Thoreau, Muir, and Jane Doe: Different Types of Private Forest Owners Need Different Kinds of Forest Management Andrew O. Finley, Department of Forest Resources, University of Minnesota, St. Paul, MN 55108; and David B. Kittredge Jr., Department of Natural Resources Conservation, University of Massachusetts, Amherst, MA 01003, and Harvard Forest, Harvard Univer- sity, Petersham, MA 01366. ABSTRACT: We present a three-phase segmentation analysis designed to highlight the heterogeneity of forest ownership values and attitudes toward government control, privacy, and environmental protection held by a sample of Massachusetts private forest owners. This case study explores private forest owner characteristics that are associated with enrollment into Massachusetts' Chapter 61 current-use forest property tax program, which requires a professionally prepared lO-year forest management plan. We suggest the key to increasing landowner participation in forest management programs is to (1) recognize this heterogeneity of the target population, and (2) tailor the program to meet segment specific needs and desires. North. J. Appl. For. 23(1):27-34. Key Words: private forest owners, segmentation analysis, attitudes, forest management. The northeastern United States, where private forest dom- inates the landscape, relies heavily on an intact, healthy, and resilient forest ecosystem. These private forests produce numerous social benefits including clean water and air, biodiversity, lumber/wood fiber, wildlife for consumptive and nonconsumptive uses, recreation, and a scenic backdrop for a rural tourism industry. Recognizing the value of pri- vate forests, several local, state, and federal agencies have designed programs to encourage private forest (PF) owners to practice forest stewardship. Forest stewardship can be defined as the judicious management of forest resources to ensure their sustainable health and productivity for future generations. Despite the development of these programs, it is increas- ingly difficult to ensure stewardship practices across the private forest landscape. An increasing PF owner popula- tion challenges efforts to secure a sustainable and ecologi- cally sound forest landscape because of parcelization, frag- mentation, and land use change (Birch 1996; Sampson and DeCoster 2000). The effectiveness of forest management programs is challenged in this developing landscape, since they provide few economically compelling alternatives to landowners faced with the option to sell their property for subdivision and eventual land use change. NOTE: Andrew O. Finley can be reached at (612) 624-1714; afinley@gis.umn.edu. Copyright @ 2006 by the Society of Amer- ican Foresters. Reprinted from Northern Journal of Applied Forestry, Vol. 23, No. We present findings of a recent PF owner study that uncovers patterns in values associated with forestland own- ership. Based on unique value orientations identified in the PF owner sample, we propose that the state's current-use property tax program (Chapter 61) could enjoy greater success if it better reflected PF owners' values. Case Study Although it is generally perceived as a center of metro- politan development, Massachusetts is roughly 62 percent forested. Seventy-eight percent of this forest landscape is controlled by 235,000 PF owners (Alerich 2000). The for- ested landscape controlled by thousands of private individ- uals provides a myriad of market-based commodities and other invaluable benefits to the state's 6.3 million citizens. Considering the pressure from Massachusetts' rapidly ex- panding urban populations, the state has a recognized the need to protect forests for their social and ecological ser- vices. To this end, the state legislature passed Chapter 61 in 1989, which provides a 95 percent reduction in the assessed value of the property resulting in a corresponding savings in property taxes. The aim of Chapter 61 is to simultaneously encourage forest stewardship and keep the private forest in active timber management. Enrollment in the Chapter 61 program requires a professionally prepared lO-year forest management plan that identifies landowner forestland ob- jectives (e.g., improving wildlife habitat, property esthetics, 27NJAF 23(1) 2006 Not for further reproduction.1, March 2006. or timber resource) and active management for those for- estland objectives. An explicit requirement of Chapter 61 is that enrolled forestland be managed to produce timber. Local towns receive an 8% severance tax from the land- owner when timber is sold. Also, once enrolled, the property cannot be withdrawn without paying a foregone back-tax penalty. Chapter 61 does not require the landowner to open their property to public access. Despite the liberal tax re- duction benefit, it is estimated that only 17% of eligible acres and 19% of eligible owners participate (unpublished data). Methods The Survey The data used in this case study were drawn from a survey distributed in the spring of 2000. The original study was designed to measure PF owners' attitudes toward eco- system management activities implemented on their forest- land (White 2001). In addition to questions concerning landownership, the survey elicited responses from landown- ers on attitudes toward a variety of themes or issues about the environment, government, and ownership decisions. The survey was mailed to 1,500 PF owners in 20 randomly selected towns in Massachusetts. The development and dis- tribution of the survey followed Dillman's (2000) Total Design Method. In an attempt to exclude those PF owners in predominately urban centers, all Metropolitan Statistical Areas and Consolidated Metropolitan Statistical Areas as defined by the United States Census were excluded. Cape Cod and all towns east of Interstate 495 (which rings the greater Boston metropolitan area) were also excluded be- cause PF owners in this area experience different environ- mental and developmental pressures related to the coastal environment. Approximately 75 PF owners in each of the sampled towns were randomly selected to receive the mail- back survey. The survey had a response rate of 47.2 percent (n = 708), and 139 observations were dropped from the analysis due to missing values. Segmentation Approach The three-phase analytical strategy defines and describes PF owner segments based on measured values and attitudes. Table 1. Rotated PC loadings, total variance explained by retained PCs, and Cronbach's Alpha for each PC. Items Reasons for owning forestland . "' ;.,., ., " ,', Income from timber n' '" 0,05 -0.07. it 0.Q7 0.57 0.61 Income from agriculture 0.04 0.13" 0.14 0.58 To pass on to my children 0.12 0.09 0.72 0.14 0.80 To preserve family and tradition 0.23 0.22 0.73 0.17 Personal recreation 0.27 0.64 0.13 0.03 0.79 6 As a place to live 0.18 0.65 0.13 0.05 7 To enjoy the scenery 0.32 0.74 0.11 -0.01 9 To protect land from development 0.66 0.27 0.26 0.08 0.82 10 To provide wildlife habitat 0.71 0.28 0.10 0.04 II To protect the environment 0.72 0.25 0.12 0.05 12 To have privacy Retained as unique variable 13 To leave land unmanaged, letting Retained as unique variable nature take its course Variance explained 1.83 1.74 1.21 0.84 1 2 3 4 5 28 NJAF 23(1) 2006 Three distinct segments of respondents were defined, based on their responses to 13 Likert scale survey items. Using a 5-point Likert scale, each item measured reasons for owning forestland (with potential responses ranging from "not at all important" to "extremely important"). First, principal com- ponents analysis (PCA) was used to reduce the dimension- ality of the survey item data set and render a parsimonious set of composite variables. Second, a cluster analysis (CA) extracted three distinct respondent segments based on the composite variables and retained original survey items. Third, a multiple discriminant analysis (MDA) identified those clustering variables that best exemplify segment dif- ferences. In addition to these three multivariate analyses, univariate tests on supplementary survey items were used to further profile the derived PF owner segments. Analysis Phase 1: Data Reduction Using Principal Components Analysis In this case study, 13 items were used from the original survey (Table I). These items measure respondents' reasons for forestland ownership. High simple correlations among several of these items warranted a data reduction procedure (Stevens 1986). PeA was performed on the 569 responses and 13 survey items. Table I displays the PC loadings and variance explained by the first four PCs. A PC loading represents the univariate correlation between the survey item and PC. These loadings are used to define and name each PC. Hair et a1. (1998) suggest that in samples greater than 50, absolute PC loadings greater than 0.50 indicate strong variable to PC association. This criterion was used here for defining each PC. However, Item 12 (To have privacy) and Item 13 (To leave land unmanaged, letting nature take its course) were not incorporated into a PC and left to stand alone in the analysis. Emerging research sug- gests different PF owners hold unique definitions for the word privacy (Rickenbach et a1. 1998); therefore, we were interested in observing the behavior of Item 12 as a unique variable. Item 13 did not adequately load on any of the four PCs, and was therefore also left as a unique variable. Rotated principal component loadings PC 1 PC2 PC3 PC4 Cronbach's Alpha Based on associated item themes, we assign the follow- ing names to the different PCs: PC 1 environmental protec- tion; PC 2 contemplative enjoyment; PC 3 family legacy; PC 4 utilitarian consumptive uses. As a measure of scale reli- ability, Cronbach's Alpha was calculated for each PC (Cronbach 1951). Because PC 4 does not meet Nunnally's (1978) suggested Cronbach Alpha minimum of 0.70, cau- tion will be used when interpreting results related to the utilitarian consumptive uses PC. Using PCA, the original 13 survey items were effectively reduced to four composite variables, and two original items (Item 12 and Item 13). PC scores serve as data observations for the new composite variables. To share a common scale with derived composite variables, Items 12 and 13 were standardized (mean = 0, variance = 1). In the second phase of the analysis, CA defines distinct respondent segments based on this reduced set of variables. Phase 2: Segment Formation Using Cluster Analysis Previous PF owner studies have effectively described commonly held reasons for landownership and attitudes concerning forest resource use (e.g., Birch 1996, Ricken- bach et al. 1998). However, they have not described the heterogeneity of these values and attitudes among survey respondents. CA is a multivariate technique that can orga- nize survey respondents into discrete segments, such that within-segment similarity is maximized and among-seg- ment similarity is minimized according to respondentS' scores on survey items. The CA technique used in this analysis produces statistically significant and discrete seg- ments of PF owners who exhibit common reasons for landownership. The k-means clustering algorithm was used to assign respondents to exclusive segments based on their response to the six clustering variables derived from the PCA. To arrive at an appropriate number of respondent segments (i.e., clusters), two-, three-, and four-cluster solutions were explored. Results from the final phase of this analysis sug- gested that the three-cluster solution was most easily inter- preted, and therefore, will serve as the basis for further analysis. Table 2. Summary of discriminant analysis and variables' contribution to segment separation. Standardized coefficients Discriminant loadings Univariate F Partial F Discriminant variables OF I . DF 2 OF 1 OF 2 Ratio Frob. Ratio Prob. PC 1 "", 0.269 -0379 0.621'LO.212 101.99 <0.01 22.76 ,<0.01 Environmental protection PC 2 0.173 0.677 0.628 0.46 II 144.31 Contemplative enjoyment -0.099 0.536 0.086 0.400 25.44 -0.007 0.296 -0.016 0.392 23.02 U87 0.227 0.953 0.212 361.72 0.334 -0.923 0.485 -0.646 147.96 Phase 3: Describing Segment Differences Using Discriminat Analysis CA defined three segments of respondents based on their scores. on six clustering variables. The third phase of the analysis describes the unique signature or profile of these respondent segments. In this analysis, MDA was used to determine which of the six clustering variables were most influential in assigning respondents to segments. This infor- mation is important because it highlights those landowner- ship values common and unique among segments. Both discriminant functions are highly significant (P < 0.01) based on an approximate F-value (see Hair et at. 1998 for approximate F-value calculations). Table 2 lists each of the discriminating variables and associated standardized canonical coefficients. To aid in interpretation, the MDA results are illustrated with a biplot (Figure 1), which depicts discriminating vari- ables as vectors and segment centroids as point symbols. The length of the vector is proportional to its discriminating power; longer vectors have greater discriminating power. The angle of the vector relates this power to the canonical axis; stated differently, the more aligned a vector is with the canonical axis the more the discriminating power is associ- ated with that axis. Furthermore, the direction of the vector mirrors the original low to high Likert scale; the variable or PC vector arrow points toward segments that scored high, and away from segments that scored low (Hair et al. 1987). Examining the discriminant loadings (Table 2) and the biplot (Figure 1), we determined that the first function (i.e., the x axis) describes a gradient dominated by the value placed in privacy. This gradient generally separates segment three from segments one and two. The second function (i.e., the y axis) describes the difference between segment one and two. Specifically, this gradient is dominated by the recognition of ownership for contemplative enjoyment ver- sus the value in ownership to preserve nature. We recognize that these value scales are not mutually exclusive (rather complimentary) and probably in most cases exist together in a landowner's set of ownership values. Our analysis sug- gests, however, that one set of values presides over the other <0.01 <0.01 <0.01 <0.01 <0.01 43.26 38.51 11.05 125.65 111.13 <0.01 <0.01 <0.01 <0.01 <0.01 NJAF 23(1) 2006 29 ... Segment 3 (Jane doe) Figure 1. deviations. Biplot of important discriminant variables (vectors) and segment centroids (~ between the segments. Specifically, landowners in segment 2 hold strong values of environmental protection, whereas landowners in segment I place higher priority on privacy and nonconsumptive uses such as recreation and scenery. Although statistically significant, the composite vari- ables offamily legacy and utilitarian uses exhibit much less influence in distinguishing between landowner segments, as indicated by the univariate F scores (Table 2). Since these are not strongly distinguishing factors, the implication is that these concepts are held more or less commonly among segments and for this reason their vectors were not included in the biplot. Segments Our analysis has identified three statistically unique seg- ments of the sampled landowner population in Massachu- setts. These segments are basis on attitudes toward environ- mental protection, privacy, and appreciative values of forests. , Henry David Thoreau is the figure we've chosen to exemplify landowners in this first segment. Like Tho- reau, these landowners highly value privacy and the Analysis of variance for survey items that measure segment specific ownership and demographic charac-Table 3. teristics. ---------- Item Henry David Thoreau (Segment 1) John Muir (Segment 2) Jane Doe (Segment 3) dJ. Frob. Forest area (acres) Respondent age (years) n GLM and Tukey's Studentized Range Test. Superscript letters attached to variable's mean denote mean separation a = 0.05, variable's standard deviation shown in parentheses. 30 NJAF 23(1) 2006 Environmental protection (PC 1) Preserve nature (Item 13) Segment 2. (John Muir) . Axes measured in units ofsymbols) standard contemplative values or benefits of forest, such as scenery, personal recreation, and a pleasing residential setting. Although their land may not have a pond, these landowners view their property as providing important "Walden"-like qualities or benefits. Like Thoreau, however, these owners are not necessarily opposed to utilizing the forest for wood products (Foster 1999), but their highest priorities are more appreciative or non- consumptive benefits. Approximately 67% of survey respondents are members of the Thoreau segment, and their average age is 58 years (Table 3). The average property size of Henry David Thoreau is 70 acres. Understandably, since they place a high priority on appreciative values such as scenery, 83% of Thoreau owners reside on their property, and only 17% live elsewhere (Table 4). In terms of future development, the vast majority of Thoreau owners (85%) have no such intentions for the next ten years (Table 5). John Muir is the historical figure we use to exemplify the second segment within the sampled landowners. Like Muir, these landowners place the highest priority . APF owner segments 69.55ab (105.67) 54.11a (60.73) 1O2.00b (167.54) 58.12a (12.64) 59.60ab (13.12) 63.46b (11.66) 2; 5660.02 2; 5660.01 Table 4. Cross-tabulation by segment for the item that measures absentee ownership. PF owner segments Frequency column Henry David Thoreau ,.' John Muir Jane Doe percent (Segment 1) (Segment 2) ;. (Segment 3) .,;; , Total Absentee 62 16.67 2923.39 3667.92 127 Non-absentee 310 83.33 9576.61 1732.08 422 Full table X- = 68.55; d.f. = 2; P < 0.01; missing values = 20. Table 5. Cross-tabulation by segment for the item that measures plans to develop property or sell for development within the next 10 years. Table 5. within the next 10 years. Henry David Thoreau Frequency column percent (Segment 1) '" No plans to development or sale 292 84.64 ' ..-v' Plans to develop or sell for development 53 15.36 Frequency column percent 'ull table K = 25.40; d.£. = 2; P < 0.01; missing values = 47. on nature, environmental quality, and protection. Im- portantly, they are in strong agreement with the Likert scale statement that they should "leave land unman- aged, letting nature take its course." An estimated 23% of respondents are members of the Muir segment, and their average age is 60 years. The mean size of Muir properties is 54 acres. Like the Thoreau, a high pro- portion of Muir owners live on their property (77%; Table 4) and a slightly higher percentage (87.5%) hav~ no plans to develop in the next 10 years (Table 5). Jane Doe is the third persona that we have identified in this segmentation analysis. We know the least about landowners in this segment (hence the anonymous name), as they appear to be indifferent to the factors that were otherwise found to be strong identifying characteristics. Unlike Thoreau and Muir, Jane Doe places little important emphasis on either environmen- tal protection or privacy and other contemplative values of forest ownership. This relatively unknown, anony- mous Jane Doe segment represents only 10% of re- spondents. Jane Doe's average age is 63 years (Table 3), which is significantly older than the Thoreauvians. Doe's average parcel size is 102 ac, which is also significantly larger than Muir's woods. Unlike Thoreau and Muir, most Jane Doe owners reside elsewhere, and not on their property (68%), which could in part ex- plain their apparent indifference to environ~ental pro- tection, privacy, and other contemplative values (Table 4). Also in contrast to Muir and Thoreau, 43% of Jane Doe owners do indeed have plans to sell or develop in the next 10 years (Table 5). This may be related in part . Table 6. Cross-tabulation by segment for the item that measures enrollment in Chapter 61. PF owner segments )}£f!~:hI'~i ';,1, . ,i,. ," , ". . Henry David Thoreau John Muir :'/.