LANDOWNER WILLINGNESS TO SUPPLY MARGINAL LAND FOR BIOENERGY PRODUCTION
IN MICHIGAN
By
Noel Hayden

A THESIS
Submitted to
Michigan State University
In partial fulfillment of the requirements
for a degree of
Agricultural, Food, and Resource Economics – Master of Science
2014

ABSTRACT
LANDOWNER WILLINGNESS TO SUPPLY MARGINAL LAND FOR BIOENERGY PRODUCTION:
IN MICHIGAN
By
Noel Hayden
Growing bioenergy crops on non-crop, marginal land offers an avenue to escape the
ethical and practical limitations of using cropland, but how much of this land are owners
actually willingly make available? A contingent valuation survey was used to examine the
willingness of landowners to supply land for bioenergy crop production. Owners of non-crop
marginal land were identified using area-frame sampling, based upon the 2010 USDA Cropland
Data Layer (CDL) of land cover. Willingness to supply land was estimated econometrically as a
survey-weighted hurdle model comprised of a participation decision probit and an acreage
commitment truncated regression.
The results reveal two significant findings. First, landowners who possess non-crop
marginal land on average own more cropland than non-crop land and given the opportunity to
rent out either land type for bioenergy crops, they preferred to rent out more cropland. This
result highlights how markets for land at the extensive margin inherently link the supply of
bioenergy crops to that of food crops. Second, even at high rental rates, less than a third of
landowners were willing to rent out their marginal land to grow bioenergy crops. This finding
suggests that the supply of marginal land for bioenergy crops is more limited than previously
believed, at least based on evidence from Michigan.

ACKNOWLEDGMENTS

This project and thesis was a great undertaking that was the work of many individuals.
Dr. Scott M. Swinton was my advisor for this thesis and provided me with much needed
support, advice, accountability, expertise, and ideas. To him I owe a great debt and I am so
appreciative of his encourage and support in all of my pursuits.
Dr. Satish Joshi and Dr. Frank Lupi, both members of my committee, were also
invaluable with their support, ideas, and guidance. In particular I owe a lot to Dr. Lupi who
helped significantly throughout the modeling stages when any problem would arise. He
equipped me and encouraged me whenever possible to find a solution on my own.
Sarah AcMoody, GISP, at the Remote Sensing and GIS Department at Michigan State
University, was extremely helpful in the GIS work she did for us. The survey used in this project
identified potential respondents through a process called area frame sampling which she
carried out by creating a variety of databases.
I also want to acknowledge my friends and family, especially my parents, who have
been such an encouragement to me in everything that I do. I greatly appreciate your loving
support and prayers throughout my life.
This work was funded by the US department of Energy Great Lakes Bioenergy Research
Center (GLBRC).
To God be the Glory.

iv

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................................................ vii
LIST OF FIGURES ......................................................................................................................................... viii
Chapter 1: Background on Bioenergy ........................................................................................................... 1
1.1

Why Bioenergy? ............................................................................................................................ 2

Chapter 2: Why Marginal Land? ................................................................................................................... 5
2.1 Land for Biomass ................................................................................................................................. 6
2.2 Defining Marginal Land for this Study ................................................................................................ 7
Chapter 3: Goals and Objectives ................................................................................................................. 10
Chapter 4: Conceptual Framework ............................................................................................................. 11
Chapter 5: Area Frame Sampling With GIS ................................................................................................. 16
5.1 Geographic Database ........................................................................................................................ 17
5.2 Sampling Method .............................................................................................................................. 17
Chapter 6: Survey Design ............................................................................................................................ 21
6.1 Questionnaire Format ....................................................................................................................... 22
6.2 Contingent Valuation Question ........................................................................................................ 23
6.3 Experimental Design ......................................................................................................................... 26
6.4 Questionnaire Review Process.......................................................................................................... 30
6.5 Survey Response Rate and Data Entry Methods .............................................................................. 30
Chapter 7: Empirical Methods .................................................................................................................... 33
7.1 Variable Specification........................................................................................................................ 36
7.2 Factor Analysis .................................................................................................................................. 38
7.3 Weighting and Scaling Model to Southern Lower Michigan ............................................................ 40
Chapter 8: Hypotheses................................................................................................................................ 43
Chapter 9: Results ....................................................................................................................................... 48
9.1 The Participation Model Results ....................................................................................................... 48
v

9.2 The Acreage Commitment Model Results ........................................................................................ 59
9.3 Scaling up Results to Southern Lower Michigan ............................................................................... 69
Chapter 10: Conclusion ............................................................................................................................... 73
APPENDIX .................................................................................................................................................... 75
REFERENCES ................................................................................................................................................ 97

vi

LIST OF TABLES

Table 1. Conceptual Model Variables ......................................................................................................... 13
Table 2. CDL 2010 - Classification of Land Use and Acreage in Southern Michigan ................................... 17
Table 3. Questionnaire Experimental Design.............................................................................................. 29
Table 4. Variable Specification for Variables Used in Analysis ................................................................... 37
Table 5. Factor Analysis for Environmental Attitudes and Landowner Concerns ...................................... 39
Table 6. Varimax Rotated Factors for Environmental Attitudes ................................................................. 39
Table 7. Varimax Rotated Factors for Landowner Concerns ...................................................................... 39
Table 8. Design for Weighting Observations .............................................................................................. 41
Table 9. Calculations for Scaling up to Southern Lower Michigan.............................................................. 42
Table 10. Likelihood Ratio Test Comparing Tobit vs. Hurdle Model ........................................................... 42
Table 11. Probit Participation Model for Cropland Rented for Prairie and Poplar .................................... 49
Table 12. Probit Participation Model for Cropland Rented for Switchgrass and Corn ............................... 50
Table 13. Probit Participation Model for Pasture and Hay Land Rented for Prairie and Poplar ................ 51
Table 14. Probit Participation Model for Pasture and Hay Land Rented for Switchgrass and Corn........... 52
Table 15. Probit Participation Model for Other Marginal Lands Rented for Prairie and Poplar ................ 53
Table 16. Probit Participation Model for Other Marginal Lands Rented for Switchgrass and Corn ........... 54
Table 17. Truncated Acreage Model for Cropland Committed to Prairie and Poplar ................................ 60
Table 18. Truncated Acreage Model for Cropland Committed to Switchgrass and Corn........................... 61
Table 19. Truncated Acreage Model for Pasture and Hay Land Committed to Prairie and Poplar ............ 62
Table 20. Truncated Acreage Model for Pasture and Hay Land Committed to Switchgrass and Corn ...... 63
Table 21. Truncated Acreage Model for Other Marginal Lands Committed to Prairie and Poplar ............ 64
Table 22. Truncated Acreage Model for Other Marginal Lands Committed to Switchgrass and Corn ...... 65

vii

LIST OF FIGURES

Figure 1. Federal Renewable Fuel Standards with Advanced Biofuels 2006-2022....................................... 3
Figure 2. Randomly Selected Michigan Counties for Sampling .................................................................. 18
Figure 3. Example of Area Frame Selection ................................................................................................ 19
Figure 4. Example of Contingent Valuation Question................................................................................. 25
Figure 5. Probability of Renting Cropland for Bioenergy Crops in Response to Rental Rate (Probit) ........ 57
Figure 6. Probability of Renting Pasture Land for Bioenergy Crops in Response to Rental Rate (Probit) .. 58
Figure 7. Probability of Renting Other Marginal Land for Bioenergy Crops in Response to Rental Rate
(Probit) ........................................................................................................................................................ 58
Figure 8. Average Acreage Offered Conditional on Renting Land for Bioenergy Crops at $100 per Acre
(Truncated Model ) ..................................................................................................................................... 68
Figure 9. Total Acreage Owned at each Percentile of Land Owners by Land Type .................................... 68
Figure 10. Average Landowner Supply of Cropland for Bioenergy Crops (Combined Participation and
Acreage Commitment Models) ................................................................................................................... 70
Figure 11. Average Landowner Supply of Pasture Land for Bioenergy Crops (Combined Participation and
Acreage Commitment Models) ................................................................................................................... 70
Figure 12. Average Landowner Supply of Other Marginal Land for Bioenergy Crops (Combined
Participation and Acreage Commitment Models) ...................................................................................... 71
Figure 13. Supply of Marginal Land (Pasture + Other Marginal Lands) for Three Bioenergy Crops in
Southern Lower Peninsula of Michigan ..................................................................................................... 72
Figure 14 Pre-survey Postcard .................................................................................................................... 76
Figure 15 Cover Letter................................................................................................................................. 77
Figure 16 Example Survey ........................................................................................................................... 78
Figure 17 Reminder Postcard ...................................................................................................................... 94
Figure 18 Second Version of Cover Letter .................................................................................................. 95

viii

Chapter 1: Background on Bioenergy

Bioenergy is any form of energy that comes from a biological source. These biological
sources are referred to as biomass. Bioenergy can come in the form of a solid or a liquid.
Historically, bioenergy came from a solid by burning biomass such as wood. Recently, the liquid
forms of bioenergy have been growing and are known as biofuels. Biofuels often serve as
substitute fuels for more conventional fuels. For example, ethanol is a biofuel substitute for
gasoline and biodiesel is a biofuel substitute for diesel.
The most plentiful biofuel in the United States is corn grain ethanol. In 2011 the United
States produced 13.9 million gallons of ethanol (Renewable Fuels Association 2012). Corn grain
ethanol is the leading focus in the U.S. because corn grows well in the region, its starches are
relatively easy to convert into ethanol, and infrastructure and harvesting equipment already
exist for the grain. Brazil is the second largest producer of ethanol and uses sugar cane as
biomass for similar reasons (Renewable Fuels Association 2012). However, ethanol can be
produced from other sources as well. The largest potential source for production is cellulosic
biomass. Cellulose is a basic organic compound comprising much of the structural material
found in plants. Cellulosic ethanol is desirable in its ability to utilize many different sources of
biomass. The current issue to mass production of cellulosic ethanol is the high infrastructure
and processing costs associated with converting cellulose into a fuel (USDA and DOE, 2000).

1

This study focuses on cellulosic biomass with the anticipation that these costs could decrease in
the future.
1.1 Why Bioenergy?
Ever since the creation of the Renewable Fuel Standard (RFS) program in 2005 U. S.
demand for bioenergy has dramatically increased. In 2007 under the Energy Independence and
Security Act (EISA), the RFS was expanded to increase the percent of ethanol blended into
gasoline. Also at that time an important distinction was made between different types of
ethanol and separate volume mandates were created for each source (EPA 2007). Advanced
biofuels such as cellulosic ethanol were specifically mandated. Figure 1 shows the previous
standards and how they were increased through 2022.
While the federal government was increasing demand for biofuels, many states have
been increasing the demand for bioenergy in the form of electricity. Thirty out of the 50 states
as well as the District of Columbia have set Renewable Portfolio Standards (RPS). These
standards mandate that a certain percentage of electricity within the state come from a
renewable source such as wind, solar, geothermal, and biomass. New York has one of the
highest and fastest goals with a RPS mandate of 25% of electricity coming from renewable
sources by 2013. Maine is also very high with an RPS set at 40% of electricity coming from
renewable sources by 2017. In the case of Michigan, where this study is focused, a RPS was
introduced that mandated 10% of Michigan electricity to be from renewable sources by 2015.

2

Figure 1. Federal Renewable Fuel Standards with Advanced Biofuels 2006-2022

What is driving this desire to increase bioenergy? There are three main perceived
benefits of bioenergy. First, bioenergy is a form of renewable energy meaning that similar to
wind and solar energy the supply can be renewed. This is a major difference from conventional
sources of energy that are often fossil fuel based and have a limited supply. Second, bioenergy
has been shown to be potentially carbon neutral through its lifecycle. Carbon neutral means
that during its lifecycle a bioenergy feedstock may add carbon dioxide to the atmosphere but it
also removes carbon dioxide, making a net neutral effect. For example in the case of cellulosic
ethanol, the amount of carbon dioxide produced when burning the fuel in an automobile is
offset by the amount of carbon dioxide sequestered through growing cellulosic biomass (Slade,
Bauen and Shah 2009). Debates exist on under what conditions this may or may not hold true
3

but according to Sedjo (2011) the literature shows that over the long term bioenergy is a
carbon neutral source of energy. The National Research Council (NRC) in 2011 also notes that
carbon emissions and reductions can vary, depending on the bioenergy source (food-based
biofuels or dedicated biofuel crops) and on where the biomass is grown. The third benefit
associated with bioenergy in the United States, is as a route to energy independence for
national security. The Energy Independence and Security Act (EISA) (110. P.L. 140) says, “(the
goal of the Act was) to move the United States toward greater energy independence and
security”. U.S. grown bioenergy feedstocks make the United States less dependent on foreign
sources of energy.

4

Chapter 2: Why Marginal Land?

Land is an essential input to produce bioenergy at a large scale and a resource with a
limited supply. It is used to produce food, feed, lumber, paper products and ecological services.
Using existing cropland to grow bioenergy crops would increase the demand for cropland.
Studies have shown that some harmful consequences exist when increasing competition for
cropland. Searchinger et al. (2008) describe how using current land that is in corn to grow
bioenergy crops could increase greenhouse emissions by 50%. Rajagopal et al. (2007) found
that in the long run by competing for cropland bioenergy production can decrease the supply of
food. In 2008, Fritsche detailed the potential greenhouse gases from direct and indirect land
use change that would occur from converting cropland to bioenergy crops. In a report to the UK
government, Gallagher (2008) echoes the concerns earlier mentioned. He notes the potential
for increases in food prices and greenhouse emissions as indirect effects of growing bioenergy
crops on existing cropland. Using cropland for biofuel would also lead to a direct connection
between prices in the energy markets and prices in the food markets as both industries would
be in direct competition for cropland (Piroli, Ciain and Kancs 2011). It can be seen then, that
increasing competition for cropland by growing bioenergy crops can have serious, negative side
effects.
Growing bioenergy crops on marginal land offers a way to avoid the problems
associated with those crops on cropland. By occupying land that is not used for crops,
5

bioenergy crop production on marginal land could ease the increasing demand for cropland and
thus has a lesser effect on food prices (Swinton et al. 2011; Campbell et al. 2008; Carroll and
Somerville 2009; Tilman et al. 2009).
2.1 Land for Biomass

Land for biomass can come from the intensive margin or the extensive margin. Biomass
from the intensive margin is any additional biomass that comes from existing cropland. There
are two ways to get more biomass from existing cropland. First, dedicated bioenergy crops
could be grown instead of existing crops. For example switchgrass is a high yielding cellulosic
based bioenergy crop that could be planted on existing cropland that is in use growing corn,
soybeans, or other crops. Alternatively, it is also possibly to get more bioenergy from the
intensive margin by increasing inputs and removing more biomass. Corn stover is a term used
for all the other parts of the corn plant besides the grain, such as husks and stalks (Sheehan et
al. 2008). Corn stover is currently left on many fields to restore nutrients the ground for the
next crop to use; however, it is cellulose-based and studies show that removal up to 50% was
not only provide biomass but also give a yield advantage (Jeschke 2011).
The extensive margin refers to expansion of biomass production onto land that is
currently not used for crop production. For example an expansion onto the extensive margin
could be clearing a piece of scrubland that is lying idle and growing a dedicated cellulosic
biofuel crop such as switchgrass on it. Land on the extensive margin may have a variety of
different types of land cover and may not be in crops for a variety of reasons. The only defining
characteristic about this land is that it is not used for crops.
6

2.2 Defining Marginal Land for this Study

As described earlier, one of the main concerns with growing bioenergy crops is that they
may compete for cropland. Therefore, when we talk about how growing bioenergy crops on
marginal land has the benefit of not competing with other crops for cropland, it is important
that we clearly define what land we are talking about.
Above we discussed the extensive margin and intensive margin. The extensive margin is
exactly the land we are defining as marginal land in this study. This marginal land is land that is
defined as such purely based upon how it is used. It is land that does not contribute to crop
production in any way and therefore its use for bioenergy crops will not affect cropland use.
This is an economic understanding of word marginal and as Laura James, notes it dates back to
Barlowe (1986) and Peterson and Galbraith (1932) (James 2010).
Throughout the history of the bioenergy discussion, the definition of marginal land has
varied; therefore, it is important to be clear about what we are not defining as marginal land.
There is another definition of marginal land that others use. It is based upon the land’s “quality”
or ability to produce agriculturally. In general this refers to land that is less fertile based upon a
number of biophysical measurements. This is land that might be sandier, rocky, contain fewer
nutrients, be susceptible to erosion or have varying elevation (Peterson Galbraith 1932;
Dangerfield and Harwell 1990; Lal 2005).

7

To understand why a distinction between these definitions of marginal land is so
important we must remember that growing biomass on cropland would raise food prices so
ideally biomass would be grown on non-cropland. There are two main reasons this lower
fertility definition of marginal land is discussed when talking about bioenergy. The first reason is
that cellulosic based bioenergy crops, when compared to typical crops, require less fertile land
and often fewer inputs making them more ideal for this type of biophysically marginal land
(Tilman et al. 2006; Perlack et al. 2011). This is definitely a positive trait of dedicated cellulosic
based bioenergy crops but not all land that is less fertile lays idle. It is important to remember
here that ideally bioenergy crops are grown on non-cropland, land not necessarily less fertile
land. Landowners vary and so do their choices. Land with the same level of fertility may be in
crops under one manager and not in crops under another. Some land may simply not be in use
because it is inaccessible by farm equipment and some land may be in crops because it is close
to a granary bringing down transportation costs. This leads to the other reason this lower
fertility definition of marginal land is often discussed when talking about bioenergy. There is a
misconception that land currently not being used for crops, is in the extensive margin solely
because it is less fertile. While it is often the case that less fertile land is not used for growing
crops, there are many other reasons land may not be crops. As mentioned earlier, landowners
make the decision what land is used to grow crops and what land is not. No matter what the
land quality may be, if the land contributes to growing crops, using it for biomass instead will
have an effect on price of food. There are a number of studies including the Billion-Ton report
completed by the Department of Energy (DOE) that cite marginal land as having the benefit of
not displacing cropland, yet still when describing marginal land they define it based upon
8

biophysical characteristics such as salinity (U.S Department of Energy 2011). This can result in a
misleading amount of land being cited as available to grow bioenergy crops without disrupting
crop prices and can cause misidentification if landowners of this type of less fertile marginal
land are solicited for a willingness to supply marginal land study.
Part of this study’s focus is to examine the possibility of growing bioenergy crops in a
way that avoids using cropland. To do this we clearly focus on eliciting landowners’ willingness
to supply land that is marginal because it is part of the extensive margin and not in crop
production. This leads us to a full description of the objectives of this study below.

9

Chapter 3: Goals and Objectives

The goal of this research is to examine the availability of marginal land to grow
bioenergy crops through eliciting landowners’ willingness to supply marginal land for bioenergy
crops. In order to reach this goal a series of objectives were identified.
1. Elicit landowners’ willingness to supply their land for bioenergy crops.
a. Identify landowners who own over 10 acres of marginal land, land not used for
crops or in forest.
b. Identify how much land under different current uses these landowners are
willing to make available for bioenergy crops.
2. Describe the potential supply of land to grow bioenergy crops.
a. Describe how willing the average landowner is to supply their land for bioenergy
crops.
b. Scale the average landowners’ response up to an area of interest to describe
how much marginal land would be available for bioenergy crops in that region.
c. Identify what factors have a significant effect on landowners’ decisions.
d. Identify which bioenergy crops may be preferred, if any.
e. Identify whether landowners have a preference on renting cropland or marginal
land.

10

Chapter 4: Conceptual Framework

The conceptual framework for this study focuses on how land use decisions are made.
Land is managed by landowners who make decisions on how it will be used. These decisions are
driven by a desire from the landowner to maximize personal utility. In 1981, Binkley modeled
household decisions on forest management citing that the land produced both timber and
other amenity values. Timmons in 2011 extended this idea to landowners’ decisions to produce
biomass. The basic concept of his work being that landowners can receive utility from both
consumption and amenities that come from land and that they will choose a combination of
both that maximizes their utility.
Utility from consumption can result from income being used to purchase consumable
goods (Timmons 2011); however, consuming purchased goods are not the only way to receive
utility. Individuals benefit from tangible and intangible amenities. For example the utility
received through a friendship is an intangible amenity, while the utility received from swimming
in a lake would be a tangible amenity.
Income for consumption can come from a variety of sources. Income can come from
salary, wages, social security, rental properties, investments, or any other income generating
source. The income of landowners does not have to solely come from land. Many households
that own land have members who work at jobs unrelated to agriculture for a primary source of

11

income. When income does come from land, it can come in a variety of ways. A landowner can
rent their land to a farmer or choose to farm the land directly for income.
Amenities from land can take a variety of forms. Some people value their land for the
scenery, for hunting or fishing, for recreational vehicle use, or for physical activities. Whatever
these amenities might be it is clear that through a land use change such as growing bioenergy
crops the value received from each amenity can change. Depending on how the amenity relates
to the type of land use, the value from that amenity could increase, decrease, or have no
change. For example when an open field used for hunting deer is changed to grow poplar trees
it may no longer attract deer or provide the sight lines necessary for hunting them thus
decreasing the potential value of hunting on the land. Like income, amenities do not solely
come from land. Landowners can get amenity value from other sources as well, such as family
or use of a public bike path.

12

Table 1. Conceptual Model Variables
Variable Description

Name

Symbol

Utility of an individual landowner

Utility

U

Income of an individual landowner
Amenities of an individual landowner
Consumption of an individual landowner
Price of consumption goods
Choice of land use
Subscript to denote from land source
Subscript to denote from other (non-land)
source
Subscript to denote base case without any land
use change into bioenergy crops
Subscript to denote land use change into
bioenergy crops

Income
Amenities
Consumption
Price
Land Use
Land subscript
Other subscript

m
a
c
Pc
LU
land
other

Base subscript

0

Land change
subscript

1

Following the landowner utility maximization models of Binkley (1981) and Timmons
(2011), utility is a function of consumption (c) and amenities (a):
max

LU

U = U [ a, c ]

(1)

Utility is maximized over the land use decision and is constrained by income (m) and the
availability of amenities, each of which can come from either land and other sources:
P * C <= m
c

land

a = a

land

+ m

(2)

other

+a

(3)

other

Income from land and amenities from land are both functions of land use:

13

m

= f(LU)

(4)

a

= g(LU)

(5)

land

land

A change in land use results in a change in income from land and therefore consumption as well
as a change in amenities from land:
∆LU → ∆m

land

→ ∆c

(6)

∆LU → ∆ a

(7)

land

Changes in consumption and amenities affect utility and thus the decision to change
land use can cause a net change in utility. Equation (8) shows the base case utility and equation
(9) shows the utility after a change in land use (U1).
U =U[c ,a ]
0

0

(8)

0

U = U [ c + ∆c , a + ∆ a]
1

0

(9)

0

If utility after the change is greater than utility in the base case, then the landowner will decide
to change the land’s use:
U = U [ c , a ] < U = U [ c + ∆c , a + ∆ a]
0

0

0

1

0

0

→

∆LU

(10)

When this conceptual model is applied to the case of growing bioenergy crops on
marginal land we can see that an individual landowner may or may not convert the land,
14

depending on the amenities received from it. While growing the bioenergy crops may prove to
be profitable on marginal land and thus raise the income of landowners, the extra consumption
this allows landowners may not provide greater utility than the amenities the land provides
them when it is not in use for bioenergy crops. In this case a utility maximizing landowner would
not change their land’s use even given their income would go up.

15

Chapter 5: Area Frame Sampling With GIS

In order to ask Michigan landowners about their willingness to supply marginal land
through a survey, landowners had to be identified. However, not all landowners own marginal
land, and there are even fewer who own significant amounts of it. Because no list of owners of
marginal lands exists, this study uses an area frame sample built from GIS databases of noncrop marginal lands.
All parcels of marginal land in Michigan create a complete area frame for the entire
population of research interest. Area frame sampling is the process of randomly selecting
landowners whose ownership parcel intersects the area frame (Cotter and Nealon, 1987). One
concern that often arises with area frame sampling is that larger parcels have a higher
probability of selection and have the potential to distort the sample if owners of large parcels
behave differently; however, this is not a concern in this study given that the objective is not to
find how the typical land owner may behave but to find the potential supply of marginal land
available for bioenergy crop production. This means that the natural effect of larger parcels of
land being more likely to be sampled is actually ideal since decisions by these landowners have
effects proportional to the amount of land they own on the potential supply of marginal land.

16

5.1 Geographic Database

For this study, area frame sampling followed a two-step procedure. First, an area frame
consisting of all current marginal lands in Michigan had to be created. A recent land use
database in Michigan, the Cropland Data Layer (CDL) of 2010, was used for this analysis.
CDL is a raster database that was created in 2010 by the United States Department of
Agriculture (USDA) from satellite imagery using spectral reflectance data (National Agricultural
Statistics Service, 2010). The database has 53 land cover categories with a crop-specific
accuracy of 86.86% for each 30m pixel. The categories that were defined as marginal land are
listed in Table 2.
Table 2. CDL 2010 - Classification of Land Use and Acreage in Southern Michigan
Land Use Classification

Percent

Acres (mil.)

Fallow / Idle Cropland
Shrubland
Grassland / Herbaceous
Pasture / Hay
Total

6%
4%
46%
44%
100%

0.17
0.13
1.4
1.3
2.97

5.2 Sampling Method

This study focused in the southern half of the Lower Peninsula, where most crop
production occurs. Michigan counties south of the county of Clare, around 43.9 degrees
latitude were considered.

17

In order to identify who owns land in Michigan, it was necessary to obtain county tax
records; however, these records can be difficult to obtain. Not all counties in Michigan have
digital records, and going through paper or pdf based maps is very time consuming. Also, many
counties charge large fees for access to the data. Give this and the fact that this study had a
limited budget, cluster sampling was done at the county level. Twelve counties in Michigan
were randomly selected from those counties south of Clare. Allegan, Barry, Branch, Ionia,
Isabella, Lenawee, Livingston, Newaygo, Saginaw, Sanilac, Tuscola, and Van Buren were
selected. Figure 2 shows the randomly selected counties. The metropolitan counties of Detroit
were excluded along with the county of Ingham, which was used for focus group pretesting of
the survey questionnaire.
Figure 2. Randomly Selected Michigan Counties for Sampling

Counties selected at random
Counties not selected
Border of non-metro counties in sampling
frame

Ten acres was treated as the minimum viable land area for bioenergy crop production.
Therefore, sampling began in each of the twelve selected counties by finding only parcels of

18

marginal land greater than ten acres. Sampling was done from the CDL 2010. CDL 2010 is a
database based on 30m pixels.
Figure 3. Example of Area Frame Selection

Sampling took place by randomly dropping points within the ten acre parcels of
marginal land and then identifying, using county tax records, who owned the parcel of land in
which the point fell. Points were limited in that they could not be dropped on the same parcel
of marginal land within 300m of each other. This was done in order to maximize the probability
of getting a different potential respondent and the probability of getting respondents with over
ten acres of marginal land. In order to select areas of land over ten acres only contiguous
parcels of 45 pixels or more of marginal land were considered. Figure 3 is an example of a
parcel of marginal land greater than ten acres. The section of speckled pixels is all marginal land
(the section is over 45 pixels or ten acres therefore it was considered for the survey). The
second picture shows the actual property lines and where that point fell. The outline is then the
piece of property whose owner was contacted for the survey.

19

The survey was targeted to 100 individuals in each of the randomly selected counties for
1200 individuals in total. However, not all of the counties selected had 100 different individuals
with parcels of over ten acres of potentially marginal land in the CDL 2010 database. After
dropping repeat individuals, individuals who owned parcels less than ten acres in size, and
parcels owned by the public and corporate sectors, the following counties had less than 100
potential respondents: Saginaw – 95, Livingston – 84, and Branch – 73. This resulted in the
survey being sent to a total of 1152 potential respondents.

20

Chapter 6: Survey Design

A mail survey was used to elicit these landowners’ willingness to supply land for
bioenergy crops. Currently, there is no market for cellulosic biomass; therefore, landowner
decisions cannot be directly observed to determine value. Champ et al (2003) show how it is
possible to value goods where a direct market does not exist for them by using a contingent
valuation survey to elicit stated preferences where observed preferences do not exist. Often
contingent valuation surveys are used to elicit a person’s willingness to pay for a good, service,
or amenity that does not have a market; however, contingent valuation surveys can be used to
elicit willingness to accept payment to supply a good or service that currently is not sold.
Swinton et al (2007) describe how it is possible to use contingent valuation surveys to elicit
farmers’ willingness to supply ecosystem services.
Throughout the survey design process many decisions were made in order to ensure
that the survey completed its goal of eliciting Michigan landowners’ willingness to supply their
marginal land for bioenergy production. The first decision was to use a mail survey. A mail
survey was preferred over an internet based or phone survey for three reasons. First, the
recipient information was gathered from county tax records which included mailing addresses.
Second, given that many landowners are older and live in more rural communities, their access
to the internet may be limited. Third, mail surveys offer an advantage over phone based
surveys in their ability to display visual information.
21

6.1 Questionnaire Format

The survey followed the tailored design method by Dillman (2009) with detailed
question design and a series of mail outs with a one dollar incentive to elicit the highest
response. The first part of the survey focused on landowners’ current land management and
land uses. They were asked to describe their land, whether they currently rented any of it,
whether they used it for any non-agricultural uses, and what their attitude was towards renting
their land. The goal of this section was to elicit how these landowners use their land.
The second section was a series of questions aimed to educate the respondents
indirectly about bioenergy and bioenergy crops. Dillman (2009) discussed difficulties with
getting respondents to read sections of text before answering questions but noted that they do
read each question as they answer it. This section asked respondents if they were aware of
certain features of bioenergy crops. The goal of this section was to allow respondents to more
fully understand the decisions they were asked to make in the following contingent valuation
section.
The part of the questionnaire after the contingent valuation question section was about
respondents’ attitudes towards the environment and what concerns they might have with
renting their land. This section followed the contingent valuation question section in order to
minimize bias to the contingent valuation questions. The goal of this section was to create
attitudinal variables from which to see if environmentalism or certain concerns increase or
decrease the probability of renting land for bioenergy crops.

22

The final section of the questionnaire elicited demographic information about
respondents. This section was aimed at helping describe respondents and creating variables to
see if demographic background plays an important role in landowner decisions.
6.2 Contingent Valuation Question

Figure 4 shows an example of the main contingent valuation question asked in each
survey. This question aimed to elicit respondents’ willingness to supply their land for bioenergy
crops. This question starts out with an information section at the top that provides the
respondent with unbiased background information on the crop so that they can make an
informed decision. The contingent valuation question is framed as their willingness to rent their
land to grow bioenergy crops. This format was used for two main reasons. First, many rural
landowners in Michigan are not involved in farming at all; therefore, if the question was
designed around them having to grow the crop themselves they would likely not be interested
or not have the capabilities to do so. Second, a rental rate is a very easy payment method to
understand. It does not involve calculating a series of costs and revenues related to farming to
deduce a profit. The only cost is their land and the only revenue is rental rate times the amount
of land they are willing to rent.
The actual decision question on supplying their land is structured as a simple binary
choice. While a binary answer does not provide a lot of information there are inherent issues in
using different question structures (Dillman 2009). An open ended question asking at what
rental rate respondents would be willing to rent their land might lead people to overstate their
actual value in an effort to affect potential future prices paid to them. In the same way when
23

given multiple options of different rental rates to choose from respondents will often overstate
their actual value or struggle from what is called median bias where a respondent believes the
middle number provided to be closest to what their land is worth and therefore select it
without considering their own true value (Dillman 2009).
The binary choice was extended in a few ways in order to obtain as much information as
possible about the respondent’s decision. It was first extended to allow respondents to state at
a given price how many acres they would be willing to make available. This was done so that
the results could be used to go beyond eliciting how many people are willing to rent to being
able to deduce how much land would actually be available. The second way the binary question
was extended was to allow respondents to explain their “no” responses. The first option “I do
not own any existing cropland” lets us know if they should even be put into the cropland
model. The other “no” options tell us whether their decision not to rent was based upon the
rental rate or on a more general disagreement with the idea of renting out their land for that
crop. The final way the contingent valuation question was extended was to allow respondents
to answer the question for each type of land they owned. Earlier in the survey respondents
were asked to classify their land into cropland, hay and pasture land, and other farmable lands.
This was done in an effort to separate out how much marginal land might truly be available for
bioenergy crops compared to how much cropland might be available.

24

Figure 4. Example of Contingent Valuation Question

25

6.3 Experimental Design

The experimental design describes how the treatment variables were structured across
the 32 versions of the survey in order to elicit landowners’ willingness to supply land for
bioenergy production. The rental rate and the contract length for a given bioenergy crop were
the design variables that varied from one questionnaire to another. The conceptual framework
set up how consumption affects utility and that consumption is bound by an income constraint.
Here we use the first variable in the experimental design to measure the amount of potential
income change. The rental rate a landowner sees multiplied by the amount of land they would
rent is their change in income. Contract length plays into the landowner utility model in a
different way. Landowners receive utility from amenities that can come from their land. These
amenities change with a land use change such as growing bioenergy crops. Given that the
contract length determines the time period for this land use change we can see that it would
directly affect amenity values as well as any other opportunity of locking the land use for the
agreed upon length of contract. The goal of this section is to explain how the different levels of
these variables were chosen and how these levels were put into the questionnaire to provide
the best potential analysis of the results.
The range of values assigned to the rental rate that respondents saw was based upon
recent rental rates in Michigan. The 2010 and 2011 Michigan Land Values and Leasing Rates
publications by Wittenberg and Harsh from the Department of Agricultural, Food and Resource
Economics at Michigan State University were consulted to provide an accurate view of current
26

Michigan rental rates. Rental rates for cropland in Michigan in 2011 vary based upon crop,
tillage practices, and irrigation but in the Southern Lower Peninsula $111 per acre was the
average rate for tiled cropland and $84 per acre was the average rate for non-tiled cropland.
Using these values as a reference points, respondents saw values of $50, $100, $200, or $300 as
the rental rate per acre. At $50 per acre, the minimum rental rate offered was around half the
typical rate. This rate was chosen because it is important to see how landowners respond to
low rental rates that might be more realistic for bioenergy crops on lands of marginal
production potential. In order to reach a level that would elicit a response from as many
respondents as possible, the upper limit was three times the current average at $300 per acre.
Respondents also saw the approximate average itself, $100 per acre, and double the average,
$200 per acre, in order to provide greater information between the minimum and maximum
rental rates offered.
The contract length varied between 5 and 10 years. The reason a varying contract length
was provided to respondents was because many bioenergy crops are perennials and require
time to grow before returning a consistent yield. Also, a land use change involving a longer
commitment could have a different opportunity cost associated with it.
Given the levels of variation for rental rate and contract length, we now consider how to
put them into the questionnaire in a way that results in the best potential analysis of the
responses. First, we must remember that each cropping system has an independent stated
choice question. That is, the cropping systems are not alternatives. The respondent is not
deciding whether they will rent out their land for switchgrass or corn but rather given the listed

27

price and contract length for switchgrass would they rent out their land to grow it. Since each
cropping system is independent, the resultant combination of all possible levels across all
factors, or full factorial design, for a given cropping system is quite small. Only two attributes
vary, rental rate and contract length. Rental rate has only four levels and contract length has
only two levels. This means that the full factorial design 4 x 2 = 8 combinations can easily be
used for each crop. Using the full factorial design for each of the four bioenergy crops ensures
orthogonality. Orthogonality means that each combination is uncorrelated, which results in
each combination providing different information than the others. Orthogonal designs allow for
independent estimation of the influence of each varying attribute, rental rate and contract
length (Keppel and Wickens 2004). Thirty two versions of the survey were created by taking
each crop and assigning it the eight full factorial combinations and then randomly pairing those
combinations with other full factorial combinations for the three additional crops systems. This
was done instead of using only eight versions of the survey to reduce potential bias created
from the order of rental rates from only a few versions (Dillman 2009). Table 3 shows the 32
different versions used and highlights the eight full factorial design combinations in each
cropping system. The other 24 combinations for each crop are simply the same eight
combinations of the full factorial design repeated three more times but just randomly placed.

28

Table 3. Questionnaire Experimental Design
Survey
Version
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32

Corn
Rental Contract
Rate
Length

Switchgrass
Rental Contract
Rate
Length

Hybrid Poplar
Rental Contract
Rate
Length

Mixed Prairie
Rental Contract
Rate
Length

50
50
100
100
200
200
300
300
200
300
200
200
50
100
200
300
50
300
300
50
50
300
100
100
100
50
200
100
100
50
300
200

50
300
50
100
50
200
200
200
50
50
100
100
200
200
300
300
300
50
300
200
200
100
200
100
300
50
300
50
100
100
100
300

300
200
50
100
50
200
50
300
200
300
200
50
300
200
100
100
50
50
100
100
200
200
300
300
100
50
300
100
100
300
200
50

300
200
50
50
100
50
50
100
300
300
200
50
300
100
200
300
300
100
50
200
100
100
200
200
50
50
100
100
200
200
300
300

5
10
5
10
5
10
5
10
5
5
10
5
5
10
10
5
10
5
10
5
10
10
5
5
10
10
5
10
5
5
10
10

5
10
10
10
10
10
5
5
5
10
5
10
5
10
5
10
5
5
10
10
5
10
10
5
5
10
10
5
10
5
5
5

29

10
5
10
10
10
10
5
5
10
5
5
5
10
5
5
10
5
10
5
10
5
10
5
10
5
5
5
10
5
10
10
10

5
10
5
10
10
5
5
10
5
5
5
10
10
5
10
10
10
5
10
5
5
10
10
5
5
10
5
10
5
10
5
10

6.4 Questionnaire Review Process

The questionnaire went through multiple levels of review. First the survey design was
presented on August 3, 2011 at the joint Michigan State University and University of Michigan
Energy and Environmental Economics Day. After further survey development, on September 3,
2011 the questionnaire was sent out to nine experts in the fields of contingent valuation survey
design, bioenergy, biology, agronomy, and crop and soil sciences. Then it was pretested during
September 14-17, 2011 in face-to-face interviews with six landowners sampled from Ingham
County in Michigan. After slight changes it was tested again with four more Ingham county
landowners on September 27-28, 2011. The questionnaire was revised for clarity and to
accommodate the fact that many rural landowners who were targeted for marginal land
ownership also owned significant tracts of cropland. In November, 2011, the questionnaire was
sent again to three professors in the field of agricultural, food, resource, and energy economics
for comments. Then in December, 2011 the questionnaire was further field tested with five
interviews in Allegan County in Michigan. Finally, the whole survey design was then presented
on February 13-14, 2012 at Great Lakes Bioenergy Research Center (GLBRC) retreat where
professors from MSU and University of Wisconsin (UW) in the fields of biology, biochemistry,
agronomy, crop and soil sciences, and chemical engineering reviewed the design. This iterative
process of testing and revision lead to a well refined questionnaire.
6.5 Survey Response Rate and Data Entry Methods

The first mail-out was sent on March 30, 2012. It was a letter informing potential
respondents that in a few days they would be receiving an important survey for which their
30

responses were highly valued. On April 6, 2012 questionnaire packets were sent out to the
sample of 1152 owners of marginal land in the southern lower peninsula of Michigan. These
packets included a cover letter introducing the questionnaire, the questionnaire itself, a one
dollar bill as incentive, and a prepaid return envelope for the questionnaire. A reminder
postcard was sent on the 17th of April, which urged recipients to respond if they had not yet.
Lastly, a second questionnaire packet was mailed out on April 27, 2012 to those who did not
respond to the first survey wave. This packet included a different cover letter and no incentive.
Examples of all of these mail outs can be seen in an appendix. By August 6, 2012, three months
after the first questionnaire wave, a total of 599 responses were collected. An additional 124
questionnaires were returned to sender by the Postal Service due to moved individuals,
deceased individuals, and address errors. This resulted in the effective response rate for the
survey being 58.3%.
Each questionnaire was coded and the data entered separately by two different
individuals using the same coding system. When differences were found between the two
coding versions, the original questionnaire was reexamined to correct any errors in coding.
Cleaning of the data also took place to ensure that responses were within the limits of the
question. All binary questions were limited to 0 and 1 and Likert scale questions were limited to
1, 2, 3, 4, and 5. Open ended integer questions were limited to positive numeric values. Very
few responses fell outside the boundaries of the questions and those that did were coded as
missing unless confirmed by both reviewers to be a clear alternative choice.

31

Michigan State University’s Institutional Review Board (IRB) found the survey to be
exempt from further review after the initial proposal. Nonetheless, to protect respondent
confidentiality, safety steps were put in place. To begin, the returned questionnaires contained
no personal contact information. Instead, a number system was used to link each questionnaire
to a respondent identity database spreadsheet that was kept only on two MSU computers that
were password protected and in a locked MSU office. While the questionnaire data was being
entered, the questionnaires were stored in this same locked MSU office and not removed until
after all data was entered. They were then moved into locked file cabinets to be stored in case
of need for future reference for three years. At the end of this time they will be shredded and
recycled.

32

Chapter 7: Empirical Methods

Through the questionnaire, landowners were asked to make two decisions for each
bioenergy crop on each type of land that they own. First, landowners were asked whether they
would be willing to rent any of their land to grow bioenergy crops. Second, if they said yes, they
were asked how much of it they would be willing to rent. To analyze this two-step decision
making process, an econometric hurdle model was used. Hurdle models allow the two decisions
to be modeled separately with the understanding that different factors may affect these
decisions.
The hurdle model was developed by Cragg in 1971 and has been used in a variety of
contingent valuation studies. The hurdle model has been used in contingent valuation studies
by Goodwin et al. (1993) to analyze hunters’ willingness to pay, Reiser and Scechter (1999) to
examine willingness to pay for environmental program benefits, Yu and Abler (2010) to infer
willingness to pay for air pollution reduction in Beijing, and Jolejole (2009), to estimate farmers’
willingness to supply ecosystem services. The hurdle model has also been widely used outside
of the willingness to accept or supply studies in areas such as household food expenditures,
consumption models, and demand for health care (Newman 2001, Jensen and Yen 1996, Yen
and Jones 1997, and Pohlmeier and Ulrich 1995).
The hurdle model separates the decision to participate from the level of participation. In
this study that means the decision to rent out land to grow bioenergy crops, referred to as the
33

participation decision, is modeled separately from the decision of how many acres to commit,
referred to as the acreage commitment model. Separating the two models allows the
explanatory variables to have different coefficient estimates between the two models. From a
theoretical standpoint this method is supported by the nature of the decisions and the
explanatory variables used. For example the amount of total cropland a landowner owns may
not have a significant effect on whether or not they will rent any cropland, but it may have a
very significant effect on how much cropland they choose to rent out.
For a model with a dependent variable censored at zero, a tobit model may suffice if
explanatory variable effects do not differ between the participation and commitment decisions.
Therefore, a tobit model was also tested for modeling these landowner decisions. However, the
likelihood ratio test comparing the tobit model with the hurdle model showed that the hurdle
model offered a significant improvement in model fit. The results of this analysis can be seen at
the end of the chapter in Table 10.
To model the participation and acreage commitment models, a probit and truncated
regression are used. A probit model is used to estimate the binary response of whether the
landowner is willing to rent land for bioenergy crops. It is estimated using standard maximum
likelihood procedure and takes the basic form below in Equation (11), where
choice of whether to rent, Pr denotes probability,
function,

is a vector of explanatory variables,

is the binary

is the normal cumulative distribution
is a vector of parameters and

is the

standard deviation for the participation model estimated by maximum likelihood, and the

34

subscript denotes an individual landowner. This framework for the hurdle is adapted from
Jolejole (2009).

(11)
A truncated regression is used to model the second step of the hurdle, the acreage
commitment model. This model takes the simple form seen in Equation (12), where
rented,

the vector of coefficients,

the explanatory variables, and

is acres

the independently

and normally distributed error term with mean zero and variance

(12)
Enrolled acres are only observed if

(

|

so our expected value of acres is,

)

(13)

where

(14)

and

where

is the standard normal probability density function and

is the standard normal

cumulative distribution function.
The hurdle model allows for the coefficients from the participation and acreage
commitment models to be different as seen in how they are labeled with different subscripts to
denote that they are from either the participation model or the acreage commitment model.

35

These two models can then be combined to show that how many acres would be expected of
an individual landowner as displayed in Equations 13 and 14.
7.1 Variable Specification

The dependent binary variable in the probit participation model indicated whether or
not a landowner was willing to rent a given type of land. If they were willing, the variable was
set to 1; if they were not willing, it was set to 0. The dependent variable in acreage
commitment model was a continuous variable simply equal to the number of acres that the
individual was willing to rent.
The explanatory variables were split into five broad categories: current land
management, income and amenity land uses, landowner opinions on the environment and
renting concerns, landowner demographic information, and the terms of the rental agreement.
A complete list of all the variables used in each category can be seen in Table 4.
Current land management practices were variables that described what the landowner’s
land looked like in terms of division between cropland, pasture, and other land covers. Income
and amenity based land uses were binary choice variables that landowners could select. The
landowner demographic variables provided a basic description in terms of age, gender, income,
and job of the landowner responding to the survey. The final set of explanatory variables
included the experimental design variables that described the rental scenario to the landowner
in terms of rental rate and contract length. Some variables that were created from the
questionnaire responses were dropped due to multicollinearity detected through F-tests; other
variables were dropped by testing for simultaneous statistical insignificance through Wald test.
36

Table 4. Variable Specification for Variables Used in Analysis
Dependent Variables
Decision to Rent Out Land

Description

Units

Whether to rent cropland, pasture, and/or other land for corn,
switchgrass, prairie, and/or poplar
Number of acres rented of cropland, pasture, and/or other
land for corn, switchgrass, prairie, and/or poplar

Binary (yes/no)

Currently Rents Land

Whether the landowner currently rents land

Binary (yes/no)

Total Cropland Owned

Total amount of cropland the landowner owns

Acres

Total Pasture Owned

Total amount of pasture/hay the landowner owns

Acres

Total Other Land Owned
Total CRP Land Owned

Total amount of other land the landowner owns
Total amount of land committed into CRP

Acres
Acres

Number of Acres Rented

Acres

Explanatory Variables
Current Land Management

Income and Amenity Land Uses
Combined Amenity –

Number of uses for the land: scenery, physical activities,
recreational vehicle use, and as a home

0, 1, 2, 3, 4

Number of uses for the land: hunting and fishing and food plots
for game

0, 1, 2

Grazing

If their land is used for grazing

Binary (yes/no)

Commercial Income

If their land is used for commercial income

Binary (yes/no)

Conservation Income

If their land is used for conservation income

Binary (yes/no)

Based Uses
Combined Hunting –
Related Uses

Landowner Opinions on the Environment and Renting Concerns
Environmental Factor –
Renewable Energy
Environmental Factor –
General Environmentalism
Concerns factor –
Agricultural Production
Concerns factor –
Renting Land to Farmer

Factor variable based upon Likert scale variables related to
opinions on renewable energy

Factor*

Factor variable based upon Likert scale variables related to
opinions on general environmentalism

Factor*

Factor variable based upon Likert scale variables related to
concerns with agricultural production

Factor*

Factor variable based upon Likert scale variables related to
concerns with renting land to a farmer

Factor*

Landowner Demographic Information
Age
Male
Farmer
Income

Landowner’s age

Age in Years

Whether the landowner’s gender is male

Binary (yes/no)

Whether or not the landowner is a farmer by trade
What level of income the landowner falls into
12.5k, 37.5k, 75k, 125k, 175k, 300k

Binary (yes/no)

The rental rate per acre per year offered for a given crop

$50, $100,

Terms of Rental Agreement
Rental Rate

$200, $300
Contract Length
The contract length offered for a given crop
5, 10 (years)
A complete view of the variables and their levels can be seen in the Mail Outs and Questionnaire section
*Factors come from factor analysis as described 7.2

37

7.2 Factor Analysis
The sections on landowners’ opinions on the environment and concerns related to
renting created a series of variables based on responses to Likert scale questions. Many of
these variables were highly correlated with one another. This led us to question whether the
variation in these variables was really just a reflection of variation in a smaller number of
unobserved variables. Factor analysis allows for the creation of new variables, based upon
linear combinations of observed variables, that best reflects the variation of the underlining
unobserved variable. By creating these new variables the total number of variables in the
dataset could be reduced adding greater degrees of freedom and a reduction in highly
correlated variables.
Tables 5-7 show the results of the factor analysis. Among both the environmental
attitudinal variables and the landowner concerns, only the first two factors are kept because
they showed eigenvalues over one (Table 5). We can also see that the first two factors explain
63% and 64% of the variation for the environmental and concern related variables, respectively.
The factors were then rotated according to a varimax rotation to ensure orthogonal factors that
are not correlated. In Tables 6 and 7, shading denotes the dominant variables in the first factor
for the environmental and concern related rotated factors. The column uniqueness in the factor
analysis displays the variance that is unique to the variable and not shared with any others.

38

Table 5. Factor Analysis for Environmental Attitudes and Landowner Concerns
Environmental Attitudes

Factor Eigenvalue
Factor1
3.2047
Factor2
1.2128
Factor3
0.8031
Factor4
0.6140
Factor5
0.5207
Factor6
0.3548
Factor7
0.2895

Cumulative
Explained
Variance
0.4578
0.6311
0.7458
0.8336
0.9079
0.9586
1

Landowner Concerns

Factor Eigenvalue
Factor1
4.4522
Factor2
1.2940
Factor3
0.9177
Factor4
0.6492
Factor5
0.4546
Factor6
0.4050
Factor7
0.2991
Factor8
0.2849
Factor9
0.2428

Cumulative
Explained
Variance
0.4947
0.6385
0.7405
0.8126
0.8631
0.9081
0.9414
0.973
1

Table 6. Varimax Rotated Factors for Environmental Attitudes
Observed Variable
Factor1
Factor2
Uniqueness
Growing crops for auto fuel is necessary
0.8681
-0.0595
0.2428
Burning renewables is worth it over coal
0.8469
-0.1565
0.2583
Humans have the right to modify the environment
0.255
0.7297
0.4026
Humankind is severely abusing the environ
0.4614
-0.671
0.3369
This ecological crisis has been exaggerated
-0.4189
0.6268
0.4316
The balance of nature is easily upset
0.328
-0.6581
0.4594
Renewable energy is not urgently needed
-0.7077
0.22
0.4507
*Highlighted variables are those contributing mostly to Factor1 or 2, respectively.

Table 7. Varimax Rotated Factors for Landowner Concerns
Observed Variable
Factor1
Factor2
Uniqueness
Smell
0.1199
0.8641
0.239
Noise
0.2236
0.8539
0.2208
Dust in air
0.2333
0.8638
0.1993
Potential legal costs
0.6195
0.4702
0.3952
Length of contract
0.7658
0.196
0.3752
Possible need for insurance
0.778
0.2216
0.3456
Having others on my land
0.7709
0.2644
0.3358
Land use changing so I can no longer use it
0.6902
0.1211
0.5089
Use of pesticide and fertilizer
0.4786
0.3702
0.6339
*Highlighted variables are those contributing mostly to Factor1 or 2, respectively.
39

The use of factor analysis is further supported here by the clear grouping of related
variables. The variables that make up the first environmental attitude factor all address
renewable energy , while the variables that make up the second environmental attitude factor
all relate to a more general environmental position as determined by the social psychology
group “The New Environmental Paradigm” (Dunlap 2008). The variables that construct the first
landowner concern factor all relate to sensory effects of agricultural production and the
variables that construct the second landowner concern factor all relate to the renting of land.
7.3 Weighting and Scaling Model to Southern Lower Michigan

In order to permit extrapolation from survey respondents to the population of the
region as a whole (scaling up), the observations were weighted according to the probability that
an observation was included given the sampling design. In this study, we sampled owners of ten
acres or more of marginal land from 12 counties; however, not all counties had the same
number of ten acre plus tracts of land. Table 8 shows the number of ten acre plus tracts of
marginal land that exist in each county according to the GIS analysis discussed in Chapter 5.
Next it shows the number of observations from each of those counties that were observed. The
probability that an observation was included given the sampling design can be seen in the next
column, which was created by dividing the number of responses for each county by the number
of ten acres plus marginal land tracts in that county. The final column simply shows the inverse
probability weights (pweights), the weights used in the analysis or inverse of the probability in
the column before. The pweights allow counties that were under sampled according to their
number of tracts of ten acres or more of marginal land to have a greater impact on the model.

40

For example Livingston county has the largest number of tracts of marginal land that are ten
acres or more, 1,210 tracts, but the second lowest number of responses, 30 responses, thus
Livingston is given the largest pweight of 40.33.
Table 8. Design for Weighting Observations
County

Allegan
Barry
Branch
Ionia
Isabella
Lenawee
Livingston
Newaygo
Saginaw
Sanilac
Tuscola
Van Buren

Number of Ten Acres or
More Marginal Land
Tracts

Number of
Responses for
the County

1028
486
67
316
620
970
1210
862
558
923
952
743

42
44
26
45
42
48
30
46
42
36
51
46

Probability that an
Observation was Included
Given the Sampling Design

pweights
(Inverse of
Probability)

0.04
0.09
0.39
0.14
0.07
0.05
0.02
0.05
0.08
0.04
0.05
0.06

24.48
11.05
2.58
7.02
14.76
20.21
40.33
18.74
13.29
25.64
18.67
16.15

Creating a supply curve for all of southern Lower Michigan involved scaling up what we
knew about our survey respondents. Combining the probit, participation model, and the
truncated, acreage commitment model, told us how many acres of marginal land an average
southern Lower Michigan landowner who owns at least ten acres of marginal land is willing to
rent. To scale up, we need to know how many tracts of at least ten acres of marginal land exist
in Southern Lower Michigan. The GIS analysis in Chapter 5 showed that 2.85 million acres of
marginal land exist in southern Lower Michigan; however, much of that land is in areas less
than ten acres. Of our 12 counties that we sampled from only 21% of the acres of marginal land
were in tracts of at least ten acres and these tracts averaged 23 acres in size. Twenty one
41

percent of 2.85 million is about 600 thousand acres. Dividing these 600 thousand acres by an
average tract size of 23 acres gives us a total of 26 thousand tracts of least ten acres existing in
Southern Lower Michigan. This number was then multiplied by the willingness to supply
marginal land of our average respondent to provide us with an estimation of the willingness to
supply of all of Southern Lower Michigan. Table 9 shows these calculations and steps.
Table 9. Calculations for Scaling up to Southern Lower Michigan
Total Number of Acres of Marginal Land in Southern Lower Michigan

2.97 Million Acres

Percent of Marginal Land in Tracts of at Least Ten Acres

21%

Total Number of Acres of Marginal Land in at Least Ten Acre Tracts in Southern Lower
Michigan
Average Size of a Tract of Marginal Land that is at Least Ten Acres

23 Acres

Estimated Number of Tracts of Marginal land that are at Least Ten Acres in Southern
Lower Michigan

Table 10. Likelihood Ratio Test Comparing Tobit vs. Hurdle Model
Prairie Poplar
Cropland
Chi Squared
Prob. Chi Squared
Pasture
Chi Squared
Prob. Chi Squared
Other Marginal Land
Chi Squared
Prob. Chi Squared

Switchgrass

Corn

3312
0.00

2820
0.00

2871
0.00

5027
0.00

4376
0.00

3647
0.00

4808
0.00

4000
0.00

4734
0.00

3344
0.00

4603
0.00

3150
0.00

42

600,000 Acres

26,000 Acres

Chapter 8: Hypotheses

The purpose of this research project is to examine the availability of marginal land for
bioenergy crops in Michigan and how landowners make the decision to change their land’s use.
To address these goals we developed a series of hypotheses to help us answer the relevant
questions. These hypotheses are stated here with the theoretical rationale for their existence.
In the results section each hypothesis is examined based upon the regression results, general
survey responses, and statistical tests.
1a. As rental rates increase, the probability of renting will increase
1b. As rental rates increase, the level of acreage committed will increase
An increase in rental rate is expected to increase the probability of renting and the
amount of land a landowner is willing to rent. Based on the conceptual model designed in
Chapter 4, landowners desire to maximize utility. As rental rate increases, the utility received
from consumption through income will rise. The larger the change in utility from income, the
greater the probability that the change will offset any utility loss that the land may have
provided in the form of amenities.

43

2. The contract length offered will influence landowner decisions.
Depending on landowner preferences, either a five year or a ten year contract might be
preferable. Some landowners might have future plans for their land or expect land rent prices
to go up in the future and therefore prefer a shorter five year contract. On the other hand,
some landowners might perceive that the rental rate they were offered was high and prefer to
get that guaranteed rental rate for as long as possible.
3. Many landowners will not rent their land even at extremely high rental rates.
It is expected that for some landowners an increase in income from their land over a
plausible range will not elicit a change in their land use. Some individuals do not like the idea of
bioenergy, others do not want anyone growing crops on their land except themselves, others
do not want crops at all, and still others are simply comfortable with their current income and
see any change as an effort not worth pursuing. Our conceptual model focuses heavily on the
idea that “rational” individuals seek utility maximization and that utility need not come only
through consumption of goods and services purchased through income but also from
amenities, amenities that can be tied to land. Therefore, a “rational” utility maximizing
landowner may decide not to rent for bioenergy crops even at a high rental rate because they
receive a very large amount of utility through amenities by leaving the land the way it is.

44

4. At the same rental rate, landowners will have a higher probability of renting out land to grow
mixed prairie or switchgrass crops rather than poplar trees or corn.
Mixed prairie and switchgrass are crops that benefit from being perennials. For this
reason, they require fewer inputs and less management. Having fewer inputs is a benefit in the
sense of input costs and because many inputs such as fertilizer and pesticide can lead to
environmental costs. Reduced land management is a benefit because it decreases the costs
associated with growing the crop. Hybrid poplar trees are also perennials and offer these same
benefits; however, they have a much larger presence on a piece of land and have a root system
that would involve extensive work and cost to remove if the land were ever to change use in
the future. Corn is not a perennial and of all the crops in this study it requires the most
agrochemical inputs and the highest level of management. Switchgrass and prairie also involve
less relative production noise and disturbance. Therefore, at the same rental rate it is expected
that mixed prairie and switchgrass would be the preferred crops.
5. At the same rental rate, landowners will rent out marginal land over cropland.
All land uses have opportunity costs. Cropland has the opportunity cost of utility from
consumption through income that the land generates to the landowner either in the form of a
rental rate or from selling the products of farming it. Cropland also has the opportunity cost of
amenities that the landowner receives from it, even given that these amenity values may be
relatively smaller. Marginal land that is not in crops often provides no income to the landowner
and therefore no monetary opportunity cost exists. However, both cropland and marginal land
have opportunity costs that come from amenities. These opportunity costs are harder to
45

measure and could range from no longer being able to hunt on the land to a change in the
desired scenery. The question then becomes when a landowner is offered the same rental rate
to grow energy crops on cropland versus marginal land, will the opportunity costs from income
and amenities on existing cropland be greater than or less than the opportunity costs from
amenities on marginal land? To answer this, we must ask why is the marginal land currently not
in agricultural production? If it is not in production because it will not produce crops profitably,
then it is likely that the opportunity cost of the marginal land is less than that of the cropland.
However, if the land is marginal because it offers greater amenities that provide the landowner
with enough utility to offset the potential utility gain from consumption due to greater income
from growing crops, then it is likely that the opportunity cost of the marginal land relative to its
potential production income is greater than that of the cropland. While both of these reasons
for marginal land are possible, evidence from landowner interviews has shown that marginal
land lays idle more often because of amenities received from it rather than a complete lack of
potential profit. Much of the idle land has characteristics that make it less desirable for growing
crops, characteristics that range from higher irrigation costs to sandier, less fertile soil. Thus, in
most cases cropland would have higher opportunity costs associated with converting it to
bioenergy crops than marginal land, meaning that landowners will prefer to rent marginal land,
given the same rental rate.
6. Most landowners who own marginal land will own cropland as well.
Rural landowners who own marginal land often own cropland. This hypothesis emerged
from our qualitative interviews with landowners and our review of geographical databases.

46

Many owners of marginal lands also told us about their cropland. When observing tracts of
marginal land in geographic databases overlaid with property parcel boundaries, it became very
apparent that parcel boundaries that included marginal land rarely were covered completely by
it. In most cases, the parcel boundaries included tracts of cropland that constituted a larger
portion of the parcel than the marginal land tracts did. These two observations gave clear
reason to expect that individuals who own marginal land also own cropland.
7. At the same rental rate, landowners who own more land will have a higher probability of
renting.
Owners of large areas of land have more land that may vary in opportunity cost to the
point where growing bioenergy crops at a given rental rate provides a benefit to them greater
than the opportunity cost on at least some part of their varying land. Also most owners of large
land areas use it as a source of income. This means that they are either familiar with renting
their land to grow crops or else they farm the land themselves. In both cases, they are more
likely to gain utility from the land via income generation than from amenities. So if their
opportunity cost for the land arises mostly from income, they would have a lower opportunity
cost for the land than a smaller landowner who receives utility from amenities of the land.
Given that large landowners have land of more heterogeneous quality and they also get lower
marginal utility from land amenities, they will be more likely to rent their land than a small
landowner at the same rental rate.

47

Chapter 9: Results

9.1 The Participation Model Results

The participation decision was modeled for each bioenergy crop on each land type. The
results of the probit models for the four different bioenergy crops, mixed prairie, hybrid poplar
trees, switchgrass, and corn, on the three different land types, cropland, pasture land, and
other marginal lands, can be seen in Tables 11-13. These results include parameter estimates
and standard errors for each explanatory variable. Across the twelve models only a few
variables were consistently significant. The participation decision is most statistically influenced
by four variables: the rental rate offered, whether the landowner currently rents any land,
whether the landowner has certain preexisting land uses, and whether the landowner has
certain concerns with renting.
The influence of the rental rate can be seen clearly. As the rental rate offered increased
from $50 per acre to $300 per acre, more landowners were willing to participate in growing
bioenergy crops on their land. This is illustrated in Tables 11-13 where the variable for rental
rate is positive and significant in each model. This result answers Hypothesis 1a showing that
rental rate has the expected effect of increasing landowner willingness to participate.

48

Table 11. Probit Participation Model for Cropland Rented for Prairie and Poplar

Experimental Variables:
Ln Rental Rate ($/acre)
Contract length (yrs)
Current Land Management:
Currently Rents Land (0/1)
Current Land Owned:
Total Cropland (acres)
Total Pasture (acres)
Total Other Land (acres)
Total CRP Land (acres)
Current Land Uses:
Group of Non-Land Based Uses
Group of Hunting Related Uses
Grazing Livestock (0/1)
Commercial Income (0/1)
Conservation Income (0/1)
Environmental Factors:
Renewable Energy
General Environmentalism
Concerns Factors:
Agricultural Based
Renting Land Based
Demographic Information:
Age (yrs)
Male (0/1)
Farmer (0/1)

Prairie
Coef.

(n=251)
Z-score

Poplar
Coef.

(n=252)
Z-score

0.62
-0.09

3.79***
-2.14**

0.81
-0.02

4.93***
-0.54

0.67

3.01***

0.35

1.39

0.00045
0.00058
0.00005
0.0040

-1.17
0.94
0.04
1.67*

0.00013
-0.00032
-0.00042
0.0021

0.47
-0.36
-0.44
0.99

0.33
0.07
-0.25
-0.02
-0.29

4.53***
0.44
-0.95
-0.08
-1.15

0.24
0.09
-0.05
0.45
-0.36

2.62***
0.52
-0.20
1.68*
-1.35

-0.08
0.08

-0.80
0.80

0.09
0.13

0.84
1.20

-0.02
-0.07

-0.16
-0.64

-0.07
-0.28

-0.67
-2.21**

0.01
0.70
-0.01
-0.74
-0.07
-0.24
0.26
0.87
-0.17
-0.71
-0.56
-2.12**
1.68EIncome (scale 1-6)
06
1.20
-1.32E-06 -0.75
Constant
-4.34
-3.49***
-4.99
-4.04***
Log Likelihood Values
-1966.49
-1740.96
Wald Chi-Squared
73.79
55.47
Probability Chi-Squared
0.00
0.00
Pseudo R-Squared
0.2613
0.237
*** - significant at 1% level, ** - significant at 5% level, * - significant at 10% level

49

Table 12. Probit Participation Model for Cropland Rented for Switchgrass and Corn
Switch
grass
Coef.

(n=247)
Z-score

Corn
Coef.

(n=266)
Z-score

Experimental Variables:
Ln Rental Rate ($/acre)
0.69
4.47***
0.93
4.35***
Contract length (yrs)
-0.02
-0.40
-0.04
-0.84
Current Land Management:
Currently Rents Land (0/1)
0.87
3.85***
1.70
6.1***
Current Land Owned:
Total Cropland (acres)
-0.00007 -0.22
0.00042
1.67*
Total Pasture (acres)
-0.00019 -0.22
0.0011
1.64
Total Other Land (acres)
0.00063
0.77
-0.0023
-1.53
Total CRP Land (acres)
-0.0023
-1.13
-0.00006
-0.03
Current Land Uses:
Group of Non-Land Based Uses 0.27
3.55***
0.29
3.31***
Group of Hunting Related Uses 0.07
0.50
0.40
2.37**
Grazing Livestock (0/1)
-0.23
-0.99
0.01
0.04
Commercial Income (0/1)
0.05
0.20
-0.12
-0.47
Conservation Income (0/1)
-0.19
-0.82
-0.79
-3.05***
Environmental Factors:
Renewable Energy
0.05
0.50
0.00
-0.02
General Environmentalism
0.06
0.56
0.07
0.75
Concerns Factors:
Agricultural Based
-0.06
-0.60
-0.29
-2.26**
Renting Land Based
-0.12
-1.09
-0.24
-1.67*
Demographic Information:
Age (yrs)
0.00
-0.03
0.00
0.52
Male (0/1)
-0.11
-0.40
0.03
0.11
Farmer (0/1)
0.09
0.43
0.23
0.96
Income (scale 1-6)
-2.44E-06 -1.82*
-1.90E-06 -1.39
Constant
-4.42
-4.01***
-6.80
-4.55***
Log Likelihood Values
-2156.46
-1716.92
Wald Chi-Squared
55.84
74.90
Probability Chi-Squared
0.00
0.00
Pseudo R-Squared
0.2015
0.4064
*** - significant at 1% level, ** - significant at 5% level, * - significant at 10% level

50

Table 13. Probit Participation Model for Pasture and Hay Land Rented for Prairie and Poplar

Prairie
Coef.

(n=274)
Z-score

Poplar
Coef.

(n=274)
Z-score

Experimental Variables:
Ln Rental Rate ($/acre)
0.44
3.65***
0.48
3.33***
Contract length (yrs)
-0.05
-1.24
0.00
0.03
Current Land Management:
Currently Rents Land (0/1)
0.20
0.99
0.09
0.42
Current Land Owned:
Total Cropland (acres)
-0.0015
-1.89*
0.00000
0.01
Total Pasture (acres)
0.0012
1.28
0.00026
0.34
Total Other Land (acres)
-0.0015
-1.16
0.00009
0.08
Total CRP Land (acres)
0.0016
1.18
-0.0022
-1.15
Current Land Uses:
Group of Non-Land Based Uses 0.09
1.17
0.05
0.56
Group of Hunting Related Uses -0.02
-0.14
-0.11
-0.73
Grazing Livestock (0/1)
-0.16
-0.71
0.02
0.07
Commercial Income (0/1)
-0.23
-1.04
-0.37
-1.46
Conservation Income (0/1)
0.01
0.05
-0.30
-1.36
Environmental Factors:
Renewable Energy
-0.03
-0.29
0.02
0.21
General Environmentalism
0.00
-0.06
0.13
1.31
Concerns Factors:
Agricultural Based
-0.06
-0.64
-0.03
-0.32
Renting Land Based
0.02
0.23
-0.17
-1.54
Demographic Information:
Age (yrs)
-0.01
-0.63
-0.01
-1.43
Male (0/1)
-0.13
-0.58
0.38
1.41
Farmer (0/1)
-0.40
-1.97**
-0.51
-2.38**
Income (scale 1-6)
2.91E-06 2.13**
-3.17E-07 -0.23
Constant
-1.76
-1.82*
-2.25
1.06
-2672.26
Log Likelihood Values
-2200.96
Wald Chi-Squared
48.90
39.12
Probability Chi-Squared
0.00
0.01
Pseudo R-Squared
0.1462
0.1283
*** - significant at 1% level, ** - significant at 5% level, * - significant at 10% level

51

Table 14. Probit Participation Model for Pasture and Hay Land Rented for Switchgrass and Corn
Switch
grass
Coef.

(n=268)
Z-score

Corn
Coef.

(n=275)
Z-score

Experimental Variables:
Ln Rental Rate ($/acre)
0.57
3.9***
0.67
4.23***
Contract length (yrs)
-0.02
-0.43
0.03
0.70
Current Land Management:
Currently Rents Land (0/1)
0.38
1.87*
0.70
3.04***
Current Land Owned:
Total Cropland (acres)
-0.00021 -0.68
0.00005
0.16
Total Pasture (acres)
0.00045
0.65
0.0015
1.79*
Total Other Land (acres)
-0.00003 -0.02
-0.00045
-0.34
Total CRP Land (acres)
-0.0022
-1.41
-0.00016
-0.10
Current Land Uses:
Group of Non-Land Based Uses 0.07
0.94
-0.01
-0.08
Group of Hunting Related Uses
-0.06
-0.41
0.15
1.12
Grazing Livestock (0/1)
-0.30
-1.37
-0.16
-0.71
Commercial Income (0/1)
-0.14
-0.61
-0.51
-2.21**
Conservation Income (0/1)
-0.11
-0.54
-0.63
-2.76***
Environmental Factors:
Renewable Energy
0.09
0.93
0.08
0.89
General Environmentalism
0.09
1.10
0.24
2.69***
Concerns Factors:
Agricultural Based
-0.09
-0.97
-0.13
-1.42
Renting Land Based
-0.03
-0.25
-0.15
-1.41
Demographic Information:
Age (yrs)
0.00
-0.29
0.00
-0.03
Male (0/1)
-0.19
-0.82
0.33
1.26
Farmer (0/1)
-0.03
-0.15
0.01
0.05
Income (scale 1-6)
-4.58E-08 -0.04
-6.73E-07 -0.49
Constant
-2.73
-2.76***
-4.31
-4.06***
Log Likelihood Values
-2684.83
-2403.75
Wald Chi-Squared
31.41
54.96
Probability Chi-Squared
0.05
0.00
Pseudo R-Squared
0.1046
0.1922
*** - significant at 1% level, ** - significant at 5% level, * - significant at 10% level

52

Table 15. Probit Participation Model for Other Marginal Lands Rented for Prairie and Poplar

Prairie
Coef.

(n=354)
Z-score

Poplar
Coef.

(n=349)
Z-score

Experimental Variables:
Ln Rental Rate ($/acre)
0.43
3.96***
0.66
5.51***
Contract length (yrs)
-0.02
-0.69
-0.03
-1.00
Current Land Management:
Currently Rents Land (0/1)
-0.01
-0.04
-0.40
-2.14**
Current Land Owned:
Total Cropland (acres)
-0.00036 -1.13
-0.00019
-0.78
Total Pasture (acres)
-0.00031 -0.49
-0.00044
-0.72
Total Other Land (acres)
0.0011
1.19
0.0017
1.72*
Total CRP Land (acres)
0.0015
0.96
-0.00052
-0.30
Current Land Uses:
Group of Non-Land Based Uses 0.01
0.24
-0.01
-0.18
Group of Hunting Related Uses 0.15
1.31
0.00
-0.03
Grazing Livestock (0/1)
-0.12
-0.63
-0.27
-1.26
Commercial Income (0/1)
0.02
0.10
0.44
2.15**
Conservation Income (0/1)
-0.57
-2.86***
-0.15
-0.74
Environmental Factors:
Renewable Energy
-0.05
-0.63
0.11
1.20
General Environmentalism
0.08
1.06
0.06
0.73
Concerns Factors:
Agricultural Based
-0.12
-1.59
0.04
0.57
Renting Land Based
0.06
0.70
-0.20
-2.36**
Demographic Information:
Age (yrs)
0.00
-0.13
-0.01
-1.50
Male (0/1)
-0.04
-0.19
0.30
1.43
Farmer (0/1)
0.02
0.13
-0.26
-1.45
Income (scale 1-6)
1.09E-06 0.90
8.67E-07
0.75
Constant
-2.20
-2.65***
-2.95
-3.39***
Log Likelihood Values
-3680.68
-3230.01
Wald Chi-Squared
43.37
69.66
Probability Chi-Squared
0.00
0.00
Pseudo R-Squared
0.1076
0.1578
*** - significant at 1% level, ** - significant at 5% level, * - significant at 10% level

53

Table 16. Probit Participation Model for Other Marginal Lands Rented for Switchgrass and Corn
Switch
grass
Coef.

(n=354)
Z-score

Corn
Coef.

(n=354)
Z-score

Experimental Variables:
Ln Rental Rate ($/acre)
0.28
2.34**
0.52
4.12***
Contract length (yrs)
0.02
0.63
0.07
2.02**
Current Land Management:
Currently Rents Land (0/1)
-0.09
-0.48
0.05
0.24
Current Land Owned:
Total Cropland (acres)
-0.00025 -0.79
0.00015
0.69
Total Pasture (acres)
-0.00094 -1.09
-0.00055
-0.05
Total Other Land (acres)
0.0023
2.32**
0.00084
0.82
Total CRP Land (acres)
-0.0035
-1.48
-0.0070
-1.52
Current Land Uses:
Group of Non-Land Based Uses 0.06
0.90
-0.06
-0.94
Group of Hunting Related Uses 0.06
0.51
0.22
1.86*
Grazing Livestock (0/1)
-0.24
-1.15
-0.32
-1.57
Commercial Income (0/1)
0.19
0.95
-0.04
-0.22
Conservation Income (0/1)
-0.56
-2.69***
-0.60
-2.73***
Environmental Factors:
Renewable Energy
0.06
0.77
0.09
1.07
General Environmentalism
0.04
0.46
0.24
2.97***
Concerns Factors:
Agricultural Based
-0.04
-0.46
-0.16
-2.1**
Renting Land Based
-0.16
-1.88*
-0.14
-1.57
Demographic Information:
Age (yrs)
0.00
-0.01
0.00
0.50
Male (0/1)
-0.30
-1.46
-0.07
-0.36
Farmer (0/1)
0.06
0.31
0.01
0.07
Income (scale 1-6)
-2.08E-07 -0.17
-3.68E-07 -0.28
Constant
-1.64
-2.03**
-3.68
-4.18***
Log Likelihood Values
-3716.60
-3171.77
Wald Chi-Squared
39.63
58.97
Probability Chi-Squared
0.01
0.00
Pseudo R-Squared
0.0916
0.1669
*** - significant at 1% level, ** - significant at 5% level, * - significant at 10% level

54

The second significant influential variable was that landowners who are currently
renting out their land are more likely to participate in renting their land out for bioenergy crops
as well. As expected, landowners who are accustomed to renting their land for an existing crop
would likely not have any amenity change if the land use changed to growing a bioenergy crop.
Also these individuals in general have shown a general preference to rent their land and that
they are comfortable with letting others manage their land, not seeing it as unduly bothersome
or a hassle.
The next consistently significant group of variables in most of the participation models
were the variables related to uses for the land. Surprisingly, landowners who generally use their
land for scenery, recreation, or physical activities are more likely to rent out their land for
bioenergy crops. The interesting feature of this group of land uses is that none of them requires
a specific land cover; they are all indirect land uses. On the other hand, landowners who use
their land for conservation income are less likely to rent their land. This is consistent with our
expectation that landowners who have alternative uses for their land that require a specific
land cover will be less likely to participate in growing bioenergy crops.
The final two variable that are consistently significant in the participation model deal
with concerns that a landowner might have when renting their land to grow bioenergy crops.
As mentioned in the empirical methods section, a factor analysis of sources of concerns divided
the concerns into two key types, those related to agricultural activities and those related to
renting land in general. In both cases the greater these concerns the less likely the landowner is
to participate in growing bioenergy crops.

55

One variable that was expected to be significant but was not was the contract length
offered. Hypothesis 2 states, “The contract length offered will influence landowner decisions.”
However, in 10 of the 12 participation models, estimated contract length was not significant
even at the 10% level. Hence, we have no support for the hypothesis that contract length
influences landowner decisions.
When the results of the participation decision are viewed graphically in Figures 5-7,
three observations are apparent that correspond to different hypotheses. First, rarely are more
than half of rural landowners willing to rent out any amount of land to grow bioenergy crops.
This result obtains regardless of the bioenergy crop, the type of land the crop is being grown
on, or the rental rate offered. At a typical crop rental rate of $100 per acre, the proportion
willing to rent out for bioenergy crops falls to between 30 and 40 percent of rural landowners.
This finding is consistent with Hypothesis 3, which states that even at high rental rates, many
landowners will not be willing to rent out land for bioenergy crops. This finding is also
consistent with the feedback from pretesting the survey with Michigan landowners. Many
landowners simply were unwilling to rent their land for bioenergy crop production, regardless
of the price offered.
The second observation is that on all three land types, landowners were most willing to
rent out their land for switchgrass and prairie (Figures 5-7). This finding partially supports
Hypothesis 4 that says, “At the same rental rate, landowners will have a higher probability of
renting out land to grow mixed prairie or switchgrass crops over poplar trees or corn.”
Landowners, particularly farmers, proved especially averse to renting land for hybrid poplar

56

trees. They were less likely to rent out any type of land for hybrid poplar trees, with the
exception of other marginal lands at high rental rates (when corn was the least preferred crop).
Hybrid poplar trees may be less desirable because they have stumps and woody root systems
that are difficult to remove if any potential future land use change is desired.
The third observation from Figures 5-7 is that landowners are only slightly more likely to
rent out their marginal land than their cropland. If we look at just at prairie and switchgrass, the
two bioenergy crops for which land is most likely to be rented, we see that the probability of
renting cropland for is only 0.26 to 0.28 at a typical rental rate of $100 per acre, while at the
same rate, the probability of renting pasture lands and other marginal lands is 0.32 to 0.38.
These results offer weak support Hypothesis 5 that states, “At the same rental rate, landowners
will rent out marginal land over cropland.”
Figure 5. Probability of Renting Cropland for Bioenergy Crops in Response to Rental Rate
(Probit)
Average Landowner's Probability of Renting Out Cropland for:
1
0.9
0.8

Probability

0.7
0.6

Corn

0.5

Switchgrass

0.4

Poplar

0.3

Prairie

0.2
0.1
0
$-

$50

$100

$150

$200

Rental Rate per Acre

57

$250

$300

Figure 6. Probability of Renting Pasture Land for Bioenergy Crops in Response to Rental Rate
(Probit)
Average Landowner's Probability of Renting Out Pasture Lands for:
1
0.9
0.8

Probability

0.7
0.6

Corn

0.5

Switchgrass

0.4

Poplar

0.3

Prairie

0.2
0.1
0
$-

$50

$100

$150

$200

$250

$300

Rental Rate per Acre

Figure 7. Probability of Renting Other Marginal Land for Bioenergy Crops in Response to Rental
Rate (Probit)
Average Landowner's Probability of Renting Out Other Marginal Land for:
1
0.9
0.8

Probability

0.7
0.6

Corn

0.5

Switchgrass

0.4

Poplar

0.3

Prairie

0.2
0.1
0
$-

$50

$100

$150

$200

Rental Rate per Acre

58

$250

$300

9.2 The Acreage Commitment Model Results
The acreage commitment model captures how many acres an individual landowner is
willing to rent, given that they have already decided to rent out some land. The results of the
truncated models for the four different bioenergy crops on the three different land types can
be seen in Tables 14-16. These results include parameter estimates and standard errors for
each model. These parameter estimates can be thought of as the change in the number of
acres a landowner is willing to rent for a unit change in the explanatory variable.
The most consistently significant influence on the acreage commitment decision was
how much land the respondent owned. This result is quite logical as the amount of land a
landowner owns directly limits how much is available to rent. However, what is interesting is
the variation in these coefficients from one land type to the next. Each additional acre of
cropland a landowner owns almost directly correlates with an additional acre of cropland that
they are willing to rent, as seen by the coefficients 0.85, .99, 1.02, and .94 prairie, poplar,
switchgrass, and corn in Tables 14a and 14b. On pasture and hay land the coefficients for this
same land type are similar at 1.05, 1.11, 0.96, and 0.61 across the same four bioenergy crops
(Tables 15a and 15b). However, on other marginal lands these coefficients fall sharply to 0.47,
0.32, 0.27, and 1.11 (Tables 16a and 16b). These results suggest that landowners in general are
willing to rent out for bioenergy crops a much higher proportion of their crop and pasture land
than their marginal land.

59

Table 17. Truncated Acreage Model for Cropland Committed to Prairie and Poplar

Experimental Variables:
Ln Rental Rate ($/acre)
Contract length (yrs)
Current Land Management:
Currently Rents Land (0/1)
Current Land Owned:
Total Cropland (acres)
Total Pasture (acres)
Total Other Land (acres)
Total CRP Land (acres)
Current Land Uses:
Group of Non-Land Based Uses
Group of Hunting Related Uses
Grazing Livestock (0/1)
Commercial Income (0/1)
Conservation Income (0/1)
Environmental Factors:
Renewable Energy
General Environmentalism
Concerns Factors:
Agricultural Based
Renting Land Based
Demographic Information:
Age (yrs)
Male (0/1)
Farmer (0/1)
Income (scale 1-6)
Constant
Log Likelihood Values
Wald Chi-Squared
Probability Chi-Squared

Prairie
Coef.

(n=70)
Z-score

Poplar
Coef.

(n=52)
Z-score

0.98
1.28

0.21
0.78

-4.13
1.93

-0.79
1.12

9.56

1.24

0.28

0.04

0.85
-0.016
-0.11
0.15

8.68***
-0.15
-1.28
1.26

0.99
-0.11
0.13
0.17

242.99***
-1.12
2.21**
2.44**

8.00
-6.18
-8.92
-18.05
9.92

1.44
-0.80
-0.70
-1.82*
0.96

-10.57
-2.99
2.05
8.54
-12.96

-3.86***
-0.86
0.25
0.75
-1.30

10.85
3.50

2.23**
0.87

3.20
8.76

0.73
2.5**

-1.22
-6.68

-0.27
-1.57

-0.93
6.77

-0.44
1.52

1.03
2.30
4.51
-2.66E-05
-103.83
-4817.92
920.59
0.00

2.35**
0.24
0.53
-0.57
-1.85*

0.42
-21.65
-7.55
9.87E-05
4.74
-3224.46
1.20
0.00

1.58
-1.71*
-0.90
3.03***
0.10

*** - significant at 1% level, ** - significant at 5% level, * - significant at 10% level

60

Table 18. Truncated Acreage Model for Cropland Committed to Switchgrass and Corn

Switch
grass
Coef.

(n=77)
Z-score

Corn
Coef.

(n=77)
Z-score

Experimental Variables:
Ln Rental Rate ($/acre)
0.00
0.00
50.75
1.50
Contract length (yrs)
-2.51
-1.47
8.19
1.02
Current Land Management:
Currently Rents Land (0/1)
31.35
2.28**
-46.26
-0.91
Current Land Owned:
Total Cropland (acres)
1.02
124.15***
0.94
23.79***
Total Pasture (acres)
0.38
3.27***
-1.48
-13.39***
Total Other Land (acres)
-0.31
-3.21***
0.56
1.7*
Total CRP Land (acres)
0.15
1.36
0.19
0.76
Current Land Uses:
Group of Non-Land Based Uses -4.54
-1.7*
-4.50
-0.36
Group of Hunting Related Uses
11.79
2.52**
10.69
0.44
Grazing Livestock (0/1)
12.23
1.31
125.96
2.57**
Commercial Income (0/1)
-14.49
-1.81*
36.23
0.78
Conservation Income (0/1)
-4.59
-0.47
84.67
1.9*
Environmental Factors:
Renewable Energy
3.03
0.64
-9.73
-0.47
General Environmentalism
-2.93
-0.77
17.03
0.91
Concerns Factors:
Agricultural Based
-2.38
-0.64
-16.92
-1.17
Renting Land Based
-2.77
-0.61
16.97
0.84
Demographic Information:
Age (yrs)
0.99
2.04**
1.86
1.16
Male (0/1)
-0.38
-0.03
5.38
0.09
Farmer (0/1)
-5.92
-0.68
-10.93
-0.37
Income (scale 1-6)
7.65E-05 0.73
2.94E-04
1.41
Constant
-88.32
-1.77*
-586.45
-2.16**
Log Likelihood Values
-4956.01
-6494.08
Wald Chi-Squared
1.90
29746.79
Probability Chi-Squared
0.00
0.00
Pseudo R-Squared
*** - significant at 1% level, ** - significant at 5% level, * - significant at 10% level

61

Table 19. Truncated Acreage Model for Pasture and Hay Land Committed to Prairie and Poplar

Experimental Variables:
Ln Rental Rate ($/acre)
Contract length (yrs)
Current Land Management:
Currently Rents Land (0/1)
Current Land Owned:
Total Cropland (acres)
Total Pasture (acres)
Total Other Land (acres)
Total CRP Land (acres)
Current Land Uses:
Group of Non-Land Based Uses
Group of Hunting Related Uses
Grazing Livestock (0/1)
Commercial Income (0/1)
Conservation Income (0/1)
Environmental Factors:
Renewable Energy
General Environmentalism
Concerns Factors:
Agricultural Based
Renting Land Based
Demographic Information:
Age (yrs)
Male (0/1)
Farmer (0/1)
Income (scale 1-6)
Constant
Log Likelihood Values
Wald Chi-Squared
Probability Chi-Squared

Prairie
Coef.

(n=102)
Z-score

Poplar
Coef.

(n=58)
Z-score

0.27
0.73

0.09
0.91

-8.02
-1.18

-1.42
-1.20

-4.64

-0.98

-4.42

-0.52

0.034
1.05
-0.13
-0.022

1.35
54.45***
-3.34***
-1.12

0.0036
1.11
-0.024
-0.22

0.60
30.52***
-0.24
-1.55

-3.75
3.37
-7.65
-1.97
4.02

-2.7***
1.00
-1.32
-0.37
0.87

-7.94
5.38
-17.61
14.98
-1.69

-2.61***
1.00
-1.68*
1.73*
-0.21

-3.38
-0.55

-1.45
-0.31

-0.85
2.38

-0.24
0.83

-1.62
-4.58

-0.73
-1.98**

5.27
-1.98

1.68*
-0.56

0.28
-8.79
2.39
-9.81E-06
-21.78
-6206.49
41739.70
0.00

1.46
-1.83*
0.66
-0.26
-0.85

0.33
-12.94
1.59
2.26E-05
32.31
-3525.63
4861.62
0.00

1.06
-1.77*
0.20
0.49
0.86

*** - significant at 1% level, ** - significant at 5% level, * - significant at 10% level

62

Table 20. Truncated Acreage Model for Pasture and Hay Land Committed to Switchgrass and
Corn
Switch
grass
Coef.

(n=88)
Z-score

Corn
Coef.

(n=79)
Z-score

Experimental Variables:
Ln Rental Rate ($/acre)
0.96
0.22
-16.38
-1.00
Contract length (yrs)
-0.05
-0.07
1.69
0.64
Current Land Management:
Currently Rents Land (0/1)
-7.75
-1.48
-15.86
-1.18
Current Land Owned:
Total Cropland (acres)
0.010
2.21**
-0.029
-1.05
Total Pasture (acres)
0.96
18.33***
0.61
6***
Total Other Land (acres)
-0.21
-3.28***
0.25
1.35
Total CRP Land (acres)
-0.27
-4.49***
0.41
4.03***
Current Land Uses:
Group of Non-Land Based Uses -0.63
-0.41
3.04
0.33
Group of Hunting Related Uses
10.10
2.14**
-6.78
-0.46
Grazing Livestock (0/1)
4.73
0.55
4.14
0.26
Commercial Income (0/1)
-6.87
-1.08
-0.43
-0.02
Conservation Income (0/1)
9.95
1.7*
29.83
1.06
Environmental Factors:
Renewable Energy
-2.86
-1.15
-0.36
-0.05
General Environmentalism
0.82
0.50
-5.72
-0.90
Concerns Factors:
Agricultural Based
-0.38
-0.21
-0.70
-0.10
Renting Land Based
-7.35
-2.61***
-18.24
-1.39
Demographic Information:
Age (yrs)
0.45
2.19**
0.78
1.36
Male (0/1)
-5.88
-1.15
-21.66
-1.05
Farmer (0/1)
6.31
1.96*
-8.60
-0.70
Income (scale 1-6)
6.39E-05 1.89*
-6.28E-05 -0.57
Constant
-49.19
-1.98**
19.07
0.25
Log Likelihood Values
-5548.03
-6288.57
Wald Chi-Squared
5172.99
2109.32
Probability Chi-Squared
0.00
0.00
Pseudo R-Squared
*** - significant at 1% level, ** - significant at 5% level, * - significant at 10% level

63

Table 21. Truncated Acreage Model for Other Marginal Lands Committed to Prairie and Poplar

Experimental Variables:
Ln Rental Rate ($/acre)
Contract length (yrs)
Current Land Management:
Currently Rents Land (0/1)
Current Land Owned:
Total Cropland (acres)
Total Pasture (acres)
Total Other Land (acres)
Total CRP Land (acres)
Current Land Uses:
Group of Non-Land Based Uses
Group of Hunting Related Uses
Grazing Livestock (0/1)
Commercial Income (0/1)
Conservation Income (0/1)
Environmental Factors:
Renewable Energy
General Environmentalism
Concerns Factors:
Agricultural Based
Renting Land Based
Demographic Information:
Age (yrs)
Male (0/1)
Farmer (0/1)
Income (scale 1-6)
Constant
Log Likelihood Values
Wald Chi-Squared
Probability Chi-Squared

Prairie
Coef.

(n=137)
Z-score

Poplar
Coef.

(n=112)
Z-score

18.16
4.43

1.27
1.22

272.14
25.23

1.33
1.06

32.81

1.77*

-94.24

-0.96

-0.10
0.50
0.47
-0.31

-0.99
2.01**
2.11**
-1.30

-0.019
1.98
0.32
-2.84

-0.29
1.31
1.07
-1.20

-11.08
9.31
19.34
-16.66
49.77

-1.9*
0.79
0.99
-0.99
2.24**

-75.21
-14.11
-94.60
102.49
104.02

-1.24
-0.36
-0.79
0.98
0.94

11.02
-0.26

1.88*
-0.04

-48.17
59.22

-1.08
1.05

-12.05
-29.23

-1.67*
-2.8***

-134.50
-65.85

-1.36
-1.55

-0.36
1.01
0.94
-1.44E-04
-166.65
-9476.71
61.72
0.00

-0.57
0.04
0.06
-1.32
-1.56

6.23
-28.22
77.26
-4.09E-04
-2189.05
-8156.93
48.96
0.00

1.34
-0.28
0.69
-1.24
-1.37

*** - significant at 1% level, ** - significant at 5% level, * - significant at 10% level

64

Table 22. Truncated Acreage Model for Other Marginal Lands Committed to Switchgrass and
Corn
Switch
grass
Coef.

(n=121)
Z-score

Corn
Coef.

(n=102)
Z-score

Experimental Variables:
Ln Rental Rate ($/acre)
-17.29
-0.66
-81.13
-0.72
Contract length (yrs)
6.75
0.83
37.61
1.12
Current Land Management:
Currently Rents Land (0/1)
29.85
0.71
-2.74
-0.02
Current Land Owned:
Total Cropland (acres)
0.034
1.32
-0.029
-0.31
Total Pasture (acres)
0.73
1.46
0.37
0.99
Total Other Land (acres)
0.27
1.86*
1.11
1.98**
Total CRP Land (acres)
0.99
1.21
2.83
0.84
Current Land Uses:
Group of Non-Land Based Uses -26.12
-1.40
112.75
1.39
Group of Hunting Related Uses 25.55
0.81
-254.20
-1.42
Grazing Livestock (0/1)
44.85
0.96
374.13
1.66*
Commercial Income (0/1)
15.99
0.40
-498.85
-1.69*
Conservation Income (0/1)
-11.50
-0.28
482.98
1.76*
Environmental Factors:
Renewable Energy
-22.18
-1.23
20.53
0.32
General Environmentalism
32.95
1.47
70.99
1.12
Concerns Factors:
Agricultural Based
-12.35
-0.77
-147.50
-1.55
Renting Land Based
-41.54
-1.45
-175.72
-1.60
Demographic Information:
Age (yrs)
0.96
0.60
6.83
1.22
Male (0/1)
-28.21
-0.47
271.99
1.04
Farmer (0/1)
49.09
1.19
44.13
0.28
Income (scale 1-6)
-1.50E-07 0.00
-2.32E-05 -0.03
Constant
-218.66
-1.03
-1461.35
-1.36
Log Likelihood Values
-8677.74
-7757.02
Wald Chi-Squared
99.83
12.45
Probability Chi-Squared
0.00
0.90
Pseudo R-Squared
*** - significant at 1% level, ** - significant at 5% level, * - significant at 10% level

65

Unlike the participation model, in the acreage commitment model the variables related
to how the current landowner uses the land were significant; however, these variables mostly
had negative coefficient estimates. This pattern suggests that the more uses an owner has for
the land, the fewer acres they are willing to enroll. This is consistent with the conceptual
model, which found that some amenities to the landowner may disappear or diminish as the
land changes use into bioenergy crops, thus causing the landowner to rent less land.
As in the participation models, so too in the acreage commitment models, landowner
concerns with renting their land for bioenergy crops were significant and reduced the number
of acres they were willing to enroll. This again supports the common idea that the more
concerned an individual is about growing bioenergy crops on their land or getting involved in a
rental contract, the less likely they are to rent out their land and the fewer acres they may be
willing to provide.
One explanatory variable that was notably insignificant in all of the 12 truncated
regressions was the rental rate offered. It appears from the results and from survey pretest
interviews with landowners that while the rental rate offered does affect their decision to rent,
it does not affect the amount they will rent once they have agreed to rent. From our
discussions with landowners, this is most likely because they perceive their land in discrete
parcels, and if they like a rental rate then they are likely to rent out the whole parcel for the
new use and not just a portion of it. This leads us to reject Hypothesis 1b that rental rate has an
effect on the area of land that landowners are willing to commit.

66

How much land is the average landowner willing to rent for cropland compared to
pasture and other marginal lands? The amount of cropland that the average landowner who
owns cropland said they would rent at a typical $100 per acre rental rate was generally over
120 acres, while the average landowner who owns pasture and other marginal lands was willing
to rent a combined total of 90 acres. This result is based upon the fact that the average land
holding of cropland was larger than that of pasture and or of marginal lands. These values and
further individual crop-based results can be seen in Figure 8.
The distribution of land holdings by area owned was highly skewed among survey
respondents. Figure 9 shows a Lorenz style curve displaying the amount of different land types
owned at each percentile. The graph shows that even though survey respondents were
targeted according to the amount of pasture land and other marginal lands they owned, the
entire group of respondents owns in total about an equal area of cropland. This finding
confirms Hypothesis 6 which states that, “Most landowners who own marginal land will own
cropland as well.” The Lorenz style curve also shows that the larger landowners own a high
proportion of the land. In fact the top 10% of cropland landowners own 80% of the potential
cropland, the top 10% of pasture and hay landowners own 70% of the potential pasture and
hay lands, and the top 10% of other marginal land landowners own 50% of the potential other
marginal land.

67

Figure 8. Average Acreage Offered Conditional on Renting Land for Bioenergy Crops at $100 per
Acre (Truncated Model )
Acres of Land an Average Committed Landowner is Willing to Rent at $100
per acre
160

Acres of Land for Commitment

140
120
100
Cropland

80

Pasture and Hay

60

Other Marginal Land
40
20
0
Corn

Switchgrass

Poplar

Prairie

Bioenergy Crop Type

Figure 9. Total Acreage Owned at each Percentile of Land Owners by Land Type
45
40

Thousands of Acres

35
30
25
Cropland
20

Pasture

15

Other

10
5
0
90%

80%

68

70%

60%

50%

40%

30%

20%

10%

Percentile of Landowners

9.3 Scaling up Results to Southern Lower Michigan

The results from the participation model and the acreage commitment model were first
combined to create a description of the average southern Lower Michigan landowner (Figures
10-12). These results show is that the average owner of marginal land, at a typical rental rate of
$100 per acre, is willing to rent out about 20 to 30 acres of cropland and about 30 acres of
marginal land (defined as the combination of pasture and other marginal land from the survey
which includes hay, pasture, scrubland, grassland, idle land, and other farmable non-crop
lands). From the truncated regression results and the Lorenz land curve (Figure 9) it is apparent
that owners of marginal land often own more cropland than marginal land. But on average they
are willing to rent similar amounts of it at the same price. However, the price elasticity of land
supply is much greater for cropland than pasture and other marginal land, meaning that a
change in price affected the supply of cropland much more than the supply of pasture or other
marginal land.

69

Figure 10. Average Landowner Supply of Cropland for Bioenergy Crops (Combined Participation
and Acreage Commitment Models)
Acres of Cropland the Average Landowner is Willing to Rent for:
$300

Rental Rate per Acre

$250
$200
Corn

$150

Switchgrass
Poplar

$100

Prairie
$50
$0

10

20

30

40

50

60

70

80

Acres of Cropland

Figure 11. Average Landowner Supply of Pasture Land for Bioenergy Crops (Combined
Participation and Acreage Commitment Models)

Acres of Pasture Lands the Average Landowner is Willing to Rent for:
$300

Rental Rate per Acre

$250
$200
Corn

$150

Switchgrass
Poplar

$100

Prairie
$50
$0

10

20

30

40

50

Acres of Pasture

70

60

70

80

Figure 12. Average Landowner Supply of Other Marginal Land for Bioenergy Crops (Combined
Participation and Acreage Commitment Models)
Acres of Other Marginal Land the Average Landowner is Willing to Rent
for:
$300

Rental Rate per Acre

$250
$200
Corn

$150

Switchgrass

$100

Prairie

$50
$0

10

20

30

40

50

60

70

80

Acres of Pasture

*Poplar omitted due to insignificance of price response to rental rate.

The supply of land for hybrid poplar other marginal lands is omitted from the graph for
the average landowner (Figure 12). As mentioned above, the rental rate coefficient estimate
was statistically insignificant in both the poplar participation model and the poplar acreage
commitment model. However, the coefficient for rental rate was quite high, over 272 for
hybrid poplar on other marginal lands (Table 16a). The large but insignificant coefficient gave
the misleading impression that renting land for poplar was very desirable at high rental rates.
The results from scaling the average individual landowner up to create a supply function
of land for bioenergy crops in Southern Lower Michigan can be seen in Figure 13 (where
Southern Lower Michigan is the southern half of the lower peninsula of Michigan). This figure
was created, as described in Chapter 7, by estimating the total number of landowners who own

71

over ten acres of marginal land and multiplying that by the average acres committed from the
combined model for pasture and other marginal lands. Figure 13 shows a maximum of around
1.2 million acres being available at very high rental rates of $300 per acre and around 0.8
million acres being available at a typical rental rate of $100 per acre. Given that the estimated
amount of marginal land in Southern Lower Michigan is around three million acres, as shown in
Chapter 5, we can see that at $100 per acre, only about 26% of existing marginal land would
actually be supplied for bioenergy crops. This result is consistent with Hypothesis 3 stating that
many landowners will be unwilling to rent out their land for bioenergy crops. Note that the
scaled up results in Figure 13 omit the land supply for hybrid poplar for the same reason that it
was omitted from the average landowner results on other marginal lands, because poplar does
not show statistically clear price response.
Figure 13. Supply of Marginal Land (Pasture + Other Marginal Lands) for Three Bioenergy Crops
in Southern Lower Peninsula of Michigan
$300

Rental Rate per Acre

$250
$200
Corn

$150

Switchgrass
Prairie

$100
$50
$0.0

0.2

0.4

0.6

0.8

1.0

1.2

Acres of Marginal Land (Millions)

*Poplar omitted due to insignificance of price response to rental rate.
72

1.4

Chapter 10: Conclusion

This thesis contributes to the literature on the potential of growing bioenergy crops on
marginal land by exploring the difference between the amount of marginal land that exists and
the amount that owners would be willing to make available for bioenergy crop production.
Marginal land is defined as rural land not currently in crops that has the potential to produce
bioenergy crops, including grassland, hay, pasture, scrubland, fallow land, and idle land.
Previous studies have shown that using marginal land to grow bioenergy crops instead of
cropland would result in reduced effects on food prices (Searchinger et al. 2008; Rajagopal et al.
2007; Fritsche 2008). Some studies have tried to measure the amount of marginal land that
would potentially be available for energy biomass production (e.g., Gelfand et al. in 2013 for
the Midwestern U.S.A.). However, no studies have yet looked into the willingness of owners of
marginal land to grow bioenergy crops. In this study we identified owners of marginal land in
southern Lower Michigan and through a survey we elicited their willingness to rent their
marginal land for bioenergy production at various rental rates.
We found that owners of marginal land in Michigan were not willing to rent all of their
marginal land, even at very high rental rates. In fact when the responses of individual
landowners were scaled up to cover Southern Lower Michigan we found we could only expect
about 26% of all marginal land to become available at a typical crop rental rate of $100 per
acre. Along with this, we discovered that owners of marginal land often own cropland as well.
73

When asked what land they would be willing to rent for bioenergy crops at specified rental
rates, they were willing to provide cropland and marginal land in similar amounts. These
findings are consistent with the conceptual framework that stated landowners maximize utility,
and the utility they receive in amenities from keeping their land use unchanged can outweigh
the gain in utility they might receive from rental income and the additional consumption that it
makes possible.
These findings point towards a number of difficulties on the road ahead for bioenergy
from marginal land. First, they show that owners of marginal land are willing to make less land
available to grow bioenergy crops than had previously been estimated by studies such as the
Billion Ton Report (U. S. Department of Energy). Second, they show that if a market to grow
bioenergy crops did exist, then landowners would choose on cropland rather than marginal
non-crop land to grow a significant portion of these bioenergy crops. In turn, this would lead to
bioenergy production having an impact on food prices. While this study of landowner
willingness to supply marginal land is limited to southern Lower Michigan, these two general
findings indicate that landowner preferences must be considered in any future estimate of
large scale bioenergy production potential. Failure to do so would result in overestimating the
amount of land available for bioenergy crops and possibly lead to exaggerated expectations for
bioenergy crop production on marginal lands. In fact, landowners will dictate where and how
much energy crops will be grown.

74

APPENDIX

75

Figure 14 Pre-survey Postcard

76

Figure 15 Cover Letter

77

Figure 16 Example Survey

78

Figure 16 (cont’d)

79

Figure 16 (cont’d)

80

Figure 16 (cont’d)

81

Figure 16 (cont’d)

82

Figure 16 (cont’d)

83

Figure 16 (cont’d)

84

Figure 16 (cont’d)

85

Figure 16 (cont’d)

86

Figure 16 (cont’d)

87

Figure 16 (cont’d)

88

Figure 16 (cont’d)

89

Figure 16 (cont’d)

90

Figure 16 (cont’d)

91

Figure 16 (cont’d)

92

Figure 16 (cont’d)

93

Figure 17 Reminder Postcard

94

Figure 17 Second Version of Cover Letter

95

REFERENCES

96

REFERENCES

Barlowe, R. (1986). Land Resource Economics: The Economics of Real Estate. Englewood Cliffs,
NJ: Prentice Hall.
Binkley, C. S. (1981). Timber supply from private nonindustrial forests: a microeconomic
analysis of landowner behavior [USA]. Yale University. School of Forestry and
Environmental Studies. Bulletin.
Campbell, J. E., Lobell, D. B., Genova, R. C., & Field, C. B. (2008). The global potential of
bioenergy on abandoned agriculture lands. Environmental Science & Technology, 42(15):
5791-5794.
Carroll, A., & Somerville, C. (2009). Cellulosic biofuels. Annual Review of Plant Biology, 60: 165182.
Champ, P. A., Boyle, K. J., & Brown, T. C. (Eds.). (2003). A Primer on Nonmarket Valuation (Vol.
3). New York: Springer.
Cotter, J., Davies, C., Nealon, J., & Roberts, R. (1987). Area Frame Design for Agricultural Surveys
(pp. 169-192). New York: Wiley.
Dangerfield Jr, C. W., & Harwell, R. L. (1990). An analysis of a silvopastoral system for the
marginal land in the southeast United States. Agroforestry Systems, 10(3): 187-197.
Dillman, D. A., Smyth, J.D., and Christian, L. M. (2009). Internet, Mail, and Mixed-Mode Surveys:
The Tailored Design Method. New York: Wiley.
Dunlap, Riley E., and Kent D. Van Liere. (2008). "The" New Environmental Paradigm". The
Journal of Environmental Education 40(1): 19-28.
Fritsche R.U. 2008. Impacts of Biofuels on Greenhouse Gas Emissions. FAO Expert Meeting
Proceedings. Food and Agriculture Organization, Rome.
Gallagher, E. (2008). The Gallagher Review of the Indirect Effects of Biofuels Production.
Ashdown House, East Sussex, UK: Renewable Fuels Agency.
Gelfand, I., Sahajpal, R., Zhang, X., Izaurralde, R. C., Gross, K. L., & Robertson, G. P. (2013).
Sustainable bioenergy production from marginal lands in the US Midwest. Nature,
493(7433): 514-517.

97

Goodwin, B. K., Offenbach, L. A., Cable, T. T., & Cook, P. S. (1993). Discrete/continuous
contingent valuation of private hunting access in Kansas. Journal of Environmental
Management, 39(1), 1-12.
James, L. (2010). Theory and Identification of Marginal Land and Factors Determining Land Use
Change. M. S. Plan B Paper, Department of Agricultural, Food, and Resource Economics,
Michigan State University.
Jeschke, M. (2011). Sustainable corn stover harvest for biofuel production. Crop Insights, 12: 1.
Jolejole, M.C., Swinton, S.M., & Lupi, F. (2009, July). Incentives to supply enhanced ecosystem
services from cropland. Selected Paper Presented at the 2009 AAEA & ACCI Joint Annual
Meeting, Milwaukwee, WI.
Keppel, G, & Wickens, T. D (2004). Design and Analysis: A Researcher’s Handbook. New Jersey:
Pearson.
Lal, R. (2005). World crop residues production and implications of its use as a biofuel.
Environment International, 31(4): 575-584.
Newman, C., Henchion, M., & Matthews, A. (2001). Infrequency of purchase and double‐hurdle
models of Irish households' meat expenditure. European Review of Agricultural
Economics, 28(4), 393-419.
Peterson, G. M., & Galbraith, J. K. (1932). The concept of marginal land. Journal of Farm
Economics, 14(2): 295-310.
Piroli, G., Ciaian, P., & Kancs, D. A. (2012). Land use change impacts of biofuels: Near-VAR
evidence from the US. Ecological Economics, 84, 98-109.
Pohlmeier, W., & Ulrich, V. (1995). An econometric model of the two-part decision making
process in the demand for health care. Journal of Human Resources, 30: 339-361.
Rajagopal, D., Sexton, S. E., Roland-Holst, D., & Zilberman, D. (2007). Challenge of biofuel: filling
the tank without emptying the stomach? Environmental Research Letters, 2(4): 044004.
Reiser, B., & Shechter, M. (1999). Incorporating zero values in the economic valuation of
environmental program benefits. Environmetrics, 10(1): 87-101.
Searchinger, T., Heimlich, R., Houghton, R. A., Dong, F., Elobeid, A., Fabiosa, J., Tokgoz, S., & Yu,
T. H. (2008). Use of US croplands for biofuels increases greenhouse gases through
emissions from land-use change. Science, 319(5867): 1238-1240.

98

Sedjo, R. (2011). Carbon Neutrality and Bioenergy: A Zero-Sum Game?. Resources for the
Future Discussion Paper #11-15.
Sheehan, J. J. (2009). Biofuels and the conundrum of sustainability. Current Opinion in
Biotechnology, 20(3): 318-324.
Slade, R., Bauen, A., & Shah, N. (2009). The commercial performance of cellulosic ethanol
supply-chains in Europe. Biotechnol Biofuels, 2(3), 1-20.
Swinton, S. M., Babcock, B. A., James, L. K., & Bandaru, V. (2011). Higher US crop prices trigger
little area expansion so marginal land for biofuel crops is limited. Energy Policy, 39(9):
5254-5258.
Swinton, S. M., Lupi, F., Robertson, G. P., & Hamilton, S. K. (2007). Ecosystem services and
agriculture: cultivating agricultural ecosystems for diverse benefits. Ecological
Economics, 64(2), 245-252.
Tilman, D., Hill, J., & Lehman, C. (2006). Carbon-negative biofuels from low-input high-diversity
grassland biomass. Science, 314(5805): 1598-1600.
Tilman, D., Socolow, R., Foley, J. A., Hill, J., Larson, E., Lynd, L., ... & Williams, R. (2009).
Beneficial biofuels—the food, energy, and environment trilemma. Science, 325(5938),
270.
Timmons, D. (2011). “The Potential Supply of Cellulosic Biomass Crops in Massachusetts.” Ph.D.
dissertation, Department of Resource Economics, University of Massachusetts Amherst.
U.S. Department of Energy. 2011. U.S. Billion-Ton Update: Biomass Supply for a Bioenergy and
Bioproducts Industry. R.D. Perlack and B.J. Stokes (Leads), ORNL/TM-2011/224. Oak
Ridge National Laboratory, Oak Ridge, TN.
Wittenberg, E., & Harsh, S. (2011). 2011 Michigan land values and leasing rates. Michigan State
University Department of Agricultural, Food, and Resource Economics Report No. 642
Yen, S. T., Jensen, H. H., & Wang, O. (1996). Cholesterol information and egg consumption in
the US: A nonnormal and heteroscedastic double-hurdle model. European Review of
Agricultural Economics, 23(3): 343-356.
Yen, S. T., & Jones, A. M. (1997). Household consumption of cheese: an inverse hyperbolic sine
double-hurdle model with dependent errors. American Journal of Agricultural
Economics, 79(1): 246-251.

99

Yu, X., & Abler, D. (2010). Incorporating zero and missing responses into CVM with open-ended
bidding: willingness to pay for blue skies in Beijing. Environment and Development
Economics, 15(5): 535-556.

100