measure of several of the “costs” associated with economic development, such as congestion. * Impacts of industries. To use the priority vector weights (w ) to rank industries, each type of impact must be measured for each industry. This measure represents how the location of an industry in a locality will contribute to each impact. When multiplied by the priority vector, the final industry score is obtained (see Equation 1). Each impact is scored using a relative scale. The industry with the largest impact receives a score of 100 (or -100 for the negative factors). The score received by each remaining industry is the percentage of that industry’s impact relative to the largest impact.

LOCAL PREFERENCES 349

Because industry targeting involves selecting among a large number of potential industries, the method used to measure industry impacts needs to be simple, replicable, and systematic. In the present study, 100 potential industries 5 were evaluated for each county. The means of measuring impacts below were used because of their simplicity and reasonably close correspondence to the concept (the impact). Measuring impact levels. The average number of jobs, average wage or salary, average level of capital investment, and average level of utility requirements are calculated for each industry using IMPLAN. IMPLAN data list total employment, total compensation to employees, payments to proprietors, and total output for each sector. Obtaining the average number of jobs per dollar in output by industry is straightforward. Total employment by industry includes local direct, indirect, and induced employment for each sector, as calculated by IMPLAN. As an example of how the employment “impact” variable was measured consider the following. For Bath County, 25.32 jobs are created in the IMPLAN Sector Knit Underwear Mills (Sector #113) per every $ one million in output. This sector has the highest total employment impact per dollar of output. For the same county, the Sector Pottery Products, NEC (Sector #241) create 18.17 jobs per $ one million output. The score for Knit Underwear Mills is 100 and for Pottery Products, NEC, it is 71.75 (18.17/25.32) (see Table A.2 for sample scores by industry for Bath County). The average wage or salary of each industry is similarly derived. The IMPLAN variable “payments to proprietors” is used to calculate the average level of capital investment of each industry per dollar of output. This variable, while an imperfect measure of investment, represents the returns on capital investment accruing to the owners, shareholders, and lenders of each firm type. IMPLAN model data were also used to calculate average local utility requirements per dollar of output for each industry. Each industry’s use of water and sewage systems is approximated by its purchases from IMPLAN’s sector 512, Other Local Government Enterprises. Sector 512 includes the local supply of sanitation, sewerage, water, gas, water transport, and terminals, airports, housing and community development, and liquor stores (Lindall and Olson 1993). Environmental impacts were measured using the 1993 Toxic Release Inventory (U.S. Environmental Protection Agency (EPA) 1995). This report lists releases of 316 chemicals from manufacturing facilities into the environment. Manufacturing firms report the amount of chemicals released into the environment, and chemicals transferred to other locations for disposal or recycling. The sum of releases and transfers is used here to represent the environmental impact of each industry. Industries using more of the chemicals included in the EPA report receive larger negative values.

350 GROWTH AND CHANGE, SUMMER 2000

These environmental data are used subject to several significant limitations. The measure used is only one of a number of possible measures of environmental quality. For example, the means by which chemicals are released into the environment could be an important consideration. A potential solution to this problem is to restructure the hierarchy by adding a level for the cleanliness of industry. In this case the sub-criterion could be the type of pollution, such as air, water, or level of solid waste. However, as mentioned, such data are limited and often dated. Other limitations of the measure used are that it does not include all chemical pollutants and that the chemicals released do not necessarily correlate directly with higher health or environmental risk. Furthermore, a number of environmental risks are not contained in the measure, as businesses can affect the environment in many ways other than chemical emissions. Additional local employment might lead to an increase in traffic and, therefore, automobilerelated emission levels. Development itself can affect the landscape, create runoff through construction, and have substantial impacts on water quality. Finally, the indicators are only available at the two-digit SIC level and do not include pollutants from non-manufacturing industries. Despite these limitations, the data represent proxies for a range of direct environmental impacts by industry. Estimating impacts of industry location on population growth and changes in property values required several steps. The indices constructed for this study are rough. More detailed estimation of population growth and property value impacts are left for further studies. The population growth index used in this study is a measure of the mismatch between occupational demand and the current labor force. It captures the demand for each of three skill levels by the industry and the supply of the same skill levels in the community. A similar index is created for the impacts on property values. Both indices increase if the industry has worker skill-level requirements that differ from the existing worker skill-level structure in the county. Such a mismatch is assumed to induce changes in property values and migration, or both. The first step in creating the population growth index is to determine the employment associated with a typical firm in each industry. The data for this calculation came from the 1992 Census of Manufacturing Preliminary Report Summary Series, the 1992 County Business Patterns for the United States and 7 the 1994-95 Virginia Statistical Abstract. The second step is to determine demand in each industry for workers of different skill levels. The average proportion of high-, semi-, and low-skilled workers for each industry is calculated using U.S. Department of Labor statistics. The Department of Labor uses seven job classifications for

LOCAL PREFERENCES 351

occupations. In order to match supply data these seven classifications are aggregated into three categories reflecting the amount of training and education required to perform tasks which are part of the occupation (Broomhall 1991). Occupations in category 1 require at least a four-year college degree. Category 2 occupations require at least some training or education beyond a high school diploma. Category 3 occupations may or may not require a high school diploma. Mij, then is the percent of workers in industry i with skill level j. This percentage represents the demand of the industry for workers of each skill level. Local “supply” of workers for each skill level is determined using educational attainment data, by county, from the 1994-95 Virginia Statistical Abstract. Skj represents the percentage of the work force in county k with skill level j. The distribution of industry labor demands for the industries are then compared to the supply of labor in each county, and the population growth index by industry by community is created as follows: 3

PGik = å (Skj − Mij) 2 * Nij, j =1

where Nij is the number of jobs needed by industry i for skill level j. It is assumed that as the proportion of skilled jobs demanded changes, the proportion of skilled people in the area will change. The more mismatch between industry 8 and county, the more migration is expected. The impact of industry location on property values is calculated in a similar fashion. As more people earn higher incomes the property values in an area will increase. Because of the different average pay received by each worker category, it is assumed that the high- and semi-skilled workers are more likely to purchase a home, while low-skilled workers are more likely to rent. Another index of mismatch is created, this time with high- and semi-skilled categories combined. The distribution of industry labor demands for high- and semi-skilled workers is compared to the distribution of high- and semi-skilled labor in each county. The difference between industry distribution and the county distribution is squared and is used to indicate the number of people who will purchase a home. Both indices are scaled into percentages of the maximum value for all industries considered for the county. It is recognized that neither index incorporates all factors affecting migration and property values.

AHP Interview and Results The AHP was applied to three rural Virginia counties: Bath, Halifax, and Montgomery Counties. These counties were chosen because they are broadly representative of county types in rural Virginia. Bath County is heavily dependent on natural resource-based tourism. It also faces economic stagnation and contains a relatively high proportion of poor households. Halifax County is a mixed agricultural-manufacturing county typical of Southside, Virginia.

352 GROWTH AND CHANGE, SUMMER 2000

County government has been fairly successful in nurturing light furniture manufacturing firms, and now faces the challenge of declining incomes due to disruption in tobacco sales. Montgomery County has grown rapidly in recent years and faces challenges of growth management. Local government and business leaders were invited to participate in the industry ranking process as representatives of their county. While it is recognized that the preferences of other county residents may vary from those invited to participate, local leaders’ involvement in the community were thought to provide a broader, more comprehensive view of the issues and constraints facing the county. For practical reasons, the group needed to be made up of four to seven people. Larger group sizes could be used if so desired; the cost of a larger group is additional time required to reach consensus. Four people from the County Planning Commission participated in the study in Bath County. Four people participated in Halifax County, including the Director of the Industrial Development Authority, two members of the Halifax County Economic Development Committee, and a representative of the local Chamber of Commerce. Five individuals participated in Montgomery County: the town manager of Blacksburg, the Chair of the Industrial Attraction Committee, the Regional Marketing Director of the New River Valley Economic Development Alliance, the Chair of the Board of Supervisors, and the Regional Director for Virginia’s Center for Innovative Technology. Participants were asked to consider every possible combination of two impacts, “When comparing impact A to impact B, how important of a consideration was one over the other with respect to the attractiveness of an industry?” A person was randomly chosen to be the first to provide a judgment. Subsequent participants responded at will. If there was immediate consensus, that judgment was entered into the matrix. If not, a discussion of the assumptions or considerations each individual used when making his or her value judgment ensued. At the end of the discussions, a person would be convinced enough to change his judgment or the group would agree on a compromise. Calculation and reevaluation of the comparison matrix. When the initial judgment matrix was filled, the priority outcomes were calculated as described above, using a program written in Gauss (Aptech Systems, Inc. 1993). The consistency ratio (see Appendix) was calculated. If the consistency ratio (CR) was above the threshold of 0.20, the judgment matrix was reexamined. Reevaluation did not mean that responses were squeezed into any predetermined pattern, rather it was used to ensure that no blatant inconsistency existed. For instance, if environmental quality is strongly favored to number of jobs and number of jobs is strongly favored to capital investment, then environmental quality should also be strongly favored to capital investment.

LOCAL PREFERENCES 353

The initial step in the reevaluation was to display the priority weights of the criteria to the group. They then identified which rankings or weights that did not make sense to them. This led the group to analyze their judgments and further discuss the assumptions behind the judgments. This process was repeated until the CR was lower than 0.2 and participants were satisfied that the ranking and weights adequately represented their preferences. Results of AHP. Experience with the interviews varied by county, but the priority weightings and rankings were reasonably consistent across counties. For two out of the three counties, reexamining the pairwise comparisons was necessary due to initial inconsistencies in the judgment matrix. The rankings are shown in Table 2. Differences in rankings highlight the location-specificity of development preferences; different counties have different preferences. These preferences should be considered explicitly when making development decisions. TABLE 2. FINAL RANKING AND WEIGHTS BY IMPACT, THREE VIRGINIA COUNTIES Bath County Rank

1

Impact

Halifax County Final Weight (%)

(%)

51

Cleanliness of Industry

16

Level of Capital Investment

Impacts on Property Values

13

Average Wage or Salary

4

No. of Jobs

6

No. of Jobs

5

Level of Capital Investment

5

Impacts of Population Growth

5

Level of Utility Requirements

4

Impacts on Property Values

2

3

6

7

CR

Cleanliness of Industry

Impact

Average Wage or Salary

Level of Utility Requirements Impacts of Population Growth

0.196

Montgomery County Final Weight (%)

Impact

Final Weight

49

Average Wage or Salary 35

23

Cleanliness of Industry

24

13

Level of Capital Investment

16

6

No. of Jobs

12

5

Impacts of Population Growth

7

3

Impacts on Property Values

4

2

Level of Utility Requirements

2

0.275

0.142

354 GROWTH AND CHANGE, SUMMER 2000

Montgomery County participants only ranked the outcomes once and did not need to go through the re-weighting process. They felt the order and weights during the first attempt adequately represented their preferences, and the CR is 0.14, well below the 0.2 threshold. In Bath County, changes had to be made following the initial weighting; respondents were dissatisfied with the initial weightings. Comparisons also needed to be reexamined since the initial comparison process yielded a CR of 0.29, above the 0.2 desired threshold. Changes were made following discussion among the participants. The final CR was 0.196, just below the CR threshold. Reexamination of the pairwise comparisons led to only slight changes in the weights, so that consistency was achieved without changing the ranking of impacts. Respondents in Halifax County felt that the rank and weight calculated for each impact adequately represented their priorities. However, the CR, at 0.27, was above the desired limit. Unfortunately, due to the two-hour time constraint, the difficulty in pinpointing possible inconsistent comparisons causing the high CR, and the respondents’ satisfaction with their original comparisons, the matrix was not reexamined. The Halifax County results should be interpreted with caution, as the underlying judgments are not within normal bounds of consistency. Preferences for outcomes. Participants in all three counties had a strong preference for a clean environment. Environmental quality was ranked a strong first in 2 out of 3 of the counties and it finished second in the third (Table 2). Bath County’s desire to maintain an attractive environment is mainly due to the county’s heavy reliance on tourism. The Hot Springs area and its associated spas and resorts form the backbone of the county economy. Halifax County, on the other hand, has aggressively recruited industry for more than a decade, and the high weight placed on industry cleanliness is surprising. Decision-makers expressed the view that “smokestack chasing” was a strategy of the past. In both counties, environmental quality received more than double the weight of the next-preferred outcome. Contrary to popular perception, the number of jobs associated with a development event is substantially less important than cleanliness of the firm and quality of jobs. This outcome was ranked only 4th most important in each county, with a weight ranging from 12 percent in Montgomery to about 6 percent in the other counties. In all counties, number of jobs is an important consideration, but job quality and environmental quality rank higher. Bath County respondents reasoned that although their unemployment rate is higher than the state average, the number of unemployed is small because of their small population. It is thus not essential to create large numbers of jobs locally. Average wage or salary was the most important consideration for Montgomery County, with a priority weight of 35 percent. Participants felt firms offering higher pay were more attractive since the presence of Virginia Tech

LOCAL PREFERENCES 355

gives Montgomery County a more-educated work force, and decision makers sought to increase the number or “head of household” type jobs. Decisionmakers in the other two counties ranked average wages highly, and stated directly that higher pay is associated with a higher quality job. Respondents in Halifax reasoned that the county had made great strides recently in increasing the number of jobs locally, and that it was time to focus on job quality over quantity. Respondents in all counties value the contribution of the development event to the local tax base. Capital investment is ranked second in Halifax (24 percent) and third in Montgomery County (16 percent). Participants considered tax revenues associated with higher capital investment to be important. Firms with large capital investments were also believed to be less likely to relocate in the future. In contrast, Bath County respondents argued that the best way to effect increases in property tax revenues was by increasing local property values. They put a low weight on capital investment, reasoning that capital investment was associated with heavy industry, and such industry might damage the tourism base of the economy. Respondents in Montgomery and Halifax Counties placed low weights on changes in property values. Impacts on utility requirements and population changes received low priority weights in all counties. In Bath County, utility requirements received a relatively low score because water, sewer, and electricity use are currently far below capacity. Montgomery and Halifax County respondents decided that if a firm had desirable characteristics, the county would expand its sewer and water capacity to meet industry needs. The impact of population growth was somewhat important to the respondents from Montgomery County (7 percent), primarily because of the increase in traffic along U.S. Highway 460, and the resulting congestion occurring in the past several years. In Bath and Halifax Counties, respondents decided the schools, roads, and other facilities are more than adequate for the current population. In all counties, population increases were not viewed favorably, but in the latter two counties such increases were accorded small weights. Scoring the impacts. The level of each impact associated with each industry is calculated as discussed above; the impact levels, converted to scores, are shown for Bath County in Table A.2. Industry scores for the other counties are found in Cox (1996). Scored impacts are multiplied by the priority weights to calculate an adjusted score for each industry and the county-specific industry rankings. Community preferences used as priority weights have a strong effect on the ranking of industries (compare Table A.2 with Table 3). The first-ranked industry in Table A.2 (Research, Development, and Testing) falls to number 11 once the rankings are introduced. This fall is mainly due to the lower average wages associated with the sector. Knit underwear falls from the top 20 industry because of its adverse effect on the environment; Fluid Power Pumps does not

356 GROWTH AND CHANGE, SUMMER 2000

TABLE 3. TOP TWENTY INDUSTRIES IDENTIFIED FOR TARGETING, BATH COUNTY

Rank

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

SIC Code

4600 4010 4810 4910 7370 15,16,17 15,16,17 3571 2085 8730 5000 7320 4720 4730 3594 15,16,17 4311

Sector Name

Pipe Lines, Exc. Natural Gas Railroads & Related Services Comm., Except Radio & TV Electric Services Comp. & Data Proces. Serv. New Gov. Facilities Maint. & Repair Oil & Gas Wells Electronic Computers Federal Gov. - Non-military Distilled Liquor, Except Brandy Research, Dev. & Testing State & Local Electric Utilities Wholesale Trade Other Business Services Arrangement Of Pass. Trans. Transportation Services State & Local Gov. - Non-ed. Fluid Power Pumps & Motors Maint. & Repair, Residential U.S. Postal Service

Average Average Value-Added No. of Jobs Wage or Salary Effect (per million $/year (total VA/$ output) output) 2 8 6 3 13 10 27 6 21 2 24 4 13 21 22 13 32 18 13 18

$58,113 $61,152 $49,516 $54,112 $33,780 $36,410 $21,658 $63,525 $47,933 $57,072 $27,336 $ 44,098 $ 36,095 $ 16,709 $ 20,393 $ 29,108 $ 31,637 $ 46,088 $ 23,717 $ 42,275

0.9011 0.7776 0.9074 0.7027 0.9108 0.7366 1.1702 0.7947 1.1401 0.9805 0.9992 0.7019 1.0206 0.9860 0.9595 0.7744 1.3026 1.1166 0.7337 0.8909

fall very much. The latter sector has high wages that offset, to a large degree, its adverse effect on the environment. Environmental quality, as shown earlier, has a major effect on industry rankings. None of the top 7 industries in Table 3 and only 3 of the top 20 industries ranked for Bath County had any adverse environmental impact (see Table A.2). Sectors that moved up in ranking were those with relatively low impacts on the environment and relatively high wages (Distilled Liquor, except Brandy; Federal Government, non-military). All of the top industries in Table 3 are highly linked to the local economy, and have favorable overall impacts. The most preferred industry for all counties is SIC 4600, Pipelines, Excluding Natural Gas (Table 4). This industry had high average wages, represented the largest proportional capital investment, and had no adverse environmental impact (see, for example, Table A.2). Other industries, such as SIC 4810 and SIC 4910 were also favorable for all counties, while some, such as Distilled Liquor, Except Brandy, were only highly ranked for single counties. Bath and Halifax County planners have been successful in attracting and retaining industries that are consistent with preferences as “measured” using the

TABLE 4. TOP TWENTY INDUSTRIES FOR EACH COUNTY Bath County

Halifax County

Montgomery County

Rank Description

Description

Description

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Pipe Lines, Exc. Natural Gas Other Business Services Maint. & Repair Oil/Gas Wells Pleating & Stitching Advertising Arrangement Of Passenger Trans. Electric Services Flavor. Extracts & Syrups Research, Dev. & Testing New Gov. Facilities Computer & Data Processing Services Comm., Except Radio & TV Maint. & Repair, Residential Maint. & Repair Other Facilities State & Local Electric Utilities Transportation Services Motor Freight Trans. & Warehousing Hardwood Dimension & Flooring Pottery Products, N.E.C. Bread, Cake, & Related Products

Pipe Lines, Exc. Natural Gas Electric Services Comm., Exc. Radio & TV Railroads & Rel. Services Electronic Computers State & Local Electric Utilities Fluid Power Pumps & Motors Federal Gov. - Non-military New Gov. Facilities Complete Guided Missiles Research, Dev. & Testing Printing Trades Machinery Industrial Patterns Instrum. To Measure Electricity Radio & TV Broadcasting U.S. Postal Service Phonograph Records & Tape Industrial Gases Wholesale Trade Special Dies & Tools & Accessories

Pipe Lines, Exc. Natural Gas Railroads & Related Services Comm., Except Radio & TV Electric Services Comp. & Data Proces. Serv. New Gov. Facilities Maint. & Repair Oil & Gas Wells Electronic Computers Federal Gov. - Non-military Distilled Liquor, Except Brandy Research, Dev. & Testing State & Local Electric Utilities Wholesale Trade Other Business Services Arrangement Of Pass. Trans. Trans. Services State & Local Gov. - Non-ed. Fluid Power Pumps & Motors Maint. & Repair, Residential U.S. Postal Service

LOCAL PREFERENCES 359

Groups of citizens could also take part in ranking industry impacts. AHP does not limit the number of people in the group or the number of groups participating in the pairwise judgements. The judgments from each group, or even from each individual, can be integrated into one set of priority weights, using methods such as voting or averaging (Saaty 1980). The size of each group would only be limited by logistical concerns. In general, industry rankings changed dramatically as preference weights were introduced. An industry that has potentially high economic impacts may in fact not be preferred because of its non-economic impacts. Impacts such as cleanliness or average wage or salary have greater effects on preferences toward an industry than the number of jobs. To illustrate, Bath County participants are concerned about raising the property tax base, while neither of the other two community groups considered an industry’s impact on property values to be important. Bath County participants consider property values very important because the county has no sales and use tax, making property taxes the only local source of revenue for the county. Identifying industries that are attractive to residents is the first step in the industry recruitment process. Community leaders need to assess what resources are required or expected by each industry, and make a decision whether to provide those resources. These resources include local infrastructure, worker skills, and local services, such as school, police, and fire protection. If a community is unable to finance the improvements necessary to successfully attract the targeted firms, state money might be made available. Some of the preference rankings common to each county indicate areas where the state can make policy changes. In each case-study area, for example, creating high-paying jobs and keeping the environment as clean as possible were considered the two most important impacts of firms. Programs should be structured so that incentives are available for the types of firms that are environmentally friendly and offer higher wages. For example, the state might create an incentive policy that specifically targets firms with pollution levels below a pre-specified limit. In order to provide more choices of industries to target, states may also choose to work with firms to make their level of waste more acceptable to the area in which the firm is thinking to locate, rather than have a community dismiss “dirty” firms out of hand. If a locality wishes to subsidize industry, it needs to be able to predict how that industry will impact its economy, public services, and the quality of life of residents. To help identify an industry’s attractiveness to a community, this research included various economic and non-economic characteristics of the firm. Specifically, the number of jobs, the average wage or salary, and the level of capital investment of a firm measure the import of economic considerations. The impacts on public service are represented by population growth, utility requirements, and change in property values. The population growth and

360 GROWTH AND CHANGE, SUMMER 2000

property value impacts may also indicate a change in quality of life in an area. The cleanliness of industry impact is also used as a measure of the changes in quality of life that may result from a firm relocation. For some of these impacts data or models for measuring an impact were not available. This lack of available data forced the use of proxies in place of measurement of the actual impact, which was the weakest part of the model. Measuring the impact of an industry location on property values, population growth, and the cleanliness of the environment are three areas of research that deserve more attention. The Community Policy Analysis System (COMPAS), an econometric community impact model proposed by Johnson and Scott (1997), helps address the information needs of this type of industry targeting, as the model has the capacity to estimate the impact of employment and income growth on population and property values. A national effort, through the Community Policy Analysis Network (CPAN), has been underway to develop these impact models (Scott and Johnson 1999). The increased capacity to estimate impacts at the local level will enhance the utility of the analysis presented here. Data on environmental impacts, or the cleanliness of an industry, are not generally available, particularly by type of industry. However, once the number of potential industries is limited using the methods described, obtaining more detailed environmental data about the top-ranked industries would be a likely extension. With a limited number of industries to consider, the cost of obtaining more specific environmental impact data would be significantly lower. Overall, including community preferences through the AHP promises to be a valuable tool for community decision support. It accommodates the diversity of communities and the inherent diversity of values within them. AHP provides a means for choosing among multiple alternatives while accommodating multiple objectives and multiple decision-makers. Finally, the use of economic impact tools provides a more comprehensive and consistent basis for comparing different sectors. NOTES 1. Each element, I*Lij, of I*L represents the jth impact that the ith firm has on community L. For example, it might be the total change in migration attributable to the location of a rubber factory in Montgomery County, VA. 2. For a more detailed description of these and other methods used to support multicriteria decision making, see Romero and Rehman (1987), Hampton et al. (1973) and Roberts (1979). For an evaluation of group decision making methods, see Srisoepardani htt p:// www.expert choice.com/support/ahpcompare/chapter6.htm 3. Two facts of matrix theory provide a basis for use of the maximum eigenvalue to calculate w. First, if λ1, . . . , λK are the eigenvalues of P that satisfy the equation Px = λx, and if pii = 1 for all i, then Σλi=K. 4. Economic Development directors were surveyed by Bailey to understand which firm attributes made the firm likely to receive an incentive package. Bailey’s results were

LOCAL PREFERENCES 361

used to define the universe of plausible firm impacts on the community and are not assumed to reflect “community values.” 5. These 100 were identified based on their total economic impact in each county. The 100 vary by county. 6. Because of the inclusion of airports, housing and community development, and liquor stores, Sector 512 might overstate utility usage. If so, it will be inflated for all industries in the study. 7. IMPLAN data were not used because the number of firms represented in IMPLAN is unknown. To measure migration, the average number of employees per firm was needed. This is not the case when the average employment (per dollar of output) was calculated earlier. 8. We do not account for potential retraining with this measure.

REFERENCES Auerbach, A.J., and M. Feldstein. 1987. Handbook of public economics, Volume II. The Netherlands: Elsevier Science Publishers B.V. Alston, J.M., G.W. Norton, and P.G. Pardey. 1995. Science under scarcity: Principles and practice for agricultural research evaluation and priority setting. Ithaca NY: Cornell University Press. Aptech Systems, Inc. 1993. Gauss system and graphics manual, Version 3.1, Volume 1. Maple Valley WA: Aptech Systems, Inc. Bailey, T. M. 1996. Analysis of firm desirability among Virginia’s economic development directors, unpublished MS Thesis, Blacksburg, VA: Department of Agricultural and Applied Economics, Virginia Tech. Broomhall, D.E. 1991. The influence of perceived employment opportunities on educational performance in Appalachia. Unpublished Ph.D. dissertation, Blacksburg, VA: Department of Agricultural and Applied Economics, Virginia Tech. Census of Manufacturing Preliminary Report Summary Series. 1992..??? Cohon, J.L. 1978. Multiobjective programming and planning. New York: Academic Press. County Business Patterns for the United States. 1992. ??? Courant, P. N. 1994. How would you know a good economic development policy if you tripped over one? Hint: Jobs don’t count, The National Tax Journal, XLVII(4): 863-881. Cox, A.M. 1996. Proactive industrial targeting: An application of the analytical hierarchy process, unpublished MS Thesis, Blacksburg, VA: Department of Agricultural and Applied Economics, Virginia Tech. Dalkey, N.C., D.L. Rourke, R. Lewis, and D. Snyder. 1972. Studies in the quality of life: Delphi and decision-making. Lexington MA: Lexington Books. Dyer, R.F., E.H. Forman, and M.A. Mustafa. 1992. Decision support for media selection using the analytical hierarchy process. Journal of Advertising. XXI(1): 59-72. Fichtner, J. 1986. On deriving priority vectors from matrices of pairwise comparisons. Socio-Economic Planning Sciences. 20(6): 341-345. Hämäläinen , R.L., and T.O. Seppäläinen. 1986. The analytical network process in energy policy planning. Socio-Economic Planning Sciences. 20(6): 399-405. Hampton, J.M., P.G. Moore, and H. Thomas. 1973. Subjective probability and its measurement. Journal of the Royal Statistical Society, Series A, 136(1): 21-42.

362 GROWTH AND CHANGE, SUMMER 2000

Harker, Patrick T. 1989. The art and science of decision making: The analytical hierarchy process. In The analytical hierarchy process, edited by B.L. Golden, E.A. Wasil, and P.T. Harker, 3:36. Heidelberg: Springer-Verlag Berlin. Isserman, A. M. 1994. State economic development policy and practice in the United States: A survey article. International Regional Science Review, 16(1 & 2): 49-100. Johnson, T.G., and J.K. Scott. 1997. The Community Policy Analysis System (COMPAS): A proposed national network of econometric community impact models. Paper presented at the 11th Federal Forecasters Conference, Washington, DC, September. 13 pp. Johnson, T.G., E.W. Wade, and R. Archambault. 1994. An economic opportunities analysis for the New River Valley. Blacksburg, VA: Department of Agricultural and Applied Economics, Virginia Tech. Lauro, G.L., and A.P.J. Vepsalainen. 1986. Assessing technology portfolios for contract competition: An analytic hierarchy process approach. Socio-Economic Planning Sciences. 20(6): 407-415. Roberts, F.S. 1979. Measurement theory with applications to decisionmaking, Utility, and the social sciences. Reading MA: Addison-Wesley Publishing Company. Romero, C., and T. Rehman. 1987. Natural resource management and the use of multiple criteria decision-making techniques: A review. European Review of Agricultural Economics, 14(): 60-89. Ross, D., and R.E. Friedman. 1990. The emerging third wave: New economic development strategies in the ‘90s. Entrepreneurial Economy Review, 9(1): 3-10. Saaty, T.L. 1977. A scaling method for priorities in hierarchical structures.” Journal of Mathematical Psychology. 15: 234-281. ———. 1980. The analytic hierarchy process. New York: McGraw-Hill. Saaty, T. L., and J. M. Alexander. 1989. Conflict Resolution: The analytical hierarchy approach. New York: Praeger Publishers. Saaty, T.L., and K.P. Kearns. 1985. Analytical planning: The organization of systems. Oxford: Pergamon Press. Scott, J.K., and T.G. Johnson. 1999. The Community Policy Analysis Network: A national infrastructure for community decision support. Western Wire. Spring: 5-7. Shaffer, R. 1989. Community economics: Economic structure and change in smaller communities. Ames IA: Iowa State University Press. Siegel, P.B, J. Alwang, and T. G. Johnson. 1995a. An input-output model: A framework for policy analysis, International Regional Science Review. 18(3): 331-353. ———. 1995b. Regional economic diversity and diversification, Growth and Change. 26(2): 261-284. Srisoepardani, K.P. 1999. Evaluation of group decision making methods. Expert Choice, Inc. http:// www expertchoice.com/support/ahpcompare/chapter6.htm. Tideman, T. Nicolaus and Gordon Tullock. 1976. A new and superior process for making social choices. The Journal of Political Economy, 84(6): 1145-1159. Venable, T. 1994. A banner year for U.S. business climates: Taxes fall, incentives fly, Site Selection Magazine: 848-858. Virginia Statistical Abstract. 1994-95. ???

LOCAL PREFERENCES 363

APPENDIX A: MEASURING INCONSISTENCY IN AHP AND ASSORTED TABLES As noted in the body of the paper, AHP is unique in that it allows a certain amount of inconsistencies in judgment. It also provides a straightforward means of examining whether observed inconsistency is within established bounds. The purpose of this appendix is to demonstrate how consistency is calculated. Inconsistency could arise for a number of reasons, such as a large number of comparisons being made, or problems with the scale of relative importance shown in the paper. An example of inconsistency is a lack of transitivity among judgments. If, for example, impact a is favored over b and b is favored over c then transitivity implies that a is favored over c. In some cases, because pairwise comparisons are being made, the AHP judgment matrix may not be consistent with transitivity. It is helpful to understand the degree of inconsistency and its implications. A measure of inconsistency is found using Consistency Index (C.I.), using λmax: (λ − K ) . (A.1) Consistency Index = max (K - 1) This value is compared to the Random Index (R.I.) value, which is created by filling a matrix of the same dimensions as is used in the study and randomly taking judgments from the Scale of Relative Importance (Saaty, 1980). The Consistency Ratio (CR) is the ratio of the C.I. to the average of the R.I. Saaty and Kearns (1985) suggest a CR between .1 and .2. Consistency Index (A.2) Consistency Ratio = Random Index If the consistency index does not fall below .20, then adjustments as described in the paper are required. TABLE A.1. RANDOM CONSISTENCY INDEX FOR MATRICES OF ORDER 1 THROUGH 15 ORDER OF MATRIX

RANDOMLY GENERATED INDEX OF CONSISTENCY

1 2 3 4 5 6 7 8 Source: Saaty 1980

0 0 .58 .90 1.12 1.24 1.32 1.41

ORDER OF MATRIX

9 10 11 12 13 14 15

RANDOMLY GENERATED INDEX OF CONSISTENCY

1.45 1.49 1.51 1.48 1.56 1.57 1.59

TABLE A.2. RAW SCORES FOR EACH INDUSTRY AND EACH IMPACT, BATH COUNTY

Raw Score Rank

Sector Name

Ave. No. of Jobs

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Research, Dev. & Testing Pipe Lines, Except Nat. Gas Maint. & Repair Oil/Gas Wells New Government Facilities Comm., Except Radio And TV Maint. & Repair, Residential Railroads And Rel. Serv. Arrangement Of Pass. Trans. Other Business Services Knit Underwear Mills Fluid Power Pumps & Motors Pleating And Stitching Electric Services Electronic Computers Comp. & Data Process. Serv. Marking Devices Phonograph Records and Tape Industrial Patterns Maint. & Repair Other Facil. Printing Trades Machinery Pottery Products, N.E.C. Transportation Services Ammun., Except Small Arms Special Dies & Tools & Accs. Spec. Prod. Sawmills, N.E.C.

71.54 6.77 88.00 31.25 19.10 39.24 24.33 67.32 62.32 100.00 53.11 68.43 8.88 19.07 38.94 66.15 35.69 45.81 36.35 28.97 71.75 38.28 39.13 43.90 58.26

RAW SCORES Ave. Level of Wage or Capital Salary Invest. 40.37 85.83 31.99 53.78 73.13 35.03 90.32 30.12 24.68 32.90 68.07 26.78 79.92 93.82 49.89 39.11 73.28 65.42 4.15 68.51 27.88 42.99 65.84 59.98 28.81

22.51 100.00 28.73 34.28 54.14 31.98 18.53 32.66 58.57 9.99 16.81 42.94 61.56 27.32 40.58 34.10 48.27 16.68 32.53 24.16 26.52 33.52 24.41 17.10 21.06

Impact of Pop. Growth

Level of Utility Require

-12.80 -3.06 -5.61 -3.87 -11.56 -3.87 -7.71 -2.27 -0.27 -1.73 -4.91 -3.37 -12.90 -4.91 -0.27 -0.19 -6.97 -4.91 -3.87 -4.91 -0.19 -2.27 -0.71 -4.91 -0.15

-12.02 -4.64 -5.05 -4.23 -7.38 -7.51 -13.25 -14.21 -14.89 -3.42 -0.82 -3.01 -30.19 -8.06 -9.84 -2.60 -1.37 -3.55 -5.19 -4.37 -6.28 -27.87 -6.28 -9.56 -1.50

Environ. Impact

Impact on Property Values

Total Score

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -8.76 -10.57 -6.10 0.00 -10.57 0.00 -15.55 -34.90 -10.57 0.00 -10.57 -10.51 0.00 -16.29 -10.57 -6.68

100.00 7.93 27.02 49.32 24.18 49.32 30.87 22.75 4.64 1.65 8.80 3.19 18.47 8.80 4.64 0.68 7.21 8.80 49.32 8.80 1.23 22.75 1.29 8.80 2.54

209.60 192.84 165.08 160.52 151.61 144.18 143.10 136.37 135.05 130.62 130.49 128.87 125.74 125.48 123.94 121.71 121.21 117.69 113.28 110.60 110.39 107.40 107.39 104.75 102.33

Rank Using Raw Score

Sector Name

26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52

Newspapers Bread, Cake, & Rel. Products Wholesale Trade Fast., Buttons, Needles, Pins Auto. Temp. Controls Misc. Publishing Dolls Hardwood Dimension & Floor Brooms And Brushes Wood Kitchen Cabinets Cutlery Inst. To Measure Electricity Cigarettes Greeting Card Publishing Wood Office Furniture Leather Goods, N.E.C. Bookbinding & Related Distilled Liquor, Exc. Brandy Wood TV & Radio Cabinets Costume Jewelry Food Products Machinery Typesetting Luggage Optical Inst. & Lenses State & Local Gov. - Non-ed. Canvas Products Wood Pallets And Skids

Ave. No. of Jobs

Ave. Wage or Salary

43.15 24.04 39.81 49.33 48.39 23.61 45.63 57.39 39.41 45.34 22.99 25.48 3.48 20.72 42.45 56.62 64.64 5.92 52.98 46.94 28.30 48.71 36.69 30.66 94.74 49.63 53.04

42.28 49.85 53.31 43.71 55.58 50.10 53.53 29.45 35.22 32.89 63.32 75.60 100.00 47.09 40.72 30.04 34.96 84.29 36.83 30.70 59.17 45.73 35.70 66.84 46.73 33.76 25.35

RAW SCORES Level of Impact Capital of Pop. Invest. Growth 42.21 45.82 18.29 26.61 22.91 50.74 17.38 24.13 41.47 29.37 41.49 21.15 47.21 50.76 27.61 24.24 17.64 16.75 16.23 37.40 22.74 19.88 33.14 18.60 0.00 14.44 16.07

-1.42 -0.18 -2.87 -0.19 -12.94 -1.42 -0.19 -0.15 -0.19 -0.15 -0.71 -12.94 -34.78 -1.42 -2.03 -0.25 -1.42 -0.18 -2.03 -0.19 -4.91 -1.42 -0.25 -12.94 -100.00 -3.37 -0.15

Level of Utility Require -8.33 -13.52 -25.41 -4.10 -5.05 -7.38 -1.78 -8.33 -3.28 -6.01 -15.30 -5.87 -7.38 -6.97 -5.05 -3.55 -7.92 -10.38 -2.73 -8.33 -12.43 -6.42 -2.32 -5.19 0.00 -4.10 -3.01

Environ. Impact -19.19 -5.65 0.00 -15.55 -18.90 -19.19 -15.55 -6.68 -15.55 -6.68 -16.29 -18.90 -19.58 -19.19 -17.02 -15.80 -19.19 -5.65 -17.02 -15.55 -10.57 -19.19 -15.80 -18.90 0.00 -6.10 -6.68

Impact on Property Values 3.43 1.15 17.84 0.68 10.20 3.43 0.68 2.54 0.68 2.54 1.29 10.20 5.69 3.43 7.50 2.37 3.43 1.15 7.50 0.68 8.80 3.43 2.37 10.20 46.18 3.19 2.54

Total Score 102.13 101.51 100.98 100.51 100.18 99.90 99.72 98.35 97.77 97.30 96.81 94.71 94.65 94.42 94.18 93.65 92.13 91.90 91.75 91.65 91.09 90.73 89.52 89.26 87.64 87.47 87.16

Raw Score Rank 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75

Sector Name Household Furniture, N.E.C. Apparel From Purch. Materials Shoes, Except Rubber Mach. Tools, Metal Cut. Types Wood Partitions & Fixtures Wood Containers Narrow Fabric Mills Complete Guided Missiles Plate Making Women’s Hosiery, Exc. Socks Musical Instruments Advertising Federal Gov. - Non-military Cut Stone & Stone Products Ophthalmic Goods State & Local Electric Utilities Misc. Fabricated Wire Products Mach. Tools, Metal Form. Types Fabric. Plate Work Millwork Furniture & Fixtures, N.E.C. Wood Products, N.E.C. Small Arms

Ave. No. of Jobs 57.01 47.58 51.15 32.31 40.58 50.83 48.76 18.56 30.98 48.10 46.36 47.18 62.53 47.98 42.96 11.61 36.80 32.57 32.83 36.50 19.89 51.25 24.04

RAW SCORES Ave. Level of Impact Wage or Capital of Pop. Salary Invest. Growth 34.43 9.88 -2.03 28.01 20.02 -3.37 32.03 19.06 -0.25 62.14 9.91 -4.91 44.66 17.24 -2.03 27.82 12.89 -0.15 36.33 13.49 -1.73 89.61 26.77 -23.59 65.77 11.30 -1.42 36.09 13.78 -1.73 35.89 27.62 -0.19 6.22 49.38 -0.27 70.79 0.00 -100.00 36.33 15.99 -0.19 48.40 24.76 -12.94 65.13 65.86 -12.90 47.12 24.06 -0.71 61.01 7.48 -4.91 56.92 13.70 -0.71 37.34 23.63 -0.15 46.10 24.10 -2.03 29.11 25.03 -0.15 53.31 24.14 -0.71

Level of Utility Require -3.69 -3.55 -3.01 -12.16 -6.15 -2.87 -5.46 -5.19 -8.33 -7.79 -14.21 -27.05 0.00 -11.34 -15.03 -69.54 -14.07 -16.80 -10.25 -18.85 -4.64 -29.51 -14.48

Environ. Impact -17.02 -6.10 -15.80 -10.57 -17.02 -6.68 -8.76 -33.50 -19.19 -8.76 -15.55 0.00 0.00 -10.51 -18.90 0.00 -16.29 -10.57 -16.29 -6.68 -17.02 -6.68 -16.29

Impact on Property Values 7.50 3.19 2.37 8.80 7.50 2.54 1.65 10.20 3.43 1.65 0.68 4.64 46.18 1.23 10.20 18.47 1.29 8.80 1.29 2.54 7.50 2.54 1.29

Total Score 86.08 85.79 85.54 85.53 84.78 84.38 84.27 82.85 82.53 81.34 80.61 80.10 79.50 79.49 79.45 78.63 78.21 77.58 77.50 74.33 73.90 71.59 71.30

Note: Scores are computed as ratios of the actual score by industry divided by the actual score of the largest score for all industries. For example, Knit Underwear Mills employ the largest number of people per firm of all industries; Small Arms firms employ on average 24 percent of the number employed in the Knit Underwear sector.