FAMILY FORESTS IN MICHIGAN:  

 

MICHIGAN’S QUALIFIED FOREST PROGRAM AS A CASE STUDY 

 

By 

 

Benjamin Michael Schram 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

A THESIS 

 

Submitted to 

Michigan State University 

in partial fulfillment of the requirements 

for the degree of 

 

Forestry – Master of Science 

 

2019 

 

 i

ABSTRACT 

 

MICHIGAN’S QUALIFIED FOREST PROGRAM AS A CASE STUDY 

FAMILY FORESTS IN MICHIGAN:  

 

By 

 

Benjamin Michael Schram 

 

Because family forest owners own the majority of forestland in the country, many states attempt 

to incentivize them to be active managers of their woodlands. Property tax incentive programs provide 

family forest owners with certain tax breaks, credits, or exemptions in exchange for developing a written 

forest management plan. Michigan’s Qualified Forest Program (QFP) is examined here because it 

provides context for analyzing landowner behavior. High non-compliance penalties provide some 

assurance that forest management practices will be completed. The ability of parcel characteristics (stand 

acres, forest type, condition (size and density), and region) to predict forest management practices at the 

stand level is evaluated. Certain variable categories are significant in predicting forest management 

practices on QFP enrolled land. Forestry incentive program managers, forestry practitioners, and 

policymakers could use these results to predict family forest owner management decisions and focus 

conservation efforts. 

 

 

 

ii 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

Copyright by 
BENJAMIN MICHAEL SCHRAM 
2019 

 

iii

ACKNOWLEDGEMENTS 

 
 
 

The path toward completing this text was filled with triumphs, tribulations, enthusiasm, and 

deliberate contemplation. It is important that I first thank my advisor, mentor, and friend, Dr. Karen 

Potter-Witter. Her patience, guidance, and expertise served as a staff that supported me through the 

entirety of this effort. 

Dr. Emily Huff played an essential role in data analysis and putting structure to the prose. Her 

understanding counsel made progress possible. I cannot thank her enough. I’d also like to thank the rest of 

my committee members, Dr. Andrew Finley and Dr. Robert Richardson. 

The support of my wife, Chelsee Schram, was unprecedented. She was able to provide the right 

proportion of motivation and commiseration throughout this endeavor. I thank my parents, Michael and 

Debra Schram, for being an enthusiastic duo of cheerleaders; the kind that perpetually inquire about 

progress. 

Finally, I thank the cast of conservationists at the Michigan Departments of Agriculture and Rural 

Development, Natural Resources, and Environmental Quality; specifically Stephen Shine, John Switzer, 

and Patricia Hines. These individuals provided the initial spark to begin a graduate program, as well as 

carried the torch to see the thing to completion. 

 

 

 

iv 

TABLE OF CONTENTS 

 
 
 

LIST OF TABLES ....................................................................................................................................... vi 
 
LIST OF FIGURES ...................................................................................................................................... x 
 
Introduction ................................................................................................................................................... 1 
Family forest ownership .................................................................................................................. 1 
Michigan real property taxation ....................................................................................................... 3 

 
Background ................................................................................................................................................... 5 
Forest ownership .............................................................................................................................. 5 
Issues facing family forest owners ................................................................................................... 8 
Incentive programs........................................................................................................................... 9 
Landowner behavior ...................................................................................................................... 12 
Implications ................................................................................................................................... 13 

 
Literature Review ........................................................................................................................................ 14 
 
What we know about family forest owners ................................................................................... 14 
 
Program evaluation ........................................................................................................................ 17 
 
Methods ...................................................................................................................................................... 22 
 
The Qualified Forest Program ....................................................................................................... 22 
Why is the Qualified Forest Program data useful for measuring behavior? .................................. 23 
 
Data collection ............................................................................................................................... 24 
 
 
Data analysis .................................................................................................................................. 25 
 
Results  ....................................................................................................................................................... 27 
Program summary results .............................................................................................................. 27 
 
 
Analysis ......................................................................................................................................... 36 
 
Discussion ................................................................................................................................................... 69 
Program summary results .............................................................................................................. 69 
 
 
Forest practices .............................................................................................................................. 71 
 
Conclusion .................................................................................................................................................. 75 
 
APPENDIX ................................................................................................................................................. 79 
 
REFERENCES ......................................................................................................................................... 132 
 

 

 

v

LIST OF TABLES 

 
 
 

Table 1: Forest condition as a percentage of stands within each forest type group. ................................... 34 
 
Table 2: Multinomial logistic regression parameter estimates for significant results. Stand acres and stand 
type as independent variables (showing results from artificial regeneration). N=20,915 unique forest 
stands enrolled in the Qualified Forest Program. ........................................................................................ 42 
 
Table 3: Multinomial logistic regression parameter estimates for significant results. Stand acres and stand 
type as independent variables (showing results from clearcut). N=20,915 unique forest stands enrolled in 
the Qualified Forest Program. ..................................................................................................................... 43 
 
Table 4: Multinomial logistic regression parameter estimates for significant results. Stand acres and stand 
type as independent variables (showing results from salvage treatment). N=20,915 unique forest stands 
enrolled in the Qualified Forest Program. ................................................................................................... 44 
 
Table 5: Multinomial logistic regression parameter estimates for significant results. Stand acres and stand 
type as independent variables (showing results from seed tree harvest). N=20,915 unique forest stands 
enrolled in the Qualified Forest Program. ................................................................................................... 45 
 
Table 6: Multinomial logistic regression parameter estimates for significant results. Stand acres and stand 
type as independent variables (showing results from shelterwood harvest). N=20,915 unique forest stands 
enrolled in the Qualified Forest Program. ................................................................................................... 46 
 
Table 7: Multinomial logistic regression parameter estimates for significant results. Stand acres and stand 
type as independent variables (showing results from selection harvest). N=20,915 unique forest stands 
enrolled in the Qualified Forest Program. ................................................................................................... 47 
 
Table 8: Multinomial logistic regression parameter estimates for significant results. Stand acres and stand 
type as independent variables (showing results from thinning). N=20,915 unique forest stands enrolled in 
the Qualified Forest Program. ..................................................................................................................... 48 
 
Table 9: Multinomial logistic regression parameter estimates for significant results. Stand acres and stand 
type as independent variables (showing results from timber stand improvement). N=20,915 unique forest 
stands enrolled in the Qualified Forest Program. ........................................................................................ 49 
 
Table 10: Multinomial logistic regression parameter estimates for significant results. Geographic region 
of Michigan as independent variable. N=20,915 unique forest stands enrolled in the Qualified Forest 
Program. ...................................................................................................................................................... 50 
 
Table 11: Multinomial logistic regression parameter estimates for significant results. Geographic region 
of Michigan as independent variable. N=20,915 unique forest stands enrolled in the Qualified Forest 
Program. ...................................................................................................................................................... 51 
 
Table 12: Multinomial logistic regression parameter estimates for significant results. Geographic region 
of Michigan as independent variable. N=20,915 unique forest stands enrolled in the Qualified Forest 
Program. ...................................................................................................................................................... 52 
 

 

vi 

Table 13: Multinomial logistic regression parameter estimates for significant results. Stand condition as 
independent variable. Eastern Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified 
Forest Program. ........................................................................................................................................... 53 
 
Table 14: Multinomial logistic regression parameter estimates for significant results. Stand condition as 
independent variable. Eastern Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified 
Forest Program. ........................................................................................................................................... 54 
 
Table 15: Multinomial logistic regression parameter estimates for significant results. Stand condition as 
independent variable. Eastern Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified 
Forest Program. ........................................................................................................................................... 55 
 
Table 16: Multinomial logistic regression parameter estimates for significant results. Stand condition as 
independent variable. Eastern Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified 
Forest Program. ........................................................................................................................................... 56 
 
Table 17: Multinomial logistic regression parameter estimates for significant results. Stand condition as 
independent variable. Northern Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified 
Forest Program. ........................................................................................................................................... 57 
 
Table 18: Multinomial logistic regression parameter estimates for significant results. Stand condition as 
independent variable. Northern Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified 
Forest Program. ........................................................................................................................................... 58 
 
Table 19: Multinomial logistic regression parameter estimates for significant results. Stand condition as 
independent variable. Northern Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified 
Forest Program. ........................................................................................................................................... 59 
 
Table 20: Multinomial logistic regression parameter estimates for significant results. Stand condition as 
independent variable. Northern Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified 
Forest Program. ........................................................................................................................................... 60 
 
Table 21: Multinomial logistic regression parameter estimates for significant results. Stand condition as 
independent variable. Southern Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified 
Forest Program. ........................................................................................................................................... 61 
 
Table 22: Multinomial logistic regression parameter estimates for significant results. Stand condition as 
independent variable. Southern Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified 
Forest Program. ........................................................................................................................................... 62 
 
Table 23: Multinomial logistic regression parameter estimates for significant results. Stand condition as 
independent variable. Southern Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified 
Forest Program. ........................................................................................................................................... 63 
 
Table 24: Multinomial logistic regression parameter estimates for significant results. Stand condition as 
independent variable. Southern Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified 
Forest Program. ........................................................................................................................................... 64 
 
Table 25: Multinomial logistic regression parameter estimates for significant results. Stand condition as 
independent variable. Western Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified 
Forest Program. ........................................................................................................................................... 65 

 

vii

Table 26: Multinomial logistic regression parameter estimates for significant results. Stand condition as 
independent variable. Western Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified 
Forest Program. ........................................................................................................................................... 66 
 
Table 27: Multinomial logistic regression parameter estimates for significant results. Stand condition as 
independent variable. Western Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified 
Forest Program. ........................................................................................................................................... 67 
 
Table 28: Multinomial logistic regression parameter estimates for significant results. Stand condition as 
independent variable. Western Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified 
Forest Program. ........................................................................................................................................... 68 
 
Table A.01: Multinomial logistic regression parameter estimates. Stand acres and stand type as 
independent variables (showing results from artificial regeneration). N=20,915 unique forest stands 
enrolled in the Qualified Forest Program. ................................................................................................... 80 
 
Table A.02: Multinomial logistic regression parameter estimates. Stand acres and stand type as 
independent variables (showing results from clearcut). N=20,915 unique forest stands enrolled in the 
Qualified Forest Program. ........................................................................................................................... 82 
 
Table A.03: Multinomial logistic regression parameter estimates. Stand acres and stand type as 
independent variables (showing results from salvage treatment). N=20,915 unique forest stands enrolled 
in the Qualified Forest Program. ................................................................................................................. 84 
 
Table A.04: Multinomial logistic regression parameter estimates. Stand acres and stand type as 
independent variables (showing results from seed tree harvest). N=20,915 unique forest stands enrolled in 
the Qualified Forest Program. ..................................................................................................................... 86 
 
Table A.05: Multinomial logistic regression parameter estimates. Stand acres and stand type as 
independent variables (showing results from shelterwood harvest). N=20,915 unique forest stands 
enrolled in the Qualified Forest Program. ................................................................................................... 88 
 
Table A.06: Multinomial logistic regression parameter estimates. Stand acres and stand type as 
independent variables (showing results from selection harvest). N=20,915 unique forest stands enrolled in 
the Qualified Forest Program. ..................................................................................................................... 90 
 
Table A.07: Multinomial logistic regression parameter estimates. Stand acres and stand type as 
independent variables (showing results from thinning). N=20,915 unique forest stands enrolled in the 
Qualified Forest Program. ........................................................................................................................... 92 
 
Table A.08: Multinomial logistic regression parameter estimates. Stand acres and stand type as 
independent variables (showing results from timber stand improvement). N=20,915 unique forest stands 
enrolled in the Qualified Forest Program. ................................................................................................... 94 
 
Table A.09: Multinomial logistic regression parameter estimates. Geographic region of Michigan as 
independent variable. N=20,915 unique forest stands enrolled in the Qualified Forest Program. ............. 96 
 
Table A.10: Multinomial logistic regression parameter estimates. Geographic region of Michigan as 
independent variable. N=20,915 unique forest stands enrolled in the Qualified Forest Program. ............. 97 
 

 

viii

Table A.11: Multinomial logistic regression parameter estimates. Geographic region of Michigan as 
independent variable. N=20,915 unique forest stands enrolled in the Qualified Forest Program. ............. 98 
 
Table A.12: Multinomial logistic regression parameter estimates. Stand condition as independent variable. 
Eastern Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program. ....... 99 
 
Table A.13: Multinomial logistic regression parameter estimates. Stand condition as independent variable. 
Eastern Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program. ..... 101 
 
Table A.14: Multinomial logistic regression parameter estimates. Stand condition as independent variable. 
Eastern Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program. ..... 103 
 
Table A.15: Multinomial logistic regression parameter estimates. Stand condition as independent variable. 
Eastern Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program. ..... 105 
 
Table A.16: Multinomial logistic regression parameter estimates. Stand condition as independent variable. 
Northern Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program. .. 107 
 
Table A.17: Multinomial logistic regression parameter estimates. Stand condition as independent variable. 
Northern Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program. .. 109 
 
Table A.18: Multinomial logistic regression parameter estimates. Stand condition as independent variable. 
Northern Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program. .. 111 
 
Table A.19: Multinomial logistic regression parameter estimates. Stand condition as independent variable. 
Northern Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program. .. 113 
 
Table A.20: Multinomial logistic regression parameter estimates. Stand condition as independent variable. 
Southern Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program. .. 115 
 
Table A.21: Multinomial logistic regression parameter estimates. Stand condition as independent variable. 
Southern Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program. .. 116 
 
Table A.22: Multinomial logistic regression parameter estimates. Stand condition as independent variable. 
Southern Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program. .. 117 
 
Table A.23: Multinomial logistic regression parameter estimates. Stand condition as independent variable. 
Southern Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program. .. 119 
 
Table A.24: Multinomial logistic regression parameter estimates. Stand condition as independent variable. 
Western Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program. .... 121 
 
Table A.25: Multinomial logistic regression parameter estimates. Stand condition as independent variable. 
Western Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program. .... 122 
 
Table A.26: Multinomial logistic regression parameter estimates. Stand condition as independent variable. 
Western Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program. .... 123 
 
Table A.27: Multinomial logistic regression parameter estimates. Stand condition as independent variable. 
Western Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program. .... 124 
 

 

ix 

LIST OF FIGURES 

 
 
 
 

Figure 1: Proportion of forest types (by area) on private land in Michigan and the Qualified Forest 
Program. 
. ................................................................................................................................................................... 28 
 
Figure 2: Proportion of forest types (by area) on private land in Michigan and the Qualified Forest 
Program in the Southern Lower Peninsula. ................................................................................................ 30 
 
Figure 3: Proportion of forest types (by area) on private land in Michigan and the Qualified Forest 
Program in the Northern Lower Peninsula. ................................................................................................ 30 
 
Figure 4: Proportion of forest types (by area) on private land in Michigan and the Qualified Forest 
Program in the Eastern Upper Peninsula. 
 .................................................................................................................................................................... 31 
 
Figure 5: Proportion of forest types (by area) on private land in Michigan and the Qualified Forest 
Program in the Western Upper Peninsula. 
 .................................................................................................................................................................... 31 
 
Figure 6: Number of stands within each condition classification, by forest type, enrolled in the Qualified 
Forest Program ............................................................................................................................................ 35 
 
Figure 7: Map of acres enrolled in the Qualified Forest Program in aspen/birch forest type group, by 
county. ....................................................................................................................................................... 123 
 
Figure 8: Map of acres enrolled in the Qualified Forest Program in elm/ash/cottonwood forest type group, 
by county. .................................................................................................................................................. 123 
 
Figure 9: Map of acres enrolled in the Qualified Forest Program in maple/beech/birch forest type group, 
by county. .................................................................................................................................................. 124 
 
Figure 10: Map of acres enrolled in the Qualified Forest Program in oak forest type group, by county.. 124 
 
Figure 11: Map of acres enrolled in the Qualified Forest Program in spruce/fir forest type group, by 
county. ....................................................................................................................................................... 125 
 
Figure 12: Map of acres enrolled in the Qualified Forest Program in white/red/jack pine forest type group, 
by county. .................................................................................................................................................. 125 
 
Figure 13: Map of acres enrolled in the Qualified Forest Program prescribed for artificial regeneration, by 
county. ....................................................................................................................................................... 126 
 
Figure 14: Map of acres enrolled in the Qualified Forest Program prescribed for clearcut, by county. ... 126 
 
Figure 15: Map of acres enrolled in the Qualified Forest Program prescribed for salvage treatment, by 
county. ....................................................................................................................................................... 127 
 

 

x

Figure 16: Map of acres enrolled in the Qualified Forest Program prescribed for seed tree harvest, by 
county. ....................................................................................................................................................... 127 
 
Figure 17: Map of acres enrolled in the Qualified Forest Program prescribed for shelterwood harvest, by 
county. ....................................................................................................................................................... 128 
 
Figure 18: Map of acres enrolled in the Qualified Forest Program prescribed for selection harvest, by 
county. ....................................................................................................................................................... 128 
 
Figure 19: Map of acres enrolled in the Qualified Forest Program prescribed for thinning, by county. .. 129 
 
Figure 20: Map of acres enrolled in the Qualified Forest Program prescribed for timber stand 
improvement, by county. .......................................................................................................................... 129 
 

 

 

xi 

 

Family forest ownership 

Introduction 

Family forest owners control the plurality (36%) of forestland in the United States (Butler et al., 

2016b). Because of this and their contributions to abundant, clean water, wildlife habitat, wood fiber, 

recreation, and other ecosystem services, it is important to study many aspects of family forest owners. 

The literature focused on the complex factors affecting family forest owners’ attitudes, behaviors, and 

motivations suggests that attitudes and motivations do not always manifest themselves in behaviors. Just 

because family forest owners say they will or will not conduct forest management does not mean they 

will or will not perform those actions (Silver et al., 2015). 

In addition to understanding the family forest owners themselves, it is important to consider the 

land they own; what it looks like, how it is managed, and why it exists in its current state. Family forests 

are a product of a feedback loop. The forests exhibit certain characteristics because of one or more owners 

that all made historic management decisions. Those management decisions were based, in part, on the 

inherited condition of the land. Land enrolled in Michigan’s Qualified Forest Program, a tax incentive 

program for family forest owners, is used as the dataset for this analysis. 

The first part of this analysis summarizes the land enrolled in Michigan’s Qualified Forest 

Program (QFP). The summary compares forest types of enrolled land to all private forestland in Michigan 

by utilizing United States Forest Service (USFS) Forest Inventory and Analysis data (USDA, 2019). 

Enrolled forest types are also summarized by region of the state; Southern Lower Peninsula, Northern 

Lower Peninsula, Eastern Upper Peninsula, and Western Upper Peninsula. The program summary 

includes an assessment of stand condition (tree size and density) by forest type. This measure can help 

evaluate the ecological maturity and the timber production potential of family forests. 

The second part of this analysis seeks to answer the question, “Can the forest parcel 

characteristics of stand acres, forest type, condition (size and density), and region be used to predict forest 

management practices at the stand level?” In this study, forest management practices refer to a set of 

silvicultural practices designed to extract timber products and regenerate forest stands. The silvicultural 

 

1

 

practices exist on a spectrum with removing no trees (no practice) to removing all trees (clearcut). These 

questions are answered using multinomial logistic regression and offer considerations as to what to credit 

for the relationships. The QFP provides an exceptional context for analyzing the nexus described above.  

The parcel characteristics analyzed in this study are readily available from the Michigan 

Department of Agriculture and Rural Development’s QFP database. They require no additional data 

acquisition, so they offer an assessment of family forest owner plans and behavior with less expense than 

landowner surveys. Steep penalties for non-compliance with forest management plans assure that planned 

forest management practices will be completed. Maps of forest types and forest management practices are 

provided to highlight certain relationships geographically. Again, I offer suggestions for what variables, 

including geography, contribute to the phenomena. 

Currently, much of the literature aimed at evaluating incentive programs focuses on counting 

tangible outputs that are easy to quantify and measure. Quantitative approaches traditionally include 

metrics that count program outcomes; the number of management plans developed or tons of phosphorus 

retained, for example. Other quantitative metrics use Likert and other rating scales to quantify attitudes 

and intentions. Qualitative evaluations describe changes in behavior or attitudes through descriptive 

interviews. No such studies assess the relationship between parcel characteristics through a taxation case 

study. 

It is hypothesized that combinations of parcel characteristics can be indicative of landowner 

behavior for landowners enrolled in a forest property taxation program. Forest types, size of holdings, and 

geographic location might all contribute to a family forest owner’s decision making. Different forest types 

have different management potential, especially when considering available markets for the material, 

which directly relates to where the forest is located. Forest types might also provoke varying levels and 

types of emotion for family forest owners in different regions. For example, family forest owners raised 

around large-diameter oak trees might attach sentimental value to the trees, and are therefore less likely to 

harvest them.  

 

2

 

Larger forest stands might make different management options more likely because of economies 

of scale. Large forest stands might also make possible the expression of certain family forest owner 

values, such as privacy or sense of place. Owners of large forest tracts might be wealthier, which impacts 

how financial motivations affect forest management decisions (Butler et al., 2016b). 

In addition to varied access to forest products markets, geography also plays a role in the social 

acceptance of forest management. In certain areas of Michigan, forest management might be interpreted 

as a positive, normal occurrence because certain silvicultural practices have been normalized, and the 

local forest-dependent industries support jobs for friends and family members. The Southern Lower 

Peninsula, where most of the state’s population exists and the least amount of its forest exist, is likely to 

support different forest management practices than the Upper Peninsula, where there are few people and a 

lot of forest. 

This study takes the first steps toward using the relationships between parcel characteristics and 

family forest owner management decisions for participants enrolled in a forest property taxation program. 

Forestry incentive program managers and forestry practitioners could use these results to help predict 

family forest owner management decisions, thereby providing justification for targeted conservation 

initiatives. It can also help refine existing taxation program or make the case for new ones. Predicting 

decisions also has implications to help the forest products industry more efficiently plan, manage, and 

utilize the resource. This study can also be used to understand the current status of a state-level property 

tax incentive program that targets family forest owners. 

Michigan real property taxation 

Because the subject of this study relates to property taxation in Michigan, it is necessary to 

outline some basic tenets regarding how real property is taxed within the state. Real property consists of 

land and any improvements to the land. Improvements include buildings and water and sewer 

infrastructure. The foundation of a Michigan property tax bill is the assessed value of a property. Per the 

Michigan Constitution, the “assessed value” (determined by a local assessment officer) is not to exceed 

50% of the true cash value, or the fair market value, of the property. An additional two adjustments, at the 

 

3

 

county and state level respectively, are applied to the assessed value to produce a state equalized value 

(SEV) (Michigan Taxpayer’s Guide, 2018). This is the basis for ad valorem taxation – taxation based on 

the value of the property. 

For a newly purchased property, the SEV is equal to the property’s taxable value (TV). The TV is 

what is actually used to calculate property taxes. Each year of ownership the TV is adjusted for inflation 

and property improvements. The TV grows annually by 5% or the rate of inflation, whichever is less, 

except for new construction. When a new owner purchases a property, the cycle begins over again, and 

the following year’s SEV becomes the new TV (Michigan Taxpayer’s Guide, 2018). 

Michigan property owners pay a millage rate on their property’s TV. One mill represents a dollar 

paid in property tax per $1,000 of TV. For example, a millage rate of 44.125 on a property with a TV of 

$100,000 means that a property owner pays 44.125 × ($100,000÷1000), or $4,412.50 annually. Millage 

rates for real property vary in Michigan from just under 34 to over just over 127, depending on the 

municipality. 

Property tax exemptions reduce the millage rate paid by the property owner. Property owners in 

Michigan may be eligible for a Principal Residence Exemption, a Qualified Agriculture Exemption, or 

Qualified Forest Exemption; all of which exempt property owners from paying their school operating 

millage. This millage is usually 18 mills, although that may be less depending on the given school 

districts.  

 

 

 

4

 

Forest ownership 

Background 

Forests provide a number of ecosystem services. They produce clean water through reducing 

nutrient and sediment loading in lakes and streams and clean air by absorbing and breaking down certain 

pollutants. Many species of wildlife spend all or a portion of their life in forest habitats for cover, food, 

and raising young. Forests also provide an economic benefit stream in the form of timber and non-timber 

forest products. They act as recreational hubs for many different activities including consumptive and 

non-consumptive uses (Ghimire et al., 2016). Forests provide aesthetics, privacy, and sense of place to 

many citizens across the United States. 

Forest ownership in the United States can be divided into classes reflecting any number of 

criteria. Often, ownership classifications reflect the legal entity presiding over the forestland, such as 

family forest owners, or an entity that has primary responsibility for managing the forestland on behalf of 

another party, such as federal and state ownership where the forestland is owned by its citizens but 

managed by an agency.  

From 1907 to 1997, the USFS recognized two major categories of forest ownership – public and 

private. Public ownership was divided into Federal and Non-Federal ownerships. Federal included 

National Forests, Bureau of Land Management, and Other Federal. Non-Federal included State, County, 

and Municipal (and other Local) forests. Private ownerships were simply divided into Forest Industry and 

Nonindustrial Private. That system appeared to ignore some modern ownership patterns including Timber 

Investment Management Organizations, Real Estate Investment Trusts, and family trusts.  

The USFS currently utilizes a revised method for classifying forest ownership. This system 

appears to move closer toward reflecting the intentions, opportunities, and constraints of forestland 

owners. This system identifies three types of public ownership: (1) federal, (2) state, and (3) local, and 

three types of private ownership: (1) family, which includes individuals and families, (2) corporate, which 

includes a variety of business structures, and (3) other private, which includes conservation and natural 

resource organizations, unincorporated partnerships and associations, and Native American tribal lands 

 

5

 

(Butler, 2014). While Native American tribal lands might make up only a small proportion of private 

ownership, the strategy to include it in “other private”, and not as its own category, might represent issues 

of environmental justice and equity that should be addressed through additional literature.  

Federal ownership includes National Forests as well as forests managed by the Bureau of Land 

Management, the National Park Service, and the Department of Defense. The literature does not describe 

what factored into the USFS categorizing land according to schema described above. The American West 

holds 75% of all public land. In the eastern United States, state and county governments manage most of 

the public lands (USFS, 2014). An important distinction between “owned” and “managed” should be 

made here. While colloquially the term “own” is often used to describe the Federal and a state 

government’s relationship to public land, it is a misnomer. This thesis has used the term “own” up to this 

point to describe the relationship mentioned above to conform to the existing body of literature. However, 

to suggest that the Federal government or a state government owns land would be incorrect. Agencies are 

tasked with managing land on behalf of its constituents. People own the land. 

Categorizing family forest owners into a seemingly homogenous group is inconvenient for 

research on family forest owners. Research on family forest owners suggests that it does not represent a 

homogenous group and is discussed in more detail later (Finley and Kittredge, 2006; Mueller, 2011). 

Butler et al. (2016a) estimate that there are approximately 816 million acres of forestland in the United 

States. Approximately 58% is held by 11.5 million private ownerships – about 473 million acres. The 

10.7 million family forest owners account for 36%, or 290 million acres, of the forestland in the United 

States (Butler et al., 2016a).  

By USFS definitions, family forests are lands that include at least one acre, are 10% stocked with 

trees, and are owned by individuals, married couples, family estates and trusts, or other groups of 

individuals who are not associated as a legal entity. Compared to non-forest owners, family forest owners 

generally have a formal education with nearly half having attained a college degree. The average age of 

the family forest owner (primary decision maker) is 62 years old. Primary decision makers of family 

forests are predominately white (95%) and male (79%) (Butler et al., 2016b). Five percent of family 

 

6

 

forest owners own between one and nine acres of woods (or forest). Family forest owners that own at 

least ten acres account for 95% of the acres of private woods (or forest) in the United States (USDA, 

2015) but only 37% of the total number of owners (Butler et al., 2016b). Given these characteristics of the 

United States’ family forest resource, one can conclude that the actions taken by family forest owners can 

have great and long-lasting impacts on the country’s forests and overall quality of life. 

The makeup of family forest owners in the United States is primarily described from the National 

Woodland Owners Survey (NWOS). The NWOS is conducted as part of the Forest Inventory and 

Analysis program within the USDA Forest Service. NWOS assists in the USDA Forest Service’s mandate 

to conduct a comprehensive inventory and analysis of the present and prospective conditions of the 

Nation’s forests (Forest and Rangeland Renewable Resources Planning Act of 1974, P.L. 93-378). The 

objective of the NWOS is to characterize the family forest owners of the United States and determine why 

they own land and what they intend to do with it (Butler and Leatherberry, 2004). 

Family forest owners owning between one and nine acres own their forest because it is a part of 

their home, beauty and scenery, privacy, wildlife habitat, and nature protection, in that order. Their top 

five issues or concerns: high property tax, vandalism or illegal dumping, keeping land intact for future 

generations, trespassing, and insects or disease (Butler et al., 2016a and 2016b). The top five reasons for 

owning forestland for owners of ten or more acres are beauty and scenery, part of their home, wildlife 

habitat, pass on to children/heirs, and privacy, in that order. Their top five issues or concerns are high 

property taxes, trespassing, keeping land intact for future generations, vandalism or illegal dumping, and 

insects or diseases, in that order (USDA, 2015). Conservation program managers may be able to align 

program outcomes with these stated reasons and concerns to address resource concerns. Timber 

harvesting and investment reasons often occur much further down the list. 

In Michigan, nearly 45% of the state’s 20 million acres of forestland is owned and maintained by 

family forest owners (Pugh, Paulson, and Butler, 2016). Communicating with this group is difficult 

because there is no complete record of forest ownership in the state. Property tax records to not indicate if 

 

7

 

a parcel is forested. Michigan is estimated to have approximately 375,000 family forest owners and nearly 

half own less than 10 wooded acres (Butler et al., 2016a).  

Issues facing family forest owners 

Family forest owners are confronted with myriad challenges in owning and managing forestland. 

Property taxes, cost of ownership, and market uncertainty all exist as challenges in maintaining ownership 

of a forested property. Increasing forest fragmentation, as well as an increasing number of family forest 

owners, creates challenges to managing forestland for specific objectives. Conway et al. (2003) show that 

many factors affect family forest owner behavior. Bequest motives, debt, and non-market activities 

(hunting, hiking, wildlife observation, etc.), as well as the usual harvesting decisions are interrelated and 

depend on landowner preferences, market, and land characteristics (Conway et al., 2003). Other factors, 

such as absenteeism, forest parcel size, and non-timber preferences are important in landowner decisions 

(Conway et al., 2003). 

As mentioned above, the decisions that family forest owners make have tremendous impacts on 

society at-large. The management of family forestland influences wood product markets, wildlife habitat 

abundance and quality, water temperature and fish habitat, storm water filtration, and clean air, to name a 

few. The management of family forestland presents a “wicked problem” (and not a “tame problem”). 

Batie (2008) suggests that “…normal science assumptions and approaches are inadequate for addressing 

the complexities of wicked problems in a policy context, but that science, including social science, 

remains crucial for the development of alternative policies.” In this context, Batie (2008) is referring 

specifically to a new approach for crafting solutions in applied economics. “…a wicked problem is not 

well understood until after formulation of a potential solution and therefore, the problem definition tends 

to change over time.” (Batie, 2008). Because wicked problems occur in a social context, and stakeholders 

may understand the problem differently than one another, solutions may be required to be active and 

adaptive. Family forest management and conservation is a wicked problem because of the number of 

stakeholders, diverging value systems, and decision makers. 

 

 

8

 

Incentive programs 

The Federal government, states, and nongovernmental organizations (NGOs) attempt to 

incentivize family forest owners to actively manage their woodlands. The benefits of forests outlined 

above can be maintained, enhanced, or protected through active forest management. Outcomes such as 

forest health, clean air, and clean water are viewed as public goods and services, which is a rationale for 

governments and NGOs to provide management incentives. These can include property tax reductions, 

forest management planning, financial assistance for implementing conservation practices, and technical 

assistance. Select programs specific to Michigan and nationwide are discussed here. 

The Agriculture Act of 2014 (Farm Bill) authorized (and reauthorized) services and programs that 

impact farmers, ranchers, and family forest owners. The Environmental Quality Incentives Program 

(EQIP) was one program reauthorized by the Farm Bill to help farmers, ranchers, and forestland owners 

voluntarily address natural resource concerns (USDA, 2019c). EQIP provides funding to the Natural 

Resources Conservation Service to pay District Conservationists to provide conservation technical 

assistance to farmers, ranchers, and forestland owners through one-on-one consultation and on-site 

evaluation. District Conservationists work with landowners to develop site-specific conservation plans 

that outline conservation practices that could be undertaken to address natural resource concerns. The 

program also makes available funding for the implementation of conservation practices such as forest 

stand improvement, brush management, windbreak establishment, and tree/shrub establishment, to name 

a few. 

Conservation districts (CDs) are special purpose subdivisions of states that are operated by a 

locally elected board of directors. They began forming slowly after President Franklin D. Roosevelt sent 

state governors proposed conservation district legislation in 1936 to incorporate into their state law. 

Currently, CD staff and directors work directly with landowners to conserve and promote healthy soils, 

water, forests, and wildlife. Nearly 3,000 conservation districts exist across the United States with more 

than 17,000 citizens serving on conservation district boards. Conservation districts help farmers, ranchers, 

and forestland owners implement conservation practices to protect soil productivity, water quality and 

 

9

 

quantity, air quality, and wildlife habitat. They also conserve wetlands, protect groundwater resources, 

and assist communities in planting trees (National Association of Conservation Districts, 2019).  

The USDA Forest Service partners with state forestry agencies, Cooperative Extension, and 

conservation districts to deliver the Forest Stewardship Program (USDA, 2019b). The Forest Stewardship 

program helps family forest owners develop a Forest Stewardship Plan (forest management plan). States 

implementing the Forest Stewardship Program are able to design and implement a program that works 

most effectively in their area. For example, Michigan’s Forest Stewardship Program utilizes federal 

dollars to offer grants to private sector consulting foresters to reduce the landowner’s cost of a forest 

management plan. Seemingly, this method works in Michigan due to a robust base of participating 

professional consulting and industrial foresters and the collective will of the forestry community to 

deliver forestry activities from the private sector. Other states choose to hire a number of government-

employed foresters that develop Forest Stewardship Plans at no cost to the landowner. Many other 

variations of the Forest Stewardship Program exist throughout the country. The program’s flexibility 

allows it to be crafted with each state’s demographics and culture in mind. The lack of consistency in the 

delivery of the Forest Stewardship Program nationwide, however, makes evaluating program 

implementation difficult. 

The State of Michigan’s Department of Agriculture and Rural Development (MDARD) 

administers a technical assistance program to encourage and incentivize family forest owners to manage 

their wooded property. Through the Forestry Assistance Program (FAP), the MDARD grants funds to 20 

conservation districts to hire full-time professional foresters to provide one-on-one technical assistance 

(MDARD, 2019a) FAP foresters conduct educational workshops and field days to teach family forest 

owners about tax incentive programs, forest management planning, forest health, income taxes, and a 

number of other topics of interest to this group. They also use the media by publishing newspaper articles 

and special interest newsletters, as well as speaking on the radio and television to promote family forest 

management. 

 

10

 

States have developed a variety of forest property tax programs over the past 300 years (Amacher 

et al., 1991) of forest taxation in the United States.  Property tax incentive programs provide family forest 

owners with certain tax breaks, credits, or exemptions in exchange for implementing forest management 

activities. Agreements often involve a family forest owner following a written forest management plan 

developed by a qualified professional (Hibbard et al., 2003). Hibbard et al. (2003) outline five major types 

of forest property tax programs. “Ad valorem laws tax forestland according to its fair market value, 

Current use programs determine the land’s taxable value according to its use in a forested condition as 

opposed to its ‘highest and best use’, Flat tax programs levy a fixed annual tax per acre, Exemption 

programs excuse forestland from property taxation altogether, and Hybrids of current use and ad valorem 

values are used to derive a taxable value for forested property.” (Hibbard et al., 2003). According to 

Hibbard et al. (2003), 66 programs existed in all 50 states at the time of their assessment; current use was 

the most popular, ad valorem and flat taxes the second. Eligibilities, application procedures, governing 

administration, and non-compliance penalties vary across forest property tax programs.  

Michigan’s history includes a number of tax incentive policies to promote forest management to 

family forest owners. An early program mirrored that available to agriculture producers. The Private 

Forest Reserve Act (or Act to Encourage Private Forestry, Public Act 86 of 1917, as amended) was 

designed to “encourage private forestry, care and management thereof and to provide for (partial) 

exemption from taxation of such private forest reserves (on farms).” It provided that owners pay a 

specific tax on forest property associated with a farm at $1 per acre combined with a 5% yield tax paid at 

the time of commercial timber harvest. The act was repealed in 2006 when the QFP, an exemption 

program, was enacted. Private Forest Reserve Act administration was stifled due to three perceived 

problems: (1) limited knowledge of the opportunity by farmers, (2) no standard record keeping 

procedures at the county and township level, and (3) poor state-county-township coordination of the act 

(Grossman and Potter, 1987). The Qualified Forest Program and the Commercial Forest Reserve Program 

are Michigan’s two current forestland tax programs. Both programs offer unique tax reductions for 

 

11

 

following a written forest management plan developed by a private sector professional forester. These 

programs are discussed in greater detail in the Methods section. 

Landowner behavior 

The developing paradigm for understanding family forest owner behaviors and motivations has 

suggested that non-monetary benefits are paramount in shaping how they view and manage their forests, 

even when there are financial benefits to be had for particular practices (Koontz 2001). Koontz (2001) 

found that financial incentives do not necessarily encouraging management activities, but rather utilize 

public funding to subsidize those that are already managing. 

The attitudes, behavior, and motivations of family forest owners have been intensively studied. 

The literature analyzes many variables including reasons for ownership, parcel characteristics, land 

tenure, and demographic characteristics. Many studies have attempted to explain why family forest 

owners make the decisions they do so that policy, technical assistance, and education/outreach can be 

more effectively designed and delivered (Kuipers and Potter Witter, 2012). “Policy tools utilized most 

prevalently in private forest policy include regulatory, incentive, and educational (or capacity-building) 

tools.” “…the choice of which tool(s) to implement as part of the objectives for various policies must be 

made strategically.” (Janota and Broussard, 2008). 

Surveys continue to serve as critical indicators of family forest owner beliefs, attitudes, intent, 

and decision-making. GC and Potter-Witter (2011) conducted a mail survey of family forest owners to 

understand factors influencing biomass and timber harvesting decisions. Amenity benefits of forests, such 

as enjoyment of scenery, protection of biologic diversity, part of home, and privacy, were cited as the 

most important reasons for forest ownership. Other studies that utilized survey data yield similar results 

(Potter-Witter, 2005; USDA, 2015; Butler, 2016b). Survey data can offer great insights into family forest 

owners and how they view and manage their land, although they can be expensive to implement and don’t 

always reflect what happens on the ground (Silver et al., 2015). 

 

 

 

12

 

Implications 

Federal and state incentive programs often regard family forest owners as a homogeneous group 

such that each acre addressed in a management plan or each acre treated equate to the same level of 

conservation. While this undoubtedly makes accounting program metrics easier, it doesn’t address the 

disproportionality that exists in natural resources. Not every forested acre is positively or negatively 

impacting soil and water quality at the same rate and not every citizen has equal access to resources and 

information. Historically, disproportionality has been described three ways within the social sciences; 

environmental justice, patterns of social sanctions and interaction, and environmental impact associated 

with different forms of social groups (Nowak et al., 2006). Interdisciplinary approaches are required to 

evaluate disproportionality effects on social and environmental systems.  

Incentive programs attempt to address challenges faced by family forest owners, but to do so 

requires program managers to have a complex understanding about characteristics of land they own, 

characteristics of the owners, and how beliefs shape their attitudes and behaviors. It is also worthwhile to 

consider program outputs, outcomes, and what it is the environmental/conservation community is striving 

for. Are they attempting to change attitudes, beliefs, or behavior? Certainly, these concepts are not 

mutually exclusive. Is it important that a family forest owner implements forest management for reasons 

outlined by natural resource professionals, or is convincing them to do it the primary objective? The same 

impacts to forest health, water quality, plant productivity, and regional economies, for example, are 

realized as long as the activity is conducted; regardless of the motivation. 

 

 

 

13

 

Literature Review 

As mentioned previously, numerous studies have attempted to acquire insight into the family 

forest owner. These studies have asked a variety of questions and have taken multiple approaches to gain 

the insight necessary to describe this group of landowners. The proceeding discussion is separated into 

four different sections germane to this study. The types of questions that have been asked regarding 

family forest owners will be discussed first, followed by a review of conservation incentive programs, 

studies aimed at evaluating these programs, and studies attempting to model family forest owner 

behavior. 

What we know about family forest owners  

Finley and Kittredge (2006) suggest that family owners in Massachusetts could be categorized 

using a market segmentation approach through cluster analysis. The study identified three groups of 

landowners; Muir, Thoreau, and Jane Doe. “Thoreau” landowners highly valued privacy and the 

contemplative values of forests, such as scenery and recreation. “Muir” landowners placed high value on 

nature and environmental protection. The final segment, “Jane Doe”, was named so because of how little 

is known about the segment. The analysis found them to be indifferent to strong identifying 

characteristics. Mueller (2011) conducted a similar study in Michigan that looked to create market 

segments of family forest owners based on reasons for ownership. The study identified three groups of 

landowner types; Timber Barons, Game Wardens, and Tenants. Mueller (2011) also suggests, in a parallel 

study, that reasons for ownership may vary regionally. Reasons for owning forestland—hunting and 

fishing, timber production, investment, non-hunting recreation, as part of a home—continue to be shown 

to have the most significant effect on family forest owner attitudes and opinions regarding forest 

management. 

Finley and Kittredge (2006) utilized data that was drawn from a survey distributed in 2000. In the 

original study, White (2001), aimed to measure private forestland owners’ attitudes toward ecosystem 

management activities implemented on their forestland. The survey followed the process as outlined in 

Dillman’s (2000) Total Response method. Finley and Kittredge (2006) also made certain to exclude data 

 

14

 

that originated from Metropolitan Statistical Areas and Consolidated Metropolitan Statistical Areas as 

defined by the United States Census. As previously mentioned, these authors took a three-phase analytical 

strategy to define and describe groups based on measured values and attitudes. A 5-point Likert scale was 

used to provide quantitative data associated with attitudes. The phases, in order, were as follows: principal 

component analysis, cluster analysis, and multiple discriminant analysis. Mueller (2011) followed the 

same approach. 

Kluender and Walkingstick (2000) performed a cluster analysis of Arkansas landowners in 16 

counties suggested that family forest owners in the area exist in four main types. Timber Managers were 

those that had sold land in the past and have definitive plans to sell timber in the future. Resident 

Conservationists lived on their land and enjoyed it for its scenic beauty. They hadn’t sold timber in the 

past and had no intention on doing so in the future. The Affluent Weekenders did not live on their 

forested property and were mostly interested in the natural amenities on the land instead of trying to make 

money from it. The Poor Rural Residents grew up in a rural environment and inherited their forestland. 

Most had conducted timber sales in the past but were unlikely to practice real timber management. These 

studies suggest that the reason a landowner owns forestland would play a crucial role in more specific 

dynamics. 

Potter-Witter (2005) sought to understand whether forest management activity reported by 

Michigan family forest owners who were enrolled in several types of incentive programs differed 

significantly by program and whether management activity was significantly explained by landowner 

demographics. The research specifically looked at five unique groups of forest management activities; 

tree planting, timber harvesting, timber stand improvement, wildlife habitat management, and soil and 

water protection. In addition to pertinent demographic characteristics about the sample populations, 

Potter-Witter (2005) demonstrated that within the groups studied, parcel size and whether or not the 

parcel was also the owner’s permanent residence were significant in explaining forest management 

activity. 

 

15

 

Potter-Witter (2005) took a unique approach in attempting to describe Michigan’s private 

landowners. Four populations of NIPF owners were studied. One population consisted of participants in a 

tax incentive program (Michigan’s Commercial Forest Reserve Program), another was a group of 

landowners engaged in a management plan development program administered by the USFS (Forest 

Stewardship Program), yet another were members of a landowner-organized information, education, and 

peer support program (Michigan Forest Association), and the last was a group consisting of a state agency 

information and education program (Two-Hearted River Watershed).  

Potter-Witter and Peterson (2006) followed this with two statewide surveys to understand forest 

interests and activity patterns among family forest owners in Michigan (Peterson and Potter-Witter, 

2006). Specifically, the study focused the following objectives items: (1) characteristics of family forest 

owners already participating in forest management programs, (2) if these landowner significantly differ 

from the overall Michigan forest landowner population, (3) the rates of forestry activity among family 

forest owners, (4) the efficacy of incentives, and (5) timber harvest attitudes and behaviors. The first 

phase included family forest owners currently participating in management activities or organizations. 

The second phase included a regionally stratified survey of randomly selected family forest owners.  

The researchers learned that most of the timber management conducted by family forest owners 

was done so by participants of incentive programs. “…tax incentives and technical advice and assistance 

may be bigger incentives for timber management than cost-sharing programs…” During the second phase 

of the study the researchers noted that less than one-third of the family forest owners that harvested 

timber had a forest management plan. Only about one-third consulted a forestry professional. Researchers 

note that many respondents to the survey indicated that they simply don’t know where to find the 

information to engage with a natural resources professional. Those in Extension and related technical 

assistance providers need to find a better way to convey this information.  

Saulnier et al. (2017) conducted a similar analysis to Potter-Witter (2005) and found similar 

results for family forest owners in Virginia. The variables income, age, forested acres owned, and forest 

management were all significantly and positively related to willingness to harvest. The researchers 

 

16

 

suggest that as demand for wood fiber in Virginia increases, the ability to predict and forecast availability 

becomes increasingly important. With most of Virginia’s forest in private ownership, as is the case in 

Michigan, family forest owners are a critical variable in the equation. 

Wolde et al. (2017) presented a survey to landowners in Virginia and Texas to identify factors 

that predict behavior, assess overlap in family forest owner motivations, and determine how respondent’s 

forest management plans and sustainability concerns correspond with their decisions. The researchers 

presented to family forest owners several different ways to supply biomass to the market. They thus 

determined that family forest owners are willing to consider multiple ways of supplying biomass 

simultaneously (Wolde et al., 2017). 

Another prolific area of research is evaluating a family forest owner’s willingness to conduct 

activities other than harvesting. Khanal et al. (2017) provide an example. Researchers examined family 

forest owner willingness to delay final timber harvest for additional carbon sequestration. The study 

suggests that participation decreases as the monetary incentive to participate decreases as compared to a 

timber harvest. Family forest owners with recreational goals for their property are the most likely group to 

participate in carbon sequestration (Khanal et al., 2017). 

Literature focused on family forest owner attitudes, behavior, and motivation has looked only at 

the propensity toward forest management without much thought as to how they felt about it. Few studies 

have attempted to describe landowner attitudes regarding forest management activities that they have 

performed on their own land. While some landowners have attitudes that reflect an opposition to forest 

management, it doesn’t necessarily mean that those landowners haven’t or will not participate in forest 

management activities at some point on their land. Because landowners that have experience in forest 

management may influence those that seek their opinions for potential work on their lands, this paradigm 

must be accompanied by a shift in how researchers interpret landowner behavior and attitudes. 

Program evaluation 

Evaluating the effectiveness and efficiency of conservation programs helps program managers, 

policymakers, and researchers craft programs that improve natural resources according to desired future 

 

17

 

conditions and outcomes. Quantitative approaches traditionally include metrics such as acres treated, 

number of management plans developed, or tons of phosphorus retained, for example. Qualitative 

evaluations aim to determine changes in behavior or attitudes in a descriptive approach. 

There are common principles for evaluating tax programs (Klemperer, 1988; Hickman, 1992). 

Hibbard et al. (2003) outline six principles. “Equity: individuals and other taxed entities are treated 

equally and in a fair and consistent manner. Efficiency: desired goals and objectives are accomplished 

with minimal involvement of nonessential efforts or expenses. Simplicity: the tax program is easy to 

understand and administer. Stability: reliably steady income accrues too the taxing authority. Adequacy: 

an acceptable level of tax income is collected. Visibility: individuals and other taxed entities can easily 

access information about the program and its provisions.” (Hibbard et al., 2003). Hibbard et al. (2003) 

also found several attributes that characterize effective tax policies: clearly articulated program goals, 

clearly demonstrated advantage, complementary with other programs, internal consistency, administrative 

consistency, adequate funding for implementation, seamless transition to new tax programs, and periodic 

evaluation for efficiency and effectiveness. 

When Wolde et al. (2017) determined that family forest owners in Virginia and Texas were 

willing to consider multiple ways of supplying biomass simultaneously, they suggested that the absence 

of programs that incentivize changes in forest management behavior in response to woody bioenergy 

represents a missed opportunity. They critique that while current woody biomass incentive programs do 

provide financial support for discrete ways of supplying material, they lack the flexibility to take 

advantage of emerging opportunities and strategies to achieve landowner objectives (Wolde et al., 2017). 

Not only are changes in forest management behavior to supply biomass discouraged, they are ineligible 

for payment and often result in penalties. 

The few woody biomass programs that existed at the time of the Wolde et al. (2017) study varied 

from one another in eligibility requirements, contract length, and payments. Navigating through the 

various options for distinct programs is confusing to family forest owners. Integration might make 

 

18

 

navigating such programs easier to eligible and interested landowners. Integration may also reduce 

communication barriers among implementing agencies and reduce the duplication of efforts. 

Examples of natural resources and conservation program evaluation exist outside of forestry and 

government programs. Karlen et al. (2014) conducted an evaluation on the Soil Science Society of 

America Soil and Water Conservation and Management Division. The paper looked at three main 

thoughts: (i) how the Soil and Water Conservation and Management Division evolved, (ii) how soil 

management research approaches have changed since the division was founded, and (iii) how division 

scientists are helping and increasing global population respond to a dynamic and changing climate 

(Karlen et al. 2014). Karlen et al. (2014) concluded, “landscape-scale soil management field research is 

vital for soil security and to meet increasing global demands for food, feed, fiber, and fuel. A new 

research paradigm with a long-term horizon is crucial and must be aggressively pursued.” 

Butler et al. (2014) examined the effectiveness of the USFS Forest Stewardship Program (FSP). 

Through a multiple analytic approach, they used annual FSP accomplishments, survey of state FSP 

coordinators, analytic comparison of family forest owners receiving and not receiving forestry practice 

assistance, and focus groups with family forest owners. The authors concluded that the FSP reaches a 

small fraction of eligible landowners and suggested that one-on-one forestry practice assistance may be 

more useful to family forest owners than other forms of assistance (e.g. development of forest 

management plans). The FSP participants are no different from other landowners with regards to future 

land use intentions and the program is not influencing inactive family forest owners to become active 

managers. The authors do, however, suggest that the current practice of state-level flexibility in the 

implementation of the program may be a positive in helping reach nationwide goals (Butler et al. 2014). 

While one-on-one technical assistance might be an effective method for influencing behavior 

(Butler et al. 2014), cost can be a concern. Using Federal Forest Stewardship Program funds to hire a state 

program coordinator might cost less to taxpayers and/or program supporters than hiring many 

professionals for one-on-one assistance, but at what cost? Decision makers have to weigh perceptions of 

efficiency against resource concerns not being addressed. 

 

19

 

As discussed in the previous chapter, disproportionality exists in how landscapes and landowners 

impact environmental quality (Nowak et al. 2006). One strategy to address the disproportionality of 

impacts on the environment is by targeting conservation efforts to the landscapes and landowners 

contributing the most pollution or with the highest risk. “Targeting” of conservation efforts is touted as an 

efficient and effective strategy to address environmental stewardship. Targeting is defined as “a set of 

spatial technologies and procedures linked to mapped variables directed to implement conservation 

management practices that take into account spatial and temporal variability across natural and 

agricultural systems,” (Berry et al. 2005, 363).  

One critique of targeting, especially in the realm of public funding, is that every taxpayer should 

have equal access and opportunity to conservation dollars. Targeting, by definition, would exclude lower 

risk individuals or landscapes in favor of addressing pollution or risk on the parcels that have the highest 

impact. Arbuckle, Jr. (2013) analyzed data from the Iowa Farm and Rural Life Poll (IFRLP), a 

longitudinal panel study of Iowa farmers. The study questioned whether there was farmer resistance to 

targeting conservation efforts. Arbuckle, Jr. (2013) concluded that most farmers support targeted 

approaches. Factors that determined whether a farmer endorses targeted conservation efforts were 

acknowledging agriculture’s environmental impacts, believing that farmers should address water quality 

issues, experiencing significant soil erosion, belief that extreme weather will become more common, 

participation in the Conservation Reserve Program, and belief that farmers who have natural resources 

issues are less likely to seek conservation assistance (Arbuckle, Jr. 2013). This analysis demonstrates that 

groups currently receiving equal access and opportunity to public funding to address conservation issues 

could be in favor of relinquishing their equal access and opportunity in favor of addressing the resource 

concern. 

The field of agriculture yields another example. A study by Jackson-Smith et al. (2010) focused 

on measuring conservation program best management practice (BMP) implementation and maintenance 

at the watershed scale. The approach was to examine the strengths and weaknesses of using USDA 

Natural Resources Conservation Service records of conservation program participation as an indicator for 

 

20

 

BMP implementation. Through farmer interviews, the authors determined that traditional record keeping 

by the agency yielded potential limitations. Limitations included (1) documentation of the incidence of 

successful BMP implementation, (2) the nature of the BMP that was/were implemented, (3) accurate 

measurement of the timing and location of BMP implementations, and (4) information about the long-

term use and maintenance of implemented BMP. They offer that records should be field-verified and that 

the agency should develop a more robust system for tracking BMP implementation and maintenance over 

periods of time (Jackson-Smith et al. 2010). 

Traditional program monitoring and assessment tell only part of the story. Evaluation is different 

because it introduces a value judgment into determining what success looks like (Kleiman et al., 2000). 

Conservation program managers should ask whether or not current metrics indicate desired outcomes, and 

if the processes implemented are the most efficient. While internal, informal program evaluations are 

common (Blackhouse et al., 1996), regular, more formal internal evaluations should be standard for 

conservation programs (Kleiman et al., 2000). Often, outputs are viewed as proxies for outcomes. 

Outcomes refer to the desired changes on the landscape. Outputs are measured in “widgets”, which are 

assumed to be directly related to the desired outcome. 

 

 

 

21

 

The Qualified Forest Program  

Methods 

Currently, Michigan has two forestland tax incentive programs: the Commercial Forest program 

(CF) and the Qualified Forest Program (QFP). Landowners enrolled in the CF program pay a specific tax 

of $1.30 (2017-2021) per acre per year in property taxes instead of ad valorem taxes, i.e. an annual rate 

applied to a property value established by the local government. Savings from enrollment in CF are often 

dramatic and become exponentially higher as acreage increases and property value increases. It is not 

uncommon for enrolled landowners to experience savings of thousands of dollars annually, depending on 

the size of the holding and the local assessed property values. Landowners enrolled in CF are required to 

implement a written forest management plan, which must include commercial timber harvesting, and 

allow public non-motorized access on their property for hunting and fishing. The ownership must be at 

least 40 contiguous acres and 50% stocked with productive forest. Per program statute, productive forest 

is defined as “capable of producing 20 cubic feet of wood per acre per year”. The Michigan Department 

of Natural Resources administers the CF program (Michigan Department of Natural Resources, 2019). 

Landowners in the QFP receive two direct benefits in exchange for enrolling. One benefit 

exempts enrollees from certain school operating taxes; usually 16 mills (Qualified Forest School Tax 

Affidavit). Millage rates are determined locally, and one mill represents a dollar paid in property taxes per 

$1,000 of taxable value. Rates vary from less than 34 to more than 127 in Michigan, depending upon the 

local government. A property’s taxable value is approximately equal to 50% of the assessed fair market 

value. This savings can represent as much as 40% of the annual property tax bill. The second benefit 

prevents the “uncapping” of the property’s taxable value to the current State Equalized Value in the event 

of a change in ownership. A new owner of Qualified Forest property will be taxed on the property value 

experienced by the previous landowner. This accounts for major property tax savings, especially if a long 

tenured owner originally enrolled the land.  

The program is funded by both General Fund support and restricted revenue generated by 

enrollment. Enrolled landowners pay a 2-mill equivalent participation fee to the Private Forestland 

 

22

 

Enhancement Fund, which is used for program administration and forestry technical assistance delivery. 

School operating funds lost through program participation are reimbursed by the State of Michigan to the 

School Aid Fund through the State’s General Fund. The Michigan Department of Agriculture and Rural 

Development administers the QFP (Michigan Department of Agriculture and Rural Development, 2019b). 

Why is the Qualified Forest Program data useful for measuring behavior? 

Landowners enrolled in the Qualified Forest Program sign and record an affidavit attesting that 

they will follow their written forest management plan. This provides an exceptional dataset for analyzing 

family forest owner behavior because steep non-compliance penalties provide some level of assurance 

that planned forest management practices will be completed.  

If the land is converted by a “change in use”, the current landowner wishes to end enrollment in 

the program, or the landowner does not comply with their written forest management plan, the landowner 

must be removed from the program. The rescinded property is subject to a recapture tax as follows: 

 

“Qualified Forest School Tax Affidavit Repayment = taxable value of property X the number of operating 

mills levied by the school district, excluding the 2-mill equivalent fee X the number of years the property 

has been under the exemption, not to exceed 7” 

 

And/or 

 

“Qualified Forest Taxable Value Affidavit Repayment = The amount of taxes that would have been paid 

had the taxable value not remained “capped” after the sale of the property, not to exceed 10 years” 

The recapture tax is doubled if a timber harvest had not been conducted on the parcel during the period of 

enrollment. 

 

 

 

 

23

 

Data collection 

Data for this study were collected in May of 2017. The data were derived from enrollment 

information regularly collected by the Michigan Department of Agriculture and Rural Development 

(MDARD). MDARD collects the information when a family forest owner voluntarily applies, and is 

accepted, for enrollment in the Qualified Forest Program. To be considered for enrollment in the program, 

one must submit a Qualified Forest Program application, a copy of the parcel’s forest management plan 

developed by a Qualified Forester (an approved professional forester in the private sector), a forest 

management practice schedule, a copy of the most recent deed and/or land contract reflecting the current 

ownership, a copy of the parcel’s tax bill, and a $50.00 application fee. The forest management practice 

schedule is the collection of forest management practices prescribed in the forest management plan in a 

format that makes data collection simpler for the agency staff. The landowner prepares the application 

materials, often with assistance from their plan writer or their local conservation district. 

The data in this study include all ownerships enrolled in the Qualified Forest Program through 

mid-2017. Data records represented 20,915 unique forest stands across all ownerships, as determined by 

the forest management plan writers. Duplicates in the dataset were common. It is possible for one forest 

management plan to serve multiple enrollments. It is common for landowners to own land in multiple tax 

collecting units of government (usually survey townships). Parcels of land in different tax collecting units 

of government are enrolled separately because of how Michigan property taxes are administered. Some 

records had to be eliminated because forest management practice dates included nonsensical data. Stand 

acres, forest type, stand condition (size and density), and region are the stand-level variables evaluated. 

These independent variables were analyzed to describe forest management practices at the stand level. 

Each forest stand enrolled in the QFP is assigned a value that describes the size 

(seedlings/saplings, poletimber, sawtimber) and density (poorly stocked, moderately stocked, well 

stocked) of the trees in the stand. The program calls this “stand condition”, and the forest management 

plan writers assign the values. Conditions are given values ranging from zero to nine. The conditions are 

as follows: non-stocked (0), seedlings/saplings – poorly stocked (1), seedlings/saplings – moderately 

 

24

 

stocked (2), seedlings/saplings – well stocked (3), poletimber – poorly stocked (4), poletimber – 

moderately stocked (5), poletimber – well stocked (6), sawtimber – poorly stocked (7), sawtimber – 

moderately stocked (8), and sawtimber – well stocked (9). Generally, seedlings/saplings are trees 1-5 

inches diameter at breast height (dbh) greater than three feet tall. Poletimber describes trees between five 

and ten inches dbh. Sawtimber describes trees with a dbh greater than ten inches. 

Data analysis 

Multinomial logistic regression was used to determine the log-odds of a forest practice occurring 

relative to another while other variable categories are held constant. In this type of analysis, one category 

of the dependent variable (forest practice) is chosen as the reference category. Each forest practice (e.g., 

artificial regeneration, clearcut) was tested compared to a reference category of “no practice”. Select other 

multinomial logistic regressions were run with other forest practices as the reference category. 

Multinomial logistic regression is a method that generalizes logistic regression to situations where there 

are more than two levels of a dependent variable. The nine potential outcomes for forest practice make 

this the appropriate statistical test. It is also the appropriate statistical test because the dataset satisfied the 

other assumptions inherent to logistic regression. Statistical tests were performed in SPSS (IBM). 

The most telling output metric in the multinomial logistic regression analysis is Exp(β), hereby 

referred to as the odds ratio. For a continuous variable (stand acres), the odds ratio determines the odds of 

selecting one category of the dependent variable over the reference category with a one-unit increase in 

the continuous variable, holding all other variables constant. For categorical variables (condition, region, 

stand type), the odds of selecting one category of the dependent variable over the reference category are 

Exp(β)-times higher than the zero parameter. The categorical variable relationship is also contingent on 

all other variables being held constant. Comparisons can be made between different categories of the 

predictor variables within a category of the independent variable by dividing the odds ratio of one 

category of the predictor variable by another. The comparison between different categories of the 

dependent variable across the same category of predictor variable utilizes the same math.  To make this 

 

25

 

comparison, divide the odds ratio of a category of predictor variable by the same predictor variable odds 

ratio in another category of the dependent variable. In the following paragraphs, a subset of results from 

the multinomial logistic regression analysis is described in an effort to highlight interesting relationships. 

The multinomial logistic regression assigns one category of each independent variable as a 

reference group. By default, the reference group is usually the variable category with the highest 

numerical value. This provides the analysis with one category in each independent variable with which to 

compare to the rest of the categories. In the tables showing the outputs of the multinomial logistic 

regression, this assignment is noted as a subscript that reads, “This parameter is set to zero because it is 

redundant.” (Table 2 through Table 28 and A.01 through A.27). 

 

 

 

26

 

Program summary results 

Results 

Applications for enrollment in the QFP are due by 1-September each year. The tax exemption is 

realized the following 1-January. At the time of data collection (2017 May), the QFP had 450,555 acres 

enrolled. Stand type is one variable considered in this summary and the analysis described in the next 

section. Stand type refers to the primary tree or other vegetative cover in a given stand. Stand type is 

interchangeable with the term forest type when considering anything broader than the forest stand. For 

purposes of this study, forest types were aggregated into groups to aid in comparative analysis. The land 

enrolled represented forest types in seven broad groups: maple/beech/birch group at 131,438 acres (29%), 

aspen/birch group at 105,195 acres (23%), spruce/fir group at 66,414 acres (15%), oak group at 45,858 

acres (10%), other forest types at 38,318 acres (9%), white/red/jack pine group at 34,032 acres (8%), and 

elm/ash/cottonwood group at 29,300 acres (7%). Forest type groupings were designed to reflect those 

used in USFS Forest Inventory and Analysis.  

Forest types enrolled in the QFP are similar to forest type proportions on all private land in 

Michigan (Figure 1). Nearly one-third of the private land in Michigan (33.18%) and land enrolled in the 

QFP (29.17%) is the maple/beech/birch group (northern hardwoods). Aspen/birch, however, is nearly 

double in the QFP (23.35%) versus the proportion of private land in Michigan (12.11%). 

 

 

 

 

 

 

 

 

 

27

 

)

%

(
 
n
o
i
t
r
o
p
o
r
P

35

30

25

20

15

10

5

0

33

29

23

12

13

7

10

7

15

11

8

6

9

3

Michigan Private Forestland

Qualified Forest Program

  
Figure 1: Proportion of forest types (by area) on private land in Michigan and the Qualified Forest 
Program. 
 

The proportion of forest types enrolled varies by region of the state. Enrollment in the Southern 

Lower Peninsula (SLP) (Figure 2), Northern Lower Peninsula (NLP) (Figure 2), and Eastern Upper 

Peninsula (EUP) (Figure 3) all exhibit similar proportions of the Northern hardwoods type 

(maple/beech/birch) – 28%, 27%, and 29%, respectively. The Western Upper Peninsula (WUP) (Figure 3) 

has a higher proportion of its private forestlands in the maple/beech/birch group (35%). The NLP, EUP, 

and WUP each contain similar proportions of the aspen/birch group (25%, 22%, 28%, respectively). 

Private forestland enrolled in the SLP exhibit a lower proportion of the aspen/birch group (12%). Oak is a 

major component in the SLP, less so on the NLP, and is nearly absent in the Upper Peninsula (0%). 

Regions were determined using existing groupings utilized by the USFS Forest Inventory and Analysis. 

In the SLP, the proportion of acreage in different forest types enrolled in QFP (Figure 2) is 

different than the proportion of forest types on all private land in that region (Figure 2) (USDA, 2019). 

The SLP contains 3,103,586 acres of forestland on private ownerships (USDA, 2019) and 51,864 acres 

enrolled in the QFP. The aspen/birch group represents only 2% of private land in the region by accounts 

for 12% of the enrolled acreage. The maple/beech/birch group offers another stark contrast. The group 

only makes up 15% of private land in the region. However, the maple/beech/birch group maintains more 

 

28

 

than a quarter (28%) of enrolled land in region. The most distinct difference in the SLP is in the oak 

group. The QFP enrolled proportion and the proportion on private land in the region is 27% and 46%, 

respectively. 

The differences between the proportion of acreage in different forest types enrolled in QFP 

(Figure 3) and the proportion of forest types on all private land (Figure 3) is subtler in the NLP region. 

The NLP contains 4,391,296 acres of forestland on private ownerships (USDA, 2019) and 200,070 acres 

enrolled in the QFP. The proportions of the maple/beech/birch group are nearly identical at 27% and 26%, 

respectively. The aspen/birch group and the oak group each exhibit significant differences between 

categories. Twenty-five percent (25%) of the regions QFP enrolled land is in the aspen/birch group. The 

type only represents 16% of private land in the region. The oak group demonstrates the opposite 

phenomenon. While only 16% of the QFP enrolled land in the region is dominated by oak, the group 

represents 27% of private land in the NLP. 

The dominant forest type on private land in the EUP is the spruce/fir group at 29% (Figure 4). 

The spruce/fir group also dominates QFP enrolled land in the region at 31% (Figure 4). The 

maple/beech/birch group has a 7% difference between private land in the region (36%) and QFP enrolled 

land (29%). The aspen/birch, elm/ash/cottonwood, and white/red/jack pine groups are similar in both 

categories. The EUP contains 2,089,390 acres of forestland on private ownerships (USDA, 2019) and 

109,854 acres enrolled in the QFP. 

The dominant forest type on private land in the WUP is the maple/beech/birch group at 60% 

(Figure 5). The maple/beech/birch group represents only 35% of the acreage enrolled in the QFP. The 

aspen/birch group is another that expresses difference between the two categories. The group comprises 

28% of the QFP enrolled land in the region but only 12% of the private land in the region. The WUP 

contains 2,991,173 acres of forestland on private ownerships (USDA, 2019) and 88,766 acres enrolled in 

the QFP. 

 

29

 

)

%

(
 
n
o
i
t
r
o
p
o
r
P

50
45
40
35
30
25
20
15
10
5
0

46

27

28

30

15

12

12

2

9

10

4

1

2

2

Michigan Private Forestland

Qualified Forest Program

 
Figure 2: Proportion of forest types (by area) on private land in Michigan and the Qualified Forest 
Program in the Southern Lower Peninsula. 
 
 
 

25

27

26

16

27

16

10

6

8

7

10

9

8

5

)

%

(
 
n
o
i
t
r
o
p
o
r
P

30

25

20

15

10

5

0

Michigan Private Forestland

Qualified Forest Program

 
Figure 3: Proportion of forest types (by area) on private land in Michigan and the Qualified Forest 
Program in the Northern Lower Peninsula. 
 
 
 
 
 
 
 

 

30

 

)

%

(
 
n
o
i
t
r
o
p
o
r
P

40

35

30

25

20

15

10

5

0

36

29

22

18

31

29

7

5

2

0

7

5

8

1

Michigan Private Forestland

Qualified Forest Program

Figure 4: Proportion of forest types (by area) on private land in Michigan and the Qualified Forest 
Program in the Eastern Upper Peninsula. 
 
 
 

 

60

35

28

12

)

%

(
 
n
o
i
t
r
o
p
o
r
P

70

60

50

40

30

20

10

0

18

15

4

5

2

0

6

5

9

1

Michigan Private Forestland

Qualified Forest Program

 
Figure 5: Proportion of forest types (by area) on private land in Michigan and the Qualified Forest 
Program in the Western Upper Peninsula. 

 

 

31

 

The condition of each stand is assigned a number, alongside the stand type, by the plan-writing 

forester at the time of application. The number represents a qualitative assessment of the size of the trees 

and the relative stocking in the stand. Stand conditions 1, 2, and 3 all represent stands in the 

seedling/sapling size class and increase in stocking from poorly stocked (1) (SS-P), to moderately stocked 

(2) (SS-M), to well stocked (3) (SS-W). Stand conditions 4, 5, and 6 represent stands in the poletimber 

class and increase in stocking from poorly stocked (4) (P-P), to moderately stocked (5) (P-M), to well 

stocked (6) (P-W). Stand conditions 7, 8, and 9 represent stands in the sawtimber class and increase in 

stocking from poorly stocked (7) (S-P), to moderately stocked (8) (S-M), to well stocked (9) (S-W). A 

stand condition “0” means the stand is non-stocked (NS). This classification is usually limited to non-

forest types, such as grass or rock. 

The plurality (46%) of forest stands enrolled in the program is middle-aged (poletimber). Nearly 

one-quarter (24%) of the stands enrolled are classified as poletimber – well stocked (Figure 6). 

Sawtimber, in all densities, represents 34% of all the forest stands enrolled. The seedling/sapling class, 

across all densities, represents 19% of the forest stands enrolled.  

Within the aspen/birch group, 37% of stands are seedling/saplings – well stocked; the most 

common condition in the group. The second most common stand condition in the group is poletimber – 

well stocked (22%). In the elm/ash/cottonwood group the majority of stands are poletimber (58%). Only 

6% of stands in the maple/beech/birch group are in the seedling/sapling condition. Forty-one percent 

(41%) of the spruce/fir group is in the poletimber-well stocked condition. Nearly half (45%) of the 

white/red/jack pine group is in the sawtimber condition. Oak is also quite “top heavy”, in that 79% of the 

group is in the sawtimber condition (Table 1). 

Maps (Appendix, Figures 11-16) illustrate the location and distribution of acres of forest types 

enrolled in QFP across the state. The aspen/birch forest type group is mostly concentrated in the 

northeastern Lower Peninsula and the central Upper Peninsula, with Alcona County having the highest 

number of acres (10,952.2) (Figure 7). The elm/ash/cottonwood group has a similar pattern to the 

aspen/birch group, except the county with the highest number of acres in the type is Delta (2,168.4) 

 

32

 

(Figure 8). Marquette County has the most acres of the maple/beech/birch group (8,254.9) (Figure 9). 

Alcona County contains the most acres in the oak forest type group (8,465.1) (Figure 10). Most of the 

spruce/fir forest type group is in the central Upper Peninsula, Delta and Menominee counties have the 

most acres with 9,190.1 and 9,458.8, respectively (Figure 11). The pine resource appears to be spread 

between two regions. The two counties with the most acres in the white/red/jack pine forest type are 

Osceola (1,869.4) and Schoolcraft (1,648.4) (Figure 12). 

Maps (Appendix, Figures 17-24) also highlight the location and distribution of acres of prescribed 

forest practices in the QFP across the state. Artificial regeneration is a practice most frequent in the 

eastern half of the NLP. Crawford (688.9) and Alpena (429.8) counties contain the most acres of artificial 

regeneration. Alcona County has the most acres of clearcut (6,342.1) prescribed, with Delta County 

having the second most (5,986.4) (Figure 14). Most of the acres prescribed selection harvest are in the 

WUP. The two counties with the most acres are Marquette (6,347.3) and Menominee (5,692.6) (Figure 

18). 

It is important to note that the mapped variables reflect only the quantity of acres of a given 

category within a county. They do not compare relative quantities or proportions within regions. For 

example, most of the acres prescribed selection harvest are in Marquette County. Marquette County also 

has a lot of acres enrolled in the program. An inappropriate interpretation would be to say that the SLP 

region does not have a high proportion of stands being prescribed a selection harvest simply because they 

have fewer acres enrolled than the other three regions.

 

33

 

Table 1: Forest condition as a percentage of stands within each forest type group. 

Forest Type 

Group 

Aspen/birch 

Elm/ash/ 

cottonwood 
Maple/beech

/birch 

Other 

Spruce/fir 

group 

White/red/ 
jack pine 

NS* 

0% 

0% 

0% 

9% 

0% 

0% 

SS-P* 

SS-M* 

SS-W* 

1% 

1% 

1% 

14% 

2% 

2% 

5% 

4% 

2% 

6% 

5% 

3% 

37% 

8% 

3% 

11% 

8% 

10% 

P-P* 

5% 

16% 

7% 

20% 

12% 

5% 

P-M* 

12% 

P-W* 

22% 

20% 

16% 

8% 

21% 

10% 

22% 

21% 

18% 

41% 

26% 

S-P* 

4% 

10% 

12% 

5% 

3% 

11% 

S-M* 

5% 

11% 

17% 

4% 

4% 

13% 

S-W* 

8% 

9% 

21% 

5% 

4% 

21% 

Oak 

0% 

1% 

26% 
*NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well stocked, P–P=Poletimber poorly 
stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, S–M=Sawtimber moderately stocked, S–W=Sawtimber well 
stocked. 
 

25% 

28% 

7% 

7% 

1% 

2% 

4% 

 

 

34

Conditions    of    Forest    Types    Enrolled    in    the    Qualified    Forest    Program    

 

    
s
d
n
a
t
s
    
f
o
    
r
e
b
m
u
N

1600 

1400 

1200 

1000 

800 

600 

400 

200 

0 

Aspen / birch group 

Elm / ash / 

Maple / beech / birch 

Other 

Spruce / fir group  White / red / jack pine 

Oak 

cottonwood group 

group 

group 

Forest    Types    

Seedlings/ 
Saplings 

Poletimber  Sawtimber 

Poorly Stocked 

Moderately Stocked 

Well Stocked 

Figure 6: Number of stands within each condition classification, by forest type, enrolled in the Qualified Forest Program.

 

 

35

 

Analysis 

The multinomial logistic regression analysis determines the odds of a particular forest 

management practice occurring relative to some category of an independent variable. The independent 

variables included region, stand condition, forest type, and stand acres. While mapping (Figures 11-24) 

displays forest types and forest practice by county, it is a difficult variable to interpret. For this reason, 

region (EUP, NLP, SLP, and WUP) was evaluated to determine its effect on forest practice. 

Stand acreage significantly increased the odds of a family forest owner selecting each forest 

practices. This analysis only evaluated the multinomial logistic odds with respect to one reference 

category; no practice. The odds of a family forest owner selecting artificial regeneration instead of no 

practice increases 0.9% with each one-acre increase in stand size (Table 2). Because the odds ratio 

describes a linear relationship, the odds of a family forest owner selecting artificial regeneration instead of 

no practice increases 9% with each ten-acre increase in stand size. The geographic region of Michigan 

can be significant in predicting forest practice. The odds of a family forest owner selecting artificial 

regeneration instead of no practice is 2.907 times higher when the stand is in the SLP and 2.093 times 

higher when the stand is in the NLP than for a stand in the WUP (Table 10). The odds of a family forest 

owner selecting artificial regeneration (tree planting) instead of no practice is 1.634 times higher when the 

stand is white spruce/fir compared to when the stand is aspen. The odds of selecting artificial regeneration 

instead of no practice are 1.886 times higher when the stand is white spruce/fir compared to when the 

stand is cedar (Table 2). The odds of selecting artificial regeneration instead of no practice are 3.496 

times higher when the stand is mixed deciduous than when the stand is white spruce/fir. The odds of 

selecting artificial regeneration instead of no practice are 3.189 times higher when the stand is red pine 

than when the stand is white spruce/fir. The odds of selecting artificial regeneration instead of no practice 

is 9.635 times higher when the stand is upland mixed conifers than when the stand is white spruce/fir 

(Table 2). No stand conditions were significant in predicting artificial regeneration instead of no practice 

in the EUP (Table 13). In the NLP, stand condition 1 is significant in predicting artificial regeneration 

instead of no practice compared to stand condition 9 (Table 17). In the SLP, no stand condition is 

 

36

 

significant in predicting artificial regeneration instead of no practice when compared to stand condition 9 

(Table 21). In the WUP, stand conditions 3 and 6 are significant in predicting artificial regeneration 

instead of no practice when compared to stand condition 9 (Table 25). 

The odds of a selecting clearcut instead of no practice increases 0.8% with each one-acre increase 

in stand size (Table 3). There is no instance where region is significant in a family forest owner selecting 

clearcut instead of no practice (Table 10). The odds of a family forest owner selecting clearcut instead of 

no practice is 1.714 times higher when the stand is aspen compared to when the stand is white spruce/fir 

(Table 3). The odds of selecting clearcut instead of no practice are 1.81 times higher when the stand is 

white spruce/fir than when the stand is northern hardwoods. The odds of selecting clearcut are 1.671 

times higher when the stand is white spruce/fir than when the stand is oak. The odds of selecting clearcut 

instead of no practice are 1.738 times higher when the stand is white spruce/fir than when the stand is red 

pine (Table 3). Stand conditions SS-P, SS-M, SS-W, P-P, and S-P are significant in predicting the 

selection of clearcut instead of no practice in the region when compared to stand condition 9 (Table 13). 

Stand conditions SS-P through S-M are significant, when compared to stand condition S-W, in predicting 

clearcut instead of no practice (Table 17). Stand conditions SS-M, SS-W, and P-P significantly predict 

clearcut instead of no practice when compared to stand condition S-W (Table 21). Stand conditions P-M, 

P-W, and S-M are significant in selecting clearcut instead of no practice when compared to stand 

condition S-W in the region (Table 25). 

The odds of selecting a salvage treatment instead of no practice increases 1.3% with each one-

acre increase in stand size (Table 4). The odds of selecting a salvage treatment instead of no practice are 

2.166 times higher if a stand is in the EUP and 3.413 times higher in the NLP than a stand in the WUP 

(Table 10). The odds of a family forest owner selecting a salvage harvest instead of no practice is 1.513 

times higher when the stand is white spruce/fir than when the stand is lowland hardwoods and 1.988 

times higher than when the stand is red pine (Table 4). The odds of selecting salvage harvest instead of no 

practice when in a white spruce/fir stand is 1.981 times higher than when in an upland mixed conifer 

stand. The odds of selecting salvage harvest instead of no practice is only 0.409 in a northern hardwoods 

 

37

 

stand compared to a white spruce/fir stand (Table 4). Stand conditions SS-W and P-M are significant in 

predicting the selecting of a salvage treatment instead of no practice in the region when compared to stand 

condition S-W (Table 14). Stand conditions SS-P through S-M are significant, when compared to stand 

condition S-W, in predicting salvage treatment instead of no practice (Table 18). Stand condition P-M 

significantly predicts a salvage treatment instead of no practice when compared to stand condition S-W 

(Table 22). No stand condition is significant in selecting a salvage treatment instead of no practice when 

compared to stand condition S-W (Table 26). 

The odds of selecting a seed tree harvest instead of no practice increases 0.6% with each one-acre 

increase in stand size (Table 5). The odds of selecting a seed tree harvest instead of no practice is 1.347 

times higher for a stand in the WUP than a stand in the EUP, 1.367 times higher than a stand in the NLP, 

and 1.533 times higher than a stand in the SLP (Table 11). The seed tree harvest practice had few 

significant results. The odds of a family forest owner selecting a seed tree harvest instead of no practice is 

9.685 times higher in a lowland poplar stand than in a white spruce/fir stand, and 18.985 times higher 

when the stand is mixed lowland conifers compared to a white spruce/fir stand. Oak is 9.141 times higher 

to have a forest practice of seed tree harvest than no practice when compared to a white spruce/fir stand. 

An upland mixed stand is 15.944 times higher to have a seed tree harvest instead of no practice than when 

the stand is white spruce/fir (Table 5). No stand conditions were significant in predicting seed tree harvest 

instead of no practice compared to stand condition S-W in the EUP (Table 14). Stand conditions SS-W 

and P-P, when compared to stand condition S-W, are significant in predicting a seed tree harvest instead 

of no practice (Table 18). No stand condition is significant in predicting a seed tree harvest instead of no 

practice when compared to stand condition S-W (Table 22). Stand conditions SS-P through P-P are 

significant in predicting a seed tree harvest instead of no practice when compared to stand condition S-W 

(Table 26). 

The odds of selecting a shelterwood harvest instead of no practice increases 0.8% with each one-

acre increase in stand size (Table 6). There is no instance where region is significant in a family forest 

owner selecting a shelterwood harvest instead of no practice (Table 11). The odds of selecting a 

 

38

 

shelterwood harvest instead of no practice are 3.856 times higher in a hemlock stand compared to a white 

spruce/fir stand (Table 6). An oak stand is 5.488 times higher to have a forest practice of a shelterwood 

harvest instead of no practice when compared to a white spruce/fir stand. Paper birch is 7.024 times 

higher than white spruce/fir of having a shelterwood harvest instead of no practice (Table 6). Stand 

conditions SS-M, SS-W, P-P, and P-W are significant in predicting a shelterwood harvest instead of no 

practice when compared to stand condition S-W (Table 15). Stand conditions SS-P through P-W are 

significant, when compared to stand condition S-W, in predicting a shelterwood harvest instead of no 

practice (Table 19). Stand conditions SS-M through P-W are significant, when compared to stand 

condition S-W, in predicting a shelterwood harvest instead of no practice in the SLP (Table 23). When 

compared to stand condition S-W, stand conditions SS-M, SS-W, and P-P significantly predict a 

shelterwood harvest instead of no practice (Table 27). 

The odds of a family forest owner selecting a selection harvest instead of no practice increases 

0.9% with each one-acre increase in stand size (Table 7). The odds of a family forest owner selecting 

selection harvest instead of no practice is 1.152 times higher and 1.264 times higher in the in the WUP 

than stands in the EUP and the NLP, respectively (Table 12). The odds of selecting selection harvest 

instead of no practice are 1.553 times higher in the SLP than the WUP (Table 12). The odds of a family 

forest owner selecting a selection harvest instead of no practice are 1.430 times higher when the stand is 

white spruce/fir compared to aspen (Table 7). A mixed deciduous stand is 7.144 times higher to have a 

selection harvest instead of no practice than a white spruce/fir stand. The northern hardwood type is 8.818 

times more likely to have a selection harvest instead of no practice than a white spruce/fir stand. The odds 

of selecting a selection harvest instead of no practice are 4.191 times higher in an oak stand than in a 

white spruce/fir stand (Table 7). Stand conditions SS-P, SS-M, SS-W, P-P, P-M, and P-W are significant 

in predicting a selection harvest instead of no practice when compared to stand condition S-W (Table 15). 

Stand conditions SS-P through S-M are significant, when compared to stand condition S-W, in predicting 

a selection harvest instead of no practice (Table 19). Stand conditions NS and SS-M through S-P are 

significant, when compared to stand condition S-W, in predicting a selection harvest instead of no 

 

39

 

practice in the region (Table 23). Stand condition SS-M through S-P significantly predict a selection 

harvest instead of no practice when compared to stand condition S-W (Table 27). 

The odds of selecting thinning instead of no practice increases by 0.8% with each one-acre 

increase in stand size (Table 8). The odds of selecting thinning instead of no practice is 3.166 times higher 

when the stand is in the NLP and 2.610 times higher in the SLP than when the stand is in the WUP (Table 

12). A white spruce/fir stand is 1.644 times higher of having a thinning practice instead of no practice 

compared to an aspen stand (Table 8). The odds of a family forest owner selecting thinning instead of no 

practice is 3.458 times higher if the stand is northern hardwoods than if the stand is white spruce/fir. An 

oak stand is 3.622 times higher of having a thinning instead of no practice than when the stand is white 

spruce/fir. Red pine is 2.235 times higher than white spruce/fir of having a thinning instead of no practice 

(Table 8). Stand conditions SS-M, SS-W, P-P, P-M, P-W, and S-M are significant in predicting thinning 

instead of no practice when compared to stand condition S-W (Table 16). Stand conditions SS-P through 

P-W are significant, when compared to stand condition S-W, in predicting a thinning instead of no 

practice (Table 20). Stand conditions SS-P through P-M and S-P are significant, when compared to stand 

condition S-W, in predicting a thinning instead of no practice (Table 24). When compared to stand 

condition S-W, stand conditions SS-M through P-M significantly predict thinning instead of no practice 

(Table 28). 

The odds of selecting selection harvest instead of thinning increases 0.3% with each one-acre 

increase in stand size (Table 7). It should be noted that stand acres might not be significant in predicting 

every forest practice if different reference categories were chosen. The odds of a family forest owner 

selecting timber stand improvement instead of no practice is 3.916 when the stand is in the NLP and 

10.914 times higher in the SLP than when the stand is in the WUP (Table 12). A family forest owner is 

42.678 times more likely to select timber stand improvement instead of no practice when the stand is 

mixed deciduous and 11.963 times more likely when the stand is northern hardwoods than when the stand 

is white spruce/fir (Table 9). Oak is 14.128 times more likely to have a timber stand improvement instead 

of no practice than when the stand is white spruce/fir. When the stand is red pine, it is 11.443 times more 

 

40

 

likely than white spruce/fir to select timber stand improvement instead of no practice (Table 9). Stand 

conditions SS-P, SS-W, P-P, P-M, P-W, and S-P are significant in predicting timber stand improvement 

instead of no practice when compared to stand condition S-W in the EUP (Table 16). Stand conditions 

SS-P through S-M are significant, when compared to stand condition S-W, in predicting a timber stand 

improvement instead of no practice in the NLP (Table 20). Stand conditions SS-M through S-P are 

significant, when compared to stand condition S-W, in predicting a timber stand improvement instead of 

no practice in the SLP (Table 24). Stand conditions SS-W, P-P, and S-M significantly predict timber 

stand improvement instead of no practice when compared to stand condition S-W (Table 28). 

Many stand types were identified as significant in predicting forest practice. Because the number 

of potential combinations of stand types and forest practices is quite high, only a few significant 

relationships are discussed. The complete set of results from this analysis is provided in the Appendix 

(Table 2 through Table 28 and A.01 through A.27). Statewide, stand condition did not statistically 

significantly relate to any forest practice measured. However, when analyzed at the regional level, stand 

condition matters in certain scenarios as described above. 

 

 

41

 

Table 2: Multinomial logistic regression parameter estimates for significant results. Stand acres and stand type as independent variables 
(showing results from artificial regeneration). N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Exp(β) 

  

  

Upper Bound 
  

1.009 
0.366 
0.114 
3.974 

0.050 
3.496 

1.005 
0.144 
0.018 
1.235 

0.001 
1.157 

1.012 
0.933 
0.703 
12.789 

4.574 
10.564 

Sig. 
0.000 
0.000 
0.035 
0.019 
0.021 

0.194 
0.027 

0.041 

0.296 

0.092 

0.951 

0.010 
0.000 

3.189 
9.635 

1.315 
3.856 

7.735 
24.075 

  

 

  

 

  

 

  

 

df 
1 
1 
1 
1 
1 

1 
1 

1 

1 
1 

0 

 

 

Forest Practicea 
Artificial 
Regeneration 
 
 

 

 

 

 

 

 

 

Intercept 
Stand Acres* 
Stand Type=Aspen* 
Stand Type=Cedar* 
Stand Type=Lowland Poplar 
(Bam)* 
Stand Type=Marsh* 
Stand Type=Mixed 
Deciduous* 
Stand Type=Mixed Lowland 
Conifers* 
Stand Type=Red Pine* 
Stand Type=Upland Mixed 
Conifers* 
Stand Type=White 
Spruce/Fir 

B 

-3.964 
0.009 
-1.005 
-2.173 
1.380 

Std. 
Error  Wald 
0.417 
0.002 
0.477 
0.929 
0.596 

90.405 
23.103 
4.437 
5.476 
5.353 

-2.989 
1.252 

2.301 
0.564 

1.688 
4.920 

-1.217 

0.596 

4.177 

1.160 
2.265 

0.452 
0.467 

6.583 
23.506 

0b 

  

a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 

 

 

 

 

  
 

 

  

 

42

 

Table 3: Multinomial logistic regression parameter estimates for significant results. Stand acres and stand type as independent variables 
(showing results from clearcut). N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

Forest Practicea 
Clearcut 

Intercept 
Stand Acres* 
Stand Type=Aspen* 
Stand Type=Bog or Muskeg* 
Stand Type=Cedar* 
Stand Type=Grass* 
Stand Type=Hemlock* 
Stand Type=Local Use (various non-
commercial or exotic)* 
Stand Type=Lowland Brush* 
Stand Type=Lowland Hardwoods* 
Stand Type=Lowland Mixed* 
Stand Type=Marsh* 
Stand Type=Mixed Deciduous* 
Stand Type=Mixed Lowland Conifers* 
Stand Type=Non-Stocked* 
Stand Type=Northern Hardwoods* 
Stand Type=Oak* 
Stand Type=Red Pine* 
Stand Type=Rock* 
Stand Type=Tamarack* 
Stand Type=Treed Bog* 
Stand Type=Upland Brush* 
Stand Type=Water* 
Stand Type=White Pine* 
Stand Type=White Spruce/Fir 
a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 

B 

-1.171 
0.008 
0.539 
-3.327 
-1.757 
-3.324 
-1.802 
-3.091 

-2.916 
-0.482 
-0.487 
-3.402 
-0.780 
-0.798 
-3.179 
-1.659 
-1.111 
-1.341 
-3.353 
-0.726 
-3.379 
-3.262 
-3.407 
-1.634 
0b 

 

Std. 
Error  Wald 
0.119 
96.608 
0.001  120.532 
18.938 
0.124 
0.493 
45.622 
0.211 
69.462 
0.255  169.498 
14.510 
0.473 
0.862 
12.854 

85.548 
9.467 
5.012 
23.317 
6.895 
27.889 
72.354 
95.482 
38.536 
47.275 
0.475 
7.719 
4.525 
15.223 
39.602 
37.768 
  

0.315 
0.157 
0.218 
0.704 
0.297 
0.151 
0.374 
0.170 
0.179 
0.195 
4.865 
0.261 
1.589 
0.836 
0.541 
0.266 
  

43

95% Confidence 
Interval for Exp(β) 
Upper 
Lower 
Bound 
Bound 

  

Exp(β) 
  

  

1.008 
1.714 
0.036 
0.173 
0.036 
0.165 
0.045 

1.007 
1.345 
0.014 
0.114 
0.022 
0.065 
0.008 

0.054 
0.029 
0.617 
0.454 
0.614 
0.401 
0.033 
0.008 
0.458 
0.256 
0.450 
0.335 
0.042 
0.020 
0.190 
0.136 
0.329 
0.232 
0.262 
0.178 
0.035  2.527E-06 
0.290 
0.484 
0.034 
0.002 
0.007 
0.038 
0.011 
0.033 
0.116 
0.195 
  
  

1.010 
2.186 
0.094 
0.261 
0.059 
0.417 
0.246 

0.100 
0.839 
0.941 
0.133 
0.820 
0.605 
0.087 
0.265 
0.468 
0.383 
484.026 
0.807 
0.767 
0.197 
0.096 
0.329 
  

Sig. 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 

0.000 
0.002 
0.025 
0.000 
0.009 
0.000 
0.000 
0.000 
0.000 
0.000 
0.491 
0.005 
0.033 
0.000 
0.000 
0.000 
  

df 
1 
1 
1 
1 
1 
1 
1 
1 

1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 

 

Table 4: Multinomial logistic regression parameter estimates for significant results. Stand acres and stand type as independent variables 
(showing results from salvage treatment). N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

Forest Practicea 
Salvage 
Treatment 

Intercept 
Stand Acres* 
Stand Type=Aspen* 
Stand Type=Black Spruce* 
Stand Type=Bog or Muskeg* 
Stand Type=Cedar* 
Stand Type=Grass* 
Stand Type=Lowland Brush* 
Stand Type=Lowland Hardwoods* 
Stand Type=Lowland Mixed* 
Stand Type=Marsh* 
Stand Type=Mixed Lowland 
Conifers* 
Stand Type=Non-Stocked* 
Stand Type=Northern Hardwoods* 
Stand Type=Oak* 
Stand Type=Red Pine* 
Stand Type=Tamarack* 
Stand Type=Upland Brush* 
Stand Type=Upland Mixed* 
Stand Type=Upland Mixed 
Conifers* 
Stand Type=Water* 
Stand Type=White Pine* 
Stand Type=White Spruce/Fir 

B 

-2.421 
0.013 
-3.289 
-1.999 
-4.502 
-2.080 
-4.515 
-2.460 
-0.719 
-1.279 
-4.566 
-1.555 

-4.534 
-0.894 
-0.988 
-4.452 
-4.152 
-1.780 
-3.098 
-3.939 

-4.460 
-1.469 
0b 

a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 

 

df 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 

1 
1 
1 
1 
1 
1 
1 
1 

1 
1 
0 

Sig. 
0.000 
0.000 
0.000 
0.001 
0.003 
0.000 
0.000 
0.000 
0.008 
0.005 
0.038 
0.000 

0.000 
0.000 
0.000 
0.000 
0.050 
0.018 
0.008 
0.040 

0.005 
0.001 
  

Std. 
Error 

Wald 

0.194  155.655 
0.001  135.277 
81.339 
0.365 
0.609 
10.797 
9.002 
1.501 
28.998 
0.386 
34.220 
0.772 
0.456 
29.101 
7.089 
0.270 
7.886 
0.455 
2.199 
4.312 
28.532 
0.291 

12.745 
13.868 
12.181 
12.173 
3.848 
5.642 
7.098 
4.225 

7.980 
11.749 
  

1.270 
0.240 
0.283 
1.276 
2.116 
0.749 
1.163 
1.917 

1.579 
0.428 
  

44

95% Confidence Interval for 

Exp(β) 

Lower 
Bound 

Exp(β) 
  

  

Upper Bound 

  

1.013 
0.037 
0.135 
0.011 
0.125 
0.011 
0.085 
0.487 
0.278 
0.010 
0.211 

0.011 
0.409 
0.372 
0.012 
0.016 
0.169 
0.045 
0.019 

0.012 
0.230 
  

1.011 
0.018 
0.041 
0.001 
0.059 
0.002 
0.035 
0.287 
0.114 
0.000 
0.119 

0.001 
0.256 
0.214 
0.001 
0.000 
0.039 
0.005 
0.000 

0.001 
0.099 
  

1.016 
0.076 
0.446 
0.210 
0.266 
0.050 
0.209 
0.827 
0.680 
0.774 
0.374 

0.129 
0.655 
0.649 
0.142 
0.996 
0.733 
0.441 
0.833 

0.255 
0.533 
  

 

Table 5: Multinomial logistic regression parameter estimates for significant results. Stand acres and stand type as independent variables 
(showing results from seed tree harvest). N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

B 
-5.027 
0.006 
2.271 

Forest Practicea 
Seed 
Tree 
Harvest 

Intercept 
Stand Acres* 
Stand Type=Lowland Poplar 
(Bam)* 
Stand Type=Mixed Lowland 
Conifers* 
Stand Type=Oak* 
Stand Type=Tamarack* 
Stand Type=Upland Mixed* 
Stand Type=White Spruce/Fir 
a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 

2.944 

2.213 
1.870 
2.769 
0b 

95% Confidence Interval 

for Exp(β) 

Std. 
Error  Wald 
0.717 
0.002 
0.858 

49.178 
12.008 
7.004 

0.724 

16.541 

0.738 
0.831 
0.779 
  

8.985 
5.059 
12.629 
  

df 
1 
1 
1 

Sig. 
0.000 
0.001 
0.008 

Lower 
Bound 

Exp(β) 

  

  

Upper Bound 
  

1.006 
9.685 

1.003 
1.802 

1.009 
52.048 

1 

1 
1 
1 
0 

0.000 

18.985 

4.596 

78.433 

0.003 
0.025 
0.000 
  

9.141 
6.487 
15.944 
  

2.151 
1.272 
3.462 
  

38.845 
33.089 
73.428 
  

 

 
 

 

45

 

Table 6: Multinomial logistic regression parameter estimates for significant results. Stand acres and stand type as independent variables 
(showing results from shelterwood harvest). N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

Forest Practicea 
Shelterwood 
Harvest 

Intercept 
Stand Acres* 
Stand Type=Aspen* 
Stand Type=Bog or 
Muskeg* 
Stand Type=Grass* 
Stand Type=Hemlock* 
Stand Type=Lowland 
Brush* 
Stand Type=Non-Stocked* 
Stand Type=Oak* 
Stand Type=Paper Birch* 
Stand Type=Water* 
Stand Type=White 
Spruce/Fir 

B 
-3.461 
0.008 
-1.546 
-3.221 

-3.130 
1.350 
-3.131 

-3.157 
1.703 
1.949 
-3.191 
0b 

Std. 
Error 
0.327 
0.001 
0.404 
1.450 

0.719 
0.449 
1.063 

1.141 
0.344 
0.548 
1.515 
  

Wald 
112.123 
35.093 
14.678 
4.931 

df 
1 
1 
1 
1 

18.947 
9.040 
8.677 

7.655 
24.494 
12.665 
4.435 
  

1 
1 
1 

1 
1 
1 
1 
0 

Sig. 
0.000 
0.000 
0.000 
0.026 

0.000 
0.003 
0.003 

0.006 
0.000 
0.000 
0.035 
  

95% Confidence Interval for 

Exp(β) 

Exp(β) 

Lower 
Bound 

  

  

  

Upper Bound 

1.008 
0.213 
0.040 

0.044 
3.856 
0.044 

0.043 
5.488 
7.024 
0.041 
  

1.006 
0.097 
0.002 

0.011 
1.600 
0.005 

0.005 
2.796 
2.401 
0.002 
  

1.011 
0.470 
0.685 

0.179 
9.295 
0.351 

0.398 
10.770 
20.553 
0.802 
  

a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 

 

 
 

 

46

 

Table 7: Multinomial logistic regression parameter estimates for significant results. Stand acres and stand type as independent variables 
(showing results from selection harvest). N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Exp(β) 

  

  

  

1.009 
0.570 
0.099 
0.111 
4.356 
0.119 
2.543 

1.007 
0.397 
0.039 
0.067 
2.722 
0.060 
1.762 

1.010 
0.817 
0.251 
0.185 
6.973 
0.234 
3.671 

Sig. 
0.000 
0.000 
0.002 
0.000 
0.000 
0.000 
0.000 
0.000 

0.000 

3.520 

2.093 

5.919 

0.001 
0.000 
0.000 
0.000 

0.000 
0.000 
0.036 
0.004 
0.000 
0.000 
0.003 
  

0.185 
7.144 
0.106 
8.818 

4.191 
4.498 
0.435 
0.119 
3.142 
0.103 
1.881 
  

0.071 
4.631 
0.050 
6.278 

2.931 
3.146 
0.200 
0.028 
1.959 
0.039 
1.242 
  

0.480 
11.023 
0.224 
12.385 

5.994 
6.430 
0.945 
0.510 
5.038 
0.269 
2.848 
  

df 
1 
1 
1 
1 
1 
1 
1 
1 

1 

1 
1 
1 
1 

1 
1 
1 
1 
1 
1 
1 
0 

Forest Practicea 
Selection 
Harvest 

Intercept 
Stand Acres* 
Stand Type=Aspen* 
Stand Type=Bog or Muskeg* 
Stand Type=Grass* 
Stand Type=Hemlock* 
Stand Type=Lowland Brush* 
Stand Type=Lowland 
Hardwoods* 
Stand Type=Lowland Poplar 
(Bam)* 
Stand Type=Marsh* 
Stand Type=Mixed Deciduous* 
Stand Type=Non-Stocked* 
Stand Type=Northern 
Hardwoods* 
Stand Type=Oak* 
Stand Type=Red Pine* 
Stand Type=Tamarack* 
Stand Type=Upland Brush* 
Stand Type=Upland Mixed* 
Stand Type=Water* 
Stand Type=White Pine* 
Stand Type=White Spruce/Fir 
a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 

-1.687  0.486 
1.966  0.221 
-2.243  0.380 
2.177  0.173 

1.433  0.183 
1.504  0.182 
-0.833  0.396 
-2.130  0.743 
1.145  0.241 
-2.277  0.491 
0.632  0.212 
  

0b 

 
 

 

 

Std. 
Error  Wald 

B 
-2.034  0.168 
0.009  0.001 
-0.562  0.184 
-2.308  0.472 
-2.198  0.260 
1.472  0.240 
-2.131  0.346 
0.934  0.187 

1.258  0.265 

145.961 
155.838 
9.366 
23.904 
71.526 
37.603 
37.911 
24.865 

22.522 

12.060 
78.985 
34.792 
157.743 

61.658 
68.001 
4.418 
8.219 
22.584 
21.474 
8.896 
  

47

 

Table 8: Multinomial logistic regression parameter estimates for significant results. Stand acres and stand type as independent variables 
(showing results from thinning). N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

Forest Practicea 
Thinning 

Intercept 
Stand Acres* 
Stand Type=Aspen* 
Stand Type=Bog or Muskeg* 
Stand Type=Cedar* 
Stand Type=Grass* 
Stand Type=Lowland Brush* 
Stand Type=Lowland Hardwoods* 
Stand Type=Lowland Poplar (Bam)* 
Stand Type=Marsh* 
Stand Type=Mixed Deciduous* 
Stand Type=Mixed Lowland 
Conifers* 
Stand Type=Non-Stocked* 
Stand Type=Northern Hardwoods* 
Stand Type=Oak* 
Stand Type=Red Pine* 
Stand Type=Tamarack* 
Stand Type=Upland Brush* 
Stand Type=Upland Mixed 
Conifers* 
Stand Type=Water* 
Stand Type=White Pine* 
Stand Type=White Spruce/Fir 

B 
-2.006 
0.008 
-1.033 
-2.724 
-1.487 
-2.617 
-2.509 
-0.521 
1.772 
-2.693 
1.340 
-1.329 

-2.668 
1.241 
1.287 
0.804 
-1.016 
-2.558 
0.822 

-2.697 
1.072 
0b 

Std. 
Error  Wald 

df  Sig.  Exp(β) 

0.167  143.627  1  0.000   
86.661  1  0.000 
0.001 
29.300  1  0.000 
0.191 
0.564 
23.313  1  0.000 
27.458  1  0.000 
0.284 
79.551  1  0.000 
0.293 
0.399 
39.600  1  0.000 
5.372  1  0.020 
0.225 
53.512  1  0.000 
0.242 
12.512  1  0.000 
0.761 
0.238 
31.609  1  0.000 
30.201  1  0.000 
0.242 

0.449 
0.176 
0.183 
0.189 
0.423 
0.900 
0.242 

0.588 
0.199 
  

35.263  1  0.000 
49.943  1  0.000 
49.520  1  0.000 
18.052  1  0.000 
5.753  1  0.016 
8.073  1  0.004 
11.559  1  0.001 

21.007  1  0.000 
28.872  1  0.000 
  

   0 

  
1.008 
0.356 
0.066 
0.226 
0.073 
0.081 
0.594 
5.882 
0.068 
3.819 
0.265 

0.069 
3.458 
3.622 
2.235 
0.362 
0.077 
2.275 

0.067 
2.920 
  

95% Confidence 
Interval for Exp(β) 
Upper 
Lower 
Bound 
Bound 

  
1.006 
0.245 
0.022 
0.130 
0.041 
0.037 
0.383 
3.659 
0.015 
2.394 
0.165 

0.029 
2.451 
2.531 
1.542 
0.158 
0.013 
1.416 

0.021 
1.975 
  

1.009 
0.517 
0.198 
0.394 
0.130 
0.178 
0.923 
9.456 
0.301 
6.093 
0.425 

0.167 
4.879 
5.184 
3.239 
0.831 
0.452 
3.653 

0.214 
4.317 
  

a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 

 

48

 

Table 9: Multinomial logistic regression parameter estimates for significant results. Stand acres and stand type as independent variables 
(showing results from timber stand improvement). N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

95% Confidence Interval for 

Exp(β) 

Lower 
Bound 

Exp(β) 

  

  

Upper Bound 

  

Sig. 
0.000 
0.001 
0.017 
0.009 
0.000 

1.005 
0.199 
5.267 
42.678 

1.002 
0.053 
1.510 
15.246 

1.007 
0.748 
18.365 
119.470 

32.169 

38.345 
31.184 
66.407 
13.550 

  

0.000 

11.963 

4.448 

0.000 
0.000 
0.000 
0.026 

14.128 
11.443 
23.261 
4.008 

5.205 
4.199 
8.148 
1.185 

  

  

  

Forest Practicea 
Timber 
Stand 
Improvement 

Intercept 
Stand Acres* 
Stand Type=Grass* 
Stand Type=Hemlock* 
Stand Type=Mixed 
Deciduous* 
Stand Type=Northern 
Hardwoods* 
Stand Type=Oak* 
Stand Type=Red Pine* 
Stand Type=Upland Mixed* 
Stand Type=Upland Mixed 
Conifers* 
Stand Type=White 
Spruce/Fir 

B 

-4.272 
0.005 
-1.615 
1.661 
3.754 

Std. 
Error  Wald 
73.510 
0.498 
10.984 
0.001 
5.708 
0.676 
0.637 
6.797 
51.081 
0.525 

2.482 

0.505 

24.179 

2.648 
2.437 
3.147 
1.388 

0.509 
0.512 
0.535 
0.622 

27.022 
22.704 
34.568 
4.989 

0b 

  

  

df 
1 
1 
1 
1 
1 

1 

1 
1 
1 
1 

0 

a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 

 

49

 

 

 

Table 10: Multinomial logistic regression parameter estimates for significant results. Geographic region of Michigan as independent 
variable. N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

Forest Practicea 
Artificial  
Regeneration 

Clearcut 

Salvage  
Treatment 

Intercept 
Region=NLP* 
Region=SLP* 
Region=WUP 
Intercept 
Region=WUP 
Intercept 
Region=EUP* 
Region=NLP* 
Region=WUP 

β 

-4.480 
0.739 
1.067 
0b 
-1.612 
0b 
-4.868 
0.773 
1.228 
0b 

Std. Error 
0.190 
0.211 
0.241 
  
0.049 
  
0.230 
0.275 
0.246 
  

Wald 

df 

555.718 
12.257 
19.596 
  
1067.827 
  
446.804 
7.898 
24.854 
  

95% Confidence Interval 

for Exp(β) 

  

  

  

Sig. 
0.000 
0.000 
0.000 
  
0.000 
  
0.000 
0.005 
0.000 
  

1 
1 
1 
0 
1 
0 
1 
1 
1 
0 

Exp(β) 

Lower Bound 
  

2.093 
2.907 
  

  

  

  

2.166 
3.413 
  

1.384 
1.812 
  

  

1.263 
2.106 
  

  

  

  

Upper 
Bound 

3.165 
4.662 
  

  

3.713 
5.530 
  

a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

50

 

 

Table 11: Multinomial logistic regression parameter estimates for significant results. Geographic region of Michigan as independent 
variable. N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

β 

-3.669 
-0.426 
-0.458 
-0.761 
0b 
-3.608 
0b 

Std. Error 
0.128 
0.197 
0.169 
0.275 
  
0.124 
  

Forest Practicea 
Seed Tree  
Harvest 

Intercept 
Region=EUP* 
Region=NLP* 
Region=SLP* 
Region=WUP 
Intercept 
Region=WUP 

Shelterwood 
Harvest 
a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

95% Confidence Interval 

for Exp(β) 

Exp(β) 

Lower Bound 
  

  

Upper 
Bound 

0.653 
0.633 
0.467 
  

  

  

0.444 
0.455 
0.272 
  

  

  

0.961 
0.881 
0.802 
  

  

Sig. 
0.000 
0.031 
0.007 
0.006 
  
0.000 
  

  

  

1 
1 
1 
1 
0 
1 
0 

Wald 

df 

827.103 
4.665 
7.339 
7.637 
  
849.011 
  

51

 

 

Table 12: Multinomial logistic regression parameter estimates for significant results. Geographic region of Michigan as independent 
variable. N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

Forest Practicea 
Selection 
Harvest 

Thinning 

Timber Stand 
Improvement 

Intercept 
Region=EUP* 
Region=NLP* 
Region=SLP* 
Region=WUP 
Intercept 
Region=NLP* 
Region=SLP* 
Region=WUP 
Intercept 
Region=NLP* 
Region=SLP* 
Region=WUP 

β 

-1.225 
-0.165 
-0.307 
0.440 
0b 
-2.499 
1.153 
0.959 
0b 
-4.316 
1.365 
2.390 
0b 

Std. Error 
0.042 
0.060 
0.054 
0.063 
  
0.073 
0.079 
0.096 
  
0.175 
0.186 
0.190 
  

a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 
 
 
 
 
 
 
 
 
 
 
 
 
 

95% Confidence Interval 

for Exp(β) 

  

  

  

Sig. 
0.000 
0.006 
0.000 
0.000 
  
0.000 
0.000 
0.000 
  
0.000 
0.000 
0.000 
  

1 
1 
1 
1 
0 
1 
1 
1 
0 
1 
1 
1 
0 

Exp(β) 

Lower Bound 
  

  

0.848 
0.736 
1.553 
  

3.166 
2.610 
  

3.916 
10.914 
  

  

  

0.753 
0.662 
1.371 
  

2.711 
2.161 
  

2.719 
7.516 
  

  

  

Upper 
Bound 

0.954 
0.817 
1.758 
  

3.698 
3.153 
  

5.639 
15.848 
  

Wald 

df 

841.816 
7.499 
32.740 
48.173 
  
1171.655 
211.762 
98.983 
  
606.582 
53.820 
157.733 
  

52

 

 
 
 
 
 
 
 
 
 
 
 

 

Table 13: Multinomial logistic regression parameter estimates for significant results. Stand condition as independent variable. Eastern 
Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program.† 

β 

-4.407 
0c 

Std. 
Error 

1.006 
  

Wald 
19.185 
  

df 

1 
0 

Sig. 
0.000 
  

  

Exp(β) 

  

  

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

  

  

  

Forest Practiceb 
Artificial 
Regen-
eration 
Clearcut 

Intercept 
S-W 

0.202 
0.751 
0.466 
0.245 
0.255 
0.407 
  

-0.851 
-2.217 
-2.030 
-1.283 
-0.587 
-1.214 
0c 

17.780 
8.712 
19.004 
27.395 
5.307 
8.882 
  

Intercept 
SS-P* 
SS-M* 
SS-W* 
P-P* 
S-P* 
S-W 
a. Region = EUP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
d. Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing. 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–
W=Seedlings/Saplings well stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber 
well stocked, S–P=Sawtimber poorly stocked, S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 

0.000 
0.003 
0.000 
0.000 
0.021 
0.003 
  

0.109 
0.131 
0.277 
0.556 
0.297 
  

1 
1 
1 
1 
1 
1 
0 

  

  

  

0.025 
0.053 
0.172 
0.337 
0.134 
  

0.475 
0.327 
0.448 
0.916 
0.660 
  

53

 

Table 14: Multinomial logistic regression parameter estimates for significant results. Stand condition as independent variable. Eastern 
Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program.† 

β 

Std. Error 

0.512 
0.874 
0.875 
  
0.588 
  

Wald 
34.795 
8.053 
5.347 
  
31.672 
  

df 
1 
1 
1 
0 
1 
0 

Sig. 
0.000 
0.005 
0.021 
  
0.000 
  

95% Confidence Interval for 

Exp(β) 

Exp(β) 

  

  

Lower 
Bound 

0.084 
0.132 
  

0.015 
0.024 
  

Upper Bound 
  

0.464 
0.735 
  

Forest Practiceb 
Salvage 
Treatment 

Intercept 
SS-W* 
P-M* 
S-W 
Intercept 
Stand Condition=9 

-3.020 
-2.481 
-2.023 
0c 
-3.308 
0c 

  

Seed Tree 
Harvest 
a. Region = EUP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
d. Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing. 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–
W=Seedlings/Saplings well stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well 
stocked, S–P=Sawtimber poorly stocked, S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 
 

  

  

  

  

  

 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

54

 

Table 15: Multinomial logistic regression parameter estimates for significant results. Stand condition as independent variable. Eastern 
Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program.† 

95% Confidence Interval for 

Exp(β) 

Forest Practiceb 
Shelterwood 
Harvest 

Selection 
Harvest 

Intercept 
SS-M* 
SS-W* 
P-P* 
P-W* 
S-W 
Intercept 
SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-W 

β 

Std. Error 

Wald 

-2.104 
-2.569 
-3.397 
-1.187 
-0.804 
0c 
0.389 
-4.150 
-4.369 
-4.186 
-2.926 
-1.582 
-1.092 
0c 

0.335 
1.059 
0.784 
0.492 
0.387 
  
0.143 
1.022 
0.728 
0.337 
0.290 
0.185 
0.162 
  

39.461 
5.883 
18.788 
5.826 
4.313 
  
7.399 
16.503 
36.025 
154.476 
102.037 
72.948 
45.263 
  

df 
1 
1 
1 
1 
1 
0 
1 
1 
1 
1 
1 
1 
1 
0 

  

  

Sig. 
0.000 
0.015 
0.000 
0.016 
0.038 
  
0.007 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
  

Exp(β) 

Lower 
Bound 

  

  

0.077 
0.033 
0.305 
0.448 
  

0.016 
0.013 
0.015 
0.054 
0.205 
0.336 
  

Upper Bound 
  

  

0.010 
0.007 
0.116 
0.210 
  

0.002 
0.003 
0.008 
0.030 
0.143 
0.244 
  

0.611 
0.156 
0.800 
0.956 
  

0.117 
0.053 
0.029 
0.095 
0.295 
0.461 
  

a. Region = EUP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
d. Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing. 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings 
well stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber 
poorly stocked, S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 
 

 
 
 
 
 

 

55

 

Table 16: Multinomial logistic regression parameter estimates for significant results. Stand condition as independent variable. Eastern 
Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program.† 

95% Confidence Interval for 

Exp(β) 

β 

Std. Error  Wald 

Exp(β) 

Lower 
Bound 

Upper Bound 
  

0.003 
0.028 
0.063 
0.206 
0.308 
0.187 
  

0.013 
0.025 
0.003 
0.040 
0.042 
0.052 
  

  

0.179 
0.119 
0.272 
0.584 
0.777 
0.809 
  

0.775 
0.151 
0.152 
0.240 
0.202 
0.642 
  

Forest Practiceb 
Thinning 

Timber 
Stand 
Improve-
ment 

Intercept 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-M* 
S-W 
Intercept 
SS-P* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-W 
a. Region = EUP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
d. Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing. 

20.381 
13.228 
59.782 
29.697 
15.879 
9.178 
6.385 
  
32.984 
4.857 
37.015 
14.283 
25.705 
35.925 
7.025 
  

-0.941 
-3.732 
-2.856 
-2.032 
-1.058 
-0.714 
-0.944 
0c 
-1.462 
-2.299 
-2.786 
-3.908 
-2.328 
-2.380 
-1.702 
0c 

0.208 
1.026 
0.369 
0.373 
0.265 
0.236 
0.373 
  
0.255 
1.043 
0.458 
1.034 
0.459 
0.397 
0.642 
  

0.024 
0.058 
0.131 
0.347 
0.490 
0.389 
  

0.100 
0.062 
0.020 
0.097 
0.093 
0.182 
  

  

  

df 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 
1 
1 
1 
1 
1 
0 

  

  

Sig. 
0.000 
0.000 
0.000 
0.000 
0.000 
0.002 
0.012 
  
0.000 
0.028 
0.000 
0.000 
0.000 
0.000 
0.008 
  

†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 
 

 

 

56

 

Table 17: Multinomial logistic regression parameter estimates for significant results. Stand condition as independent variable. Northern 
Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program.† 

Forest Practiceb 
Artificial 
Regen-
eration 

Clearcut 

Intercept 
SS-P* 
S-W 
Intercept 
SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-M* 
S-W 
a. Region = NLP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
d. Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing. 

0.000 
0.000 
  
0.000 
0.000 
0.000 
0.000 
0.000 
0.019 
0.000 
0.000 
0.039 
  

Std. Error 
0.326 
0.404 
  
0.100 
0.398 
0.284 
0.152 
0.223 
0.138 
0.120 
0.154 
0.147 
  

Wald 
108.005 
35.567 
  
39.694 
14.368 
38.531 
97.385 
72.193 
5.544 
25.105 
22.310 
4.272 
  

β 
-3.384 
2.409 
0c 
-0.630 
-1.510 
-1.764 
-1.502 
-1.893 
-0.326 
0.601 
-0.727 
-0.305 
0c 

df 
1 
1 
0 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 

Sig. 

Exp(β) 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

  

  

  

  

5.040 
  

0.101 
0.098 
0.165 
0.097 
0.551 
1.442 
0.358 
0.552 
  

24.553 
  

0.482 
0.299 
0.300 
0.233 
0.947 
2.307 
0.654 
0.984 
  

  

  

11.124 
  

0.221 
0.171 
0.223 
0.151 
0.722 
1.824 
0.483 
0.737 
  

†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 
 
 
 
 
 
 
 
 

 

57

 

Table 18: Multinomial logistic regression parameter estimates for significant results. Stand condition as independent variable. Northern 
Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program.† 

Forest Practiceb 
Salvage 
Treatment 

Intercept 
SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-M* 
S-W 
Intercept 
SS-W* 
P-P* 
S-W 
a. Region = NLP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
d. Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing. 

88.881 
9.949 
20.994 
71.538 
38.295 
55.795 
63.143 
44.147 
37.656 
  
117.299 
12.146 
6.328 
  

0.000 
0.002 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
  
0.000 
0.000 
0.012 
  

Std. Error 
0.119 
0.832 
0.728 
0.442 
0.571 
0.346 
0.345 
0.248 
0.259 
  
0.291 
0.564 
0.691 
  

β 
-1.121 
-2.623 
-3.335 
-3.736 
-3.531 
-2.583 
-2.739 
-1.651 
-1.592 
0c 
-3.149 
-1.966 
-1.738 
0c 

df 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 
1 
0 

Seed Tree 
Harvest 

Wald 

Sig. 

Exp(β) 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

  

  

0.014 
0.009 
0.010 
0.010 
0.038 
0.033 
0.118 
0.122 
  

0.046 
0.045 
  

  

0.370 
0.148 
0.057 
0.090 
0.149 
0.127 
0.312 
0.338 
  

0.423 
0.681 
  

  

0.073 
0.036 
0.024 
0.029 
0.076 
0.065 
0.192 
0.204 
  

0.140 
0.176 
  

  

  

†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 
 

 

 

58

 

Table 19: Multinomial logistic regression parameter estimates for significant results. Stand condition as independent variable. Northern 
Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program.† 

Forest Practiceb 
Shelter-
wood 
Harvest 

Selection 
Harvest 

Intercept 
SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-W 
Intercept 
SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-M* 
S-W 

β 
-2.479 
0.754 
-2.059 
-2.670 
-1.658 
-1.036 
-1.212 
0c 
0.054 
-1.950 
-2.478 
-2.768 
-2.654 
-1.288 
-1.021 
-0.883 
-0.250 
0c 

Std. Error 
0.212 
0.385 
0.777 
0.535 
0.482 
0.364 
0.366 
  
0.082 
0.356 
0.282 
0.171 
0.222 
0.134 
0.121 
0.126 
0.119 
  

Wald 
136.694 
3.831 
7.023 
24.876 
11.846 
8.092 
11.002 
  
0.424 
30.022 
77.242 
262.101 
142.750 
92.145 
71.360 
48.758 
4.445 
  

df 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Sig. 

Exp(β) 

  

  

0.000 
0.050 
0.008 
0.000 
0.001 
0.004 
0.001 
  
0.515 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
0.035 
  

  

  

2.125 
0.128 
0.069 
0.191 
0.355 
0.297 
  

0.142 
0.084 
0.063 
0.070 
0.276 
0.360 
0.414 
0.779 
  

  

  

0.999 
0.028 
0.024 
0.074 
0.174 
0.145 
  

0.071 
0.048 
0.045 
0.046 
0.212 
0.284 
0.323 
0.617 
  

4.522 
0.585 
0.198 
0.490 
0.725 
0.609 
  

0.286 
0.146 
0.088 
0.109 
0.359 
0.456 
0.530 
0.983 
  

a. Region = NLP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
d. Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing. 

†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 
 

 

 

59

 

Table 20: Multinomial logistic regression parameter estimates for significant results. Stand condition as independent variable. Northern 
Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program.† 

Forest Practiceb 
Thinning 

Timber 
Stand 
Improve-
ment 

Intercept 
SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-W 
Intercept 
SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-M* 
S-W 

β 
-2.479 
0.754 
-2.059 
-2.670 
-1.658 
-1.036 
-1.212 
0c 
0.054 
-1.950 
-2.478 
-2.768 
-2.654 
-1.288 
-1.021 
-0.883 
-0.250 
0c 

Std. Error 
0.212 
0.385 
0.777 
0.535 
0.482 
0.364 
0.366 
  
0.082 
0.356 
0.282 
0.171 
0.222 
0.134 
0.121 
0.126 
0.119 
  

Wald 
136.694 
3.831 
7.023 
24.876 
11.846 
8.092 
11.002 
  
0.424 
30.022 
77.242 
262.101 
142.750 
92.145 
71.360 
48.758 
4.445 
  

df 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Sig. 

Exp(β) 

  

  

0.000 
0.050 
0.008 
0.000 
0.001 
0.004 
0.001 
  
0.515 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
0.035 
  

  

  

2.125 
0.128 
0.069 
0.191 
0.355 
0.297 
  

0.142 
0.084 
0.063 
0.070 
0.276 
0.360 
0.414 
0.779 
  

  

  

0.999 
0.028 
0.024 
0.074 
0.174 
0.145 
  

0.071 
0.048 
0.045 
0.046 
0.212 
0.284 
0.323 
0.617 
  

4.522 
0.585 
0.198 
0.490 
0.725 
0.609 
  

0.286 
0.146 
0.088 
0.109 
0.359 
0.456 
0.530 
0.983 
  

a. Region = NLP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
d. Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing. 

†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 
 

 

 

60

 

Table 21: Multinomial logistic regression parameter estimates for significant results. Stand condition as independent variable. Southern 
Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program.† 

Forest Practiceb 
Artificial 
Regen-
eration 
Clearcut 

Intercept 
S-W 

Intercept 
SS-M* 
SS-W* 
P-P* 
S-W 

β 
-3.090 
0d 

Std. Error 
0.569 
  

-0.796 
-2.025 
-2.185 
-0.968 
0d 

0.213 
0.652 
0.479 
0.320 
  

Wald 

29.500 
  

13.980 
9.655 
20.802 
9.136 
  

df 
1 
0 

1 
1 
1 
1 
0 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Sig. 

Exp(β) 

  

  

0.000 
  

0.000 
0.002 
0.000 
0.003 
  

  

  

  

  

  

  

0.132 
0.112 
0.380 
  

0.037 
0.044 
0.203 
  

  

0.474 
0.288 
0.712 
  

a. Region = SLP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
d. Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing. 

†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

61

 

Table 22: Multinomial logistic regression parameter estimates for significant results. Stand condition as independent variable. Southern 
Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program.† 

Forest Practiceb 
Salvage 
Treatment 

β 
-3.943 
1.881 
0d 
-3.027 
0d 

Std. Error 
0.860 
0.903 
  
0.552 
  

Intercept 
P-M* 
S-W 
Intercept 
Seed Tree 
Harvest 
S-W 
a. Region = SLP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
d. Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing. 

20.999 
4.341 
  
30.073 
  

0.000 
0.037 
  
0.000 
  

  

  

6.558 
  

Wald 

Sig. 

Exp(β) 

df 
1 
1 
0 
1 
0 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

  

  

1.118 
  

  

  

38.468 
  

  

  

  

†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

62

 

Table 23: Multinomial logistic regression parameter estimates for significant results. Stand condition as independent variable. Southern 
Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program.† 

95% Confidence Interval 

for Exp(β) 

Forest Practiceb 
Shelterwood 
Harvest 

Selection 
Harvest 

Intercept 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-M* 
S-W 
Intercept 
NS* 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-W 

β 
-0.683 
-2.683 
-4.076 
-4.199 
-3.475 
-2.347 
-2.609 
-1.705 
0d 
1.071 
-4.238 
-3.457 
-3.481 
-2.990 
-1.663 
-1.596 
-1.244 
0d 

Std. Error 
0.205 
0.825 
1.043 
1.074 
0.766 
0.448 
0.609 
0.442 
  
0.138 
1.976 
0.521 
0.356 
0.290 
0.206 
0.196 
0.211 
  

Wald 

11.107 
10.574 
15.262 
15.279 
20.554 
27.471 
18.351 
14.914 
  
60.700 
4.599 
43.943 
95.523 
106.041 
65.098 
66.049 
34.609 
  

df 
1 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 
1 
1 
1 
1 
1 
1 
0 

0.001 
0.001 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
  
0.000 
0.032 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
  

Sig. 

Exp(β) 

Lower 
Bound 

  

  

0.068 
0.017 
0.015 
0.031 
0.096 
0.074 
0.182 
  

0.014 
0.002 
0.002 
0.007 
0.040 
0.022 
0.076 
  

Upper Bound 
  

0.344 
0.131 
0.123 
0.139 
0.230 
0.243 
0.432 
  

  

  

  

1.443E-02 
0.032 
0.031 
0.050 
0.189 
0.203 
0.288 
  

0.000  0.694413539 
0.088 
0.011 
0.062 
0.015 
0.089 
0.028 
0.127 
0.284 
0.298 
0.138 
0.436 
0.190 
  
  

a. Region = SLP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
d. Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing. 

 
 
 
 
 
 

 

63

 

Table 24: Multinomial logistic regression parameter estimates for significant results. Stand condition as independent variable. Southern 
Lower Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program.† 

Forest Practiceb 
Thinning 

Timber 
Stand 
Improve-
ment 

Intercept 
SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M* 
S-P* 
S-W 
Intercept 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-W 

β 
-0.427 
6.196 
-1.915 
-1.940 
-2.308 
-0.739 
-0.573 
0d 
-0.588 
-2.163 
-2.351 
-2.068 
-1.214 
-0.547 
-1.118 
0d 

Std. Error 
0.189 
3.084 
0.528 
0.373 
0.416 
0.267 
0.282 
  
0.199 
0.628 
0.465 
0.409 
0.314 
0.264 
0.341 
  

Wald 

5.110 
4.037 
13.152 
26.987 
30.784 
7.695 
4.130 
  
8.765 
11.854 
25.532 
25.605 
14.923 
4.302 
10.755 
  

df 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 
1 
1 
1 
1 
1 
0 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Sig. 

Exp(β) 

  

  

  

  

0.024 
0.045 
0.000 
0.000 
0.000 
0.006 
0.042 
  
0.003 
0.001 
0.000 
0.000 
0.000 
0.038 
0.001 
  

490.585 
0.147 
0.144 
0.099 
0.477 
0.564 
  

0.115 
0.095 
0.126 
0.297 
0.579 
0.327 
  

  
206821.405 
0.415 
0.299 
0.225 
0.805 
0.980 
  

  

0.394 
0.237 
0.282 
0.550 
0.970 
0.638 
  

1.164 
0.052 
0.069 
0.044 
0.283 
0.324 
  

0.034 
0.038 
0.057 
0.160 
0.345 
0.168 
  

a. Region = SLP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
d. Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing. 

†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 

 

 

64

 

Table 25: Multinomial logistic regression parameter estimates for significant results. Stand condition as independent variable. Western 
Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program.† 

95% Confidence Interval 

for Exp(β) 

Wald 

Sig. 

Exp(β) 

Lower 
Bound 

Clearcut 

Forest Practiceb 
Artificial 
Regen-
eration 

Intercept 
SS-W* 
P-P* 
S-W 
Intercept 
P-M* 
P-W* 
S-M* 
S-W 
a. Region = WUP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
d. Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing. 

19.448 
8.655 
6.619 
  
20.380 
7.240 
13.807 
8.259 
  

0.000 
0.003 
0.010 
  
0.000 
0.007 
0.000 
0.004 
  

Std. Error 
0.608 
0.927 
1.019 
  
0.491 
0.502 
0.499 
0.555 
  

β 
-2.681 
-2.727 
-2.621 
0c 
-2.216 
1.351 
1.854 
1.596 
0c 

df 
1 
1 
1 
0 
1 
1 
1 
1 
0 

  

  

  

  

0.065 
0.073 
  

3.861 
6.386 
4.932 
  

Upper Bound 
  

  

0.011 
0.010 
  

1.443 
2.402 
1.661 
  

0.402 
0.536 
  

10.331 
16.981 
14.646 
  

†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 
 

 

 

65

 

Table 26: Multinomial logistic regression parameter estimates for significant results. Stand condition as independent variable. Western 
Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program.† 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Sig. 

Exp(β) 

Wald 

Forest Practiceb 
Salvage 
Treatment 
Seed Tree 
Harvest 

β 
-5.698 
0c 
-2.523 
2.166 
-2.303 
-2.938 
-1.859 
0c 

Std. Error 
2.663 
  
0.565 
0.661 
1.119 
0.914 
0.803 
  

Intercept 
S-W 
Intercept 
SS-P* 
SS-M* 
SS-W* 
P-P* 
S-W 
a. Region = WUP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
d. Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing. 

4.579 
  
19.953 
10.732 
4.235 
10.336 
5.355 
  

0.032 
  
0.000 
0.001 
0.040 
0.001 
0.021 
  

  

  

df 
1 
0 
1 
1 
1 
1 
1 
0 

  

  

  

  

  

  

8.720 
0.100 
0.053 
0.156 
  

2.387 
0.011 
0.009 
0.032 
  

  

31.859 
0.896 
0.318 
0.752 
  

†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 
 

 

 

66

 

Table 27: Multinomial logistic regression parameter estimates for significant results. Stand condition as independent variable. Western 
Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program.† 

Forest Practiceb 
Shelterwood 
Harvest 

Selection 
Harvest 

Intercept 
SS-M* 
SS-W* 
P-P* 
S-W 
Intercept 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-W 

β 
-2.275 
-2.557 
-2.754 
-1.886 
0c 
0.861 
-3.645 
-3.706 
-3.225 
-1.817 
-1.105 
-0.704 
0c 

Std. Error 
0.504 
1.092 
0.768 
0.719 
  
0.184 
0.402 
0.271 
0.284 
0.214 
0.203 
0.235 
  

Wald 

20.364 
5.483 
12.859 
6.889 
  
21.987 
82.384 
186.975 
129.105 
72.331 
29.595 
8.960 
  

df 
1 
1 
1 
1 
0 
1 
1 
1 
1 
1 
1 
1 
0 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Sig. 

Exp(β) 

  

  

0.000 
0.019 
0.000 
0.009 
  
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
0.003 
  

  

  

0.078 
0.064 
0.152 
  

0.026 
0.025 
0.040 
0.162 
0.331 
0.494 
  

  

  

0.009 
0.014 
0.037 
  

0.012 
0.014 
0.023 
0.107 
0.222 
0.312 
  

0.659 
0.287 
0.620 
  

0.057 
0.042 
0.069 
0.247 
0.493 
0.784 
  

a. Region = WUP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
d. Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing. 

†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 
 

 

 

67

 

Table 28: Multinomial logistic regression parameter estimates for significant results. Stand condition as independent variable. Western 
Upper Peninsula. N=20,915 unique forest stands enrolled in the Qualified Forest Program.† 

Forest Practiceb 
Thinning 

Timber 
Stand 
Improvement 

Intercept 
SS-M* 
SS-W* 
P-P* 
P-M* 
S-W 
Intercept 
SS-W* 
P-P* 
S-M* 
S-W 

β 
-1.265 
-3.154 
-2.612 
-2.077 
-0.848 
0c 
-2.760 
-2.059 
-2.557 
1.768 
0c 

Std. Error 
0.328 
0.855 
0.464 
0.475 
0.372 
  
0.631 
0.819 
1.105 
0.696 
  

Wald 

14.863 
13.611 
31.709 
19.137 
5.201 
  
19.141 
6.325 
5.350 
6.444 
  

df 
1 
1 
1 
1 
1 
0 
1 
1 
1 
1 
0 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Sig. 

Exp(β) 

  

  

0.000 
0.000 
0.000 
0.000 
0.023 
  
0.000 
0.012 
0.021 
0.011 
  

  

  

0.043 
0.073 
0.125 
0.428 
  

0.128 
0.078 
5.858 
  

  

  

0.008 
0.030 
0.049 
0.207 
  

0.026 
0.009 
1.496 
  

0.228 
0.182 
0.318 
0.888 
  

0.635 
0.677 
22.938 
  

a. Region = WUP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
d. Floating point overflow occurred while computing this statistic. Its value is therefore set to system missing. 

†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 

 

 

 

68

 

Program summary results 

Discussion 

The forest types enrolled in the QFP are represented at a similar proportion to all private 

forestland in Michigan. The maple/beech/birch group characterizes the highest proportion of enrolled 

forest types (29%). The maple/beech/birch group also characterizes the highest proportion of forest types 

on all private land in Michigan (33%). This forest type often exists on productive sites and can offer the 

highest and most frequent return on investment in the form of income from timber products. Because of 

the strict QFP requirement to conduct forest management activity according to the written plan, family 

forest owners with productive, high-value forest types might be more inclined to participate. Potentially, 

these high-quality forest types have the lowest risk for non-compliance due to a lower likelihood of 

having non-marketable timber. 

The elm/ash/cottonwood group ranks the second most common forest type on private forestland 

in Michigan (13%), but seventh in the QFP (7%). The elm/ash/cottonwood group, commonly referred to 

as lowland hardwoods, is a forest type not known for high-value forest products. This low value is 

compounded by the mortality and stand degradation caused by the Emerald Ash Borer (Agrilus 

planipennis) (Klooster et al., 2014), an exotic invasive pest of North American ash trees, and Dutch Elm 

Disease (Loo, 2009), a vascular disease of elm (Ulmus), caused by a fungus (Ceratocystis ulmi). Thus, 

these forest types might not meet the productivity standard defined in statute of capable of producing 20 

cubic feet of wood per acre per year.  

The aspen/birch group (Figure 1) makes up a proportion that is 11% less on QFP enrolled land 

than on private land statewide. The forest products market for the aspen/birch type throughout much of 

the state suggests that enrollment would not risk compliance issues. In 2016, aspen represented 20% of 

the state’s total cords processed by primary mills and 14% of the total board foot volume processed by 

primary mills (MDNR, 2018). Michigan is home to pulp-producing mills (Packaging Corporation of 

America, Verso Corporation) and composite board-producing mills (Decorative Panels International, 

Louisiana-Pacific Corporation, and Weyerhaeuser) that utilize large volumes of aspen, specifically. 

 

69

 

The maple/beech/birch group (28%) and oak group (27%) constitute the majority of the enrolled 

land in the Southern Lower Peninsula (SLP). Much of the region is dominated by an agricultural 

landscape, which makes the preponderance of the maple/beech/birch group counterintuitive. I expected 

the maple/beech/birch type to be enrolled at a lower level than it’s proportion on private land statewide. 

As previously mentioned, the northern hardwoods type often grows on high-quality sites. Much of the 

land that would have produced this forest type within the region was clearcut and cultivated because it 

also produced the best cropland.  

The maple/beech/birch group (27%), aspen/birch group (25%), and oak group (16%) dominate 

the land enrolled in the Northern Lower Peninsula (NLP). Not only is this likely dictated by climate and 

soil characteristics, but forest industry may also play a role in the maintenance of these proportions. The 

NLP is home to a number of wood-consuming mills that primarily utilize aspen (Populus spp.) and other 

hardwood pulp. The demand for this material and the silvics of aspen lend itself to a perpetuation of the 

forest type. 

Similar to the NLP, the Eastern Upper Peninsula (EUP) has significant acres in the 

maple/beech/birch group (29%) and aspen/birch group (22%), although the spruce/fir group represents the 

highest proportion (31%). This region boasts a similarly strong market for aspen pulp, which could be 

additive to the climatic effects driving the abundance of the type. The spruce/fir group is preponderant 

likely because the amount of low, poor growing sites in the region. 

The Western Upper Peninsula (WUP) also contains maple/beech/birch (35%), aspen/birch (28%), 

and spruce/fir (18%) as the highest proportions of enrolled forest type groups. The region is known for 

producing high quality sugar maple (Acer saccharum) sawlogs and veneer logs. The region also falls into 

a shared procurement zone between pulp mills and sawmills in Northeastern Wisconsin and the rest of the 

Upper Peninsula. 

Statewide, the majority (52%) of stands enrolled are classified as well stocked. Only 19% are 

classified as poorly stocked. This speaks to the health of enrolled forest stands. As previously mentioned 

in regard to proportion of forest types, the reason that the program exhibits healthy, well-stocked stands 

 

70

 

might be related to the statutory expectation of enrollees to conduct forest management in order to receive 

the tax exemption. Potential participants and plan-writing foresters might see poorly stocked stands as 

risky for enrollment. If they are not able to market the material (due to lack of volume or any other 

reason) they risk removal from the program and are forced to pay steep recapture penalties. 

The breakdown of stand condition within each species group is telling (Table 2). For example, 

37% of the aspen/birch group is well-stocked seedlings/saplings. Only 5% of the group is sawtimber – 

well stocked and only 22% is poletimber – well stocked. This could lead to a figurative “pig in the 

python” scenario where production from QFP lands on this forest type is relatively low for a number of 

years and then, as the type reaches maturity, a much greater amount is put up for sale in a small cluster of 

years. Another example is the condition of the oak resource on QFP land. More than three-quarters (79%) 

of the type is in the sawtimber size class. As these stands are harvested, there may be societal demands for 

oak stands to regenerate for long-term financial, wildlife, and biodiversity concerns. The lack of stands in 

the two smaller size classes may cause similar short and medium-term issues. The majority (58%) of 

stands in the elm/ash/cottonwood group are in the poletimber size class. The future of this type is clouded 

with forest health issues. Ash (Fraxinus) and elm (Ulmus) species are both troubled by non-native insects 

and diseases such as emerald ash borer (Agrilus planipennis) and Dutch elm disease (Ascomycota) that 

puts future forest cover in jeopardy. 

Forest practices 

The size of a forest stand was significant in predicting all forest practices on QFP enrolled land 

(Tables 2.01a – 2.08a). The largest odds ratio for stand acres is when selecting a salvage treatment (1.3% 

with each one-acre increase in stand size) (Table 4). Salvage treatments are conducted in an effort to take 

advantage of available wood fiber directly after an event that causes the loss of wood volume and/or 

quality. For example, many salvage treatments in Michigan occur after a forest health issue, such as an 

insect infestation, kills many trees on a large landscape. The relatively elevated odds of selecting a 

salvage treatment with an increase in stand size on QFP enrolled land makes intuitive sense. The quality 

and volume of wood fiber per unit of area is generally lower for stands qualifying for salvage treatments, 

 

71

 

by definition. Because of this, one might argue that a greater amount of volume is required to make a 

timber sale attractive to a potential buyer. Professional forest management plan writers determine forest 

stands. The program requires forest management to be conducting on any given parcel. Markets and 

access often drive forest management. It is reasonable, then, to posit that foresters delineate stands and 

prescribe forest management practices that are economically viable as well as ecologically justified. 

The odds of a family forest owner choosing artificial regeneration instead of no practice is 2.907 

times higher in the SLP than in the WUP. The WUP is almost entirely forested, while the SLP is a mosaic 

of forests, grasslands, farms, and urban environments. It’s possible that there are simply more acres 

available for tree planting in the SLP because it isn’t all already in forest cover. The nomenclature of the 

practice can cause confusion. Artificial regeneration suggests that the practice is an attempt at reforesting 

a stand after some other management activity. It is possible that some participants are using the term 

artificial regeneration as a means of afforestation.  

The availability of financial assistance to conduct tree/shrub plantings through the USDA Natural 

Resources Conservation Service (NRCS) is a related justification. While this funding is available to all 

farmers, ranchers, and forestland owners, farmers commonly participate because of their familiarity with 

the involved federal agencies and programs. Another forest conservation practice funded through the 

NRCS is forest (timber) stand improvement. That might be why the odds of selecting that practice are 

10.914 times higher in the SLP than the WUP. 

Another explanation is the fragmented nature of forests in the SLP compared to the WUP. This 

might play a role in the ability to market the harvested material removed in a traditional timber stand 

improvement. Removal of the same material in the WUP might make its way into a normal commercial 

timber harvest, which might prompt the plan-writing forester to use another silvicultural term, such as 

selection harvest. 

The multinomial logistic regression with stand type as the independent variable was conducted at 

a statewide scope. As one considers the “normal” silviculture associated with each forest type, many of 

the results are intuitive. For example, aspen is a forest type not normally propagated by planting, rather by 

 

72

 

even-aged methods of overstory removal, which allows for the aspen root suckers to grow into new stems. 

The odds ratio for aspen predicting artificial regeneration is 0.366 (Table 2) meaning the likelihood of an 

aspen stand selecting artificial regeneration is negative (odds less than one). However, red pine stands are 

often created and/or regenerated through planting. The elevated odds ratio (3.189) of a red pine stand 

selecting artificial regeneration instead of no practice is logical.  

Some stand types yielded less intuitive results. The lowland poplar (mostly Populus balsamifera) 

type had an odds ratio of 3.974 (Table 2) for selecting artificial regeneration instead of no practice. This is 

difficult to understand because the lowland poplar type usually exists in wet areas that are not well suited 

for easy planting, and the trees are somewhat short-lived and grow quite fast. The wood is not particularly 

sought after by the forest products industry. One justification for this phenomenon is the low number of 

observations for this stand type (n=213). 

Clearcutting is a common practice across the regions and in may forest types (Table 3, Table 10, 

Figure 14). The odds ratio of aspen selecting clearcut is 1.714, meaning the forest type is 71.4% more 

likely to select clearcut than no practice (Table 3). I originally expected a higher likelihood of this type 

and practice, but after reviewing the proportion of aspen stands in the seedling/sapling size class, it is 

logical. Over the maximum 20-year forest management-planning period many aspen stands may not be 

economically mature enough to warrant a timber harvest. The statewide county mapping for forest 

practice and forest type demonstrated that Alcona County has the highest number of acres in the 

aspen/birch type (10,952.2, Figure 7) as well as the highest number of acres prescribed clearcut (6,342.1, 

Figure 14). 

Another example is the odds of a northern hardwoods forest type predicting selection harvest 

instead of no practice (Table 7). A northern hardwoods stand is 781.8% more likely to have a selection 

harvest than no practice. This is a common silvicultural practice in a northern hardwood type. The result 

is likely indicative of both forestry norms and the societal desire to see big trees and minimal site 

disturbance in forested ecosystems. Current challenges in northern hardwood regeneration might begin to 

skew this type to another, more uneven-aged management technique in an effort to successfully recruit 

 

73

 

shade intolerant species as well as combat the increased white-tailed deer (Odocoileus virginianus) 

numbers in the state.  

Oak is a forest type that can be, and is being, managed along the entire continuum of silvicultural 

strategies. Oak on QFP is being managed by even-aged techniques such as a seed tree harvest (odds ratio 

of 9.141) (Table 5) and an uneven-aged technique of selection harvest (odds ratio of 4.498) (Table 7). 

Each technique is used in pursuit of a different outcome. Seed tree harvests in oak stands are primarily 

used as a regeneration strategy because of oak’s relative shade intolerance and slow growth. The selection 

harvest method might be used to increase diameter, therefore timber quality, and vigor among existing 

high-quality trees in a middle-aged stand. 

Stand conditions did not appear to be significant in a statewide analysis. However, when the data 

were aggregated regionally, certain stand conditions did become significant in selecting certain forest 

practices (Table 13 through 7.4 and A.12 through A.27). Again, this might have to do with available 

markets for forest products within a region. Intuitively, stands with a higher overall volume relative to the 

size of trees within the stand (conditions 3, 6, and 9) might be more likely to be harvested because the 

higher volume could be seen as attractive for timber producers. Also, stands with relatively smaller trees 

(conditions 1-6) can only be marketed when a use for products that use smaller-diameter roundwood 

exists within the region. For example, the SLP contains virtually no market for pulpwood and other small-

diameter tree products. It can be difficult for family forest owners interested in forest management to find 

markets for their small diameter material, which can reduce their ability to conduct forest management in 

the absence of technical and financial assistance. 

 

 

74

 

Conclusion 

This study informs the growing literature on family forest owner behavior and participation in 

conservation programs. First, this analysis explored the relationship between forest types enrolled in the 

QFP and those on private land in Michigan statewide and by region. Second, The results have 

demonstrated that, in certain cases, stand and parcel characteristics can predict forest practice. 

Characteristics such as size, condition, type, and region each play a unique role in influencing the 

outcome of a forest management decision. This study has taken the first steps toward using the 

relationships between parcel characteristics and family forest owner management decisions for 

participants enrolled in a forest property taxation program. This study can be used to understand the 

current status of a state-level property tax incentive program that targets family forest owners. 

There are 31 acceptable forest types for plan-writing foresters to describe a stand. This appears to 

contribute a lot of “noise” into the data. Is the level of specificity necessary? In the program summary 

results I attempted to aggregate forest types into a few larger groups to facilitate the comparison to the 

Forest Inventory and Analysis data for private lands. This aggregation might help the QFP staff have 

simpler data that is easier to interpret, manipulate, and analyze. It would also make statewide and 

nationwide program comparisons possible and may avoid confusion. For example, it is difficult to 

interpret the odds ratio for a forest practice when the stand type is water, or the difference between 

northern hardwoods, upland mixed, and mixed deciduous stand types. 

Results from this study can be used to make observations about QFP participants, but its 

application to non-enrolled family forest owners is still uncertain. As participation in QFP grows one 

might argue that it either more closely represents all family forest owners because the program will make 

up a greater proportion, or that the tax break and available technical assistance through Michigan’s 

conservation districts are changing behavior such that similarities cannot be drawn. Both of these 

hypotheses are outside the scope of this research. A similar analysis of family forests outside the QFP 

might help to strengthen those claims in either direction, but the difficulty of acquiring data outside of a 

voluntary government program might be prohibitive. Cutting-edge forestry technologies might make 

 

75

 

acquiring this data for all private land more feasible in the future. The parcel characteristics outlined in 

this analysis can be determined through objective observation, and do not necessarily require interaction 

with the landowner. Aerial photo interpretation and other remote sensing technologies might be used to 

acquire the data quickly and inexpensively, although these methods might bring up other concerns, such 

as privacy. 

Other variables could be helpful in an analysis of family forest owner behavior. The year that 

enrolled stands have a planned harvest would provide some further insight to the management of 

forestland enrolled in the QFP. The data is easily accessible from MDARD’s QFP database. The planned 

harvest year could be categorized by the following conditions: “harvest shortly after plan development”, 

“harvest in near future”, and “harvest in distant future”. Dividing harvest years into these categories might 

tell us something about whether family forests are at currently at harvestable volumes and values, or 

whether enrolled lands are relatively “young” compared to non-enrolled family forests or those in other 

ownerships. It might also indicate the level of management intensity that family forest owners are willing 

to engage in. This data is available for each forest stand enrolled in the QFP. It was not included in this 

analysis because it is outside of the focus of evaluating parcel characteristics with respect to forest 

management practices.  

Similar analysis could be combined with forest management plan writer interviews to determine 

if harvest year reflects forest age or some social reason instead. These data are not currently available. 

More time might be required, thus more data to work with, to make meaningful claims about these topics. 

A more thorough examination of the significant relationships might yield stand and parcel 

characteristics that could be selectively utilized in unique regression equations. The forest management 

plan writer is another variable that could have substantial effects on forest practice; both at the time of 

plan development and implementation. Forest management plan writer data are collected in the program’s 

database and assigned to each parcel at the time of application. 

Critics of the QFP often claim that the program provides a tax incentive only to those already 

managing forestland. It is unclear to them as to whether the program has enticed family forest owners that 

 

76

 

were not already practicing forest management. This is difficult to infer, and might require a combination 

of analyses, including surveys to participants and non-participants. 

The number of variables and categories created a volume of results difficult to fully analyze. 

Future analyses should take the results provided in this research and begin to “connect the dots” by 

pairing significant results for each category of forest practice to create odds ratio combinations to serve as 

the next step in developing a model to predict family forest owner behavior. 

In 2006, Michigan Public Acts 378, 379, and 380 created the QFP. At the time, it was 

administered by the Michigan Department of Natural Resources. In 2013, Public Acts 42, 43, 44, 45, 49, 

and 50 made major revisions to the program, including agency administration (Michigan Department of 

Agriculture and Rural Development), established a restricted revenue source for long-term funding, and 

reduced steep recapture penalties. Approximately 90,000 acres were enrolled over the first seven years. In 

the period between 2013 and 2017, more than 450,000 acres of forestland were enrolled. 

The mechanism for this steep growth curve is likely a combination of factors. The General Fund 

support that arrived with the change in agency administration allowed the Michigan Department of 

Agriculture and Rural Development to dedicate full-time equivalent positions that the Michigan 

Department of Natural Resources was unable to support. Over time, the Private Forestland Enhancement 

Fund is expected to be sufficient to fund the delivery of both the QFP and the Forestry Assistance 

Program. Based on a 2013 economic impact study conducted by Minnesota IMPLAN Group on behalf of 

the Michigan Department of Agriculture and Rural Development, the program requires an enrollment of 

2.5 million acres to fully fund both programs (Harlow). Prior to the 2013 Public Acts, the recapture tax 

for withdrawing land from the program was much steeper than it is currently. The reduced recapture tax 

may make the program less intimidating and more accessible to family forest owners. The new recapture 

taxes are simpler to calculate and more attainable for short-term financial saving. Likely, all of these 

factors contributed to the steep increase in enrollment. 

Near the end of developing this paper, the QFP experienced a change in law that will take effect 

on 3 March 2019. Among the changes, one was the removal of the maximum number of acres a 

 

77

 

landowner can enroll per tax-collecting unit of government, which was previously 640 acres. This has the 

potential to increase enrollment rapidly, as large landowners are able to enroll the remaining portions of 

their ownerships that were previously excluded. Another major change was the increase of the statewide 

program maximum acreage to 2.5 million acres from 1.2 million acres. 

In this study, describing what the enrolled forests look like, where they are located, and what 

management decisions are being implemented has summarized enrollment in the QFP. Determining that 

stand size, condition, type, and region can be used to predict forest management decisions on land 

enrolled in the QFP was successful. Family forests make up the largest proportion of forest ownership in 

the county. It is important that academic institutions and government agencies continue to study this 

ownership group because of its large impact on forest conservation and natural resource management. 

Conservation programs similar to the QFP have the potential to conserve and protect forest landscapes 

that are important economically, for communities that rely on the forest products we use every day; 

environmentally, to preserve critical ecosystem functions and rare plant communities; and socially, to 

maintain forests that positively affect humans and their networks. With this contribution to the literature, 

practitioners have a new tool to aid in evaluating the conservation programs that promote these values.

 

78

 

 

APPENDIX 

 

79

 

Table A.01: Multinomial logistic regression parameter estimates. Stand acres and stand type as independent variables (showing results 
from artificial regeneration). N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

Forest Practicea 
Artificial 
Regen-
eration 

Intercept 
Stand Acres* 
Stand Type=Agricultural 
Stand Type=Aspen* 
Stand Type=Black Spruce 
Stand Type=Bog or Muskeg 
Stand Type=Cedar* 
Stand Type=Grass 
Stand Type=Hemlock 
Stand Type=Jack Pine 
Stand Type=Local Use 
(various non-commercial or 
exotic) 
Stand Type=Lowland Brush 
Stand Type=Lowland 
Hardwoods 
Stand Type=Lowland 
Mixed 
Stand Type=Lowland 
Poplar (Bam)* 
Stand Type=Marsh* 
Stand Type=Mixed 
Deciduous* 
Stand Type=Mixed 
Lowland Conifers* 
Stand Type=Non-Stocked 
Stand Type=Northern 
Hardwoods 
Stand Type=Oak 

 

 

B 

-3.964 
0.009 
-3.076 
-1.005 
0.502 
-3.020 
-2.173 
-0.346 
-0.670 
0.284 
-45.644 

Std. 
Error  Wald 
0.417 
90.405 
23.103 
0.002 
0.275 
5.861 
0.477 
4.437 
0.789 
0.565 
3.184 
1.692 
5.476 
0.929 
0.464 
0.555 
0.368 
1.105 
0.152 
0.728 
0.000 
  

-0.655 
-0.518 

0.562 
0.564 

1.357 
0.842 

-0.632 

0.821 

0.593 

1.380 

0.596 

5.353 

-2.989 
1.252 

2.301 
0.564 

1.688 
4.920 

-1.217 

0.596 

4.177 

-0.871 
0.604 

0.613 
0.445 

2.022 
1.842 

0.378 

0.483 

0.611 

df 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 

1 
1 

1 

1 

1 
1 

1 

1 
1 

1 

95% Confidence Interval 

for Exp(B) 

Lower 
Bound 

Exp(B) 

  

  

Upper Bound 
  

1.009 
0.046 
0.366 
1.652 
0.049 
0.114 
0.708 
0.511 
1.328 
1.503E-
20 

1.005 
4.730E-07 
0.144 
0.546 
0.002 
0.018 
0.285 
0.059 
0.319 
1.503E-20 

1.012 
4500.868 
0.933 
4.997 
1.346 
0.703 
1.758 
4.463 
5.537 
1.503E-20 

Sig. 
0.000 
0.000 
0.600 
0.035 
0.374 
0.074 
0.019 
0.456 
0.544 
0.697 
  

0.244 
0.359 

0.519 
0.596 

0.172 
0.197 

0.441 

0.531 

0.106 

1.564 
1.801 

2.656 

0.021 

3.974 

1.235 

12.789 

0.194 
0.027 

0.050 
3.496 

0.001 
1.157 

4.574 
10.564 

0.041 

0.296 

0.092 

0.155 
0.175 

0.419 
1.830 

0.126 
0.765 

0.435 

1.459 

0.566 

0.951 

1.390 
4.378 

3.760 

 

 

 

 

 

 

 

 

80

 

 Table A.01 (cont’d) 

 

Stand Type=Paper 
Birch 
Stand Type=Red 
Pine* 
Stand Type=Rock 
Stand Type=Sand 
Dunes 
Stand 
Type=Tamarack 
Stand Type=Treed 
Bog 
Stand 
Type=Upland 
Brush 
Stand 
Type=Upland 
Mixed 
Stand 
Type=Upland 
Mixed Conifers* 
Stand Type=Water 

Stand Type=White 
Pine 
Stand Type=White 
Spruce/Fir 
Stand Type=White 
Spruce/Fir 

-3.737 

8.664 

0.186 

1  0.666 

0.024 

1.005E-09 

565049.106 

1.160 

0.452 

6.583 

1  0.010 

3.189 

1.315 

7.735 

-2.935 
-2.929 

15.857 
18.308 

0.034 
0.026 

1  0.853 
1  0.873 

0.053 
0.053 

1.690E-15  1.67116E+12 
1.393E-17  2.05228E+14 

-2.547 

2.189 

1.353 

1  0.245 

0.078 

0.001 

5.720 

-2.965 

5.179 

0.328 

1  0.567 

0.052 

2.013E-06 

1320.153 

-0.196 

0.834 

0.055 

1  0.814 

0.822 

0.160 

4.212 

1.102 

0.568 

3.763 

1  0.052 

3.011 

0.989 

9.171 

2.265 

0.467 

23.506 

1  0.000 

9.635 

3.856 

24.075 

-1.921 

1.090 

3.104 

1  0.078 

0.146 

0.017 

0.280 

0.551 

0.259 

1  0.611 

1.324 

0.450 

0b 

0b 

  

  

  

  

0 

0 

  

  

  

  

  

  

1.241 

3.897 

  

  

a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 

 

 

 

81

 

Table A.02: Multinomial logistic regression parameter estimates. Stand acres and stand type as independent variables (showing results 
from clearcut). N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

Forest Practicea 
Clearcut 

Intercept 
Stand Acres* 
Stand Type=Agricultural 
Stand Type=Aspen* 
Stand Type=Black Spruce 
Stand Type=Bog or Muskeg* 
Stand Type=Cedar* 
Stand Type=Grass* 
Stand Type=Hemlock* 
Stand Type=Jack Pine 
Stand Type=Local Use (various 
non-commercial or exotic)* 
Stand Type=Lowland Brush* 
Stand Type=Lowland 
Hardwoods* 
Stand Type=Lowland Mixed* 
Stand Type=Lowland Poplar 
(Bam) 
Stand Type=Marsh* 
Stand Type=Mixed Deciduous* 
Stand Type=Mixed Lowland 
Conifers* 
Stand Type=Non-Stocked* 
Stand Type=Northern 
Hardwoods* 
Stand Type=Oak* 
Stand Type=Paper Birch 
Stand Type=Red Pine* 
Stand Type=Rock* 

 

B 

-1.171 
0.008 
-3.487 
0.539 
-0.026 
-3.327 
-1.757 
-3.324 
-1.802 
0.077 
-3.091 

Std. 
Error 
0.119 
0.001 
1.799 
0.124 
0.185 
0.493 
0.211 
0.255 
0.473 
0.226 
0.862 

Wald 
96.608 
120.532 
3.758 
18.938 
0.020 
45.622 
69.462 
169.498 
14.510 
0.116 
12.854 

df 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 

-2.916 
-0.482 

0.315 
0.157 

85.548 
9.467 

-0.487 
0.149 

0.218 
0.253 

-3.402 
-0.780 
-0.798 

0.704 
0.297 
0.151 

-3.179 
-1.659 

0.374 
0.170 

-1.111 
-0.173 
-1.341 
-3.353 

0.179 
0.429 
0.195 
4.865 

5.012 
0.349 

23.317 
6.895 
27.889 

72.354 
95.482 

38.536 
0.163 
47.275 
0.475 

82

1 
1 

1 
1 

1 
1 
1 

1 
1 

1 
1 
1 
1 

Sig. 
0.000 
0.000 
0.053 
0.000 
0.889 
0.000 
0.000 
0.000 
0.000 
0.733 
0.000 

0.000 
0.002 

0.025 
0.555 

0.000 
0.009 
0.000 

0.000 
0.000 

0.000 
0.687 
0.000 
0.491 

95% Confidence Interval 

for Exp(B) 

Lower 
Bound 

Upper 
Bound 

  

Exp(B) 
  

  

1.008 
0.031 
1.714 
0.975 
0.036 
0.173 
0.036 
0.165 
1.080 
0.045 

0.054 
0.617 

0.614 
1.161 

0.033 
0.458 
0.450 

0.042 
0.190 

0.329 
0.841 
0.262 
0.035 

1.007 
0.001 
1.345 
0.679 
0.014 
0.114 
0.022 
0.065 
0.694 
0.008 

0.029 
0.454 

0.401 
0.707 

0.008 
0.256 
0.335 

0.020 
0.136 

1.010 
1.039 
2.186 
1.399 
0.094 
0.261 
0.059 
0.417 
1.681 
0.246 

0.100 
0.839 

0.941 
1.907 

0.133 
0.820 
0.605 

0.087 
0.265 

0.232 
0.363 
0.178 
2.527E-06 

0.468 
1.951 
0.383 
484.026 

 

Table A.02 (cont’d) 

 

-3.347 
-0.726 
-3.379 
-3.262 
-0.036 
-0.297 

Stand Type=Sand Dunes 
Stand Type=Tamarack* 
Stand Type=Treed Bog* 
Stand Type=Upland Brush* 
Stand Type=Upland Mixed 
Stand Type=Upland Mixed 
Conifers 
Stand Type=Water* 
Stand Type=White Pine* 
Stand Type=White Spruce/Fir 
a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 
* Indicates significant results 

5.617 
0.261 
1.589 
0.836 
0.227 
0.228 

0.541 
0.266 
  

0.355 
7.719 
4.525 
15.223 
0.026 
1.701 

39.602 
37.768 
  

1 
1 
1 
1 
1 
1 

1 
1 
0 

0.551 
0.005 
0.033 
0.000 
0.873 
0.192 

0.000 
0.000 
  

0.035 
0.484 
0.034 
0.038 
0.964 
0.743 

0.033 
0.195 
  

5.819E-07 
0.290 
0.002 
0.007 
0.618 
0.476 

0.011 
0.116 
  

2126.254 
0.807 
0.767 
0.197 
1.504 
1.161 

0.096 
0.329 
  

  

  

  

  

  

  

  

-3.407 
-1.634 
0b 

 

 

 

83

 

Table A.03: Multinomial logistic regression parameter estimates. Stand acres and stand type as independent variables (showing results 
from salvage treatment). N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

Forest Practicea 
Salvage 
Treatment 

Intercept 
Stand Acres* 
Stand Type=Agricultural 
Stand Type=Aspen* 
Stand Type=Black Spruce* 
Stand Type=Bog or Muskeg* 
Stand Type=Cedar* 
Stand Type=Grass* 
Stand Type=Hemlock 
Stand Type=Jack Pine 
Stand Type=Local Use (various 
non-commercial or exotic) 
Stand Type=Lowland Brush* 
Stand Type=Lowland 
Hardwoods* 
Stand Type=Lowland Mixed* 
Stand Type=Lowland Poplar 
(Bam) 
Stand Type=Marsh* 
Stand Type=Mixed Deciduous 
Stand Type=Mixed Lowland 
Conifers* 
Stand Type=Non-Stocked* 
Stand Type=Northern 
Hardwoods* 
Stand Type=Oak* 
Stand Type=Paper Birch 
Stand Type=Red Pine* 
Stand Type=Rock 

 

B 
-2.421 
0.013 
-4.615 
-3.289 
-1.999 
-4.502 
-2.080 
-4.515 
-3.935 
0.040 
-56.864 

Std. 
Error 
0.194 
0.001 
5.423 
0.365 
0.609 
1.501 
0.386 
0.772 
2.313 
0.365 
0.000 

Wald 
155.655 
135.277 
0.724 
81.339 
10.797 
9.002 
28.998 
34.220 
2.893 
0.012 
  

df 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 

-2.460 
-0.719 

0.456 
0.270 

29.101 
7.089 

-1.279 
-1.328 

0.455 
0.740 

-4.566 
-0.168 
-1.555 

2.199 
0.386 
0.291 

-4.534 
-0.894 

1.270 
0.240 

0.283 
-0.988 
8.629 
-5.364 
-4.452 
1.276 
-4.515  16.032 

7.886 
3.216 

4.312 
0.190 
28.532 

12.745 
13.868 

12.181 
0.386 
12.173 
0.079 

84

1 
1 

1 
1 

1 
1 
1 

1 
1 

1 
1 
1 
1 

Sig. 
0.000 
0.000 
0.395 
0.000 
0.001 
0.003 
0.000 
0.000 
0.089 
0.913 
  

0.000 
0.008 

0.005 
0.073 

0.038 
0.663 
0.000 

0.000 
0.000 

0.000 
0.534 
0.000 
0.778 

95% Confidence Interval for 

Exp(B) 

Exp(B) 

Lower 
Bound 

  

  

  

Upper Bound 

1.013 
1.011 
0.010  2.398E-07 
0.037 
0.018 
0.041 
0.135 
0.001 
0.011 
0.059 
0.125 
0.011 
0.002 
0.000 
0.020 
1.041 
0.509 
2.015E-25  2.015E-25 

0.085 
0.487 

0.278 
0.265 

0.010 
0.845 
0.211 

0.011 
0.409 

0.035 
0.287 

0.114 
0.062 

0.000 
0.397 
0.119 

0.001 
0.256 

1.016 
408.969 
0.076 
0.446 
0.210 
0.266 
0.050 
1.821 
2.129 
2.015E-25 

0.209 
0.827 

0.680 
1.131 

0.774 
1.801 
0.374 

0.129 
0.655 

0.372 
0.214 
0.005  2.114E-10 
0.012 
0.001 
0.011  2.470E-16 

0.649 
103632.527 
0.142 
484796543537.9 

 

Table A.03 (cont’d) 

 

Stand Type=Sand Dunes 
Stand Type=Tamarack* 
Stand Type=Treed Bog 
Stand Type=Upland Brush* 
Stand Type=Upland Mixed* 
Stand Type=Upland Mixed 
Conifers* 
Stand Type=Water* 
Stand Type=White Pine* 
Stand Type=White Spruce/Fir 

-4.506  18.510 
-4.152 
2.116 
5.146 
-4.577 
0.749 
-1.780 
-3.098 
1.163 
1.917 
-3.939 

0.059 
3.848 
0.791 
5.642 
7.098 
4.225 

-4.460 
-1.469 
0b 

1.579 
0.428 
  

7.980 
11.749 
  

1 
1 
1 
1 
1 
1 

1 
1 
0 

0.808 
0.050 
0.374 
0.018 
0.008 
0.040 

0.005 
0.001 
  

0.011  1.937E-18 
0.016 
0.000 
0.010  4.286E-07 
0.039 
0.169 
0.045 
0.005 
0.000 
0.019 

0.012 
0.230 
  

0.001 
0.099 
  

6.29857E+13 
0.996 
246.751 
0.733 
0.441 
0.833 

0.255 
0.533 
  

a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 
* Indicates significant results 

 

  

  

  

  

  

  

  

85

 
 

 

 

Table A.04: Multinomial logistic regression parameter estimates. Stand acres and stand type as independent variables (showing results 
from seed tree harvest). N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

Forest Practicea 
Seed 
Tree 
Harvest 

Intercept 
Stand Acres* 
Stand Type=Agricultural 
Stand Type=Aspen 
Stand Type=Black Spruce 
Stand Type=Bog or Muskeg 
Stand Type=Cedar 
Stand Type=Grass 
Stand Type=Hemlock 
Stand Type=Jack Pine 
Stand Type=Local Use (various 
non-commercial or exotic) 
Stand Type=Lowland Brush 
Stand Type=Lowland 
Hardwoods 
Stand Type=Lowland Mixed 
Stand Type=Lowland Poplar 
(Bam)* 
Stand Type=Marsh 
Stand Type=Mixed Deciduous 
Stand Type=Mixed Lowland 
Conifers* 
Stand Type=Non-Stocked 
Stand Type=Northern 
Hardwoods 
Stand Type=Oak* 
Stand Type=Paper Birch 
Stand Type=Red Pine 
Stand Type=Rock 

B 
-5.027 
0.006 
-2.031 
0.537 
0.344 
-1.993 
1.317 
-1.923 
0.440 
0.297 
-54.262 

Std. 
Error  Wald 
0.717 
0.002 
6.138 
0.742 
1.010 
1.856 
0.763 
1.054 
1.249 
1.253 
0.000 

49.178 
12.008 
0.110 
0.524 
0.116 
1.153 
2.977 
3.327 
0.124 
0.056 
  

-1.908 
0.951 

1.418 
0.785 

1.190 
2.271 

0.879 
0.858 

1.811 
1.468 

1.834 
7.004 

-1.957 
1.717 
2.944 

2.460 
0.887 
0.724 

0.633 
3.748 
16.541 

-1.943 
0.943 

1.503 
0.752 

0.738 
2.213 
1.435 
1.278 
-0.239 
0.933 
-1.937  16.471 

1.672 
1.571 

8.985 
0.793 
0.065 
0.014 

df 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 

1 
1 

1 
1 

1 
1 
1 

1 
1 

1 
1 
1 
1 

Sig. 
0.000 
0.001 
0.741 
0.469 
0.734 
0.283 
0.084 
0.068 
0.725 
0.812 
  

0.178 
0.226 

0.176 
0.008 

0.426 
0.053 
0.000 

0.196 
0.210 

0.003 
0.373 
0.798 
0.906 

95% Confidence Interval 

for Exp(B) 

Lower 
Bound 

1.003 
7.824E-07 
0.400 
0.195 
0.004 
0.836 
0.018 
0.134 
0.116 
2.717E-24 

Upper Bound 
  

1.009 
21981.489 
7.317 
10.219 
5.178 
16.659 
1.154 
17.942 
15.688 
2.717E-24 

Exp(B) 

  

  

1.006 
0.131 
1.710 
1.410 
0.136 
3.732 
0.146 
1.552 
1.346 
2.717E-24 

0.148 
2.588 

3.286 
9.685 

0.141 
5.567 
18.985 

0.143 
2.568 

9.141 
3.589 
0.788 
0.144 

0.009 
0.556 

0.587 
1.802 

0.001 
0.979 
4.596 

0.008 
0.588 

2.389 
12.051 

18.387 
52.048 

17.547 
31.654 
78.433 

2.725 
11.224 

2.151 
0.215 
0.126 

38.845 
59.790 
4.906 
1.376E-15  1.50843E+13 

 

86

 

Table A.04 (cont’d) 

0.010 
5.059 
0.128 
0.404 
12.629 
0.497 

1.044 
0.625 
  

1 
1 
1 
1 
1 
1 

1 
1 
0 

0.919 
0.025 
0.720 
0.525 
0.000 
0.481 

0.307 
0.429 
  

0.145 
6.487 
0.144 
0.159 
15.944 
2.059 

0.139 
2.001 
  

9.452E-18  2.21478E+15 
33.089 
5800.138 
45.843 
73.428 
15.339 

1.272 
3.577E-06 
0.001 
3.462 
0.276 

0.003 
0.358 
  

6.116 
11.178 
  

-1.971 
0.694 
0b 

1.930 
0.878 
  

  

  

  

  

  

  

  

-1.933  19.014 
1.870 
0.831 
5.410 
-1.938 
2.889 
-1.837 
2.769 
0.779 
1.025 
0.722 

 

Stand Type=Sand Dunes 
Stand Type=Tamarack* 
Stand Type=Treed Bog 
Stand Type=Upland Brush 
Stand Type=Upland Mixed* 
Stand Type=Upland Mixed 
Conifers 
Stand Type=Water 
Stand Type=White Pine 
Stand Type=White Spruce/Fir 
a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 
* Indicates significant results 

 

 
 

 

87

 

Table A.05: Multinomial logistic regression parameter estimates. Stand acres and stand type as independent variables (showing results 
from shelterwood harvest). N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

95% Confidence Interval for 

Exp(B) 

Exp(B) 

Lower 
Bound 

  

  

  

Upper Bound 

Forest Practicea 
Shelterwood 
Harvest 

Intercept 
Stand Acres* 
Stand Type=Agricultural 
Stand Type=Aspen* 
Stand Type=Black Spruce 
Stand Type=Bog or 
Muskeg* 
Stand Type=Cedar 
Stand Type=Grass* 
Stand Type=Hemlock* 
Stand Type=Jack Pine 
Stand Type=Local Use 
(various non-commercial or 
exotic) 
Stand Type=Lowland 
Brush* 
Stand Type=Lowland 
Hardwoods 
Stand Type=Lowland Mixed 
Stand Type=Lowland Poplar 
(Bam) 
Stand Type=Marsh 
Stand Type=Mixed 
Deciduous 
Stand Type=Mixed Lowland 
Conifers 
Stand Type=Non-Stocked* 
Stand Type=Northern 
Hardwoods 

B 
-3.461 
0.008 
-3.272 
-1.546 
0.243 
-3.221 

0.472 
-3.130 
1.350 
-2.669 
-55.471 

Std. 
Error 
0.327 
0.001 
5.038 
0.404 
0.471 
1.450 

0.373 
0.719 
0.449 
2.038 
0.000 

Wald 
112.123 
35.093 
0.422 
14.678 
0.267 
4.931 

1.604 
18.947 
9.040 
1.716 
  

-3.131 

1.063 

8.677 

0.470 

0.375 

1.568 

-0.742 
0.206 

0.670 
0.671 

-3.186 
0.876 

1.973 
0.486 

1.228 
0.094 

2.608 
3.251 

-0.532 

0.404 

1.736 

-3.157 
-0.080 

1.141 
0.368 

7.655 
0.048 

df 
1 
1 
1 
1 
1 
1 

1 
1 
1 
1 
1 

1 

1 

1 
1 

1 
1 

1 

1 
1 

 

88

Sig. 
0.000 
0.000 
0.516 
0.000 
0.605 
0.026 

0.205 
0.000 
0.003 
0.190 
  

1.008 
0.038 
0.213 
1.276 
0.040 

1.603 
0.044 
3.856 
0.069 
8.112E-25 

1.006 
1.951E-06 
0.097 
0.506 
0.002 

0.772 
0.011 
1.600 
0.001 
8.112E-25 

0.003 

0.044 

0.005 

0.211 

1.600 

0.767 

0.268 
0.759 

0.106 
0.071 

0.476 
1.229 

0.041 
2.401 

0.128 
0.330 

0.001 
0.927 

0.188 

0.587 

0.266 

0.006 
0.827 

0.043 
0.923 

0.005 
0.448 

1.011 
737.087 
0.470 
3.214 
0.685 

3.329 
0.179 
9.295 
3.762 
8.112E-25 

0.351 

3.337 

1.769 
4.579 

1.975 
6.223 

1.296 

0.398 
1.899 

 

Table A.05 (cont’d) 

 

Stand Type=Oak* 
Stand Type=Paper Birch* 
Stand Type=Red Pine 
Stand Type=Rock 
Stand Type=Sand Dunes 
Stand Type=Tamarack 
Stand Type=Treed Bog 
Stand Type=Upland Brush 
Stand Type=Upland Mixed 
Stand Type=Upland Mixed 
Conifers 
Stand Type=Water* 
Stand Type=White Pine 
Stand Type=White 
Spruce/Fir 

0.344 
1.703 
0.548 
1.949 
0.612 
0.373 
-3.136  13.642 
-3.130  15.752 
0.524 
0.313 
4.452 
-3.163 
2.342 
-3.045 
0.189 
0.571 
0.469 
0.723 

-3.191 
-0.099 
0b 

1.515 
0.472 
  

24.494 
12.665 
2.696 
0.053 
0.039 
0.357 
0.505 
1.691 
0.110 
2.373 

4.435 
0.044 
  

1 
1 
1 
1 
1 
1 
1 
1 
1 
1 

1 
1 
0 

0.000 
0.000 
0.101 
0.818 
0.842 
0.550 
0.477 
0.193 
0.741 
0.123 

0.035 
0.834 
  

5.488 
7.024 
1.845 
0.043 
0.044 
1.368 
0.042 
0.048 
1.208 
2.060 

0.041 
0.906 
  

2.796 
2.401 
0.888 

10.770 
20.553 
3.831 
1.061E-13  17803821030.679 
1.710E-15 
1.11794E+12 
3.823 
0.490 
260.270 
6.870E-06 
4.686 
0.000 
0.395 
3.698 
5.169 
0.821 

0.002 
0.360 
  

0.802 
2.283 
  

a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 
* Indicates significant results 

 

  

  

  

  

  

  

  

89

 
 

 

 

Table A.06: Multinomial logistic regression parameter estimates. Stand acres and stand type as independent variables (showing results 
from selection harvest). N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

Forest Practicea 
Selection 
Harvest 

Intercept 
Stand Acres* 
Stand Type=Agricultural 
Stand Type=Aspen* 
Stand Type=Black Spruce 
Stand Type=Bog or Muskeg* 
Stand Type=Cedar 
Stand Type=Grass* 
Stand Type=Hemlock* 
Stand Type=Jack Pine 
Stand Type=Local Use (various 
non-commercial or exotic) 
Stand Type=Lowland Brush* 
Stand Type=Lowland 
Hardwoods* 
Stand Type=Lowland Mixed 
Stand Type=Lowland Poplar 
(Bam)* 
Stand Type=Marsh* 
Stand Type=Mixed Deciduous* 
Stand Type=Mixed Lowland 
Conifers 
Stand Type=Non-Stocked* 
Stand Type=Northern 
Hardwoods* 
Stand Type=Oak* 
Stand Type=Paper Birch 
Stand Type=Red Pine* 
Stand Type=Rock 

Std. 
Error  Wald 

B 
-2.034  0.168 
0.009  0.001 
-2.361  1.578 
-0.562  0.184 
0.038  0.257 
-2.308  0.472 
-0.219  0.211 
-2.198  0.260 
1.472  0.240 
-0.452  0.379 
-54.556  0.000 

-2.131  0.346 
0.934  0.187 

0.366  0.248 
1.258  0.265 

-1.687  0.486 
1.966  0.221 
-0.276  0.200 

145.961 
155.838 
2.238 
9.366 
0.022 
23.904 
1.074 
71.526 
37.603 
1.422 
  

37.911 
24.865 

2.179 
22.522 

12.060 
78.985 
1.916 

95% Confidence Interval 

for Exp(B) 

Lower 
Bound 

Upper 
Bound 

Exp(B) 

  

  

  

1.009 
0.094 
0.570 
1.039 
0.099 
0.803 
0.111 
4.356 
0.636 
2.025E-24 

1.007 
0.004 
0.397 
0.628 
0.039 
0.531 
0.067 
2.722 
0.303 
2.025E-24 

1.010 
2.079 
0.817 
1.719 
0.251 
1.215 
0.185 
6.973 
1.338 
2.025E-24 

0.119 
2.543 

1.442 
3.520 

0.185 
7.144 
0.759 

0.106 
8.818 

4.191 
1.575 
4.498 
0.109 

0.060 
1.762 

0.887 
2.093 

0.071 
4.631 
0.513 

0.050 
6.278 

0.234 
3.671 

2.343 
5.919 

0.480 
11.023 
1.122 

0.224 
12.385 

2.931 
0.613 
3.146 
2.588E-05 

5.994 
4.050 
6.430 
455.701 

df 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 

1 
1 

1 
1 

1 
1 
1 

1 
1 

1 
1 
1 
1 

Sig. 
0.000 
0.000 
0.135 
0.002 
0.881 
0.000 
0.300 
0.000 
0.000 
0.233 
  

0.000 
0.000 

0.140 
0.000 

0.001 
0.000 
0.166 

0.000 
0.000 

0.000 
0.346 
0.000 
0.602 

-2.243  0.380 
2.177  0.173 

34.792 
157.743 

1.433  0.183 
0.454  0.482 
1.504  0.182 
-2.220  4.256 

61.658 
0.889 
68.001 
0.272 

 

90

-2.214  4.914 
-0.833  0.396 
-2.250  1.395 
-2.130  0.743 
1.145  0.241 
0.193  0.281 

-2.277  0.491 
0.632  0.212 
  

0b 

0.203 
4.418 
2.600 
8.219 
22.584 
0.473 

21.474 
8.896 
  

1 
1 
1 
1 
1 
1 

1 
1 
0 

0.652 
0.036 
0.107 
0.004 
0.000 
0.492 

0.000 
0.003 
  

0.109 
0.435 
0.105 
0.119 
3.142 
1.213 

0.103 
1.881 
  

7.178E-06 
0.200 
0.007 
0.028 
1.959 
0.699 

0.039 
1.242 
  

1663.015 
0.945 
1.624 
0.510 
5.038 
2.106 

0.269 
2.848 
  

  

  

  

  

  

  

  

 

Table A.06 (cont’d) 

 

Stand Type=Sand Dunes 
Stand Type=Tamarack* 
Stand Type=Treed Bog 
Stand Type=Upland Brush* 
Stand Type=Upland Mixed* 
Stand Type=Upland Mixed 
Conifers 
Stand Type=Water* 
Stand Type=White Pine* 
Stand Type=White Spruce/Fir 
a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 
* Indicates significant results 

 

 
 

 

91

 

Table A.07: Multinomial logistic regression parameter estimates. Stand acres and stand type as independent variables (showing results 
from thinning). N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

95% Confidence Interval 

for Exp(B) 

Forest Practicea 
Thinning 

Intercept 
Stand Acres* 
Stand Type=Agricultural 
Stand Type=Aspen* 
Stand Type=Black Spruce 
Stand Type=Bog or Muskeg* 
Stand Type=Cedar* 
Stand Type=Grass* 
Stand Type=Hemlock 
Stand Type=Jack Pine 
Stand Type=Local Use 
(various non-commercial or 
exotic) 
Stand Type=Lowland Brush* 
Stand Type=Lowland 
Hardwoods* 
Stand Type=Lowland Mixed 
Stand Type=Lowland Poplar 
(Bam)* 
Stand Type=Marsh* 
Stand Type=Mixed 
Deciduous* 
Stand Type=Mixed Lowland 
Conifers* 
Stand Type=Non-Stocked* 
Stand Type=Northern 
Hardwoods* 
Stand Type=Oak* 

Std. Error  Wald 

df  Sig. 

Exp(B) 

Upper 
Bound 

B 
-2.006 
0.008 
-2.775 
-1.033 
-0.309 
-2.724 
-1.487 
-2.617 
-0.418 
-0.449 

0.167  143.627  1  0.000   
1.009 
86.661  1  0.000 
0.001 
2.726 
2.073  1  0.150 
1.927 
0.517 
0.191 
29.300  1  0.000 
1.272 
1.215  1  0.270 
0.280 
0.198 
23.313  1  0.000 
0.564 
0.394 
0.284 
27.458  1  0.000 
0.130 
79.551  1  0.000 
0.293 
1.424 
1.127  1  0.289 
0.394 
1.416  1  0.234 
1.337 
0.377 
-14.227  337.748 
0.002  1  0.966  6.626E-07  2.135E-294  2.05E+281 

Lower Bound 
  
1.008 
0.062 
0.356 
0.734 
0.066 
0.226 
0.073 
0.658 
0.639 

  
1.006 
0.001 
0.245 
0.424 
0.022 
0.130 
0.041 
0.304 
0.305 

-2.509 
-0.521 

0.399 
0.225 

39.600  1  0.000 
5.372  1  0.020 

-0.540 
1.772 

0.316 
0.242 

2.911  1  0.088 
53.512  1  0.000 

-2.693 
1.340 

0.761 
0.238 

12.512  1  0.000 
31.609  1  0.000 

0.081 
0.594 

0.583 
5.882 

0.068 
3.819 

0.037 
0.383 

0.314 
3.659 

0.015 
2.394 

0.178 
0.923 

1.084 
9.456 

0.301 
6.093 

-1.329 

0.242 

30.201  1  0.000 

0.265 

0.165 

0.425 

-2.668 
1.241 

0.449 
0.176 

35.263  1  0.000 
49.943  1  0.000 

0.069 
3.458 

0.029 
2.451 

0.167 
4.879 

1.287 

0.183 

49.520  1  0.000 

3.622 

2.531 

5.184 

 

92

 

Table A.07 (cont’d)  

 

Stand Type=Paper Birch 
Stand Type=Red Pine* 
Stand Type=Rock 
Stand Type=Sand Dunes 
Stand Type=Tamarack* 
Stand Type=Treed Bog 
Stand Type=Upland Brush* 
Stand Type=Upland Mixed 
Stand Type=Upland Mixed 
Conifers* 
Stand Type=Water* 
Stand Type=White Pine* 
Stand Type=White Spruce/Fir 

-1.019 
0.804 
-2.651 
-2.646 
-1.016 
-1.632 
-2.558 
0.160 
0.822 

-2.697 
1.072 
0b 

0.881 
0.189 
5.201 
6.005 
0.423 
1.031 
0.900 
0.300 
0.242 

0.588 
0.199 
  

1.337  1  0.247 
18.052  1  0.000 
0.260  1  0.610 
0.194  1  0.660 
5.753  1  0.016 
2.508  1  0.113 
8.073  1  0.004 
0.286  1  0.593 
11.559  1  0.001 

21.007  1  0.000 
28.872  1  0.000 
  

   0 

0.361 
2.235 
0.071 
0.071 
0.362 
0.195 
0.077 
1.174 
2.275 

0.067 
2.920 
  

0.064 
1.542 
2.638E-06 
5.487E-07 
0.158 
0.026 
0.013 
0.652 
1.416 

2.029 
3.239 
1887.828 
9173.359 
0.831 
1.474 
0.452 
2.113 
3.653 

0.021 
1.975 
  

0.214 
4.317 
  

a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 

* Indicates significant results 
 
 

 

  

  

  

  

  

  

  

 

93

 

Table A.08: Multinomial logistic regression parameter estimates. Stand acres and stand type as independent variables (showing results 
from timber stand improvement). N=20,915 unique forest stands enrolled in the Qualified Forest Program. 

95% Confidence Interval for 

Exp(B) 

Lower 
Bound 

Exp(B) 

  

  

Upper Bound 

  

Sig. 
0.000 
0.001 
0.637 
0.761 
0.788 
0.131 
0.395 
0.017 
0.009 
0.768 
0.118 

0.085 
0.115 

1.005 
0.165 
1.172 
0.801 
0.170 
0.559 
0.199 
5.267 
1.298 
3.037 

0.237 
2.378 

1.002 
9.188E-05 
0.420 
0.159 
0.017 
0.146 
0.053 
1.510 
0.229 
0.754 

0.046 
0.809 

1.007 
295.669 
3.270 
4.032 
1.699 
2.136 
0.748 
18.365 
7.349 
12.228 

1.219 
6.991 

7.594 

Forest Practicea 
Timber 
Stand 
Improvement 

Intercept 
Stand Acres* 
Stand Type=Agricultural 
Stand Type=Aspen 
Stand Type=Black Spruce 
Stand Type=Bog or Muskeg 
Stand Type=Cedar 
Stand Type=Grass* 
Stand Type=Hemlock* 
Stand Type=Jack Pine 
Stand Type=Local Use 
(various non-commercial or 
exotic) 
Stand Type=Lowland Brush 
Stand Type=Lowland 
Hardwoods 
Stand Type=Lowland 
Mixed 
Stand Type=Lowland 
Poplar (Bam) 
Stand Type=Marsh 
Stand Type=Mixed 
Deciduous* 
Stand Type=Mixed 
Lowland Conifers 
Stand Type=Non-Stocked 
Stand Type=Northern 
Hardwoods* 
Stand Type=Oak* 

B 

-4.272 
0.005 
-1.803 
0.159 
-0.222 
-1.772 
-0.581 
-1.615 
1.661 
0.261 
1.111 

Std. 
Error  Wald 
73.510 
0.498 
10.984 
0.001 
0.222 
3.823 
0.523 
0.092 
0.072 
0.825 
2.276 
1.174 
0.722 
0.684 
0.676 
5.708 
6.797 
0.637 
0.087 
0.885 
0.711 
2.443 

-1.440 
0.866 

0.836 
0.550 

2.969 
2.479 

0.707 

0.674 

1.102 

1.349 

0.717 

3.541 

-1.736 
3.754 

1.546 
0.525 

1.262 
51.081 

-0.456 

0.614 

0.550 

-1.732 
2.482 

0.960 
0.505 

3.250 
24.179 

2.648 

0.509 

27.022 

df 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 

1 
1 

1 

1 

1 
1 

1 

1 
1 

1 

 

94

0.294 

2.028 

0.542 

0.060 

3.853 

0.945 

15.707 

0.261 
0.000 

0.176 
42.678 

0.009 
15.246 

0.458 

0.634 

0.190 

0.071 
0.000 

0.177 
11.963 

0.027 
4.448 

0.000 

14.128 

5.205 

3.645 
119.470 

2.114 

1.163 
32.169 

38.345 

 

Table A.08 (cont’d) 

 

Stand Type=Paper Birch 
Stand Type=Red Pine* 
Stand Type=Rock 
Stand Type=Sand Dunes 
Stand Type=Tamarack 
Stand Type=Treed Bog 
Stand Type=Upland Brush 
Stand Type=Upland Mixed* 
Stand Type=Upland Mixed 
Conifers* 
Stand Type=Water 
Stand Type=White Pine 
Stand Type=White 
Spruce/Fir 

-2.478 
5.597 
2.437 
0.512 
-1.735  10.240 
-1.732  11.821 
-1.285 
1.475 
3.371 
-1.719 
0.714 
0.850 
0.535 
3.147 
1.388 
0.622 

0.196 
22.704 
0.029 
0.021 
0.759 
0.260 
1.417 
34.568 
4.989 

-1.755 
1.063 
0b 

1.220 
0.579 
  

2.072 
3.367 
  

1 
1 
1 
1 
1 
1 
1 
1 
1 

1 
1 
0 

0.658 
0.000 
0.865 
0.884 
0.384 
0.610 
0.234 
0.000 
0.026 

0.150 
0.067 
  

0.084 
11.443 
0.176 
0.177 
0.277 
0.179 
2.339 
23.261 
4.008 

0.173 
2.895 
  

4871.931 
1.444E-06 
31.184 
4.199 
3.387E-10 
91819418.900 
1.534E-11  2040656707.531 
4.981 
132.623 
9.480 
66.407 
13.550 

0.015 
0.000 
0.577 
8.148 
1.185 

0.016 
0.930 
  

1.887 
9.012 
  

a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 
* Indicates significant results 

 

  

  

  

  

  

  

  

95

 

 

 

Table A.09: Multinomial logistic regression parameter estimates. Geographic region of Michigan as independent variable. N=20,915 
unique forest stands enrolled in the Qualified Forest Program. 

95% Confidence Interval 

for Exp(β) 

  

  

  

Sig. 
0.000 
0.865 
0.000 
0.000 
  
0.000 
0.894 
0.208 
0.190 
  
0.000 
0.005 
0.000 
0.134 
  

1 
1 
1 
1 
0 
1 
1 
1 
1 
0 
1 
1 
1 
1 
0 

Exp(β) 

Lower Bound 
  

1.045 
2.093 
2.907 
  

1.009 
1.078 
0.896 
  

2.166 
3.413 
1.640 
  

  

  

0.628 
1.384 
1.812 
  

0.883 
0.959 
0.760 
  

1.263 
2.106 
0.858 
  

  

  

  

Upper 
Bound 

1.741 
3.165 
4.662 
  

1.153 
1.211 
1.056 
  

3.713 
5.530 
3.136 
  

Wald 

df 

555.718 
0.029 
12.257 
19.596 
  
1067.827 
0.018 
1.584 
1.721 
  
446.804 
7.898 
24.854 
2.242 
  

Forest Practicea 
Artificial  
Regeneration 

Clearcut 

Salvage  
Treatment 

Intercept 
Region=EUP 
Region=NLP* 
Region=SLP* 
Region=WUP 
Intercept 
Region=EUP 
Region=NLP 
Region=SLP 
Region=WUP 
Intercept 
Region=EUP* 
Region=NLP* 
Region=SLP 
Region=WUP 

β 

-4.480 
0.044 
0.739 
1.067 
0b 
-1.612 
0.009 
0.075 
-0.110 
0b 
-4.868 
0.773 
1.228 
0.495 
0b 

Std. Error 
0.190 
0.260 
0.211 
0.241 
  
0.049 
0.068 
0.059 
0.084 
  
0.230 
0.275 
0.246 
0.331 
  

a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 
*  Indicates significant results 
 
 
 
 
 
 
 
 
 
 

 

96

 

 

Table A.10: Multinomial logistic regression parameter estimates. Geographic region of Michigan as independent variable. N=20,915 
unique forest stands enrolled in the Qualified Forest Program. 

Forest Practicea 
Seed Tree  
Harvest 

Shelterwood 
Harvest 

Intercept 
Region=EUP* 
Region=NLP* 
Region=SLP* 
Region=WUP 
Intercept 
Region=EUP 
Region=NLP 
Region=SLP 
Region=WUP 

β 

-3.669 
-0.426 
-0.458 
-0.761 
0b 
-3.608 
0.160 
-0.265 
0.106 
0b 

Std. Error 
0.128 
0.197 
0.169 
0.275 
  
0.124 
0.165 
0.158 
0.198 
  

Wald 

df 

827.103 
4.665 
7.339 
7.637 
  
849.011 
0.940 
2.825 
0.284 
  

95% Confidence Interval 

for Exp(β) 

  

  

Sig. 
0.000 
0.031 
0.007 
0.006 
  
0.000 
0.332 
0.093 
0.594 
  

1 
1 
1 
1 
0 
1 
1 
1 
1 
0 

Exp(β) 

Lower Bound 
  

0.653 
0.633 
0.467 
  

1.174 
0.767 
1.111 
  

  

0.444 
0.455 
0.272 
  

0.849 
0.563 
0.754 
  

  

  

Upper 
Bound 

0.961 
0.881 
0.802 
  

1.623 
1.045 
1.639 
  

a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 
*  Indicates significant results 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

97

 

 

Table A.11: Multinomial logistic regression parameter estimates. Geographic region of Michigan as independent variable. N=20,915 
unique forest stands enrolled in the Qualified Forest Program. 

Forest Practicea 
Selection 
Harvest 

Thinning 

Timber Stand 
Improvement 

Intercept 
Region=EUP* 
Region=NLP* 
Region=SLP* 
Region=WUP 
Intercept 
Region=EUP 
Region=NLP* 
Region=SLP* 
Region=WUP 
Intercept 
Region=EUP 
Region=NLP* 
Region=SLP* 
Region=WUP 

β 

-1.225 
-0.165 
-0.307 
0.440 
0b 
-2.499 
0.107 
1.153 
0.959 
0b 
-4.316 
0.326 
1.365 
2.390 
0b 

Std. Error 
0.042 
0.060 
0.054 
0.063 
  
0.073 
0.099 
0.079 
0.096 
  
0.175 
0.226 
0.186 
0.190 
  

a. The reference category is: No Practice. 
b. This parameter is set to zero because it is redundant. 
*  Indicates significant results 
 
 
 
 
 
 
 
 
 
 

95% Confidence Interval 

for Exp(β) 

  

  

  

Sig. 
0.000 
0.006 
0.000 
0.000 
  
0.000 
0.279 
0.000 
0.000 
  
0.000 
0.149 
0.000 
0.000 
  

1 
1 
1 
1 
0 
1 
1 
1 
1 
0 
1 
1 
1 
1 
0 

Exp(β) 

Lower Bound 
  

0.848 
0.736 
1.553 
  

1.113 
3.166 
2.610 
  

1.386 
3.916 
10.914 
  

  

  

0.753 
0.662 
1.371 
  

0.917 
2.711 
2.161 
  

0.890 
2.719 
7.516 
  

  

  

  

Upper 
Bound 

0.954 
0.817 
1.758 
  

1.350 
3.698 
3.153 
  

2.158 
5.639 
15.848 
  

Wald 

df 

841.816 
7.499 
32.740 
48.173 
  
1171.655 
1.170 
211.762 
98.983 
  
606.582 
2.083 
53.820 
157.733 
  

98

 

 

Table A.12: Multinomial logistic regression parameter estimates. Stand condition as independent variable. Eastern Upper Peninsula. 
N=20,915 unique forest stands enrolled in the Qualified Forest Program. † 

Forest Practiceb 
Artificial 
Regen-
eration 

β 

Std. 
Error 

-4.407 
-16.903 
0.646 
1.526 
-1.095 
-0.271 
0.462 
-0.042 
0.837 
0.730 
0c 
-0.851 

-2.217 
-2.030 
-1.283 
-0.587 
-0.176 
0.079 
-1.214 
-0.383 
0c 

Clearcut 

1.006 
0.000 
1.427 
1.090 
1.231 
1.232 
1.087 
1.087 
1.235 
1.235 
  
0.202 
-17.514  3076.338 
0.751 
0.466 
0.245 
0.255 
0.230 
0.217 
0.407 
0.311 
  

Intercept 
NS 
SS-P 
SS-M 
SS-W 
P-P 
P-M 
P-W 
S-P 
S-M 
S-W 
Intercept 
NS 
SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M 
P-W 
S-P* 
S-M 
S-W 
a. Region = EUP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
*  Indicates significant results 
 

Wald 
19.185 
  
0.205 
1.959 
0.791 
0.048 
0.180 
0.001 
0.459 
0.350 
  
17.780 
0.000 
8.712 
19.004 
27.395 
5.307 
0.582 
0.133 
8.882 
1.510 
  

df 

1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 

Sig. 
0.000 
  
0.651 
0.162 
0.374 
0.826 
0.671 
0.969 
0.498 
0.554 
  
0.000 
0.995 
0.003 
0.000 
0.000 
0.021 
0.445 
0.715 
0.003 
0.219 
  

Exp(β) 

  
4.560E-08 
1.907 
4.598 
0.335 
0.763 
1.587 
0.959 
2.310 
2.076 
  

  
2.476E-08 
0.109 
0.131 
0.277 
0.556 
0.839 
1.082 
0.297 
0.682 
  

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

  

  

4.560E-08 
0.116 
0.543 
0.030 
0.068 
0.188 
0.114 
0.205 
0.185 
  

4.560E-08 
31.243 
38.943 
3.733 
8.526 
13.368 
8.069 
26.013 
23.352 
  

  

  

0.000 
0.025 
0.053 
0.172 
0.337 
0.534 
0.708 
0.134 
0.371 
  

.d 
0.475 
0.327 
0.448 
0.916 
1.317 
1.655 
0.660 
1.256 
  

99

 

 
Table A.12 (cont’d) 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 

 

 

100

 

Table A.13: Multinomial logistic regression parameter estimates. Stand condition as independent variable. Eastern Upper Peninsula. 
N=20,915 unique forest stands enrolled in the Qualified Forest Program. † 

95% Confidence Interval for 

Exp(β) 

Forest Practiceb 
Salvage 
Treatment 

β 

Std. Error 

Seed Tree 
Harvest 

-3.020 
-17.838 
-17.514 
-18.865 
-2.481 
-1.657 
-2.023 
0.076 
-0.549 
0.260 
0c 
-3.308 
-17.527 
-17.204 
-18.555 
-1.500 
-0.676 
-0.126 
0.112 
-0.955 
-0.368 
0c 

Intercept 
NS 
SS-P 
SS-M 
SS-W* 
P-P 
P-M* 
P-W 
S-P 
S-M 
S-W 
Intercept 
NS 
SS-P 
SS-M 
SS-W 
P-P 
P-M 
P-W 
S-P 
S-M 
S-W 
a. Region = EUP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
*  Indicates significant results 
 

0.512 
0.000 
4388.911 
5467.087 
0.874 
0.876 
0.875 
0.549 
0.881 
0.689 
  
0.588 
0.000 
4339.871 
5406.000 
0.773 
0.775 
0.670 
0.629 
1.166 
0.926 
  

 

 

Exp(β) 

  

  

1.792E-08 
2.475E-08 
6.410E-09 
0.084 
0.191 
0.132 
1.079 
0.577 
1.297 
  

Lower 
Bound 

1.792E-08 
0.000 
0.000 
0.015 
0.034 
0.024 
0.368 
0.103 
0.336 
  

  

  

  

2.443E-08 
3.375E-08 
8.741E-09 
0.223 
0.509 
0.882 
1.119 
0.385 
0.692 
  

2.443E-08 
0.000 
0.000 
0.049 
0.111 
0.237 
0.326 
0.039 
0.113 
  

Upper Bound 
  

1.792E-08 
.d 
.d 
0.464 
1.061 
0.735 
3.163 
3.247 
5.008 
  

2.443E-08 
.d 
.d 
1.015 
2.321 
3.278 
3.836 
3.784 
4.252 
  

Wald 
34.795 
  
0.000 
0.000 
8.053 
3.581 
5.347 
0.019 
0.388 
0.143 
  
31.672 
  
0.000 
0.000 
3.765 
0.762 
0.035 
0.032 
0.670 
0.158 
  

df 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 

Sig. 
0.000 
  
0.997 
0.997 
0.005 
0.058 
0.021 
0.890 
0.533 
0.705 
  
0.000 
  
0.997 
0.997 
0.052 
0.383 
0.851 
0.858 
0.413 
0.691 
  

101

 

Table A.13 (cont’d) 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 

 

 

102

 

Table A.14: Multinomial logistic regression parameter estimates. Stand condition as independent variable. Eastern Upper Peninsula. 
N=20,915 unique forest stands enrolled in the Qualified Forest Program. † 

Forest Practiceb 
Shelterwood 
Harvest 

β 

Std. Error 

Wald 

95% Confidence Interval for 

Exp(β) 

Exp(β) 

  

  

Lower 
Bound 

Upper Bound 
  

1.352E-08 
0.381 
0.077 
0.033 
0.305 
0.635 
0.448 
0.346 
0.519 
  

  

  

9.045E-09 
0.016 
0.013 
0.015 
0.054 
0.205 
0.336 
0.659 
0.884 
  

0.000 
0.080 
0.010 
0.007 
0.116 
0.292 
0.210 
0.092 
0.170 
  

0.000 
0.002 
0.003 
0.008 
0.030 
0.143 
0.244 
0.427 
0.589 
  

  

.d 
1.819 
0.611 
0.156 
0.800 
1.380 
0.956 
1.308 
1.586 
  

.d 
0.117 
0.053 
0.029 
0.095 
0.295 
0.461 
1.017 
1.326 
  

df 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 

Sig. 
0.000 
0.998 
0.227 
0.015 
0.000 
0.016 
0.252 
0.038 
0.118 
0.250 
  
0.007 
0.995 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
0.059 
0.550 
  

Intercept 
NS 
SS-P 
SS-M* 
SS-W* 
P-P* 
P-M 
P-W* 
S-P 
S-M 
S-W 
Intercept 
NS 
SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P 
S-M 
S-W 

Selection 
Harvest 

-2.104 
-18.119 
-0.964 
-2.569 
-3.397 
-1.187 
-0.454 
-0.804 
-1.060 
-0.656 
0c 
0.389 
-18.521 
-4.150 
-4.369 
-4.186 
-2.926 
-1.582 
-1.092 
-0.418 
-0.124 
0c 

0.335 
7788.446 
0.797 
1.059 
0.784 
0.492 
0.396 
0.387 
0.678 
0.570 
  
0.143 
2737.192 
1.022 
0.728 
0.337 
0.290 
0.185 
0.162 
0.221 
0.207 
  

39.461 
0.000 
1.462 
5.883 
18.788 
5.826 
1.314 
4.313 
2.444 
1.324 
  
7.399 
0.000 
16.503 
36.025 
154.476 
102.037 
72.948 
45.263 
3.557 
0.358 
  

103

a. Region = EUP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
*  Indicates significant results 
 

 

 

 

Table A.14 (cont’d) 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 

 

 

104

 

Table A.15: Multinomial logistic regression parameter estimates. Stand condition as independent variable. Eastern Upper Peninsula. 
N=20,915 unique forest stands enrolled in the Qualified Forest Program. † 

Forest Practiceb 
Thinning 

β 

Std. Error  Wald 

Intercept 
NS 
SS-P 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P 
S-M* 
S-W 
Intercept 
NS 
SS-P* 
SS-M 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-M 
S-W 
a. Region = EUP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
 

-0.941 
-18.229 
-17.906 
-3.732 
-2.856 
-2.032 
-1.058 
-0.714 
0.112 
-0.944 
0c 
-1.462 
-19.288 
-2.299 
-20.316 
-2.786 
-3.908 
-2.328 
-2.380 
-1.702 
-18.096 
0c 

0.208 
4599.402 
1887.099 
1.026 
0.369 
0.373 
0.265 
0.236 
0.300 
0.373 
  
0.255 
0.000 
1.043 
5180.650 
0.458 
1.034 
0.459 
0.397 
0.642 
1987.331 
  

Timber 
Stand 
Improve-
ment 

 

 

95% Confidence Interval for 

Exp(β) 

Exp(β) 

  

  

Lower 
Bound 

Upper Bound 
  

1.211E-08 
1.673E-08 
0.024 
0.058 
0.131 
0.347 
0.490 
1.119 
0.389 
  

0.000 
0.000 
0.003 
0.028 
0.063 
0.206 
0.308 
0.622 
0.187 
  

.d 
.d 
0.179 
0.119 
0.272 
0.584 
0.777 
2.013 
0.809 
  

  

  

  

4.200E-09 
0.100 
1.503E-09 
0.062 
0.020 
0.097 
0.093 
0.182 
1.383E-08 
  

4.200E-09 
0.013 
0.000 
0.025 
0.003 
0.040 
0.042 
0.052 
0.000 
  

4.200E-09 
0.775 
.d 
0.151 
0.152 
0.240 
0.202 
0.642 
.d 
  

df 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 

Sig. 
0.000 
0.997 
0.992 
0.000 
0.000 
0.000 
0.000 
0.002 
0.708 
0.012 
  
0.000 
  
0.028 
0.997 
0.000 
0.000 
0.000 
0.000 
0.008 
0.993 
  

20.381 
0.000 
0.000 
13.228 
59.782 
29.697 
15.879 
9.178 
0.140 
6.385 
  
32.984 
  
4.857 
0.000 
37.015 
14.283 
25.705 
35.925 
7.025 
0.000 
  

105

 

Table A.15 (cont’d) 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 

 

 

106

 

Table A.16: Multinomial logistic regression parameter estimates. Stand condition as independent variable. Northern Lower Peninsula. 
N=20,915 unique forest stands enrolled in the Qualified Forest Program. † 

Forest Practiceb 
Artificial 
Regen-
eration 

Intercept 
NS 
SS-P* 
SS-M 
SS-W 
P-P 
P-M 
P-W 
S-P 
S-M 
S-W 
Intercept 
NS 

Clearcut 

β 
-3.384 
24.095 
2.409 
-0.527 
-0.884 
-0.571 
-0.488 
-0.096 
0.149 
0.399 
0c 
-0.630 
-1.592 

Std. Error 
0.326 
0.000 
0.404 
0.638 
0.455 
0.512 
0.480 
0.422 
0.424 
0.418 
  
0.100 
0.000 

SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-M* 
S-W 
a. Region = NLP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
 

-1.510 
-1.764 
-1.502 
-1.893 
-0.326 
0.601 
-0.727 
-0.305 
0c 

0.398 
0.284 
0.152 
0.223 
0.138 
0.120 
0.154 
0.147 
  

 

 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Sig. 

Exp(β) 

0.000 
  
0.000 
0.409 
0.052 
0.265 
0.309 
0.819 
0.726 
0.340 
  
0.000 
  

0.000 
0.000 
0.000 
0.000 
0.019 
0.000 
0.000 
0.039 
  

  

  
2.911E+10  2.911E+10 
5.040 
0.169 
0.169 
0.207 
0.240 
0.397 
0.505 
0.657 
  

11.124 
0.590 
0.413 
0.565 
0.614 
0.908 
1.160 
1.490 
  

  

  

2.035E-01 

0.204 

0.221 
0.171 
0.223 
0.151 
0.722 
1.824 
0.483 
0.737 
  

0.101 
0.098 
0.165 
0.097 
0.551 
1.442 
0.358 
0.552 
  

  

2.911E+10 
24.553 
2.062 
1.008 
1.543 
1.572 
2.077 
2.665 
3.381 
  

  
0.20354829
4 
0.482 
0.299 
0.300 
0.233 
0.947 
2.307 
0.654 
0.984 
  

Wald 
108.005 
  
35.567 
0.682 
3.774 
1.240 
1.034 
0.052 
0.123 
0.910 
  
39.694 
  

14.368 
38.531 
97.385 
72.193 
5.544 
25.105 
22.310 
4.272 
  

df 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 

1 
1 
1 
1 
1 
1 
1 
1 
0 

107

 

Table A.16 (cont’d) 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 

 

 

108

 

Table A.17: Multinomial logistic regression parameter estimates. Stand condition as independent variable. Northern Lower Peninsula. 
N=20,915 unique forest stands enrolled in the Qualified Forest Program. † 

Forest Practiceb 
Salvage 
Treatment 

Seed Tree 
Harvest 

Intercept 
NS 
SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-M* 
S-W 
Intercept 
NS 
SS-P 
SS-M 
SS-W* 
P-P* 
P-M 
P-W 
S-P 
S-M 
S-W 
a. Region = NLP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
 

β 
-1.121 
-3.260 
-2.623 
-3.335 
-3.736 
-3.531 
-2.583 
-2.739 
-1.651 
-1.592 
0c 
-3.149 
-1.694 
-1.233 
-1.474 
-1.966 
-1.738 
-0.653 
-0.672 
0.348 
0.496 
0c 

Std. Error 
0.119 
0.000 
0.832 
0.728 
0.442 
0.571 
0.346 
0.345 
0.248 
0.259 
  
0.291 
0.000 
1.163 
0.832 
0.564 
0.691 
0.448 
0.430 
0.365 
0.367 
  

 

 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Sig. 

Exp(β) 

0.000 
  
0.002 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
  
0.000 
  
0.289 
0.076 
0.000 
0.012 
0.145 
0.119 
0.341 
0.177 
  

  

  

  

3.841E-02 
0.073 
0.036 
0.024 
0.029 
0.076 
0.065 
0.192 
0.204 
  

3.841E-02 
0.014 
0.009 
0.010 
0.010 
0.038 
0.033 
0.118 
0.122 
  

3.841E-02 
0.370 
0.148 
0.057 
0.090 
0.149 
0.127 
0.312 
0.338 
  

  

  

  

1.837E-01 
0.291 
0.229 
0.140 
0.176 
0.520 
0.511 
1.416 
1.641 
  

0.184  0.183700594 
2.845 
0.030 
0.045 
1.170 
0.423 
0.046 
0.681 
0.045 
0.216 
1.252 
1.187 
0.220 
2.897 
0.692 
3.370 
0.799 
  
  

Wald 

88.881 
  
9.949 
20.994 
71.538 
38.295 
55.795 
63.143 
44.147 
37.656 
  
117.299 
  
1.125 
3.138 
12.146 
6.328 
2.127 
2.436 
0.907 
1.822 
  

df 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 

109

 

Table A.17 (cont’d) 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 

 

 

110

 

Table A.18: Multinomial logistic regression parameter estimates. Stand condition as independent variable. Northern Lower Peninsula. 
N=20,915 unique forest stands enrolled in the Qualified Forest Program. † 

Forest Practiceb 
Shelter-
wood 
Harvest 

Selection 
Harvest 

Intercept 
NS 
SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P 
S-M 
S-W 
Intercept 
NS 

SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-M* 
S-W 

β 
-2.479 
-2.083 
0.754 
-2.059 
-2.670 
-1.658 
-1.036 
-1.212 
-0.347 
-0.178 
0c 
0.054 
-2.278 

-1.950 
-2.478 
-2.768 
-2.654 
-1.288 
-1.021 
-0.883 
-0.250 
0c 

Std. Error 
0.212 
0.000 
0.385 
0.777 
0.535 
0.482 
0.364 
0.366 
0.308 
0.309 
  
0.082 
0.000 

0.356 
0.282 
0.171 
0.222 
0.134 
0.121 
0.126 
0.119 
  

Wald 
136.694 
  
3.831 
7.023 
24.876 
11.846 
8.092 
11.002 
1.267 
0.333 
  
0.424 
  

30.022 
77.242 
262.101 
142.750 
92.145 
71.360 
48.758 
4.445 
  

df 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 

1 
1 
1 
1 
1 
1 
1 
1 
0 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Sig. 

Exp(β) 

0.000 
  
0.050 
0.008 
0.000 
0.001 
0.004 
0.001 
0.260 
0.564 
  
0.515 
  

0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
0.035 
  

  

  

  

1.245E-01 
2.125 
0.128 
0.069 
0.191 
0.355 
0.297 
0.707 
0.837 
  

1.245E-01 
0.999 
0.028 
0.024 
0.074 
0.174 
0.145 
0.386 
0.457 
  

  

  

1.024E-01 

0.102 

0.142 
0.084 
0.063 
0.070 
0.276 
0.360 
0.414 
0.779 
  

0.071 
0.048 
0.045 
0.046 
0.212 
0.284 
0.323 
0.617 
  

1.245E-01 
4.522 
0.585 
0.198 
0.490 
0.725 
0.609 
1.293 
1.533 
  

  
0.10244028
4 
0.286 
0.146 
0.088 
0.109 
0.359 
0.456 
0.530 
0.983 
  

a. Region = NLP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
 

 

 

111

 

Table A.18 (cont’d) 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 

 

 

112

 

Table A.19: Multinomial logistic regression parameter estimates. Stand condition as independent variable. Northern Lower Peninsula. 
N=20,915 unique forest stands enrolled in the Qualified Forest Program. † 

Forest Practiceb 
Thinning 

Timber 
Stand 
Improve-
ment 

Intercept 
NS 
SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P 
S-M 
S-W 
Intercept 
NS 

SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-M* 
S-W 

β 
-2.479 
-2.083 
0.754 
-2.059 
-2.670 
-1.658 
-1.036 
-1.212 
-0.347 
-0.178 
0c 
0.054 
-2.278 

-1.950 
-2.478 
-2.768 
-2.654 
-1.288 
-1.021 
-0.883 
-0.250 
0c 

Std. Error 
0.212 
0.000 
0.385 
0.777 
0.535 
0.482 
0.364 
0.366 
0.308 
0.309 
  
0.082 
0.000 

0.356 
0.282 
0.171 
0.222 
0.134 
0.121 
0.126 
0.119 
  

Wald 
136.694 
  
3.831 
7.023 
24.876 
11.846 
8.092 
11.002 
1.267 
0.333 
  
0.424 
  

30.022 
77.242 
262.101 
142.750 
92.145 
71.360 
48.758 
4.445 
  

df 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 

1 
1 
1 
1 
1 
1 
1 
1 
0 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Sig. 

Exp(β) 

0.000 
  
0.050 
0.008 
0.000 
0.001 
0.004 
0.001 
0.260 
0.564 
  
0.515 
  

0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
0.035 
  

  

  

  

1.245E-01 
2.125 
0.128 
0.069 
0.191 
0.355 
0.297 
0.707 
0.837 
  

1.245E-01 
0.999 
0.028 
0.024 
0.074 
0.174 
0.145 
0.386 
0.457 
  

  

  

1.024E-01 

0.102 

0.142 
0.084 
0.063 
0.070 
0.276 
0.360 
0.414 
0.779 
  

0.071 
0.048 
0.045 
0.046 
0.212 
0.284 
0.323 
0.617 
  

1.245E-01 
4.522 
0.585 
0.198 
0.490 
0.725 
0.609 
1.293 
1.533 
  

  
0.10244028
4 
0.286 
0.146 
0.088 
0.109 
0.359 
0.456 
0.530 
0.983 
  

a. Region = NLP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
 

 

 

113

 

Table A.19 (cont’d) 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 

 

 

114

 

Table A.20: Multinomial logistic regression parameter estimates. Stand condition as independent variable. Southern Lower Peninsula. 
N=20,915 unique forest stands enrolled in the Qualified Forest Program. † 

Wald 

Sig. 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Exp(β) 

Forest Practiceb 
Artificial 
Regen-
eration 

Clearcut 

  

  

  

4.323E+05 
.c 
5.477 
1.851 
5.945 
2.801 
2.710 
8.148 
4.974 
  

β 
-3.090 
-2.942 
-12.652 
-0.146 
-1.768 
0.472 
-0.555 
-0.524 
0.776 
0.093 
0d 
-0.796 
-3.291 
-24.345 
-2.025 
-2.185 
-0.968 
-0.300 
0.253 
-0.523 
-0.025 
0d 

  

  
5.276E-02  6.438E-09 
0.000 
0.136 
0.016 
0.433 
0.118 
0.129 
0.579 
0.242 
  

0.000 
0.864 
0.171 
1.604 
0.574 
0.592 
2.172 
1.098 
  

Std. Error 
0.569 
29.500 
8.122  0.131208647 
0.000 
0.024 
2.113 
0.499 
0.471 
0.456 
1.323 
0.015 
  
13.980 
1.133 
  
9.655 
20.802 
9.136 
1.139 
0.981 
2.704 
0.007 
  

Intercept 
NS 
SS-P 
SS-M 
SS-W 
P-P 
P-M 
P-W 
S-P 
S-M 
S-W 
Intercept 
NS 
SS-P 
SS-M* 
SS-W* 
P-P* 
P-M 
P-W 
S-P 
S-M 
S-W 
a. Region = SLP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 

df 
1 
0.000 
1  0.717182149 
0.999 
1 
1 
0.877 
0.146 
1 
0.480 
1 
0.493 
1 
1 
0.499 
0.250 
1 
0.904 
1 
0 
  
0.000 
1 
0.287 
1 
  
1 
1 
0.002 
0.000 
1 
0.003 
1 
0.286 
1 
1 
0.322 
0.100 
1 
0.933 
1 
0 
  

0.000  15.93535024 
0.000 
0.000 
0.037 
0.474 
0.288 
0.044 
0.712 
0.203 
1.285 
0.427 
0.781 
2.124 
1.106 
0.318 
1.741 
0.547 
  
  

8052.235 
0.942 
1.216 
0.669 
0.809 
0.776 
0.674 
0.771 
  
0.213 
3.092 
0.000 
0.652 
0.479 
0.320 
0.281 
0.255 
0.318 
0.296 
  

  
3.722E-02 
0.000 
0.132 
0.112 
0.380 
0.741 
1.288 
0.593 
0.976 
  

 

115

 

Table A.21: Multinomial logistic regression parameter estimates. Stand condition as independent variable. Southern Lower Peninsula. 
N=20,915 unique forest stands enrolled in the Qualified Forest Program. † 

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Exp(β) 

  

  

  

Forest Practiceb 
Salvage 
Treatment 

Seed Tree 
Harvest 

Std. Error 
0.860 

Wald 

1.039E-11 
0.000 
0.071 
0.002 
0.178 
1.118 
0.036 
0.492 
0.142 
  

β 
-3.943 
-2.783 
-23.837 
0.013 
-2.118 
0.300 
1.881 
-0.826 
1.188 
0.252 
0d 
-3.027 
-3.756 
-24.810 
-2.974 
-3.091 
-1.259 
-0.227 
-1.160 
0.334 
-1.037 
0d 

Intercept 
NS 
SS-P 
SS-M 
SS-W 
P-P 
P-M* 
P-W 
S-P 
S-M 
S-W 
Intercept 
NS 
SS-P 
SS-M 
SS-W 
P-P 
P-M 
P-W 
S-P 
S-M 
S-W 
a. Region = SLP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly 
stocked, S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 

20.999 
11.484  0.058726441 
  
0.000 
0.983 
0.084 
4.341 
0.424 
1.505 
0.050 
  
30.073 
0.101 
  
0.990 
2.195 
1.720 
0.097 
1.696 
0.227 
1.008 
  

0.000  256624758.2 
0.000 
0.000 
17.914 
0.000 
0.001 
2.712 
1.863 
0.043 
3.322 
0.191 
1.796 
0.055 
0.354 
5.510 
2.683 
0.047 
  
  

Sig. 

df 
1 
0.000 
1  0.808520294 
1 
  
0.992 
1 
0.321 
1 
0.772 
1 
0.037 
1 
1 
0.515 
0.220 
1 
0.823 
1 
0 
  
0.000 
1 
0.750 
1 
1 
  
0.320 
1 
1 
0.138 
0.190 
1 
0.755 
1 
0.193 
1 
1 
0.634 
0.315 
1 
0 
  

0.000 
1.359 
2.136 
1.034 
0.903 
1.268 
0.968 
1.126 
  
0.552 
11.796 
0.000 
2.990 
2.086 
0.960 
0.728 
0.890 
0.701 
1.032 
  

3.683E+08 
0.000 
14.537 
7.918 
10.236 
38.468 
5.257 
21.879 
11.704 
  

6.186E-02 
0.000 
1.013 
0.120 
1.350 
6.558 
0.438 
3.280 
1.287 
  

2.337E-02 
0.000 
0.051 
0.045 
0.284 
0.797 
0.314 
1.396 
0.355 
  

  

  

  

 
 

 

 

116

 

Table A.22: Multinomial logistic regression parameter estimates. Stand condition as independent variable. Southern Lower Peninsula. 
N=20,915 unique forest stands enrolled in the Qualified Forest Program. † 

Forest Practiceb 
Shelterwood 
Harvest 

Intercept 
NS 

β 
-0.683 
-5.172 

Std. Error 
0.205 
7.418 

Selection 
Harvest 

SS-P 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-M* 
S-W 
Intercept 
NS* 
SS-P 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-M 
S-W 

-26.226 
-2.683 
-4.076 
-4.199 
-3.475 
-2.347 
-2.609 
-1.705 
0d 
1.071 
-4.238 
-25.292 
-3.457 
-3.481 
-2.990 
-1.663 
-1.596 
-1.244 
-0.149 
0d 

0.000 
0.825 
1.043 
1.074 
0.766 
0.448 
0.609 
0.442 
  
0.138 
1.976 
0.000 
0.521 
0.356 
0.290 
0.206 
0.196 
0.211 
0.192 
  

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Sig. 

Exp(β) 

0.001 
0.4856484
6 
  
0.001 
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
  
0.000 
0.032 
  
0.000 
0.000 
0.000 
0.000 
0.000 
0.000 
0.437 
  

  

  

  

5.673E-03  2.752E-09 

1.169E+04 

0.000 
0.068 
0.017 
0.015 
0.031 
0.096 
0.074 
0.182 
  

0.000 
0.014 
0.002 
0.002 
0.007 
0.040 
0.022 
0.076 
  

0.000 
0.344 
0.131 
0.123 
0.139 
0.230 
0.243 
0.432 
  

  

  

  

1.443E-02 
0.000 
0.032 
0.031 
0.050 
0.189 
0.203 
0.288 
0.861 
  

0.000  0.694413539 
0.000 
0.000 
0.088 
0.011 
0.062 
0.015 
0.089 
0.028 
0.127 
0.284 
0.298 
0.138 
0.436 
0.190 
0.591 
1.255 
  
  

Wald 

11.107 
0.48615286
6 
  
10.574 
15.262 
15.279 
20.554 
27.471 
18.351 
14.914 
  
60.700 
4.599 
  
43.943 
95.523 
106.041 
65.098 
66.049 
34.609 
0.603 
  

df 
1 
1 

1 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 
1 
1 
1 
1 
1 
1 
1 
1 
0 

a. Region = SLP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
 

 

 

117

 

Table A.22 (cont’d) 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 

 

 

118

 

Table A.23: Multinomial logistic regression parameter estimates. Stand condition as independent variable. Southern Lower Peninsula. 
N=20,915 unique forest stands enrolled in the Qualified Forest Program. † 

Forest Practiceb 
Thinning 

Intercept 
NS 

β 
-0.427 
-3.475 

Std. Error 
0.189 
2.824 

Timber 
Stand 
Improve-
ment 

SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W 
S-P* 
S-M 
S-W 
Intercept 
NS 

SS-P 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-M 
S-W 

6.196 
-1.915 
-1.940 
-2.308 
-0.739 
0.247 
-0.573 
-0.260 
0d 
-0.588 
-3.735 

-24.789 
-2.163 
-2.351 
-2.068 
-1.214 
-0.547 
-1.118 
0.070 
0d 

3.084 
0.528 
0.373 
0.416 
0.267 
0.227 
0.282 
0.272 
  
0.199 
3.470 

0.000 
0.628 
0.465 
0.409 
0.314 
0.264 
0.341 
0.272 
  

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Sig. 

Exp(β) 

0.024 
0.21841659
3 
0.045 
0.000 
0.000 
0.000 
0.006 
0.278 
0.042 
0.339 
  
0.003 
0.282 

  
0.001 
0.000 
0.000 
0.000 
0.038 
0.001 
0.797 
  

  

  

  

3.095E-02 

1.222E-04 

7.839E+00 

490.585 
0.147 
0.144 
0.099 
0.477 
1.280 
0.564 
0.771 
  

1.164 
0.052 
0.069 
0.044 
0.283 
0.819 
0.324 
0.452 
  

  

  

2.388E-02 

0.000 

0.000 
0.115 
0.095 
0.126 
0.297 
0.579 
0.327 
1.072 
  

0.000 
0.034 
0.038 
0.057 
0.160 
0.345 
0.168 
0.629 
  

206821.405 
0.415 
0.299 
0.225 
0.805 
1.999 
0.980 
1.314 
  

  
21.4794808
7 
0.000 
0.394 
0.237 
0.282 
0.550 
0.970 
0.638 
1.827 
  

Wald 

5.110 
1.51474416
8 
4.037 
13.152 
26.987 
30.784 
7.695 
1.176 
4.130 
0.915 
  
8.765 
1.158 

  
11.854 
25.532 
25.605 
14.923 
4.302 
10.755 
0.066 
  

df 
1 
1 

1 
1 
1 
1 
1 
1 
1 
1 
0 
1 
1 

1 
1 
1 
1 
1 
1 
1 
1 
0 

a. Region = SLP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
 

 

119

 

Table A.23 (cont’d) 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 
 

 

 

120

 

Table A.24: Multinomial logistic regression parameter estimates. Stand condition as independent variable. Western Upper Peninsula. 
N=20,915 unique forest stands enrolled in the Qualified Forest Program. † 

95% Confidence Interval 

for Exp(β) 

Std. Error 
0.608 

Wald 

Sig. 

Exp(β) 

Lower 
Bound 

Forest Practiceb 
Artificial 
Regenera
tion 

Clearcut 

  

  

Upper Bound 
  

9.000E+22 
0.000 
1.310 
0.402 
1.086 
1.056 
0.536 
2.602 
7.996 
  

β 
-2.681 
-3.978 
-39.578 
-1.608 
-2.727 
-1.473 
-1.434 
-2.621 
-0.720 
0.578 
0c 
-2.216 
-1.542 
-23.733 
-0.706 
0.217 
0.210 
1.351 
1.854 
0.685 
1.596 
0c 

  

  
1.873E-02  3.897E-27 
0.000 
0.031 
0.011 
0.048 
0.054 
0.010 
0.091 
0.397 
  

0.000 
0.200 
0.065 
0.229 
0.238 
0.073 
0.487 
1.783 
  

Intercept 
NS 
SS-P 
SS-M 
SS-W* 
P-P 
P-M 
P-W* 
S-P 
S-M 
S-W 
Intercept 
NS 
SS-P 
SS-M 
SS-W 
P-P 
P-M* 
P-W* 
S-P 
S-M* 
S-W 
a. Region = WUP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 

19.448 
28.996  0.018817855 
  
2.816 
8.655 
3.443 
3.567 
6.619 
0.709 
0.570 
  
20.380 
0.050 
  
1.289 
0.182 
0.160 
7.240 
13.807 
1.533 
8.259 
  

df 
1 
0.000 
1  0.890889952 
  
1 
1 
0.093 
0.003 
1 
0.064 
1 
0.059 
1 
1 
0.010 
0.400 
1 
0.450 
1 
  
0 
1 
0.000 
0.823 
1 
  
1 
0.256 
1 
1 
0.670 
0.689 
1 
0.007 
1 
0.000 
1 
1 
0.216 
0.004 
1 
0 
  

0.000  156677.4199 
0.000 
0.000 
1.669 
0.146 
0.458 
3.370 
3.446 
0.441 
10.331 
1.443 
16.981 
2.402 
0.671 
5.874 
14.646 
1.661 
  
  

  
2.139E-01 
0.000 
0.494 
1.243 
1.233 
3.861 
6.386 
1.985 
4.932 
  

0.000 
0.958 
0.927 
0.794 
0.759 
1.019 
0.855 
0.766 
  
0.491 
6.890 
0.000 
0.621 
0.509 
0.524 
0.502 
0.499 
0.554 
0.555 
  

 
 

 

 

121

 

Table A.25: Multinomial logistic regression parameter estimates. Stand condition as independent variable. Western Upper Peninsula. 
N=20,915 unique forest stands enrolled in the Qualified Forest Program. † 

95% Confidence Interval 

for Exp(β) 

Exp(β) 

Lower 
Bound 

  

  

Upper Bound 
  

Forest Practiceb 
Salvage 
Treatment 

Seed Tree 
Harvest 

Std. Error 
2.663 

Wald 

8.559E-31 
0.000 
0.000 
0.003 
0.005 
0.065 
0.014 
0.097 
0.021 
  

8.471E+28 
0.000 
473.096 
190.482 
319.051 
2353.961 
628.916 
3798.587 
1685.974 
  

β 
-5.698 
-1.312 
-47.455 
-1.089 
-0.278 
0.192 
2.514 
1.105 
2.953 
1.785 
0c 
-2.523 
-3.288 
2.166 
-2.303 
-2.938 
-1.859 
-1.085 
0.544 
-0.611 
0.395 
0c 

Intercept 
NS 
SS-P 
SS-M 
SS-W 
P-P 
P-M 
P-W 
S-P 
S-M 
S-W 
Intercept 
NS 
SS-P* 
SS-M* 
SS-W* 
P-P* 
P-M 
P-W 
S-P 
S-M 
S-W 
a. Region = WUP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 

4.579 
34.654  0.001433478 
  
0.087 
0.010 
0.005 
0.881 
0.164 
1.197 
0.384 
  
19.953 
0.030 
10.732 
4.235 
10.336 
5.355 
2.645 
0.855 
0.623 
0.289 
  

0.000  5.58784E+14 
31.859 
2.387 
0.011 
0.896 
0.318 
0.009 
0.752 
0.032 
1.249 
0.091 
0.544 
5.461 
2.474 
0.119 
6.273 
0.351 
  
  

Sig. 

df 
1 
0.032 
1  0.969798273 
1 
  
0.768 
1 
0.922 
1 
0.946 
1 
0.348 
1 
1 
0.685 
0.274 
1 
0.535 
1 
0 
  
0.000 
1 
0.863 
1 
0.001 
1 
1 
0.040 
0.001 
1 
0.021 
1 
0.104 
1 
1 
0.355 
0.430 
1 
0.591 
1 
0 
  

0.000 
3.698 
2.820 
2.844 
2.678 
2.724 
2.699 
2.880 
  
0.565 
19.003 
0.661 
1.119 
0.914 
0.803 
0.667 
0.588 
0.774 
0.735 
  

2.693E-01 
0.000 
0.337 
0.757 
1.212 
12.355 
3.018 
19.154 
5.961 
  

3.731E-02 
8.720 
0.100 
0.053 
0.156 
0.338 
1.723 
0.543 
1.485 
  

  

  

  

 

122

 

Table A.26: Multinomial logistic regression parameter estimates. Stand condition as independent variable. Western Upper Peninsula. 
N=20,915 unique forest stands enrolled in the Qualified Forest Program. † 

Forest Practiceb 
Shelterwood 
Harvest 

Selection 
Harvest 

Intercept 
NS 
SS-P 
SS-M* 
SS-W* 
P-P* 
P-M 
P-W 
S-P 
S-M 
S-W 
Intercept 
NS 
SS-P 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W* 
S-P* 
S-M 
S-W 

β 
-2.275 
-3.553 
-49.695 
-2.557 
-2.754 
-1.886 
-0.300 
-0.476 
-0.276 
-0.874 
0c 
0.861 
-4.243 
-50.386 
-3.645 
-3.706 
-3.225 
-1.817 
-1.105 
-0.704 
0.294 
0c 

Std. Error 
0.504 

Wald 

20.364 
19.153  0.034405754 
  
5.483 
12.859 
6.889 
0.299 
0.733 
0.184 
0.923 
  
21.987 
0.549 
  
82.384 
186.975 
129.105 
72.331 
29.595 
8.960 
1.331 
  

0.000 
1.092 
0.768 
0.719 
0.549 
0.556 
0.645 
0.910 
  
0.184 
5.728 
0.000 
0.402 
0.271 
0.284 
0.214 
0.203 
0.235 
0.254 
  

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

Upper 
Bound 

Exp(β) 

  

  

  

2.865E-02 
0.000 
0.078 
0.064 
0.152 
0.741 
0.621 
0.759 
0.417 
  

1.425E-18 
0.000 
0.009 
0.014 
0.037 
0.253 
0.209 
0.214 
0.070 
  

5.760E+14 
0.000 
0.659 
0.287 
0.620 
2.171 
1.847 
2.683 
2.481 
  

  

  

  

1.436E-02 
0.000 
0.026 
0.025 
0.040 
0.162 
0.331 
0.494 
1.341 
  

0.000  1077.857084 
0.000 
0.000 
0.012 
0.057 
0.042 
0.014 
0.069 
0.023 
0.247 
0.107 
0.222 
0.493 
0.784 
0.312 
2.208 
0.815 
  
  

Sig. 

df 
1 
0.000 
1  0.852846385 
1 
  
0.019 
1 
0.000 
1 
0.009 
1 
0.584 
1 
1 
0.392 
0.668 
1 
0.337 
1 
0 
  
0.000 
1 
0.459 
1 
1 
  
0.000 
1 
0.000 
1 
0.000 
1 
0.000 
1 
1 
0.000 
0.003 
1 
0.249 
1 
0 
  

a. Region = WUP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 

 

123

 

Table A.27: Multinomial logistic regression parameter estimates. Stand condition as independent variable. Western Upper Peninsula. 
N=20,915 unique forest stands enrolled in the Qualified Forest Program. † 

Forest Practiceb 
Thinning 

Timber 
Stand 
Improvement 

Intercept 
NS 
SS-P 
SS-M* 
SS-W* 
P-P* 
P-M* 
P-W 
S-P 
S-M 
S-W 
Intercept 
NS 
SS-P 
SS-M 
SS-W* 
P-P* 
P-M 
P-W 
S-P 
S-M* 
S-W 

β 
-1.265 
-3.377 
-49.519 
-3.154 
-2.612 
-2.077 
-0.848 
-0.010 
0.206 
0.593 
0c 
-2.760 
-3.699 
-49.841 
-2.234 
-2.059 
-2.557 
-1.343 
-1.475 
0.273 
1.768 
0c 

Std. Error 
0.328 

Wald 

14.863 
10.622  0.101069353 
  
13.611 
31.709 
19.137 
5.201 
0.001 
0.278 
1.986 
  
19.141 
0.020 
  
3.328 
6.325 
5.350 
2.996 
3.455 
0.136 
6.444 
  

0.000 
0.855 
0.464 
0.475 
0.372 
0.350 
0.391 
0.421 
  
0.631 
26.234 
0.000 
1.225 
0.819 
1.105 
0.776 
0.794 
0.741 
0.696 
  

Sig. 

df 
1 
0.000 
1  0.750550123 
1 
  
0.000 
1 
0.000 
1 
0.000 
1 
1 
0.023 
0.977 
1 
0.598 
1 
0.159 
1 
0 
  
0.000 
1 
0.888 
1 
1 
  
0.068 
1 
0.012 
1 
0.021 
1 
1 
0.083 
0.063 
1 
0.713 
1 
0.011 
1 
0 
  

Exp(β) 

  
3.416E-02 
0.000 
0.043 
0.073 
0.125 
0.428 
0.990 
1.229 
1.810 
  

  
2.476E-02 
0.000 
0.107 
0.128 
0.078 
0.261 
0.229 
1.314 
5.858 
  

95% Confidence Interval 

for Exp(β) 

Lower 
Bound 

  
3.106E-11 
0.000 
0.008 
0.030 
0.049 
0.207 
0.498 
0.571 
0.793 
  

Upper Bound 
  

3.756E+07 
0.000 
0.228 
0.182 
0.318 
0.888 
1.967 
2.645 
4.132 
  

  

  

0.000  5.29581E+20 
0.000 
0.000 
0.010 
1.181 
0.635 
0.026 
0.677 
0.009 
0.057 
1.194 
1.084 
0.048 
5.621 
0.307 
22.938 
1.496 
  
  

a. Region = WUP 
b. The reference category is: No Practice. 
c. This parameter is set to zero because it is redundant. 
†NS=Non-stocked, SS–P=Seedlings/Saplings poorly stocked, SS–M=Seedlings/Saplings moderately stocked, SS–W=Seedlings/Saplings well 
stocked, P–P=Poletimber poorly stocked, P–M=Poletimber moderately stocked, P–W=Poletimber well stocked, S–P=Sawtimber poorly stocked, 
S–M=Sawtimber moderately stocked, S–W=Sawtimber well stocked. 

 

124

 

Figure 7: Map of acres enrolled in the Qualified Forest Program in aspen/birch forest type group, 
by county. 

 

Figure 8: Map of acres enrolled in the Qualified Forest Program in elm/ash/cottonwood forest type 
group, by county. 
 

 

 

125

 

Figure 9: Map of acres enrolled in the Qualified Forest Program in maple/beech/birch forest type 
group, by county. 
 

 

Figure 10: Map of acres enrolled in the Qualified Forest Program in oak forest type group, by 
county. 
 

 

 

126

 

Figure 11: Map of acres enrolled in the Qualified Forest Program in spruce/fir forest type group, 
by county. 
 

 

Figure 12: Map of acres enrolled in the Qualified Forest Program in white/red/jack pine forest type 
group, by county. 
 

 

 

127

 

Figure 13: Map of acres enrolled in the Qualified Forest Program prescribed for artificial 
regeneration, by county. 
 

 

 
Figure 14: Map of acres enrolled in the Qualified Forest Program prescribed for clearcut, by 
county. 
 

 

128

 

Figure 15: Map of acres enrolled in the Qualified Forest Program prescribed for salvage treatment, 
by county. 
 

 

Figure 16: Map of acres enrolled in the Qualified Forest Program prescribed for seed tree harvest, 
by county. 
 

 

 

129

 

 
Figure 17: Map of acres enrolled in the Qualified Forest Program prescribed for shelterwood 
harvest, by county. 
 

Figure 18: Map of acres enrolled in the Qualified Forest Program prescribed for selection harvest, 
by county. 
 

 

 

130

 

 
Figure 19: Map of acres enrolled in the Qualified Forest Program prescribed for thinning, by 
county. 
 

Figure 20: Map of acres enrolled in the Qualified Forest Program prescribed for timber stand 
improvement, by county.

 

 

131

 

 

REFERENCES

132

 

REFERENCES 

 
 
 
 

Amacher, G.S., R.J. Brazee, and T.A. Thomson. 1991. The effect of forest productivity taxes on timber 

stand investment and rotation length. Forest Science; 37(4): 1099-118. 

 
Arbuckle, Jr., J.G. 2013. Farmer attitudes toward proactive targeting of agriculture conservation 

programs. Society & Natural Resources: An International Journal; 26(6): 625-641. 

 
Batie, S.S. 2008. Wicked problems and applied economics. American Journal of Agriculture Economics. 

90:5, 1186-1191. 

 
Beach, R.H., S.K. Subhrendu, J. Yang, B.C. Murray, and R.C. Abt. 2005. Econometric studies of non-
industrial private forest management a review and synthesis. Forest Policy and Economics; 7, 261-
281. 

 
Berry, J.K., J.A. Delgado, F.J. Pierce, and R. Khosla. 2005. Applying spatial analysis for precision 

conservation across the landscape. J. Soil Water Conserv. 60(6): 363-370. 

 
Backhouse, G. N., T. W. Clark, and R. L. Wallace. 1996. Reviewing recovery programs for endangered 
species: some considerations and recommendations. Pages 170–179 in S. Stephens and S. Maxwell, 
editors. Back from the brink: refining the threatened species recovery process. Surrey Beatty & Sons, 
Chipping Norton, New South Wales, Australia. 

 
Butler, B.J. 2014. Forest Ownership. Forest Ownership Map of the Coterminous United States. Research 

Highlight 628. https://www.fs.fed.us/research/highlights/highlights_display.php?in_high_id=628  

 
Butler, B.J., and E.C. Leatherberry. 2004. America’s family forest owners. Journal of Forestry; Vol. 102, 

No. 7, pp. 4-14. 

 
Butler, B.J., J.H. Hewes, B.J. Dickinson, K. Andrejczyk, M. Markowski-Lindsay, and S.M. Butler. 2016a. 

USDA Forest Service National Woodland Owner Survey: National, regional, and state statistics for 
family forest and woodland ownerships with 10+ acres, 2011-2013. A technical document supporting 
the Forest Service update of the 2010 RPA assessment. USDA For. Serv., Resour. Bull. NRS-99, 
Northern Research Station, Newtown Square, PA. 48 p. 

 
Butler, B.J., J.H. Hewes, B.J. Dickinson, K. Andrejczyk, M. Markowski-Lindsay, and S.M. Butler. 

2016b. Family forest ownerships of the United States, 2013: findings from the USDA Forest 
Service’s National Woodland Owner Survey. Journal of Forestry; 114(6): 638-647. 

 
Butler, B.J., M. Markowski-Lindsay, S. Snyder, P. Catanzaro, D.B. Kittredge, K. Andrejczyk, B.J. 

Dickinson, D. Eryilmaz, J.H. Hewes, P. Randler, D. Tadle, and M.A. Kilgore. 2014. Effectiveness of 
landowner assistance activities: an examination of the USDA Forest Service’s Forest Stewardship 
Program. Journal of Forestry; 112 (2): 187-197. 

 
Conway, C.M., G.S. Amacher, J. Sullivan, and D. Wear. 2003. Decisions nonindustrial forest landowners 

make: an empirical examination. Journal of Forest Economics: 9, pp. 181-203 

 
Dillman, D.A. 1991. The design and administration of mail surveys. Annu. Rev. Sociol. 17: 225-249. 

 

133

 

Dillman, D.A. 2000. Mail and Internet Surveys: The Tailored Design Method. 2nd ed. John Wiley & Sons, 

New York, NY. 400p. 

 
Dillman, D.A., J.D. Smyth, and L.M. Christian. 2009. Internet, mail and mixed-mode surveys: the 

tailored design method, 3rd edition. John Wiley: Hoboken, NJ. 499p. 

 
Finley, A.O. and D.B. Kittredge. 2006. Thoreau, Muir, and Jane Doe: Different types of private forest 

owners need different kinds of forest management. Northern Journal of Applied Forestry: 22 (1), 27-
34. 

 
GC, Shivan and K. Potter-Witter. 2011. Status of Michigan’s forest products industry in the changing 

market environment. Forest Products Journal: 61(1), 77-82. 

 
Ghimire, R., G.T. Green, N.C. Poudyal, and H.K. Cordell. 2016. Who Recreates Where: Implications 

from a National Recreation Household Survey. Journal of Forestry; 114(4): 458-465. 

 
Grossman, G.H. and K.L. Potter. 1987. Status of Michigan’s Private Forest Reserve Act. National 

Woodlands. July-August 1987. 

 
Hair, J.F., B. Black, B. Babin, R.E. Anderson, and R.L. Tatham. 2006. Multivariate Data Analysis (6th 

edition). Pearson Education: Upper Saddle River, New Jersey. 

 
Harlow, Richard. 22 April 2019. Personal interview. 
 
Hibbard, Calder M., Michael A. Kilgore, and Paul V. Ellefson. 2003. Property taxation of private forests 

in the United States: A national review. Journal of Forestry. 101.3: 44-49. 

 
Hickman, C.A. 1992. Property tax laws as an incentive to forest management – National overview. 

Proceedings of the Society of American Foresters National Convention. Bethesda, MD: Society of 
American Foresters. 

 
Jackson-Smith, D.B., M. Halling, E. de la Hoz, J.P. McEvoy, and J.S. Horsburgh. 2010. Measuring 

conservation program best management practice implementation and maintenance at the watershed 
scale. Journal of Soil and Water Conservation; 65(6): 413-423. 

 
Janota, J.J. and S.R. Broussard. 2008. Examining private forest policy preferences. Forest Policy and 

Economics: 10(3), pp. 89-97. 

 
Johnson, R.B., A.J. Onwuegbuzie, L.A. Turner, 2007. Toward a definition of mixed methods research. 

Journal of Mixed Methods Research: 1(2), pp. 112-133. 

 
Khanal, P.N., D.L. Grebner, I.A. Munn, S.C. Grado, R.K. Grala, J.E. Henderson. 2018. Evaluating non-

industrial private forest landowner willingness to manage for forest carbon sequestration in the 
southern United States. Forest Policy and Economics; 75: 112-119. 

 
Kleiman, D.G., R.P. Reading, B.J. Miller, T.W. Clark, J.M. Scott, J. Robinson, R.L. Wallace, R.J. Cabin, 
and F. Felleman. 2000. Improving the Evaluation of Conservation Programs. Conservation Biology; 
14(2): 356-365. 

 

 

134

 

Klemperer, W.D. 1988. Revising forest property taxes – Recent trends and policy questions. Forest 

taxation: adapting in an era of change, ed. M.P. Hamel. Madison, WI: Forest Products Research 
Society. 

 
Klooster, W.S., D.A. Herms, K.S. Knight, C.P. Herms, D.G. McCullough, A. Smith, K.J.K. Gandhi, and 

J. Cardina. 2014. Ash (Fraxinus spp.) mortality, regeneration, and seed bank dynamics in mixed 
hardwood forests following invasion by emerald ash borer (Agrilus planipennis). Biological Invasion; 
16(4): 859-873. 

 
Kluender, R.A. and T.L. Walkingstick. 2000. Rethinking how nonindustrial landowners view their lands. 

South. Jour. Appl. For. 24: 150-158. 

 
Koontz, T.M. 2001. Money talks-but to whom? Financial verses nonmonetary motivations in land use 

decisions. Society of Natural Resources: 14 (1), pp. 51-66. 

 
Korpi, K.W. 2010. Trends in Michigan’s forest products industry 2000-2010. Michigan Forest Product 

Council.  

 
Kuipers, B.T. and K. Potter-Witter. 2012. Non Industrial Private Forest Landowner Communication 

Analysis in North Eastern Michigan: A Segmentation Approach. Thesis. Michigan State University. 

 
Leatherberry, E.C., N.P. Kingsley, and T.W. Birch. 1998. Private Timberland Owners of Michigan, 1994. 

USDA Forest Service, North Central Forest Experiment Station Resource Bulletin NC-191 

 
Loo, J.A. 2009. Ecological impacts of non-indigenous invasive fungi as forest pathogens. Biological 

Invasions; 11(1):81-96. 

 
Michigan Department of Agriculture and Rural Development, Forestry Assistance Program. 2019a. 

www.michigan.gov/mifap 

 
Michigan Department of Agriculture and Rural Development, Qualified Forest Program. 2019b. 

www.michigan.gov/QFP  

 
Michigan Department of Natural Resources, Commercial Forest Lands. 2019. 

https://www.michigan.gov/dnr/0,4570,7-350-79136_79237_80945_83262---,00.html  

 
Michigan Department of Natural Resources. 2018. Michigan Primary Mill Survey 2016. 

https://www.michigan.gov/documents/dnr/IC4002_PrimaryMillsSurvey_633444_7.pdf  

 
Michigan Legislature. 2018. Michigan Taxpayer’s Guide. 

http://www.legislature.mi.gov/documents/Publications/TaxpayerGuide-2018.pdf. 17p. 

 
Mueller, L. 2011. Nonindustrial private forest owners in Michigan: Examination at a finer scale. Thesis. 

Michigan State University, East Lansing, MI, USA. 70p. 

 
National Association of Conservation Districts. About NACD. 2019. https://www.nacdnet.org/about-

nacd/  

 
Nowak, Pete, Sarah Bowen, and Perry E. Cabot. 2006. Disproportionality as a framework for linking 

social and biophysical systems. Society & Natural Resources. 19:2, 153-173. 

 

 

135

 

Oswalt, Sonja N., and W B. Smith, editors. U.S. Forest Resource Facts and Historical Trends. FS-1035 

ed., United States Forest Service, 2014, 
www.fia.fs.fed.us/library/brochures/docs/2012/ForestFacts_1952-2012_English.pdf. Accessed 24 
July 2018. 

 
Peterson, G. and K. Potter-Witter. 2006. Stalking the elusive family forest owner in Michigan. 2006 SAF 

National Convention Proceedings. Bethesda, Md.: Society of American Foresters. 

 
Potter-Witter, K. 2005. A cross-sectional analysis of Michigan nonindustrial private forest landowners. 

North. Jour. Appl. For. 22(2): 132-138. 

 
Pugh, S.A., C. Paulson, and B.J. Butler. 2016. Forests of Michigan, 2015. Resource Update FS-79. 

Newtown Square, PA: U.S. Department of Agriculture, Forest Service, Northern Research Station. 
4p. 

 
Saulnier, W.J., M.C. Bolding, S.M. Barrett, and S.P. Prisley. 2017. Characteristics of Virginia’s private 

forest landowners and their attitudes toward harvesting. Forest Products Journal: 67 (1/2), 69-80. 

 
Silver, Emily J, J. Leahy, A.R. Weiskittel, C.L. Noblet, D.B. Kittredge. 2015. An Evidence-Based 

Review of Timber Harvesting Behavior among Private Woodland Owners. Journal of Forestry; 
Bethesda Vol. 113, Iss. 5: 490-499. 

 
USDA, Forest Service Northern Research Station. Who Owns America’s Trees, Woods, and Forests? 

NRS-INF-31-15. March 2015 

 
USDA Forest Service, Forest Inventory and Analysis Program, Fri Feb 08 00:13:07 GMT 2019a. Forest 

Inventory EVALIDator web-application Version 1.8.0.00. St. Paul, MN: U.S. Department of 
Agriculture, Forest Service, Northern Research Station. [Available only on internet: 
http://fsxopsx1056.fdc.fs.usda.gov:9001/Evalidator/evalidator.jsp]  

 
USDA Forest Service, Forest Stewardship Program, Tue April 16, 2019b. 
https://www.fs.fed.us/managing-land/private-land/forest-stewardship/  

 
USDA Natural Resources Conservation Service, Environmental Quality Incentives Program. Tue April 
16, 2019c. https://www.nrcs.usda.gov/wps/portal/nrcs/main/national/programs/financial/eqip/  
 
White, S.C. 2001. Massachusetts NIPF owner’s attitudes and preferences towards ecosystem 

management: a reluctance to participate. Thesis. The University of Massachusetts, Amherst. 147p. 

 
Wolde, B., P. Lal, and P. Burli. 2017. Forestland owner’s willingness to consider multiple ways of 

supplying biomass simultaneously: Implications for biofuel incentive policies. Energy Policy: 105 
(2017), 183-190. 

 

 

 

 

136