THE ANALYSIS OF THE FORESTRY PRACTICES REFERENDA IN THE STATE OF MAINE IN 1996 AND 1997

This research project describes and explains the 1996 and 1997 Maine forestry practices referenda, which came about amid heightened concern about threats to Maine's large tracts of privately-owned forestlands. part of a larger area known as the Northern Forest. It explores the factors influencing how people cast their votes in the two referenda and the ties between these factors and the spatial patterns of the votes. This research shows that in addition to concern for the environment, economic considerations of individuals were an important factor with regard to how people cast their votes. Attempts to influence voter opinion by the different sides in this issue with media messages also played an important. although hard to quantify role in the ultimate defeat of any changes to existing forestry practices regulations . Analysis of voting patterns suggests an important role for grassroots environmental and property rights groups in influencing the outcome of these referenda. The ballot questions put before Maine voters are of particul ar importance in deciding the future of the Northern Forest. an important economic, recreational and ecological resource that the proposed changes to forestry practices regulations would have affected. This resource stretches into several neighboring states who may use this research to gauge citizen reaction to changes in forestry regulations.

This situation has generated a great deal of concern among environmentalists and forest users in general as forest products companies become more aggressive in finding ways to maximize returns to shareholders. One way the companies do this is to shift capital by liquidating forest landholdings through outright sales or by increasing the timber harvest to generate cash.
One way to increase harvest efficiency that is of particular concern to environmentalists is the practice of clearcutting, where all of the commercially valuable trees in large contiguous areas are removed. Many view clearcutting as at least unsightly if not environmentally unsound. While this practice has been used long before globalization of the economy, it has recently become just one issue in the larger debate on the future of the Northern Forest.
In 1995 the Maine Green party led by Jonathan Carter, frustrated by existing forestry practices regulations, began a petition drive to gather enough signatures to place an initiative to ban the practice of clearcutting in Maine's unincorporated territories on the ballot in the 1996 statewide election. Maine's governor, Angus King, forest products industry representatives, and some mainstream environmental groups responded by developing a compromise ballot choice called "the Compact for Maine· s Forests" . A third choice, required by Maine ' s Constitution, was no change to existing forestry practices regulation, which already placed restrictions on clearcutting, although not enough to adequately protect the forest in the view of Carter.
2 These ballot choices and the debate surrounding them polarized the population generally into those who feared that further restrictions to clearcutting would cause a loss of jobs and slow the state's economy, which was still recovering from a recession in the late 1980s, and those who feared that global economic forces acting on large forest landholders would lead to increasingly unsustainable forestry practices and loss of traditional multiple use of the forest resource. In addition, many property rights groups were opposed to any kind of additional regulations on the use of private property and saw both the ban clearcutting and compromise option as unacceptable government interference. All sides of the issue waged a fierce campaign to influence voters and ultimately none of the three ballot choices received more than the 50% required for passage. Because the Compact received the most votes in 1996, it became subject to a runoff election where voters were given the choice of accepting or rejecting the Compact.
In 1997 the Compact was defeated by a close margin. This essentially left the issue of a ban on clearcutting unresolved and likely to resurface as a ballot choice or as legislation in the future.

Objectives of the Research
The research has three objectives. First, it provides the reader with the necessary economic, social and geographic context for understanding the complex debate surrounding changes to forestry practices regulation that were put before Maine 's voters.
Second, it traces the history of initiatives and referenda both generally and specific to the 1996 and 1997 elections and shows the spatial distribution of the votes for the various ballot choices in both years. Third, it examines the relationship between demographic 3 and economic variables, and the results of the 1996 and 1997 votes descriptively and quantitatively to determine if they support the research hypotheses described below.

Research Hypotheses
The first set of hypotheses deal with the relationship between the 1996 and 1997 votes on the compromise option: the Compact for Maine 's Forests. The null hypothesis for this relationship is that the votes for the Compact in 1996 and 1997 are independent of each other. The alternative hypothesis is that a vote of yes for the Compact in 1997 was significantly related to a vote of yes for the Compact in 1996, more specifically. that in counties where the majority voted for the Compact in 1996 there was a significant likelihood that the majority would vote for it again in 1997 , indicating the presence and location of areas of core support for the Compact.
The second hypothesis deals with the role of media campaigns of the proponents and opponents of the three options in the 1996 election and proponents and opponents of the Compact in the 1997 election. The null hypothesis is that media campaigns did not influence the voting results in either year. The alternative hypothesis is that media campaigns had a strong influence on the voting results in both 1996 and 1997.
Two closely related variables, percent of the age 25+ population that graduated high school ( 1990) and percent of the age 25+ population that graduated college ( 1990), were used as indicators of social status, which has been demonstrated as a relevant variable in referendum elections (Hahn and Kamieniecki 1986, 93-113). Voting preferences of low-status and high-status voters, as indicated by level of education, may be formed by contrasting perceptions of what is in the public interest or general welfare 4 of the community. In some conceptualizations of the relationship between status and voting, higher status (more educated) voters "express their tastes. values and lifestyles through the acceptance or rejection of referendum proposals unrelated to major financial considerations" (Hahn and Kamieniecki 1986, 49-50). Along these lines, the null hypothesis is that there is no relationship between votes for any of the 1996 referendum choices and level of education, both high school and college. Alternatively, the research hypothesized that voters in counties with higher education levels will be more likely to vote to ban clearcutting and less likely to support the Compact or None of the Above choices because these voters generally place a higher value on environmental protection and can make this decision independent of financial considerations. Similarly, the null hypothesis for the 1997 vote is that there is no relationship between votes for the 1997 referendum choices and level of education, both high school and college. The alternative hypothesis for 1997 is that counties with higher education levels will be significantly more likely to vote for the Compact than those with lower education levels.
The null hypothesis for per capita income, another indicator of social status , is that it has no significant relationship to the votes in the 1996 or 1997 referenda.
Alternatively, the research hypothesized that counties with higher per capita income would be significantly more likely to vote for the Clearcut Ban and against the Compact or none of the above option in 1996 and for the Compact in 1997.
The variable population density, as measured in persons per square mile in 1995, was also analyzed as an independent correlate to the 1996 and 1997 referendum votes.
This variable serves as an indication of the "ruralness" of the individual counties. The concept of this variable is that there is a land use ethic and lifestyle associated with rural 5 living that shapes voter preference. This ethic and lifestyle promotes a strong sense of individualism, self reliance. and accompanying opposition to government intervention into private land use decisions such as changes to forestry practices regulations. In this conceptualization, voters in more rural areas will vote more on the basis of opposition to new government regulation than voters from more urban areas. For this variable the null hypothesis is that there is no significant relationship between the votes for any of the 1996 referendum choices and population density. Alternatively. the research hypothesized that voters in counties with higher population densities would be more likely to vote to ban clearcutting and less likely to vote for the Compact or none of the above in 1996. The null hypothesis remains essentially the same for the 1997 vote (no relationship) while the alternative hypothesis is that counties with higher population densities will be more likely to vote for the Compact.
Another variable closely related to population density is area under the jurisdiction of Maine's Land Use Regulatory Commission (LURC). This is because the unincorporated areas over which LURC has jurisdiction tend to have very low populations. This variable serves as a rough proxy for the amount of land owned as commercial woodlots and is also closely related to forest products industry employment, the next variable to be discussed (Figure 3.2). This is because the majority of commercial timberlands lie in unincorporated territories under LURC jurisdiction (Bradbury 1996, 1 ).
The null hypothesis for this variable is that there is no relationship between the county votes for any of the choices in 1996 and area under LURC jurisdiction. Alternatively, the research hypothesizes that voters in counties with greater area under LURC jurisdiction will be more likely to vote against a ban on clearcutting, against the Compact, and for the 6 none of the above option in 1996. The null hypothesis remains essentially the same for the 1997 (no relationship) vote. while the alternative hypothesis is that voters in counties with greater area under LURC jurisdiction will be more likely to vote against the Compact in 1997. The concept here is that voters in and near areas of extensive LURC jurisdiction will tend to vote against additional state regulation. seeing land use controls in general as a local issue, and forestry practices regulation as a threat to local economic health.
The next two variables and their relationship to the 1996 and 1997 votes lie at the core of the research. which is that in the 1996 and 1997 referenda. voter preference was largely a function of financial considerations. That is, voters in counties that are more economically dependent on the forest products industry were significantly more likely to vote against any regulation that might hinder the industry and cause an accompanying decline in employment. The null hypothesis is that there is no relationship between county votes for any of the three choices in 1996 and forest products industry employment. Alternatively this research hypothesizes that voters in counties with higher forest products industry employment will be more likely to vote against a ban on clearcutting, against the Compact. and for none of the above options. The null hypothesis remains essentially the same for the 1997 vote (no relationship) while the alternative hypothesis is that voters in counties with greater forest products industry employment will be more likely to vote against the Compact in 1997.
Like forest products industry employment, this research hypothesizes that county unemployment would have a similar correlation to the votes as forest products industry employment, although unemployment figures are not necessarily a measure of people 7 who had been working in the forest products industry that lost their jobs. The concept of this \·ariable is that many voters equate clearcutting restrictions with a loss of jobs and that \'Oters in areas already experiencing high unemployment relative to the rest of the state would have a concern that the unemployment level would increase. The null hypothesis is that there is no relationship between any of the choices in both the 1996 and 1997 votes and level of unemployment. Alternatively, voters in counties with higher unemployment will be more likely to vote against the clearcut ban and Compact and for the none of the above and then against the Compact in 1997.

Research Methodologies
Three general methodologies were employed in this research: descriptive analysis, quantitative statistical analysis and spatial analysis. Descriptive analysis was used to assess the role of the media in influencing the 1996 and 1997 election results and the correlation between county characteristics and voting results. Quantitative statistical analysis techniques in the form of chi-square was used to test the hypotheses concerning the correlation between votes for the Compact in 1996 and 1997. Multiple regression analysis was used to test the hypotheses on the correlation between the votes and education level. area under LURC jurisdiction, income, population density, unemployment and forest products industry employment. Spatial analysis was used to examine voting patterns and their proximity to concentrated areas of forest products industry employment and other significant social and geographic features. 8

Significance of the Research
The forestry practices ballot questions put before Maine voters were of great potential importance in deciding future land use in the Northern Forest. An understanding of voter response to the choices presented in 1996 and 1997 and the reasons behind them is especially important because the issue was not resolved and is likely to surface again in the form of a ballot question or in proposed legislation. Another reason that this research is significant is because the Northern Forest stretches into several neighboring states who may use this research to gauge citizen reaction to future proposed changes in forestry and other land use regulations .

Introduction
Chapter Two

Literature Review
The purpose of this chapter is to review literature that is relevant to an understanding of Maine's population, economy and geography. This chapter will also review the general literature on referenda. This literature review falls into four distinct categories: 1. Maine's economic and demographic characteristics at the state, regional and county levels; 2. Maine land use, with a particular emphasis on the densely forested northern half of the state, part of what is known as the "Northern Forest"; 3. General characteristics of referenda and initiatives. Included in this topical area are the role of politics and the media, and writings on grassroots property rights and environmental organizations; and 4. Correlation between the referenda votes and economic and demographic characteristics of areas.

Maine's Forest-Based Economy
This section, describing aspects of Maine ' s economy, is divided by different regional scales, from the global and national level down to the municipal level. At the state level, Maine Business Online ranks the paper industry first in both employment and value of product produced in the manufacturing sector of the economy. The lumber and JO wood products industry ranked third. In combination these two forest products industries have nearly two times the employment of the next largest manufacturing employer, transportation equipment (Haugen 1996, 2 Forest counties are critical to this study because of its central hypothesis that people in counties that are more dependent on the forest products industry will have a tendency to desire the least restrictions on forestry practices and will vote accordingly.

Maine Land Use
The Northern Forest Klyza and Trombulak, in

Ownership
With regard to land ownership patterns. Klyza and Trombulak (1994) characterizes the portion of the Northern Forest that lies in Maine as large ( 15 million acres), and comprising a center of large industrial and private property ownership. They characterize landowners as falling into four categories.
( 1) small , local landowners who live year-round on their land; (2) small landowners who use their land as a second vacation home; and (3) large corporate and family landowners who actively harvest and work the land industrially (Klyza and Trombulak 1994 Practices Act which placed restrictions on clearcutting in 1986. Klyza and Trombulak ( 1994) describe the forest products industry harvesting practices that impact the environment as: • "Cutting and removing trees from large areas" (i.e . clearcutting); • "Use of heavy machinery in harvesting operations"; • "Herbicide spraying to reduce unwanted tree species"; • "Replanting areas with monocultures of commercially desirable species"; and • "Repeated cutting at short intervals" (Klyza and Trombulak 1994, 20-21

General Background on Referenda and Initiatives
The initiative and referenda literature ranges widely from broad treatments to analyses of specific ballot issues, and the use of referenda and initiatives for political change in America has a long and complex history. Referenda have been a means for citizen participation in government since colonial times (Schmidt 1989, 3). Their use brings up fundamental debate about the roles of representative democracy, as manifested by state and federal legislatures and executive officers, and direct democracy, as manifested by citizen generated initiatives and referenda put on the ballot by legislatures.
There is an important distinction between referendum and initiative described in David Schmidt's Citizen Lawmakers. An initiative is a new idea initiated by citizen petition or by a legi slature and, in most states, a minimum number of voter signatures is required for an initiative to be placed on the ballot. A referendum is initiated by citizens or legislators, but must be approved by the legislature before being presented to the public for approval. Common usage has led to referendum and initiative meaning essentially the same thing for most people, i.e. anything put on the ballot for a vote. Most states allow their representatives to place referendums on the ballot. but only 23, including Maine, allow citizens. through the petition process, to place initiatives on the ballot (Galvin 1992. 3505). Maine, one of the earlier states to allow initiatives and referenda, voted to put the initiative process in place in 1908 (Schmidt 1989, 16 Some of the pros and cons of initiatives and referenda that have been described by various authors are summarized below (Cronin 1989, 61-62 and207-209 , Schmidt 1989, 26 and McManus 1997 20-2 I):

Summary of Advantages of Initiatives and Referenda
• Encourages more accountable government: May provoke legislators into action; • Greater Citizen Participation: Creates perception that ordinary citizens make a difference; • Sometimes lead to greater voter turnout in elections; •

Summary of Disadvantages of Initiatives and Referenda
• Initiati ves poorly written and often unconstitutional: They can't be modified like a bill , and mistakes can be made: • The side that spends the most money wins: Initiatives and referenda serve special interests. They enhance minority rule because many voters do not vote on them ; • Problems with dropoffs i.e. people vote for the candidates but not for the referendum question(s); • People unable to vote intelligently on complex issues , not informed or motivated enough to learn about complicated policy issues; • Process of signature collection for initiatives, which must be done quickly, is not conducive to information dissemination ; • Voters selfishly "vote with their pocketbooks" ; • Ballot access for initiatives is not easy, you must be organized enough to get thousands of signatures ; • Signature drive deceptions such as issue oversimplification : For example: "Do you want to avoid environmental catastrophe" ; and • Initiatives and referenda cause a weakening of state legislatures.

Referenda, Initiatives and Campaign Spending
Many authors have studied the reasons for the success or failure of ballot question campaigns both initiative and referenda. Some, in the specific instance of grassroots versus business interests, have attributed the wealth and organization of business interests against less organized and well financed grassroots groups as being the deciding factor in 20 success or failure. One specific case where this occurred was in a vote to shut down the Maine Yankee Nuclear Power Plant in Wiscassett where business interests spent nearly three times more than opponents and won. although by a narrower margin than the pending imbalance would suggest (Lyden berg I 983 , 53 ). This is ironic because at least in some places, initiatives, originally viewed as a mechanism by which ordinary citizens could overcome powerful business interests, are now viewed as tool s of power and resourceful business interests (Dwyre, et al 1994, 46). Convincing arguments can be made both for and against ballot questions. regardless of how they got on the ballot, and in 27 states without an initiative process the arguments against have prevailed.

Media Influence on Referenda and Initiative Voting
Campaign spending translates , in large part, to spending on consulting, petition gathering, and actual spots in the print, radio and television media (McManus 1997. 20-21 ). Spending large sums of money by itself doesn 't automatically guarantee the outcome of a ballot issue campaign, but it buys resources such as sophisticated public opi nion tracking surveys, telephone banks , get-out-the-vote drives. experienced public relations and media consultants, mass mailings and targeted media ads. These resources often provide the edge needed for success in a ballot issue campaign. "So called big money has only about a 25% success rate in promoting ballot issues . . . . However when big money (usually, though not always, business money) opposes a poorly funded ballot measure, the evidence suggests that the wealthier side has about a 75 percent or better chance of defeating it" (Cronin 1989, 109).
Media spots can also have the effect of confusing voters, who are more likely to vote against a ballot question if they are unsure of the benefits and uneasy about the risks .
They can also have the effect of swinging conditional voters , particularly in the final few day before an election. which is often when voters decide how they will vote on ballot questions. as opposed to partisan candidate elections (McManus 1997, 20-21 ). Some sources state that it is easier to defeat a referendum than to win one, and that many voters will adopt an attitude of "when in doubt, vote no" (Cronin 1989, 85).

Correlation Between the Referenda Votes and Economic and Demographic Characteristics of Areas
In their paper entitled Property Rights and Property Culture: State Property Rights Bills and the Districts whose Legislators Support Th em Pendall , et al use legislator votes on property rights bills to study geographic, socio-economic, and political locations of property culture in various states (Pendall ,et. al I 998). This study specificall y attempts to find out the clusters of geographic, socio-economic, and political characteristics that correlate with anti-regulatory movements.
In this paper the authors hypothesize that several land-use and ownership characteristics correlate with a district's representative's voting behavior. For instance, a higher proportion of forested land should correlate with support for property rights legislation because it is in these areas that environmental restrictions would be an economic threat and run counter to beliefs about how land should be used. Along these lines, more urban land may contain more residents who wish to retain or strengthen environmental controls, and that this would be reflected in the votes of their 22 representatives in the legislature. This study uses a similar approach, with the differences being that the actual vote can be observed rather than the votes of representatives to the state general assembly, which may or may not reflect the majority view of the residents in their districts. The forestry practices ballot questions are similar to property rights bills introduced in the Maine Legislature in that the options involve differing degrees of land regulation. In the case of the forestry practices referenda the options range from banning an important tree harvesting method (clearcutting) to not changing existing regulations.
with a compromise option in between. Pendall, et al concluded that the urban-rural distinction is very important, with urban representatives tending to vote against legislation that would give more power to property owners (Pendall , et. al. 1998). The authors also concluded that forest cover was significant in some of the votes, citing an example in Maine.

Findings
After careful review of the literature relevant to this research, several themes emerged that carry through the rest of this paper. The first theme is the importance of the forest products industry to the state economy in general, the manufacturing sector in particular and to the Northern Forest counties, which are more dependent on these industrial groups than other areas of the State. A sub-theme within the forest products industry is cyclical nature of lumber and paper product production and its dependence on the national and international markets , which tend to have more of an impact on the Northern Forest areas of the state. There also exists strong regional differences in the contribution of non-manufacturing industries to the economy, particularly fishing, 23 hunting, hiking and other forest-related recreational activities. In general. most sources emphasize that the economies of the Northern Forest counties are more closely linked to the natural resource base than those in the south and coastal regions. especially the manufacturing sector.

Introduction
Chapter Three

Maine in Profile
The purpose of this chapter is to provide the reader with background on the state of Maine and its counties to facilitate an understanding the outcome of 1996 and 1997 referenda. This was done by reviewing the general characteristics of Maine including its geography, land use, demographics. economy, and government, and by applying economic and demographic characterization methodologies. The resulting profiles of both the state and counties set the stage for analysis of the correlation between county characteristics and how these counties voted in the 1996 and 1997 referenda. Other useful sources in book form introduce one of the most significant natural features in the state, the so-called "Northern Forest" which is defined by the types of ecosystems it 25 contains, by its economic structure, and by regional policymaking (Klyza andTrombulak 1994 andDobbs andOber 1995).

Sources of Information
Demographic data aggregated at differing levels of spatial detail are a useful and important source of information for creating profiles of local areas (Myers 1993). Some of the more common types of census data used for profiling include population.
household or per capita income, and ethnicity or diversity. This view of the importance of population and its dynamics is particularly relevant for this study because of their potential application for explaining demographic/social factors behind how votes are cast in a certain area. The tendency of census data to be more richly detailed at the regional level than at the local and sub-county level has been noted by some sources and data availability at differing spatial scales is an important factor in the development of analysis methodologies (Myers 1993 andKlosterman 1990).
As for economic analysis of areas , one of the most important steps is the identification of the boundaries of the study. Some of the most useful boundaries are counties and multi-county regions due in large part to the availability of reliable economic data at regular intervals (Klosterman 1990).
Information on employment, a commonly used unit for economic analysis. is obtainable for counties each year in U.S. Bureau of Census County Business Patterns.
Payroll is another useful measurement of economic activity because it accounts for things such as overtime and seasonal employment, and can give a rough estimate of an industry's contribution to the local economy (Klosterman 1990 This type of data has useful application to assessing the contribution of the forest product industry to the state economy because it captures some of the "ripple effects" that would be missed by just looking at employment and payroll in County Business Patterns, although this data is often only available at the state level (Klosterman 1990

Overview of Maine Forest La.nd Use and Ownership
In total, Maine contains 17.5 million acres of woodlands representing 90 percent of its total land area. This vast forest cover makes it the most heavily forested state in the nation. Almost half of the forest acreage. 8.1 million acres, is owned by eight large paper corporations, sawmills and other manufacturing concerns. This gives Maine the highest concentration of "industrially-owned" forest in the country. Industrial ownership refers to companies that own manufacturing facilities. including pulp and paper product mills.
Large "non-industrial" landowners who se ll wood to the forest products industry own another 3.1 million forest acres in tracts larger than 5,000 acres. "Approximately one hundred thousand small landowners control 5.4 million acres, and the rest. 900,000 acres, is publicly owned -at 5 percent, the smallest percentage of public land in the Northeast" (Dobbs andOber 1995 117-118, Irland 1995, 13 and  Ownership considers a smaller area as woodland than Dobbs and Ober or Irland but provides a helpful illustration of these land use patterns when represented spatially ( Figure 3.2). With such large landholdings , large forest products companies are a major force in the management and ultimately the character of much of Maine's forested areas (Klyza and Trombulak 1994, 36). Agriculture in contrast, only accounts for 7.6% of total private land usage, following a trend of decline for at least a decade (CBER 1998,  The Northern Forest provides the raw materials which power the forest products industry, a large presence in the state, regional and local economies, and the management of timberlands is closely tied to the global economy, which experiences cycles of demand for forest products. The Northern Forest also provides a place for recreational activities, which also makes a significant contribution to the economy at different geographic scales. At the same time, this area is prized by conservationists for its large contiguous tracts of relatively intact forest. The above described situation has resulted in competing and conflicting uses of the Northern Forest. Government at the federal state and local level struggle to balance conflicts between timber harvesting, recreation and conservation activities which are in turn influenced by complex social and economic forces (Klyza and Trombulak 1994, 66). All Woodlands 17.5 100 (Dobbs and Ober 1995 , 117-118)

Demographics
Maine's 1997 population stood 1.224 million , a -0.3 % change from 1990 when the population was l.228 mi Ilion, and even lower than the 1980 population of l. 125 million. Its population is forecast to increase 6.3 % to 1.30 l million by 2006 (Table 3.2).  which contains the capitol city of Augusta, and where state government would be a significant employer with a large contribution to the county's economy, or in some of the coastal counties with large tourist industries , and thus many self-employed workers.
However, since the focus of this paper is on the forest products industry, which is covered under the SIC manufacturing sector category and more detailed major groups. these shortcomings will be overlooked for this research . The following table shows the most basic SIC divisions, or sectors, and some specific major groups related to the forest products industry used in this paper to characterize the economy of individual counties (Table 3.3). Under the manufacturing sector, the major group lumber and wood products ... "includes establishments engaged in cutting timber and pulpwood: merchant saw mill s, lath mills , shingle mill s, cooperage stock mills , planin g mill s, plywood mill s, and veneer mills engaged in producing lumber and wood basic material s: and establishments engaged in manufacturing finished articles made entirely of wood or related material s" (Office of Management and Budget 1987, I 07). Another major group of interest, Logging, which is a sub-category of lumber and wood products, includes . .. "establishments primarily engaged in cutting timber and in producing rough, round, hewn , or riven primary forest or wood raw material s, or in producing wood chips in the field ." This major group doesn't include activities such as the trucking of timber or maple sugaring (Office of Management and Budget 1987, 107). The major group paper and allied products 37 includes ... "establishments primarily engaged in manufacturing pulp from wood or from other materials, such as rags, !inters, wastepaper, and straw" (Office of Management and Budget 1987, 119). Together these major groups cover most of the forest products industry's activities with the exception of trucking. In 1995 Maine's entire manufacturing sector, which includes the major groups described above, accounted for 20.9 percent of state employment and 27.3 percent of payroll. The lumber and wood products, and paper and allied products groups made up 12.7 and 15.2 percent of state manufacturing employment respectively, although they represented less than 6% of the overall state employment. These figures belie the fact that when ranked by the value of the products produced, the paper and allied products major group alone ranked first by far among manufacturing industry groups in value of product produced at $3.7 billion (Haugen 1996, 2).
Forest-based industries covered under these major groups have increased their share of manufacturing production in Maine since 1905 during a time when manufacturing's share of total employment has declined. This runs counter to normal economic growth where the share of resource-dependent industries shrinks as other nonresource-dependent industries grow. This is due to productivity increases in the forest products industry (Irland 1998, 19). Lumber and paper product production tends to be cyclical depending on demand in the national and international markets. "From 1986 to 1991, fully half of the increase in Maine's manufacturing output was due to increases in lumber and paper sales volumes. In 1992, the top three employers in manufacturing were paper ( 16,489), transportation equipment (mostly shipbuilding) ( 13,067), and lumber and wood" ( 10,794) (Irland, 1998 p.19). In 1992 the average hourly earnings in the pulp and 38 paper industries was $16.25 versus $11.38 for all manufacturing and $9.25 for lumber and wood products industries (Irland, 1998 p.19).
The economies of the Northern Forest counties (Oxford, Franklin, Somerset, Piscataquis, Aroostook, Penobscot, Hancock and Washington) are more closely linked to the natural resource base than those in the south and coastal regions , especially the manufacturing sector. The timber resources of these northern counties provide the primary input for the forest products industry. Large paper mills , many owned by multinational corporations, process the region's trees into products such as regular white paper. newspapers, specialty papers, and cardboard boxes. More numerous sawmills cut logs into lumber products that are used for home construction, kitchen cabinets and furniture (Klyza and Trombulak 1994. 52).
Maine's unemployment has shown greater variation than that of New England but in general, the southern portion of the state has experienced unemployment levels significantly lower than the rest of the state.
The service sector dominates Maine's economy, providing one out of every three jobs based on 1995 data. The health services major group leads the service sector.
accounting for almost half of the service employment (Figure 3.4 and Table 3.4). Retail trade accounts for roughly one quarter of state employment led by eating and drinking establishments and food stores, common staples in any economy. The rest of the major standard industrial classes listed in table 3.4 contain the remainder of the state's jobs, with no sector accounting for more than 6% of the total (CBER 1998).
Maine's judicial branch of government operates at three levels. including district judges, a superior court, and a supreme court. The sixteen counties provide the governmental structure for the superior court system, law enforcement. and land records.
Counties are also responsible for some road maintenance and construction functions, however as is typical of New England states, county government is relatively weak compared to southern and western states, and many land-use regulatory functions such as zoning are performed at the municipal level (Brittanica, 1998).

General Overview of Maine's Regulatory Structure
At the direct regulation level, the Department of Environmental Protection (DEP) exercises considerable power over land use and planning through the Site Location of Development Law, which requires a permit for developments with potential for substantial impact on the environment. This law covers any development that occupies more than 20 acres and subdivisions , structures, mining, or excavation of natural resources with an area greater than 60,000 square feet. The DEP also has the power to regulate dredging and filling of coastal wetlands (Growsmart 1996, 2). U.S Army Corps of Engineers and the U.S. Environmental Protection Agency also regulate certain types of wetland alteration (Irland 1995, 77 and Table 3.5).  (Table 3.5). Its provisions include: • Standards for tree regeneration; • Performance standards (e.g. maximum clearcut size -250 acres); • Harvesting plans required for clearcuts over 50 acres;

46
• Provisions for variances: and • Provisions regarding transfer or sale of property (Maine Revised Statutes 1997) Local La.nd Use Regulation The state allows, but does not require, municipalities to adopt comprehensive plans and growth management programs, which must be enacted by the municipal legislature to be recognized as valid (Growsmart 1996, 4). Cities and towns may adopt zoning ordinances and land use controls, including timber harvesting practices. and these must also be enacted by the municipal legislature to be valid. State law mandates that the zoning ordinance be consistent with the comprehensive plan , if one exists. The state also mandates zoning and land use controls in shoreline areas (Growsmart 1996, 4-5 • Increasing polarization among forest user groups; • Rising property taxes, causing loss of land from natural resource uses; • Pressure for development of high-value areas near shorelines and scenic places; • Jobs lost to competition from other regions and countries; • Incomplete knowledge of land management techniques to maintain or enhance biological diversity; • Lack of funding and clear priority-setting for public land and easement acquisition; • Insufficient attention to and funding for public land management; • Fear of losing public recreational opportunities and access to private lands; • Loss of respect for the traditions of private ownership and uses of private land; and • Failure to consider forest land as a whole, as an integrated landscape (Northern Forest Lands Council 1994, 11-12).
The NFLC made the following specific recommendations to address these issues: • Change tax structure away from taxation for highest and best use (i.e. development) and toward current value and current use; 49 • Encourage sustainable forest management through education and an assessment of forest practices and programs impacts; • Increase funding for public land management agencies geared towards land acquisition planning programs; and • Encourage market cooperatives, networks, and direct assistance to natural resource-   Table 3.6).

51
The service sector accounts for 36% of Androscoggin county's total employment of 39,431 followed by retail trade which accounts for another approximately one third.
Manufacturing made about a quarter of the county's 1995 employment (9.4 11 workers).
Of this figure the lumber and wood products and paper and allied products together account for about 15 % of the manufacturing jobs. None of the sectors make up more than 5% of employment (Appendix A).

Aroostook County
Aroostook County, the northernmost county, ranks first among counties in land area with 6,819 square miles. trade. with 8% of employment reflects Portland's role as a regional distribution center.
Cumberland County relies less on manufacturing employment than any other county at 12% of the total employment. Of this 12 percent only 15 % comes from the lumber and wood or paper and allied products categories, although in absolute numbers these categories employ more workers in these categories than in many of the counties.
particularly in the south. These figures are based on an estimate for paper and allied products (Appendix A) .

Franklin County
Franklin County, located in the western mountain region , ranks seventh among counties in land area with 1,744 square miles. civilian labor force at 15 ,530 workers , was one of the smallest county workforces in the state. The county's labor force enjoyed the fourth lowest unemployment rate in the state at 6.0% and a relatively higher per capita income of $20,483 in 1993 (Table 3.6).
The service sector accounts for 32% of Lincoln county's total employment of 6,886 followed by retail trade which accounts for another approximately one third.
Manufacturing made up only 1 1 % of the county's 1995 employment of 788 workers and of this figure , only 3. 17 % or 25 jobs can be attributed to lumber and wood products, none to paper and allied products. The only other sector making up more than 5% is construction at 6.56% (Appendix A).    Table 3.6).

Oxford County
The service sector accounts for almost 33 % of Waldo county's total 1995 employment of 5,783 followed by retain trade and manufacturing which each contribute about a third of the county's employment. Of this figure the lumber and wood products and paper and allied products together account for about 16% of the manufacturing jobs.
nearly all in lumber and wood products, although some of these figures are estimated so the actual number may vary (see explanation in Appendix A). Transportation and public utilities is the only other sector to employ more than 5% of the workforce (Appendix A). York County's service sector employs slightly less than one third of jobs followed closely by retail trade at 28 %, a reflection of the numerous tourist related restaurants and miscellaneous retail establishments of this recreation al area (Appendix A).

Washington County
Manufacturing makes up about one quarter of the county employment. Less than 5%, or 353 workers total. of manufacturing employment comes from the lumber and wood or paper and allied products categories, indicating that these activities do not contribute a significant portion of the economy. Thi s is an estimate for paper and allied products. No other major sector category makes up more than 5% of the employment (Appendix A).   This chapter also makes an attempt to assess the role of the media in influencing the outcome of the 1996 election. Following a similar structure to the 1996 referendum , the runoff election that resulted from the 1996 vote is discussed in terms of campaigning, geographic distribution of the votes, and the role of the media in influencing the voting results.

The Diamond International Deal
In 1988 Diamond International Corporation, experiencing pressure to divest some of its timber landholdings in the wake of a hostile takeover, started a complicated chain of events that led to the 790,000-acre portion of their 970,000 of timberland holdings in the Northern Forest of Maine ultimately being sold to a combination of the State. The Nature Conservancy and several paper companies. This alarmed both environmentalists, who feared that this was the beginning of fundamental land use changes that would be detrimental to Northern Forest ecosystems, and the forest products industry whose leaders feared that large tracts of land would be removed from timber production. making it harder for the industry to cheaply obtain the raw materials needed to make lumber. paper, and related products (NFLC 1994). Another fear of environmentalists was that timber landholders would increase harvests in anticipation of selling land to development speculators. land that many environmentalists and others believed was already being overharvested.
These fears led to the creation of The Northern Forest Lands Study. whose charge was to gather economic, biological and social data on the Northern Forest. and later the Council of Governors and The Northern Forest Lands Council , whose charges from Congress was to make recommendations based on an intensive study of the complex land use issues concerning in this region. Regulations to limit clearcutting had been in place since passage of the Forestry Practices Act in 1989. The Forestry Practices Act placed a limit on the size of clearcuts at 250 acres, required buffers between clearcuts, and required owners to plant trees if the clearcut area did not regenerate naturally. "Since the law went into effect, annual clearcutting has averaged about 55,000 acres statewide. That means that 13 percent of all the acres from which trees are taken are clear-cut. The remaining 87 percent of acres are harvested with partial cuts, in which loggers remove some trees and leave others for the future " (Bradbury 1996, 2-3). Many environmentalists, believing that these regulations were not enough to protect Northern Forest ecosystems, advocated tougher regulation. A source well informed about forest products industry, on the other hand, stated it was difficult to get permission to clearcut over 70 acres, much less the 250 acre maximum and that the regulations were already stringent (Whitney 1999).
In the late 1980s sentiment against large clearcuts increased both inside and outside the forest products industry. Foresters and commercial woodland managers within the industry feared a backlash against clearcuts with the potential to turn the public against forest management in general. Ecologists and environmentalists outside the industry feared that increased clearcuts and the monoculture of commercially valuable softwoods that replace them would make the next outbreak of spruce budworm. a destructive insect that kills commercially valuable trees, worse than the disastrous one experienced previously. "Environmentalists were becoming more vocal, and aerial photographs of sprawling clearcuts in Maine started appearing in regional publications" (Dobbs and Ober 1995 , 126).
A 1995 Maine Forest Service study concluded that in the period of 1991-1993 , that for the most part, industrial landowners were using techniques that encouraged healthy forest growth on the large majority of their land, indicating proper management.
However other reports, some also by the Maine Forest Service, painted a conflicting picture of the 1995 assessment. One notable environmentalist critic Mitch Lansky, author of Beyond the Beauty Strip stated that many cut areas failed to meet US Forest Service guidelines for healthy forest growth (Bradbury 1996. 3).

The 1996 and 1997 Forestry Practices Referenda
The "Ban Clearcutting" Initiative  Bradbury 1996, and Maine Constitution). Carter characterized the initiative as pro-jobs because it would encourage the growth of a healthier forest (Bradbury 1996, 3). In the larger context of environmental groups active in Maine, the Green Party was more aggressive in advocating forestry regulation than the more conservative and mainstream groups like the Nature Conservancy and the Sierra Club. The Green Party's 1996 Platform Statement on natural resources, in addition to advocating a ban on clearcutting, proposed that the pulp and paper industry phase out completely the use of chlorine and chlorine-based chemicals by the year 2003 and "encourage fully sustainable. labor-intensive and chemical-free forestry practices" (Maine Green Party 1996). These positions were considered "extreme" by the forest products industry and were more radical than many mainstream environmental groups could tolerate.
The Ban Clearcutting initiative specifically said: • Clearcutting would be prohibited in unorganized territories: • Landowners would be required to leave more trees standing after a conventional harvest; • In a 15-year period. no more than one third of the trees could be removed from an acre of timberland; • The essential mixture of tree species could not be altered: • Tree limbs must be left near where they are cut; • The state may grant exceptions after a landowner proves hardship; and • The effective date would be April I st 1997. (Bradbury 1996, 1-2) According to Article IV , Section 18 of the Maine Constitution, in order for a group to gain access to the statewide ballot, they must first obtain an excess of ten percent of the total vote cast in the preceding gubernatorial election (Maine Constitution

A Response to the Initiative
The Green Party's success prompted immediate concern and debate among property rights organizations. the Governor's office, the forest products industry, mainstream environmental groups, sportsmen, labor unions , and ordinary citizens (Bradbury 1996, I). The forest products industry in Maine, with 30,000 employees and $5 billion in annual generated product value prior to the vote, represented a significant part of the state economy and this was a major concern for all of these groups , directly or indirectly. A more specific concern, that the ban would result in closure of small and independent mills and result in the loss of a large number of jobs. troubled many of these groups. Governor King's position on the issue was that it would stifle the economy and would be extremely expensive to enforce, especially without provision of funding for regulation. King estimated that the referendum's passage would necessitate the hiring of 10 foresters at a cost of $500,000 to $750,000 per year (Bradbury 1996. 1-2). His view of economic hardship was echoed by an article written in a Canadian newspaper that estimated a loss of 15 ,600 jobs, a $1.3 billion drop in economic output. a nearly 20% reduction in the wood supply, and a steep increase in the price that mills would have to pay for raw materials (Poitras 1996). The State Forest Service estimated that a clearcut ban would decrease the wood harvested in the unorganized territories by 58 percent, at a time when these areas supplied 62 percent of Maine's annual wood harvest (Bradbury 1996, 4). Although the ballot question that would be put to the voters were simple, the legislation behind them was complex and would give the state legislature a large mandate for placing further restrictions on timber harvesting (Whitney 1999).

A Third Option
A third option, "none of the above", was put on the ballot in accordance with the Maine Constitution, which states: "The measure thus proposed, unless enacted without change by the legislature at the session at which it is presented shall be submitted to the electors together with any amended form , substitute. or recommendation of the legislature, and in such manner that the people can choose between the competing measures or reject both" (Maine Constitution).

Property rights activists were at least in part represented by the voice of Mary
Adams of Common Sense for Maine's Forests and Alliance for America. She voiced the opinion of many property rights groups that the compact was a "land grab" by environmental organizations. in particular because it referred to set asides of "ecological forest reserves ", and that small tract owners, those with less than I 00,000 acres , would be hurt by the Compact, which was negotiated with the large timber tract holders. A complete ban on clearcutting was not considered an option by property rights activists who tended to support the " none of the above" option (Ell sworth American 1997 andSettler's Advocate 1996). A trade publication Pulp and Paper, stated after the election that the third option "none of the above" had most of its votes delivered by private property rights proponents.

The 1996 Referendum Campaign
A unique characteristic of this referendum/initiative is that while area that the three forestry practices ballot options would affect lies in the sparsely populated unincorporated territories, mostly within the Northern Forest, the outcome would be decided by the majority of voters residing in the more urbanized south (Bradbury 1996).
There were over half a million registered voters before the referendum in November 1996, but only 15 percent of the voting age population lived in the Northern Forest counties of Aroostook, Piscataquis, Somerset and Washington where the regulations would cover the majority of the land area (Bradbury 1996, 3). As is common with referenda. the initial response to the ban clearcutting initiative, before the other two options became available. was positive. with polls showing that over 70% of voting age population supported the initiative.
An article in the Portland Press Herald characterized support for the clearcut ban, as of March before the election. as being heavy in the southern and more urban areas, carrying a 71 % approval rating for the state as a whole according to one poll (Bradbury 1996, 4). Voter sentiment changed in the months leading up to the election as the proponents of the three referendum options intensified their media campaigns to sway voter opinion in their favor. As is common in referendum and initiative campaigns, the original instinctive support of the idea to ban clearcutting was replaced by a more negative view of its possible consequences by many voters (Cronin 1989. 84).
Some portrayed the media campaign for the Compact for Maine's Forests as a conspiracy by Governor King and multinational forest products companies to keep the industry self-regulated, and characterized the professionals hired by compact supporters as "initiative-crushers" (Huber 1998 ). Others painted grim pictures of what would happen if clearcutting was banned outright, including the loss of thousands of jobs and an over one billion dollars decrease in the state's economic output (Poitras 1998). This uncertainty may have kept any of the ballot options from gaining a clear majority.

The 1996 Referendum Results
As mentioned in the previous section, some counties had much greater importance in terms of number of voters (

Geographic Information Systems Data and Methodology
Geographic Information Systems (GIS) are a tool for collecting, storing, retrieving, transforming, and displaying spatial data for a particular purpose. GIS can be 84 used to handle complex spatially referenced data in a way that it can be generalized for analysis (Burrough and McDonnell 1998). Two basic data types are used in GIS; graphic.
consisting of points, lines and polygons , and attributes consisting of non-graphical data.
These two fundamental types of data are linked by geographic location identifiers (Kaiser and Godschalk 1995). Economic, demographic and voting tabulation data provide the material to create attribute tables which can then be linked to spatially oriented graphic elements such as county and municipal boundaries and presented visually (Kaiser and Godschalk 1995). There are many sources and types of GIS datasets available for this research project. which has economic social , political, and land use elements. The sources include: • Statewide political boundaries compiled from I :64,500 scale United States Geological Survey (USGS) quadrangle maps including standard town names, identifiers for county, and identifiers for whether the town is organized or unorganized. (Maine Office of GIS 1998); • Voting data, which can be merged with the previously described data coverages, (Maine Bureau of Corporations, Elections and Commissions I 996 and 1997); • Economic data which can be merged with the previously described data coverages (U.S. Bureau of Census 1995); and • Demographic data which can be merged with the previously described data coverages (U.S. Bureau of Census 1996). 85 These data can be analyzed with GIS to identify the spatial correlation between economic and demographic variables and how votes were cast on specific referenda.
The following section will discuss the voting results and their spatial variation, displayed using GIS, from the 1996 and 1997 referenda.  (4) a higher population density; (5) a lower percentage of land under LURC jurisdiction;

Distribution of the Votes
and (6) and are less dependent on the forest products industry for employment than the four counties least supportive of a ban on clearcutting. County which showed the least support for the Clearcutting Ban, the Clearcut Ban won.
By county, support for the Compact ranged from 39.6 to 53.8 percent with the greatest support in the Northern Forest counties of Aroostook, Franklin , and Oxford. and Androscoggin County, and the least support in the three easternmost coastal counties of Waldo. Hancock, and Washington, and the Northern Forest county of Piscataquis (Table   4.3 and Figure 4.2). In terms of how the counties most supportive of the Compact differ from those that are least supportive the Compact by the characteristics (   Like the counties that were most supportive of the clearcut ban, the counties most supportive of the none of the above option differ from those that are least supportive in terms of the characteristics described in the county profiles in Chapter Three, only in a different way. The four counties most supportive of the none of the above option without exeption have (I) a lower percentage of high school and college educated residents; (2) higher unemployment rates; (3) a lower per capita income; (4) a lower population density: and (5)  These results show that counties exhibiting the most support for the clearcutting ban have nearly opposite characteristics from the counties most supportive of the none of the above option and that most and least supportive counties for the Compact do not exhibit distinct differences . These results and their significance will be di scussed in more detail in Chapter Five.

The Runoff Election
According to Maine's Constitution, when there are competing referenda and none receives a majority of the votes, the one receiving the most votes is put on the ballot in the next statewide election. provided that it received more than one third of the vote (Maine Constitution). The votes on the Compact satisfied these requirements and it was slated to be put on the ballot in November 1997 with the option of "yes" or "no" for the Compact. The Compact for Maine's Forests was listed as Question I: Carry-over measure and was worded: "Do you want the Compact for Maine's Forests to become law to promote sustainable forest management practices throughout the state ?" (Maine 1997 and Grenzke, et. al. 1998).
This election, in addition to the carry-over measure, contained measures for three bond issues, a con stitutional amendment and a referendum question dealing with funding for improvements to the Maine Turnpike. Unlike the 1996 referenda vote, which was part of a general election, the 1997 referendum vote was part of a special election.
Specifically the carry-over measure contained the following provi sions: • The Commissioner of Conservation must adopt statewide rules overseeing timber harvesting, subject to subsequent review by the Legislature.
• Requires the establishment of a voluntary audit program.
• Landowners must get a permit from the Commissioner of Conservation before doing any clearcutting. The landowner would have to justify reasons for clearcutting.
• Permitted clearcuts subject to size, proximity to other clearcuts and total area under one ownership restrictions .
• The Commissioner must to adopt rules regarding forest regeneration after a timber harvest.
• Municipalities may adopt timber harvesting ordinances that are more restrictive than the State's rules.
• It would make certain state-owned land off limits to timber harvesting.
• Placed harvest restrictions on land held for less than ten years.

The Role of the Media in the 1997 Election
According to some polls , many voters were confused by the Compact even though it had been available for examination for more than a year. This confusion , which some sources describe as media-induced may have been crucial for the surprise defeat of the Compact (Grenzke, Swope and Carter 1998 , 33 ). Media influence on the 1997 election will be discussed in greater detail in Chapter Fi ve.

Voting Results
The 1997    The areas of highest support for the Compact at the municipal le vel appears to be very "patchy" although concentrated in the southwestern and extreme northern portions of the state. The least support tended to be on or near the borders of unincorporated territories where, as mentioned earlier, people vote in the nearest town in the appropriate house or senate district.

After Defeat
Governor King characterized the opposition to the compact as "the most bizarre coalition in the history of Maine politics, "a blend of "far-end" enviro groups and conservative property rights activists ... " (Greenwire 1998).

Both Jonathan Carter of the Green Party and Mary Adams of Common Sense for
Maine's Forests (a property rights advocacy group) felt that the defeat of the Compact was a victory for them. Carter, thinking that it sent a clear message to the seven largest paper companies that clearcutting will not be tolerated and Adams thinking that it showed that the people in Maine didn't want more regulation (Ellsworth American 1997).

Findings
Initiatives and referenda have long been used for political change and many types of issues have been put to the public for a vote. There exists a long-standing debate over the value of using initiatives and referenda to elicit political change over the traditional lawmaking process. Proponents of initiatives and referenda maintain that they make for a more accountable government and encourage citizen involvement in important issues.
Opponents maintain that many issues are too complex for the average voter to understand and that most voters will selfishly vote with their pocketbooks i.e. what is best for them economically and not society as a whole. Another criticism of initiatives and referenda is that they are too easily influenced by campaign spending, often in the form of biased media campaigns, and that these media campaigns often confuse voters rather than making them more knowledgeable of the issues.
The 1996  Descriptive analysis of the correlation between support and opposition for the three ballot options and the characteristics of counties produced the following results: • Counties most supportive of the clearcut ban were more educated, experienced lower unemployment, had a higher per capita income, a higher population density, a lower IOI I area of land under LURC jurisdiction and were less dependent on the forest products industry for employment than those counties least supportive of the ban.
• Counties most supportive of the Compact were not clearly different from those least supportive in terms of education, unemployment per capita income, population density, area of land under LURC jurisdiction or dependence on the forest products industry for employment.
• Counties most supportive of the none of the above option were less educated, experienced higher unemployment, had a lower per capita income. a lower population density, and were more dependent on the forest products industry for employment than those counties least supportive of the none of the above option.
• Counties most supportive of the clearcutting ban exhibited characteristics that were directly opposite to those most supportive of the none of the above option.
With no one option receiving more than 50% of the vote statewide, a runoff election was required to be held the next year as required by Maine· s Constitution. As in the 1996 election , the various sides used media extensively to try and influence the outcome of the referendum , in some cases confusing voters about the economic and environmental impact of the Compact. As is often the case in off-year elections. voter turnout dropped off substantially, in the range of 40% statewide.
Descriptive analysis of the correlation between support and opposition for the

Methods for Using Inferential Statistics for Hypothesis Testing
Chi square tests are a popular way to test hypotheses , especially in soc ial research, because of their versatility. Chi-squares are especially useful where the \'ariables of interest have more than two categories and where there are more than two samp les. Chisquares enab le the researcher to measure the degree to which the relationship between two variables is random (Healy 1996, 250).
The testing of research questions can be accomplished using a null hypothesis, based on the assumption of data randomness . Expected frequencies are developed based on the randomness assumption and can be compared to the observation frequencies.
In this study an attempt was made to determine quantitatively whether there is a significant correlation between support for the Compact in the 1996 and 1997 elections.

}
Another useful statistical method for the testing of research questions, stepwise multiple regression , was employed to quantitatively analyze the separate effects of each independent variable on the dependent variable and to determine the combined effect of all of the independent variables on the dependent variable (Healy 1996, 438). The research hypotheses are outlined in the introduction and below in Table 5.1.

Research Hypotheses
As mentioned in Chapter One, the purpose of this study is to examine social, economic, and other variables as possible independent correlates to the 1996 and 1997 Maine forestry practices referendum votes. The variables used in this examination and described in more detail in Chapter One, are listed in Table 5.1.

Statistical correlation between the 1996 and 1997 referendum votes
Chi square analysis was performed as shown in Tab le 5.2 to determine if a statistically significant relationship exists between majority support for the Compact in the 1996 three-way vote and support for the Compact in the 1997 Compact runoff election. As mentioned previously, this research hypothesized that counties that voted for the Compact in 1996 would more likely to vote for it again in 1997. In order to make the comparison between the 1996 vote, in which there were three choices (Ban Clearcutting, the Compact, and None of the Above), and the 1997 vote where there were two choices (yes or no for the Compact ). the research combined Ban Clearcutting and None of the Above into the "No" vote category for the 1996 vote.

Findings
The analysis of the relationship of the votes for the Compact at the county level in the two years indicates that the relationship between the variables was not statistically different from what would be expected from a random distribution. However. data at the J county level may not be a sufficiently meaningful level of aggregation for looking at the relationship between the votes in these two years, particularly when using chi squares for analysis . This is because with only 16 counties, most of the possible combinations contain less than the desired minimum of five observations. While this deficiency was corrected for statistically, it is desirable to have a greater degree of certainty that there was no relationship between the votes in the two years.
In order to obtain a better analysis of the relationship between 1996 and 1997 voter support for the Compact. this research analyzed the votes by the 506 municipalities 108 for which voting data was available for both years as shown in Table 5.3 , with the same aggregation as used for the county votes (majority for "Ban Clearcutting" plus "None of the Above"= No). There were only three municipalities that met the latter criterion.

Spatial Analysis
When To get a more detailed look at the influence of forest products indu try employment on the votes. this research examined the ten largest paper manufacturing facilities in Maine and the municipalities in which they are located. The rational behind examining these municipalities is that wood products manufacturing facilities are likely to be their largest employer. As such, these facilities are likely to have a significant impact on the local economy, both directly through payroll and purchase of supplies and services, and indirectly through secondary spending by businesses and individuals . The voting patterns in these municipalities should represent most clearly the economic influence that the presence of the forest products industry has on voter preferences.

}
Of the largest six municipalities in terms of employment in single paper products making facilities , all of them were areas of core Compact support. In fact. of the ten largest. seven are in municipalities that voted for the Compact both times.
Many of these paper facility municipalities stand out on the map because they are surrounded by municipalities where the majority voted against the Compact in the 1997 Unlike the region of core Compact support along the coast. the large block of core opposition in the center of the state represents only a smal I portion of the total vote statewide but may represent the area where property rights gro ups are the strongest or at least where sentiment against government regulation is the strongest.

Statistical Correlation Between Referenda Votes and Social and Demographic Variables
In order to obtain a more rigorous analysis of the relationship between the independent variables in the research hypotheses and the voting results in the 1996 and 1997 referenda, a quantitative approach was extended to include multiple regression.

I Chi Square Analysis
The difficulty encountered with the use of chi square analysis is that when crosstabulating votes for 16 counties with an ordinal or interval level ranking of variables, is that resulting combinations often produced cells that had an expected frequency of less than 5. "When sample size is small , one can no longer assume that the sampling distribution of all possible test statistics is accurately described by the chi square distribution. In the case of the chi square test, a small sample is defined as one where a high percentage of cells have expected frequencies of 5 or less .... In the case of 2x2 tables, the value of x 2 (obtained) can be adjusted by applying Yates correlation for continuity ... " (Healy 1996, 261 ). For tables larger than 2x2, which is what is obtained from the above described analysis. there is no formula for correcting the obtained chi square value for possibilities with less than five observations. Categories of variables can be combined to avoid this problem if there is a clear theoretical justification (Healy 1996. 261 ). I determined that combining the variables in thi s way would not result in meaningful information to answer the research question s posed.

Multiple Regression Analysis
In order to address this problem, this study used the data for the dependent and independent variables in their interval form. For example, instead of aggregating unemployment into a two category ranking (i.e., counties with above mean state percent unemployment and counties with below mean state percent unemployment), this research used the actual value for each county, which makes the data interval level. For the 117 J referendum votes, this research used actual percentage of the vote for a particular ballot choice. This allows the use of a stepwise multiple regression analysis to determine collectively which independent variables make a significant contribution to the dependent variab les (the 1996 and 1997 referendum votes). It also allows a determination of how much of the variation in the dependent variable can be explained by significant independent contributors.
:\ multiple regression analysis was done for four dependent variables: At the 95% confidence level, all of the independent variables were removed from the regression equation except for forest products industry employment. Results show an inverse relationship between percent voting to ban clearcutting and forest products industry employment, that is counties with a higher level of forest products industry employment had a significantly lower percentage of votes to ban clearcutting. In fact 78.1 % of the variation in the 2a vote at the county level can be attributed to the level of I forest products industry employment (Table 5.5).
As discussed in Chapter Three, due to the policy of the U.S. Census Bureau to give estimates in many cases to protect the privacy of individual employers, many of the aggregated figures for county forest products industry employment contained at least one component figure that was an estimate.
However, in most instances the uncertainty in the estimate does not alter the employment percent significantly. For example, for Cumberland County, the total of the 120 three forest products industry employment categories was 1.9% of total employment using the halfway method of determining a value for an estimated range for category 2600; paper and allied products. Assuming that actual employment was at the very bottom of the range for this category, the result would be total forest products industry

Agreement Between Results of Descriptive and Quantitative Analysis
The question that arises when comparing the results of descriptive and quantitative analysis is why didn ' t all of the variables that appeared to be correlated to voting results in the descriptive analysis appear as statistically significant in the quantitative analysis? The probable reason for this is that a 95% confidence level was used as the threshold for statistical significance in the quantitative analysis. With only 16 counties. the strength of the correlation must be very strong to pass this threshold. The quantitative analysis did not contradict any of the descriptive analysis but rather puts the correlation found to be significant on more solid footing.

Findings
The original reason for aggregating the data at the county level was because of the availability of forest products industry employment and payroll , which were central to my research questions described previously. In general, because there are only sixteen counties in the State of Maine. this is a rather broad analysis of the data although geographical analysis at the municipal level partly compensates for thi s deficiency.
To summarize the previous section, I determined the following findings: • Municipalities that supported the Compact in 1996 were significantly likel y to support it again in 1997.
• Geographical analysis shows that the main areas of Compact support were located along the densely populated southern coast in York, Cumberland, Sagadahoc and Lincoln Counties, in municipalities hosting large paper making facilities, and in Aroostook county near Caribou.

I
• Higher forest products industry employment was associated with opposition to the 1996 Ban Clearcutting option.
• None of the independent variables showed a significant correlation to the 1996 Compact vote.
• Lower population density was associated with support for the 1996 None of the Above option.
• Higher per capita income was associated with support for the Compact in 1997.

Discussion
The The hypothesis that higher population density correlates with support for the clearcut ban in 1996 and support for the Compact in 1997 at the cou nty le\'el and conversely that lower population density correlates with support for the none of the above option are supported by the results of descriptive analysis . Lower population density also showed a significant correlation with support for the 1996 none of the above option at the 95 % confidence level in the quantitative analysis. This variable served as an indication of the "ruralness" of the individual counties. The concept of this variable is that there is a land use ethic and lifestyle associated with rural living which includes a strong sense of individualism. self reliance, and accompanying opposition to government intervention into private land use decisions such as a ban on clearcutting. In this conceptualization, 126 I voters in more rural areas will vote more on the basis of opposition to new government regulation than voters from more urban areas.
Greater area under LURC jurisdiction only appeared to be correlated with opposition to the clearcut ban in the 1996 descriptive analysis and did not appear significant for both years in the quantitative analysis. For this variable the hypothe is was that voters in counties with greater area under LURC jurisdiction will be more likely to vote against the Compact in 1997. The concept for this variable was that voters in and near areas of extensive LURC jurisdiction will tend to vote against additional state regulation, seei ng land use controls in general as a local issue. and a ban on clearcutting as a threat to local economic health.
The variable and its relationship to the 1996 and 1997 votes lie at the core of my thesis. which is that in the 1996 and 1997 referenda, voter preference was largely a function of financial considerations, that is voters in counties that are more economically dependent on the forest products industry were significantly more likely to vote against any regulation that might hinder the industry and cause an accompanying decline in employment. Voters in areas of high industry dependence will tend to "vote with their pocketbooks" and reject regulation that might potentially lessen the individual' s economic prospects. Accordingly the hypotheses were that voters in counties with higher forest products industry employment were more likely to vote against a ban on clearcutting and the Compact and for none of the above options in 1996 and that voters in counties with greater forest products industry employment were more likely to vote against the Compact in 1997. These research hypotheses are supported by descriptive analysis at the county level which showed that the counties most supportive of the 127 I clearcut ban in 1996 had greater dependent on the forest products industry for employment than counties least supportive of the ban and that counties most supportive of the none of the above option were more dependent on the forest products industry for employment than those counties least supportive of the none of the above option. Higher forest products industry employment was found to be significantly correlated with opposition to the 1996 Ban Clearcutting option in the quantitative analysis.

The Role of the Media in the 1996 and 1997 Elections
The media certainly played a large role in the 1996 referendum. This role is difficult to quantitatively assess with available data other than the rough measure of total campaign spending and anecdotal accounts. One source placed spending by Jonathan Carter's campaign at $870,000 (Greenwire 11/5/98). The Associated Press stated that King, the paper industry and some environmental groups raised in excess of $5.5 million dollars to support the compact (Associated Press 1996). Another source placed this figure at 56 million (Grenzke, Swope and Carter 1998, 33 Council , allowed Compact proponents to feature their endorsement in mailings and television commercials (Grenzke et. al. 1998 , 2). They attributed what they claimed as their "v ictory'' to a geographically targeted message that carefully avoided areas where a pro-environmental message might cause people to vote for the Compact (Grenzke et. al. 1998, 1 ). According to the Grenzke. et. al. account, the forest products industry claimed that the Compact would strictly limit clearcutting but certain anti-Compact forces claimed that " ... the Compact was 27 pages of technical language and generalities that would have allowed clearcutting to triple ... " over the next ten years (Grenzke et. al. 1998, 2). Polling prior to the 1997 Compact vote indicated "confusion" over Compact from conflicting advertisements in the 1996 campaign. where the majority believed that the Compact would reduce clearcutting thus making it the pro-environment choice. Thi s perception by voters was dam ag in g to the anti-Compact fo rces in the 1997 campaign who fe lt that the Compact was the anti-environment choice. With thi s in mind, their strategy was to : 1) Reach a pro-environment base with the "true" information about the Compact (i.e. that it would actually increase clearcutting).
2) Avoid reaching property rights voters who " .. . were opposed to any kind of government regulation and therefore were already against the Compact" (Grenzke, Swope and Carter, 1998 p.3).
129 J " .. . polling data demonstrated that a significant proportion of property rights voters would move from opposing the Compact to supporting the Compact if they heard our message about the Compact allowing more clearcutting" (Grenzke et. al. 1998. 3). "Fortunately, our polling showed that our potential block of environmental voters were mostly in the Portland medial market and in selected coastal zip codes in the Bangor media market.
Property-rights and pro-paper corporation voters were most likely in the less populated northern counties of the Bangor media market. This geographic separation allowed us to target our television , radio. mail and Get-Out-The-Vote (GOT) efforts to voters we needed to reach , and to avoid communicating an environmental message which could encourage property rights voters to support the Compact" (Grenzke et. al. 1998 , 3). The spatial analysis, which shows the Bangor area as a cluster of Compact opposition generally supports this account.
The anti-Compact forces portrayed the Compact as a trick question in their communications to selected markets containing " pro-environment" voters. They claimed that this is what made the difference in the narrow defeat of the Compact by only a 17 ,000 vote margin (Grenzke et. al. 1998 , 3). As mentioned in Chapter Four it is easier to defeat a referendum than it is to win one, especially if voters are unsure about the benefits and uneasy about the risks. and that was the strategy that anti-Compact forces adopted.
The previous account attributes the outcome to the ironic (and fortunate for anti-Compact forces) confluence of property rights group opposition to any regulation and at least some voters being persuaded that the Compact was not the best option for furthering environmental protection. Thi s account also suggests that there is a block of voters 130 / supportive of the message of grassroots groups who felt that the Compact was inadequate and that a complete clearcut ban was preferable in unincorporated territories.

Conclusion
The vote was largely about balancing environmental concerns with concerns about economic well being. People in more urban and more affluent areas have greater ability to place more importance on the environmental side of the equation and tend to vote accordingly. This brings up a broader and historically deep rooted debate over who gets to decide how land is used. There also appears to be a general property rights sentiment, often appearing in sparsely populated areas, against additional government regulation that seems to be important enough to influence how people vote.