EARLY CONTRACTUAL RELATIONSHIPS IN THE RENEWABLE ENERGY INDUSTRY:
ASSESSING PARTIES’ PREFERENCES FOR TIMELY SUSTAINABLE
GROWTH
By
Guilherme Signorini

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Agricultural, Food and Resource Economics – Doctor of Philosophy
2015

ABSTRACT
EARLY CONTRACTUAL RELATIONSHIPS IN THE RENEWABLE ENERGY INDUSTRY:
ASSESSING PARTIES’ PREFERENCES FOR TIMELY SUSTAINABLE GROWTH
By
Guilherme Signorini
The motivation for this dissertation comes from early attempts to establish long-lasting
relationships within the emerging supply chain for biomass-based electricity in the Midwest
United States. Despite effort exerted by policy makers and government leaders (i.e. federal
mandates and market facilitation), renewable electricity has yet to take off. Failure to perform is
partially related to the inability of trading parties to devise effective mechanisms and incentive
structures to govern biomass transactions in upstream links of the supply chain. In three essays,
the present dissertation identifies deterring factors preventing the biomass-based electricity
industry from developing and formulates mechanisms that are both effective and implementable
from the perspective of supply chain participants.
Essay 1 studies biomass transactions in the Upper Peninsula of Michigan and examines
why logging firms (i.e. biomass producers) and a dedicated renewable energy generator (i.e.
biomass processors) have failed to accommodate private interests under specification contracts.
The Essay borrows a theoretical framework from the incomplete contracts literature to explain that
idiosyncratic assets and unforeseen market conditions led previously devised contracts out of the
self-enforcing range. As a result, a significant potential was created for the energy generator to
hold up logging firms by threatening to reduce its purchases unless logging firms reduced biomass
prices. The problem is diagnosed as a hold-up problem (Klein 1996) and a feasible solution is
presented. The use of credible commitments and collective action among logging firms could have
reduced hold-up potential and led parties to exchange biomass more efficiently.

Combining the case study evidence to other deterring factors previously discussed in the
literature, essay 2 and essay 3 examine preference data from agricultural producers – potential
biomass suppliers – to elaborate mechanisms capable of motivating sustainable relationships.
Essay 2 examines whether hesitation towards energy crops, presumably expressed by agricultural
producers, emerges from stated preferences for crop attributes. Results indicate that producers are
interested in growing crops holding key attributes also found in switchgrass (e.g. lower intensity
of production practices when compared to corn/soybeans rotation systems), and that regional
characteristics and farming capabilities influence agricultural producers’ willingness to convert
farmland into acres of switchgrass.
Essay 3 tackles the transaction problem identified in essay 1. The paper examines whether
agricultural producers distinguish market situations with high hold-up potential from market
situations with low hold-up potential; and whether contracts better equipped to deal with hold-up
problems are preferred when such problems are credible. Results indicate that producers tend to
misinterpret market conditions and prefer acreage-based contracts regardless of the hold-up
potential. There is an indication, however, that propensity to adopt contracts capable of minimizing
the negative effects of hold-up problems increases when these problems constitute credible threats.
In light of the results obtained throughout this dissertation, implications to energy crops
and bioelectricity initiatives are offered to practitioners as they devise entry strategies in the
biomass-based energy industry.

ACKNOWLEDGMENTS

I would first like to thank the guidance committee for the intellectual support during this
research project: Dr. Christopher Peterson, Dr. Brent Ross, Dr. Soren Anderson, and Dr. David
Ortega. Special thanks to Dr. Peterson, my major advisor, who gave me the opportunity to pursue
a doctoral degree in the first place. I hope I was able to meet his expectations over the last five and
one-half years. Thanks must be extended to Dr. Jay Johnson, regional director of the USDA/NASS,
and his staff, Mr. Marty Saffell and Ms. Rachel Bakowski. The second and third Essays of this
dissertation would not have been successfully completed without their engagement and dedication.
I must also thank all Midwestern agricultural producers who have put some thought into our
questionnaires. This research work would have proved unfeasible without their anonymous
participation.
I cannot miss the chance to thank all friends who made my life in Michigan more pleasant
despite of the miserable winters. Each and every one of you will certainly be remembered, causing
unexpected chuckles here and there as good memories come back to mind. A big thank you must
be extended to my parents and family for providing strength and guiding me through perseverance
to pursue this professional achievement. Special thanks goes to my wife, Marília, for her constant
support, and for keeping my life balanced between work and leisure. I am forever grateful for her
unrestricted support and friendship.

iv

TABLE OF CONTENTS

LIST OF TABLES ........................................................................................................................ vii
LIST OF FIGURES ..................................................................................................................... viii
INTRODUCTION .......................................................................................................................... 1
REFERENCES ............................................................................................................................... 6
ESSAY 1: ASSESSING TRANSACTION EFFICIENCY FOR BIOMASS: A CASE STUDY
OF THE UPPER PENINSULA OF MICHIGAN ........................................................................ 11
1. Introduction........................................................................................................................ 11
2. Analytical Methodology and Primary Data ....................................................................... 15
3. Vertical coordination for biomass transactions ................................................................. 16
3.1. Transactions between small-scale private landowners (SPLs) and logging firms.. 20
3.2. Transactions between large-scale commercial landowners (LCLs) and logging
firms ................................................................................................................................ 23
3.3. Transactions between public parks and logging firms ............................................ 25
3.4. Transactions between large-scale commercial landowners (LCLs) and biomass
processors ....................................................................................................................... 27
3.5. Transactions between logging firms and biomass processors ................................ 30
4. Private Enforcement Capital Model (Klein 1996) ............................................................. 33
4.1. The Private Enforcement Capital Model, Applied .................................................. 36
4.2. The Use of a Credible Commitment Mechanism: a necessary but not sufficient
condition ......................................................................................................................... 42
5. Conclusion ......................................................................................................................... 46
REFERENCES ...................................................................................................................... 48
ESSAY 2: ASSESSING FARMERS’ WILLINGNESS TO PRODUCE ENERGY CROPS IN
THE U.S. MIDWEST ................................................................................................................... 52
1. Introduction........................................................................................................................ 52
2. Identifying Regional Categories ........................................................................................ 55
3. Identifying Categories of Farming Operations .................................................................. 60
4. Research Methodology ...................................................................................................... 61
4.1. The Model ................................................................................................................ 67
5. Empirical Results ............................................................................................................... 73
6. Implications for Energy Crops........................................................................................... 82
7. Conclusions........................................................................................................................ 85
APPENDICES ....................................................................................................................... 88
APPENDIX A Sample choice scenario used in the unlabeled experiment ........................... 89
APPENDIX B Summary statistics – aggregate counties and individual county categories.. 90
REFERENCES ...................................................................................................................... 96

v

ESSAY 3: DOES THEORY PREDICT ADOPTION OF PRODUCTION CONTRACTS?
EVIDENCE FROM THE DEVELOPING BIOENERGY INDUSTRY IN U.S. MIDWEST ... 103
1. Introduction...................................................................................................................... 103
2. Theoretical Framework .................................................................................................... 106
2.1. Industrial Organization Approach to Hold-up Problems ..................................... 106
2.2. Institutional Economic Approach to Hold-up Problems ....................................... 109
3. Research Methodology .................................................................................................... 110
3.1. Hypotheses ............................................................................................................. 115
3.2. Survey design, data collection, and data ............................................................... 116
3.3. The Model .............................................................................................................. 121
3.4. Refinements of the Multinomial Logit Model ........................................................ 123
3.5. The Empirical Models ........................................................................................... 126
4. Empirical Results ............................................................................................................. 128
5. Implications for Bioelectricity Initiatives ........................................................................ 141
6. Conclusions...................................................................................................................... 144
APPENDIX.......................................................................................................................... 148
Example of choice scenarios ............................................................................................... 149
REFERENCES .................................................................................................................... 151
CONCLUSION ........................................................................................................................... 156
REFERENCES ........................................................................................................................... 160

vi

LIST OF TABLES

Table 1: Predicted versus adopted governance structure and the underlying Williamson’s
attributes (1996) ............................................................................................................................ 22
Table 2: Summary of data ............................................................................................................. 56
Table 3: Expected Net Margin from corn-soybean rotation by State and by category (2014-2019).
Frequency of county categories per state ...................................................................................... 58
Table 4: Null hypotheses: Willingness-to-grow (WTG) .............................................................. 60
Table 5: Attributes and levels used in the choice experiment ...................................................... 62
Table 6: Demographic statistics .................................................................................................... 66
Table 7: Random Parameter Logit Estimates for Cropping Systems ........................................... 73
Table 8: Scaled Multinomial Logit Estimates for Cropping Systems .......................................... 74
Table 9: Mean Willingness-to-accept (WTA) and 95% confidence intervals .............................. 78
Table 10: Summary statistics - all counties and separate county categories ................................ 90
Table 11: Hypothetical market conditions .................................................................................. 111
Table 12: Attributes, levels and hypotheses tested in the choice experiment ............................. 114
Table 13: Demographic statistics ................................................................................................ 119
Table 14: Parameter estimates from RPL models (main effects vs. two-way interactions) ....... 131
Table 15: Parameter estimates from S-MNL and G-MNL models with two-way interactions .. 133
Table 16: Mean Willingness-to-accept (WTA) and 95% confidence intervals .......................... 135
Table 17: Mean Willingness-to-accept (WTA) and 95% confidence intervals .......................... 138

vii

LIST OF FIGURES

Figure 1: Scheme of the wood value chain in Michigan .............................................................. 18
Figure 2: Probability of hold-up potential .................................................................................... 35
Figure 3: Scenario preceding contract performance ..................................................................... 38
Figure 4: Scenario once contracts started performing .................................................................. 40
Figure 5: Potential scenario if hostages (ω) were ex ante implemented ....................................... 43
Figure 6: County categories per state............................................................................................ 59
Figure 7: Sample choice scenario used in the unlabeled experiment ........................................... 89
Figure 8: Primary hypothesis, illustrated .................................................................................... 116
Figure 9: Example of a choice scenario under information treatment 1 ..................................... 149
Figure 10: Example of a choice scenario under information treatment 2 ................................... 150

viii

INTRODUCTION

Over the past several years countless mass media have released articles about
environmental disasters or human health threats related to production and use of fossil energy.
Such articles shape public opinion, which seems to agree that renewable inputs must partially
substitute for traditional sources of energy worldwide. Government leaders also seem to agree on
the latter statement. Minutes of international meetings (UNFCCC 1998, UN 2002, UN 2012)
indicate broad commitment of nations to promote production and consumption of renewable
energy as well as reducing greenhouse gases (GHG) emissions. To mention a few initiatives,
European Union (EU) countries are committed to reach 20% share of energy from renewable
sources by 2020. In the United States (US), the Renewable Portfolio Standards (RPS) defines
proportion goals of renewable electricity for 30 states with multiple deadlines.
Numerous scientific articles have analyzed the effectiveness of federal policies in reaching
the goals intended by government leaders and desired by public opinion (Berry and Jaccard 2001,
Menz 2005, Palmer and Burtraw 2005, Wiser et al. 2007, Carley 2009, Bird et al. 2011). In the
US, these policy analyses provide great contribution but seldom explain why the federal RPS
mandate or state-level programs reach mixed results (i.e. ability to reduce GHG emissions,
participation of renewable inputs, ability to motivate economic development, job creation).
Discussing whether policy outcomes are aligned with initial intentions is by all means important,
but capturing what specific factors are preventing the bioenergy industry from gaining the expected
momentum takes us to a higher level of comprehension. Therefore, it is safe to assert that the
research agenda in the realm of the developing renewable energy industry in the US is far from
complete.
1

Some scholars have taken the first steps to discuss critical factors associated with the
development of sustainable relationships in bioenergy markets (Roos et al. 1999, Goldemberg
2004, Alexander et al. 2012). Results point out that stable transactions, dedicated transportation
systems, sophisticated information flows, and economies of scale and scope in existing fossil
energy industries constitute barriers to the establishment of renewable energy supply chains. Other
authors add input substitution challenges and entry barriers for feedstock producers in upstream
links of the supply chain (e.g. transactions between suppliers of biomass feedstock and generators
of bioelectricity or advanced biofuels) to the list of establishment barriers (Sherrington et al. 2008).
It is in this context that the first essay of this dissertation emerges. Essay 1 examines early
attempts of logging firms and a bioelectricity generator to trade biomass feedstock under contract
arrangements. The primarily goals are to diagnose whether transaction challenges represent
barriers to development and to propose a theory-based mechanism to overcome those challenges
as the industry grows. Essay 1 uses a case study research methodology (Yin 2009) to analyze
unsuccessful attempts of biomass producers and processors to conduct feedstock transactions in
the Michigan Upper Peninsula. The article argues that logging firms are susceptible to hold-up
problems (Williamson 1975, Goldberg 1976, Hart and Moore 1988) if complex production
contracts or more vertically integrated governance strategies are not put into use; and that credible
commitments (Williamson 1983) might introduce enough complexity to minimize the probability
of hold-up potential and improve contract efficiency.
The literature in place, nevertheless, indicates that agricultural producers – perhaps the
largest group of decision makers who can dedicate assets to energy crops – seem skeptical to
growing energy crops. A recent study conducted in 12 southeastern states indicates that ‘hesitation’
has significant effects on willingness to grow energy crops and on farmland conversion (Qualls et

2

al. 2012). Using a similar methodology, Hayden (2013) finds little interest of Michigan producers
in renting out marginal land for production of energy crops even when the rental rate offered
exceeds market rates for conventional field crops. Together these studies suggest that unsuccessful
experiences lived by some biomass suppliers might have spread doubts and skepticism over
broader populations of potential producers. Skepticism and hesitation might in fact have become
two of the most important barriers to development.
The second and third essays of this dissertation search mechanisms and incentive structures
capable of overcoming these developing barriers. Essay 2 proposes an innovative approach to
examine whether agronomic or technical characteristics of energy crops underlie the hesitation
argument discussed in the literature. The study starts from the working assumption that cropping
systems are primary substitutes, fulfilling similar purposes of agricultural producers (e.g. seeking
profit, maximizing utility). Each cropping system is interpreted as a fixed set of attributes with
varying levels from which agricultural producers derive value. The paper conducts a discrete
choice experiment in which agricultural producers make production decisions and are assumed to
derive utility expectations as in the random utility maximization (RUM) model (McFadden 1974,
1981).
Differently from other studies that attempt to estimate agricultural producers’ willingness
to change land use, Essay 2 refines mean-based assumptions commonly made in simulation
exercises and does not generalize sample specific findings. More specifically, some scholars
simulate production decisions for a ‘representative farm’ and extrapolate results to the entire region
of interest with little consideration for regional characteristics (Scheffran and BenDor 2009,
Bocquého and Jacquet 2010, Sherrington and Moran 2010, Griffith et al. 2012). This paper

3

recognizes that regional particularities as well as producers’ heterogeneity regarding operational
capabilities are important factors affecting production decisions.
Essay 3 complements Essay 2 and assesses whether hesitation towards energy crops is
related to transaction challenges. This study seeks to measure (i) whether agricultural producers
distinguish market conditions with high hold-up potential from those with low hold-up potential,
and (ii) whether preference for contract provisions change as market conditions create more or less
exposure to hold-up problems. A specific objective of Essay 3 is to examine whether contracts
better equipped to deal with hold-up problems are preferred when such problems are credible.
The modern contract theory (Bolton and Dewatripont 2005), nevertheless, shows that the
collective use of credible commitments – as discussed in Essay 1 – is not the only mechanism
available to minimize probability of hold-up potential and consequently enhance efficiency of
supply chain relationships. In addition to private enforcement capital mechanisms (Klein 1996),
parties might draft specific performance contracts as presented in Aghion et al. (1994) or option
contracts as in Nöldeke and Schmidt (1995). The literature in place also contends that the existence
of effective mechanisms to overcome transaction challenges does not guarantee implementation
(Peterson et al. 2001). From the empirical standpoint, the question that remains unanswered is
whether offered solutions to hold-up problems sustain in real-world settings; and if so, which one
is preferred by agricultural producers.
Taking these theory developments as motivation, Essay 3 uses preference data collected
from agricultural producers located in Midwestern states to examine which (if any) theory-based
solution to hold-up problems are likely to prevail. Due to major limitations regarding access to
field data on contractual relationships and strategic alliances (Just and Wu 2009), Essay 3 relies

4

on the random utility maximization (RUM) model (McFadden 1974, 1981) and estimates
agricultural producers decisions from stated choice data.
Essay 3 contributes to the development of bioenergy markets on three fronts. On the
empirical front, it digs deeper into contract provisions and highlights the contract specification that
is more likely to motivate engagement of agricultural producers with energy crops. It also discusses
whether preference for contract provisions varies as market conditions set the state for varying
levels of hold-up potential. On the theoretical front, it provides valuable evidence to whether
theory-driven solutions to hold-up problems suffice in the developing bioenergy industry. On the
outreach front, results of the choice experiment contribute to extension programs of land grant
universities by exposing bounded rational behavior of agricultural producers. With such
information, effectiveness of orientation programs might be enhanced to educate agricultural
producers of how to settle long-lasting relationships within the emerging bioenergy industry even
if market conditions carry high probability of hold-up potential.
This dissertation proceeds as follows. Essay 1 examines early attempts of biomass suppliers
and utilities of generation to engage in contractual partnerships to trade biomass feedstock. Essay
2 analyzes agricultural producers’ preferences for cropping system attributes, ties results to
characteristics of switchgrass, and identifies sub-regions within the Midwest United States where
farmers are more likely to convert farmland into acres of switchgrass. Essay 3 examines producers’
preferences for contract provisions and discusses whether the preferable structures are likely to
sustain long-lasting relationships under market conditions with high hold-up potential or low holdup potential. Finally, a concluding chapter offers closing remarks.

5

REFERENCES

6

REFERENCES

Aghion, Philippe, Mathias Dewatripont, and Patrick Rey. "Renegotiation Design with Unverifiable
Information." Econometrica 62, no. 2 (1994): 257-82.

Alexander, Corinne, Rasto Ivanic, Stephanie Rosch, Wallace Tyner, Steven Y. Wu, and Joshua R.
Yoder. "Contract Theory and Implications for Perennial Energy Crop Contracting." Energy
Economics 34, no. 4 (2012): 970-79.

Berry, Trent, and Mark Jaccard. "The Renewable Portfolio Standard: Design Considerations and
an Implementation Survey." Energy Policy 29, no. 4 (2001): 263-77.

Bird, Lori, Caroline Chapman, Jeff Logan, Jenny Sumner, and Walter Short. "Evaluating
Renewable Portfolio Standards and Carbon Cap Scenarios in the U.S. Electric Sector."
Energy Policy 39, no. 5 (2011): 2573-85.
Bolton, Patrick, and Mathias Dewatripont “Contract Theory” Cambridge, US: The MIT Press,
2005

Bocquého, G., and F. Jacquet. "The Adoption of Switchgrass and Miscanthus by Farmers: Impact
of Liquidity Constraints and Risk Preferences." Energy Policy 38, no. 5 (2010): 2598-607.

Carley, Sanya. "State Renewable Energy Electricity Policies: An Empirical Evaluation of
Effectiveness." Energy Policy 37, no. 8 (2009): 3071-81.

Goldberg, Victor P. "Regulation and Administered Contracts." The Bell Journal of Economics 7,
no. 2 (1976): 426-48.

Goldemberg, José, Suani Teixeira Coelho, Plinio Mário Nastari, and Oswaldo Lucon. "Ethanol
Learning Curve—the Brazilian Experience." Biomass and Bioenergy 26, no. 3 (2004): 30104.

7

Griffith, Andrew P., James A. Larson. "Analysis of Contracting Alternatives for Switchgrass as a
Production Alternative on an East Tennessee Beef and Crop Farm." AgBioForum 15, no.
2 (2012): 206-216.

Hart, Oliver, and John Moore. "Incomplete Contracts and Renegotiation." Econometrica 56, no. 4
(1988): 755-85.

Hayden, Noel J., "Landowner Willingness to Supply Marginal Land for Bioenergy Production: A
Case Study from Michigan." M.S. Thesis, Agricultural Economics, Michigan State
University, 108 pp, 2013.

Just, David R., and Steven Y. Wu. "Experimental Economics and the Economics of Contracts."
American Journal of Agricultural Economics 91, no. 5 (2009): 1382-88.

Klein, Benjamin. "Why Hold-Ups Occur: The Self-Enforcing Range of Contractual
Relationships." Economic Inquiry 34, no. 3 (1996): 444-63.

McFadden, Daniel L. "Conditional Logit Analysis of Qualitative Choice Behavior" In: Frontiers
in Econometrics, Zarembka, P. (ed). Academic Press, 1974; 105-142
———. "Structural Discrete Probability Models from Theories of Choice." In: Structural Analysis
of Discrete Data and Econometric Applications, Manski, C.F., McFadden, D.L. (eds). The
MIT Press, 1981; 198-272.

Menz, Fredric C. "Green Electricity Policies in the United States: Case Study." Energy Policy 33,
no. 18 (2005): 2398-410.

Nöldeke, George, and Klaus M. Schmidt. "Option Contracts and Renegotiation: A Solution to the
Hold-up Problem." RAND Journal of Economics 26, no. 2 (1995): 163-79.

Peterson, H. Christopher, Allen Wysocki, and Stephen B. Harsh. "Strategic Choice Along the
Vertical Coordination Continuum." The International Food and Agribusiness Management
Review 4, no. 2 (2001): 149-66.

Palmer, Karen, and Dallas Burtraw. "Cost-Effectiveness of Renewable Electricity Policies."
Energy Economics 27, no. 6 (2005): 873-94.
8

Qualls, D. Joshua, Kimberly L. Jensen, Christopher D. Clark, Burton C. English, James A. Larson,
and Steven T. Yen. "Analysis of Factors Affecting Willingness to Produce Switchgrass in
the Southeastern United States." Biomass and Bioenergy 39, no. 0 (2012): 159-67.

Roos, Anders, Robin L. Graham, Bo Hektor, and Christian Rakos. "Critical Factors to Bioenergy
Implementation." Biomass and Bioenergy 17, no. 2 (1999): 113-26.
Scheffran, Jurgen, Todd BenDor. 2009. “Bioenergy and Land Use: a Spatial-agent dynamic model
of energy crop production in Illinois”. International Journal of Environment and Pollution
(39) 1: 4-27.

Sherrington, Chris, Justin Bartley, and Dominic Moran. "Farm-Level Constraints on the Domestic
Supply of Perennial Energy Crops in the Uk." Energy Policy 36, no. 7 (2008): 2504-12.

Sherrington, Chris, and Dominic Moran. "Modelling Farmer Uptake of Perennial Energy Crops in
the Uk." Energy Policy 38, no. 7 (2010): 3567-78.

UNFCCC (Ed.), Kyoto Protocol to the United Nations Framework Convention on Climate Change
[internet],
United
Nations,
New
York
(1998).
Available
from:
http://unfccc.int/resource/docs/convkp/kpeng.pdf [cited 2013 Oct 15].

UN (Ed.), Report of the World Summit on Sustainable Development [internet], United Nations,
Johannesburg (2002). Available from: http://daccess-ddsny.un.org/doc/UNDOC/GEN/N02/636/93/PDF/N0263693.pdf?OpenElement [cited 2013
Oct 15].

UN (Ed.), Doha Climate Change Conference: Reports [internet], United Nations, Doha (2012).
Available from:
http://unfccc.int/meetings/doha_nov_2012/meeting/6815/php/view/reports.php [cited
2013 Oct 15].

Williamson, O. E. "Credible Commitments - Using Hostages to Support Exchange." American
Economic Review 73, no. 4 (1983): 519-40.
______. “Markets and Hierarchies” New York, US: Free Press, 1975.
9

Wiser, Ryan, Christopher Namovicz, Mark Gielecki, Robert Smith. "Renewables Portfolio
Standards: A Factual Introduction to Experience from the United States. " Ernest Orland
Lawrence Berkeley National Laboratory (2007): LBNL-62569.
Yin, R.K. “Case Study Research: Design and Methods” Thousand Oaks, CA: Sage, 2009

10

ESSAY 1: ASSESSING TRANSACTION EFFICIENCY FOR BIOMASS: A CASE
STUDY OF THE UPPER PENINSULA OF MICHIGAN

1. Introduction

Renewable energy has been extensively promoted for its potential to contribute to energy
security, stable energy prices, and climate change mitigation (Khanna et al. 2008). In the state of
Michigan such promotion has been pushed through the Renewable Energy Plan (REP1), enacted
in 2008 by the Michigan Public Service Commission (MPSC) in concordance with the federal
Renewable Portfolio Standards (RPS) program. The plan requires every distribution utility in the
state to have 10% of the total supply generated from renewable sources (i.e. solar, wind, hydro,
and biomass) by 2015 (MPSC 2012).
Among other implications the REP has motivated the establishment of new transaction
flows within the wood value chain of Michigan. A great part of this development has taken place
in the Upper Peninsula (UP) where the logging industry is predominant, enjoying synergies
associated with economies of scale and scope as well as complementarity (Milgrom and Roberts
1992) of physical assets and human capital. To put into perspective, logging firms have identified
the emerging biomass market as an opportunity for diversification. As these firms trade regularly
with landowners and public parks to harvest trees, managers at logging firms have decided to
invest in productive assets (i.e. chippers) for collection of forest residue2, a source of energy that
was commonly left in the field before the REP. In 2007, a year before the enactment of REP, there

1

Specific information on this can be found at http://www.michigan.gov/mpsc/0,1607,7-159-16393---,00.html
Forest residue is for instance trees rejected during logging operations (on account of trunk diameter and wood
quality), standing wood decay, tree bark, and branches. Economic agents refer to forest residue as biomass.
2

11

were 23 commercial logging firms in the UP of Michigan. After four intense years of entry and
exit, there were 25 logging firms in operation in 20113 (Manufacturers News Inc., 2013).
The decision of investing in new capacity was not exclusive to logging firms. Based on a
positive interpretation of the REP, biomass processors have also invested in equipment (i.e. boilers,
steam turbines, and generation units) to either reduce production costs (in the case of
manufacturing firms) or profit from agreements with distribution utilities (in the case of
independent power generators). Total capacity of power generation based on biomass increased
from 433.6MW in 2007 to 475.6MW in 2011 – year in which the MPSC released the first report
to the Michigan Senate and House of Representatives4. In the case of the UP of Michigan, installed
capacity increased 40.7% – from 68.8MW in 2007 to 96.8MW in 2011 (EIA, 2013). Although
such capacity growth is considerable, it came from the expansion of a single firm, probably
indicating high barriers to entry due to expensive initial investments at the biomass processor’s
link of the value chain.
Performance figures, however, reveal a market that has not gained the momentum
policymakers were expecting from the REP. Supply of renewable electricity has barely reached
half of the mandatory goal with a little over one year to its deadline. In February 2014 the
Commission reported that 5.4% of the electricity distributed across the state in 2012 was renewable
(Quackenbush et al. 2014). Such slow development seems to derive from transaction problems
within the emerging biomass value chain, which has consequently led to slow growth of biomass
harvesting and processing capacity.

3

During this time period eight firms entered the industry in the UP of Michigan and six firms exited. These figures
capture the commercial loggers but may not account for individual loggers who are sole proprietors and do not have
any permanent employees (Manufacturers News Inc., 2013).
4
One firm entered the industry, four firms expanded, and one firm retrenched operation (EIA, 2013).

12

This paper focuses on biomass transactions in the UP of Michigan, aiming to identify the
bottlenecks that are preventing the biomass-based electricity industry from developing faster in
the state. This article conducts a qualitative analysis using a deductive approach (i) to examine
trading relationships through the lenses of transaction cost economics (TCE), (ii) to interpret the
underlying causes of transaction problems through the private enforcement capital model (Klein
1996), and (iii) to propose plausible alternatives to assist market development.
Our research work begins with an extensive and rather rich qualitative primary data
collected from key agents of the woody biomass value chain (i.e. landowners, logging firms, and
biomass processors) via focus group interviews. The data allows us to sketch the wood value chain
structure in Michigan and identify five important biomass transactions. Interview data for each set
of transactions are then examined using Williamson’s framework (1996) to predict how trading
parties should coordinate relationships in order to devise an efficient market. The paper proceeds
by comparing theory predictions to observed coordination strategies. Data collected in focus group
interviews are used as supporting arguments throughout the analysis.
After identifying a misalignment between theory prediction and reality, Klein’s private
enforcement capital model (1996) is put forward as a refinement of the classic Williamson’s
transaction cost model. The application of Klein’s model takes our discussion beyond the
traditional coordination strategy question and explains why logging firms and biomass processors
face hold-up problems (Williamson 1975, Goldberg 1976, Hart and Moore 1988); an usual type
of transaction problem in emerging markets when parties draft, to the best of their knowledge,
incomplete contracts (Bolton and Dewatripont 2005). The article concludes with a proposed
alternative strategy to alleviate hold-up problems. It is argued that a credible commitment
mechanism (or hostages as defined by Schelling 1956 and Williamson 1983) might lead logging

13

firms and biomass processors towards long-term sustainable partnerships. That would occur,
however, only if managers at logging firms set aside some competitive behavior and engage in
collaborative action.
This study is expected to contribute to the agribusiness literature on three important fronts:
(i) it provides a comprehensive analytical assessment of transactions in the emerging biomass
value chain; (ii) it diagnoses transaction problems based on two well known TCE studies
(Williamson 1996, Klein 1996), task seldom seen in prior work; and (iii) it proposes a mechanism
to alleviate hold-up problems while aligning suggested strategies to predictions of Williamson’s
model (1985; 1996) as well as placing transactions back on Klein’s (1996) self-enforcing range of
agreements.
The reminder of this paper is organized as follows. Next section describes the analytical
methodology and primary data. Section three presents the biomass value chain, examines focus
groups data based on TCE concepts (i.e. uncertainty, asset-specificity, complexity, and frequency),
and compares the adopted coordination strategies to efficient coordination mechanisms as
predicted through Williamson’s framework (1996). Section four presents Klein’s private
enforcement capital model (1996) and its applied version to further explain why specification
contracts expose logging firms and biomass processors to hold-up problems. In section four, a
credible commitment mechanism is presented as an alternative to alleviate hold-up problems.
Section five offers concluding remarks.

14

2. Analytical Methodology and Primary Data

This article uses a case study research methodology (Yin 2009) to identify the facts
influencing coordination decisions in Michigan’s emerging biomass value chain, interpret the
problems associated with these decisions, and propose mechanisms to improve transaction
efficiency. Case study corresponds to a research methodology that provides guidance for rigorous
data collection, presentation and analysis. As Yin (2009) suggests, case study fits best for
qualitative analysis if: (i) research focuses on why or how questions, (ii) research is interested in
contemporary context, and (iii) investigator has no control over the set of events analyzed. Yin
(2009) adds that case studies are appropriate for situations in which multiple sources of evidence
and prior theoretical propositions are considered to guide data collection and analysis.
The methodology used in this article matches all three conditions mentioned above. As
outside observers, we analyze how economic agents engage in transactions and what problems
have caused concerns. Sequentially, TCE theory (Williamson 1985, 1996; Klein 1996) is used to
examine how partnerships can overcome current transaction limitations and minimize probability
of hold-up potential. Following the inductive approach (Corbin and Strauss 1990, Strauss and
Corbin 1990, Peterson 2011) our analysis indicates that credible commitments are powerful
mechanisms to solve some intractable misalignments, if specified correctly and collectively
adopted.
Primary data were collected in two focus group sessions and one set of phone call
interviews: one session with small-scale private landowners (SPLs), one session with large-scale
commercial landowners (LCLs), and three phone call interviews with managers at biomass
processing plants; all conducted in August 2011 (complete questionnaires are available from the
authors upon request). Questionnaires were previously developed and tested with experts (i.e.
15

scholars, educators, and consultants) of the Michigan forestry industry. The main goal of the
interviews was to examine the perceptions of participants about the challenges and opportunities
in trading biomass.

3. Vertical coordination for biomass transactions

Before the enactment of the REP in 2008, harvesting rights for trees was the only service
traded between landowners (SPLs, LCLs, and Public parks) and logging firms in Michigan. Once
harvest was complete logging firms (or whoever maintains ownership over cut trees) would trade
timber with either paper and pulp companies or saw milling firms. Paper and pulp companies
would dedicate clean tree trunks (trunks without the natural bark) to pulp extraction and paper
manufacture. Saw mills would process logs into lumber, which would be manufactured further
into a broad array of outputs (e.g. pallets, boards for construction and furniture, veneer, windows
and doors, heating pellets). In the state of Michigan, firms conducting fine processing are either
independent or vertically integrated with saw milling firms (focus group interview, 2011).
Processed lumber would then be traded with manufacturers of wooden goods (e.g. decorative
panels, furniture, building contractors, power pole producers).
In 2008 the REP incentivized a set of new transactions between logging firms and
landowners, and between logging firms and biomass processors (e.g. paper and pulp companies
and independent power generators). Such transactions started to take place at the margin of existing
log transactions in order to enjoy synergies and complementarity of specific assets. Transactions
between logging firms and landowners, for instance, started to include rights to harvest forest
residue, agreements between logging firms and paper and pulp companies included provisions for
biomass (i.e. quantity, minimum quality, delivery schedule), and relationships between logging
16

firms and independent power generators designed a new flow of feedstock that were unimportant
before 2008.
Figure 1 sketches the structure of the wood value chain in Michigan once biomass
transactions started to perform. Blue solid arrows represent transactions of log-derived
throughputs. Green dashed arrows represent transactions of biomass feedstock. The dark arrow
represents service agreements between LCLs and logging firms where timber and biomass
ownership is maintained by LCLs after harvest. The grey shaded area indicates that saw mills and
fine processing firms might be vertically integrated. Each box includes, to the best of the data
available, the number of operating facilities and the aggregated capacity of value generation as of
2011 (NAICS5, 2011). Secondary transactions of saw milling residue were deliberately omitted.

5

North American Industrial Classification System. A single firm might own or control multiple facilities.

17

Figure 1: Scheme of the wood value chain in Michigan
Biomass Processors
Independent power

Logging firms

LCLs

Resawing & planing
Facilities: 44
Value generation: $ 250,932

Board for Constr. & Furniture
Facilities:
Value generation:

Veneer producers
Saw Mills
Facilities: 138
Value generation:
$547,780

Paper & Pulp companies
Facilities: 21
Value generated: $ 2,576,866

Prefabric. Wood building
Facilities: 27
Value generation:

Other millwork & flooring
Facilities: 65
Value generation: $ 160,470

Window & door

Logging f irms

SPLs

Federal and
state parks

Facilities:
Value generation:

generators
Facilities: 7
Value generation: $98,000

Facilities: 37
Value generation: $ 148,575

Furniture
Facilities: 358
Value generation:

Wood preservation
Facilities: 13
Value generation: $ 96,200

Wood Pallets

Construction
Facilities: 96
Value generation:

Facilities: 97
Value generation: $ 207,265

Landowners

Wood Pellets
Facilities:
Value generation:

Power Poles
Facilities:
Value generation:

Source: Developed by the authors. Dollar figures in thousands.
For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of this
dissertation

18

The emergence of biomass feedstock related transactions – those represented as green
dashed and dark solid arrows in figure 1 – requests research on whether parties have chosen to
govern transactions as TCE theory would predict. For this study, Williamson’s framework (1996)
is employed and begins with a comprehensive assessment of the underlying attributes of
transactions (i.e. asset-specificity, uncertainty, complexity, and frequency).
The concept of asset specificity, the central concept in the TCE literature, has been
understood as the degree at which an asset can be redeployed to alternative uses or by alternative
users without sacrifice of productive value (Peterson et al. 2001). Uncertainty refers to the fact that
economic agents live only by knowing something about the future (Knight 1921); and that the
occurrence of certain events cannot be modeled through the assumption of a probability
distribution. The concept of complexity refers to the difficulty level of programming activities
associated in the transaction (Mahoney 1992). As complexity increases, exchange partners tend to
be more vulnerable to contract breaches and therefore, are likely to incur additional safeguards in
order to protect specific assets. Masten (1984) suggests that the greater the complexity of a
transaction, the greater the likelihood of a party being bound to an inappropriate action of the
counterparty. Frequency, the fourth attribute of transaction modeled by Williamson (1985, 1996),
refers to how often a given transaction happens.
Based on a discrete categorization of these attributes, TCE predicts the coordination
mechanism that is efficient in the sense of minimizing transaction costs as well as safeguarding
specific assets against expropriation of quasi-rents (Klein et al. 1978; Shelanski and Klein 1995,
Williamson 1998). Theory predictions are then compared to the strategies adopted in the market
place.

19

3.1. Transactions between small-scale private landowners (SPLs) and logging firms
The first attribute to be considered is the specificity of productive assets. From the SPLs
side, forestland is the asset dedicated to production of biomass. Focus group interviewees indicate
that land has low specificity because any potential change of activities compromises little
productive value. Interviewees in fact highlight that recreation (i.e. hunting games, vacation
property) tends to be the primary use of SPLs’ land, and that little effort is put into production of
wood and biomass by the owners (focus group interviews, 2011). From the logging firms’
perspective, chippers have low specificity as these assets might be utilized for numerous
transactions. In other words, reallocating chippers from a transaction with SPL A to a transaction
with SPL B incurs little cost to logging firms.
Uncertainty, the second attribute of transaction, is associated with procurement efforts,
price discovery, and log/biomass productivity. As transactions are infrequent for the reasons
discussed below, parties must exert some effort to discover potential partners and stumpage6 prices,
which tends to influence uncertainty of supply. Focus group representatives of logging firms also
indicate that there is a considerable variation of production yield for both logs and biomass in
SPLs’ land. The low effort dedicated to production has been argued as the main factor causing
production variation. Focus group interviewees, nevertheless, mention that there is little
uncertainty of demand for the biomass harvested at SPLs’ land, and that is primarily associated
with the relative small amount of biomass collected. Hence, uncertainty seems to reach moderate
levels for transactions between SPLs and logging firms.
Transactions were identified as of low complexity. Once trading parties agree on service
schedule and payment methods, logging firms have little difficulties to program operational

Stumpage refers to standing trees selected for harvest. Professional “foresters” (consultant agents) are often consulted
by landowners to assist on price discovery and valuation.
6

20

activities. Putting it differently, once parties settle an agreement, harvesting activities become taskprogrammable and can be executed without major concerns to either trading party. Frequency of
transactions between SPLs and logging firms, however, are occasional. According to focus group
responses, SPLs are “unlikely to trade harvest rights more than once or twice in their lifetime”.
Some participants believe that infrequent transactions are closely related to the nature of forest
plants and the size of SPLs’ properties in the UP of Michigan. As trees take many years to mature
and sophisticated management systems are unfeasible on small properties, frequency of
transactions between SPLs and logging firms turns out to be occasional.
Based on Williamson’s framework, it is plausible to predict that a spot market is the most
efficient strategy to govern transactions in which assets are nonspecific, uncertainty reaches
moderate levels, complexity is low, and frequency is occasional. Looking to reality, SPLs and
logging firms seem to adopt the same coordination strategy: “most agreements [between SPLs
and] loggers are one-time formal written agreements although handshake agreements are not
uncommon” (focus group interview 2011). Trading parties, therefore, choose the most efficient
coordination mechanism as predicted through Williamson’s transaction cost economics model
(1985 p.79, 1996). Table 1 summarizes the transaction between SPLs and logging firms and
highlights both predicted and observed governance strategies.

21

Table 1: Predicted versus adopted governance structure and the underlying Williamson’s attributes (1996)
Parties
SPLs
LCLs
F & S parks
LCLs
Logging
companies

AssetSpecificity

Uncertainty

Complexity

Logging
Low
Moderate
companies
Logging
Moderate
Low
companies
Logging
Low
Low/moderate
companies
Biomass
Moderate/high
Low
processors
Biomass
High/low
Low/moderate
processors

22

Low

Frequency

Governance Structure
Predicted
Observed

Occasional Spot Mkt.

Spot Mkt.
Prod. Ctr. or
Spot Mkt.
Spot Mkt.
and Prod. Ctr.

Moderate

Recurrent

Prod. Ctr.

Moderate

Moderate

Spot Mkt.

High

Recurrent

Str. Alliance Prod. Ctr.

High

Moderate

Str. Alliance Prod. Ctr.

3.2. Transactions between large-scale commercial landowners (LCLs) and logging firms
Transactions between LCLs and logging firms are considerably different from transactions
between SPLs and logging firms. Although chippers continue to be categorized as nonspecific
assets for the reasons mentioned above – redeployment to an alternative exchange would cause
little loss of productive value – forestland is now moderately specific to wood and biomass
production. Differently from above, LCLs adopt numerous agricultural practices and exert high
effort to enhance production efficiency. Focus group interviewees agree, therefore, that
reallocating LCLs’ forestland to a different use sacrifices productive value but such move would
not reach prohibitive values. In that sense, LCLs forestland reaches a moderate degree of asset
specificity.
Transactions between logging firms and LCLs are exposed to low levels of uncertainty.
Focus group interviewees suggest that certain characteristics of LCLs’ forestland such as high
productivity and large extensions are factors lowering uncertainty of supply. Productivity of wood
and biomass tends to be high and consistent because of the management practices often employed
in LCLs’ land and absent in SPL’s land. Large production scale also facilitates procurement and
discovery of counterparties. Additionally, both logging firms and LCLs are frequently well
informed about stumpage prices, facilitating negotiation and decreasing uncertainty of supply
(focus group interviews 2011).
Higher efficiency and expertise of LCLs, however, tends to increase complexity of
transactions. Logging firms must follow a set of administrative and operational procedures if wood
and biomass are to be harvested from LCLs’ land. Focus group representatives of logging firms,
however, demonstrate preference in trading with LCLs rather than SPLs even though the
programmability of activities is more complex. “Loggers want to harvest from LCLs forestland in

23

order to decrease uncertainty exposure associated with procurement” (...), even if a set of
operational standards are required (focus group interview 2011).
Finally, property size also affects frequency of transactions. Logging firms and LCLs
conduct recurrent transactions because production plots and operations are managed to allow
frequent cuts. With moderate asset idiosyncrasy, low uncertainty, moderate complexity, and
recurrent transactions Williamson’s framework (1985 p. 79, 1996) plausibly suggests that
production contracts would suffice to minimize transaction costs and safeguard moderately
specific assets. Production contracts are superior because parties must devise enforceable
conditions for exchange in order to coordinate moderately complex transactions and to prevent
expropriation of quasi-rents.
Focus group interviewees indicate that actual transactions are often governed through
production contracts, but spot market relationships are not unusual. Large-scale commercial
landowners tend to maintain ownership over cut trees and biomass regardless of the coordination
mechanism adopted (focus group interviews 2011). The use of spot markets might be explained
by implementability constraints (Peterson et al. 2001) faced by LCLs. Although LCLs understand
that production contracts return higher transaction efficiency, the absence of compatible partners
(due to low operational skills or inability to follow operational standards, for example) prevents
LCLs from using the optimal mechanism for all transactions with logging firms. Table 1
summarizes the underlying characteristics of transactions between LCLs and logging firms as well
as the predicted and observed coordination strategies

24

3.3. Transactions between public parks and logging firms
When stumpage and biomass are traded between public parks and logging firms, asset
specificity returns to low levels. Public land is primarily dedicated to environmental protection
and recreation, so wood and biomass production are not prioritized (focus group interview 2011).
In that sense, land as a productive asset has low specificity because any failure to trading wood
and biomass would have minor impact on the value of public parks. Chippers, the piece of
equipment used by logging firms to cut and gather biomass, are also unspecific as redeployment
to a different harvest job would cause little loss of rents.
Uncertainty reaches low-to-moderate levels for transactions between public parks and
logging firms. Due to the large extension of public parks in the UP of Michigan7, harvest services
are often in need. Discovery of stumpage prices require low effort as transactions are relatively
frequent for agents on either trading side, leading to low uncertainty. Productivity and quality of
stumpage, however, are variable and induce some uncertainty of supply. Once again, these facts
are related to the low use of agricultural practices in public parks given that wood and biomass
production are secondary objectives for this sub-group of landowners. Uncertainty of demand is
of little concern because managers at logging firms are often willing to collect biomass: “the
additional cost of running the chipper is lower than the value generated from selling biomass”
(focus group interview 2011).
Considerable attention is given to operational procedures. When wood harvest is necessary
such service must be performed with great care to avoid anthropic impact on protected areas:
“requests for logging firms to use specific routes and forest tracks are not unusual” (focus group
interview 2011). Harvest activities are programmable but certain requests to minimize

7

State and federal land account for 49% of the total forest land in the UP of Michigan.

25

environmental interference introduce some degree of complexity. Finally, frequency of
transactions is not as recurrent as transactions with LCLs, but tends to be more frequent than
transactions between loggers and SPLs due to the extension of public parks.
Transaction cost economics indicates that spot markets should be used to coordinate
transactions when assets carry little specificity, trades are relatively recurrent, and uncertainty is
low-to-moderate. When transactions are complex, however, theory suggests that greater
involvement should be preferred so parties may monitor ex post behavior and avoid unwanted
outcomes (i.e. excessive environmental disturbance). Interestingly, actual transactions show that
parties devise agreements that match theory predictions. Public parks appoint logging firms
through reverse auctions, a classic example of spot market coordination, but draft performance
contracts in which monitoring provisions are clearly specified.
Transactions unfold as follows: pre-specified contracts are firstly made available to logging
firms; auctions are sequentially held among those who agree on contract terms; the logging firm
with the lowest bid is the auction winner; the winner and park manager formally sign the harvesting
agreement. Logging firms have incentive to behave as determined in contracts because park
managers can monitor behavior and block participation of deviators in future auctions (focus group
interviews 2011). The coordination mechanism adopted by public parks and logging firms is
therefore unlikely to inflict problems to the development of the biomass value chain in Michigan
as TCE prediction and parties’ coordination strategies are aligned. Table 1 summaries transaction
attributes, predicted governance mechanisms and implemented coordination strategies.

26

3.4. Transactions between large-scale commercial landowners (LCLs) and biomass
processors
Specificity of assets takes a different set of arguments when biomass processors are
involved in the transaction. If agreements with large-scale biomass suppliers breach, independent
power generators incur losses associated with two main facts. First, the inability to generate steady
supply of renewable electricity triggers expensive penalties imposed by distribution utilities.
Because the latter group has a tight schedule to distribute clean electricity through the state,
agreements often count with incentive provisions (fines) to motivate optimal behavior (personal
interview 2011). Second, without stable access to biomass – coming primarily from LCLs – the
value of assets used for electricity generation would be drastically expropriated. Power generation
units tend to be meticulously designed to the type of input used: material density, granularity,
moisture content, and temporal availability of biomass are some of the technical factors affecting
specificity. This second fact alone indicates that power generating units are dedicated assets, a type
of investment that would not be undertaken but for the prospect of selling significant amount of
electricity.
Similar reasons would trigger losses at manufacturing firms (e.g. paper and pulp
companies) that have invested in renewable electricity generators. Although these companies do
not usually have downstream partners to trade electricity, and hence do not face potential penalties
in case of short electricity supply, they also require steady generation of power. Their reason is
rather maintain efficient levels of industrial operations and minimize usage of more expensive
inputs (i.e. electricity acquired from distribution utilities).
Once biomass is the adopted input for in-house power generation, managers at
manufacturing plants rely on its supply and optimize operational activities accordingly. But if

27

unstable supply of biomass leads to recurrent power shortages, operations become compromised
as manufacturing equipment lose productive efficiency (personal interview 2011). Managers may
still consider backup plans such as generating coal-based electricity or acquiring electricity from
the local utility. Such plans are, however, more expensive than generating continuous flows of
biomass-based electricity. Moreover, neither plan is easily implementable because switching from
one electricity source to the next involves a precise procedure that may damage manufacturing
equipment if not properly performed (personal interview 2011). In the other end of the transaction,
forestland dedicated to production of wood and biomass in LCLs’ properties continues to be of
moderate specificity for the reasons mentioned above. In that sense, the assets utilized in these
transactions reach moderate-to-high degrees of specificity.
Uncertainty levels are low for biomass transactions between LCLs and processors.
Quantity and quality of biomass tend to be consistent because of high efforts exerted by LCLs.
Agricultural practices, for instance, lead to higher production efficiency and lower uncertainty of
supply when compared to SPLs and public parks. Uncertainty of demand has been identified as a
minor issue by interviewees because of the performance incentives in place (i.e. incentive
provisions in contracts with independent power generators and the argument of industrial
optimization in the case of manufacturing firms).
Continuous generation of power and steady supply of biomass also influence complexity
of transactions between processors and LCLs. Incentive mechanisms are often devised to
guarantee stable supply as independent power generators must meet precise contractual terms with
distribution utilities. Manufacturing companies also attempt to devise these mechanisms to
minimize acquisition of power from distribution grids – often more expensive (focus group
interview 2011). Therefore, harvest activities and transportation schedules must follow specific

28

programs, leading to complex transactions. Representatives of biomass processors indicate that
incentives and monitoring mechanisms are well implemented when counterparties are LCLs.
Economies of scale and better managerial skills when compared to small suppliers of biomass (i.e.
logging firms) are plausible reasons for this argument (remote interviews 2011).
Finally, transactions are frequent between LCLs and biomass processors. That can be
explained by the size of operations (and demanding levels of efficiency) in both ends of the
transaction. With moderate to high degrees of idiosyncrasy, low uncertainty, high levels of
complexity, and recurrent transactions, Williamson’s framework would suggest that efficient
transactions require hybrid strategies (Williamson 1998), such as relation-based alliance. Peterson
et al. (2001) argue that hybrid coordination strategies are in order when ex post control are of high
importance and courts of justice are incapable to verify parties’ behavior, even if provisions are
clearly specified in bilateral contracts. Such argument fits well for transactions between LCLs and
biomass processors. Parties must not only match quality and quantity specifications but also follow
precise schedules in order to maintain steady supply of biomass and generation of electricity.
Courts would be unable to verify performance of parties regarding these provisions, even though
they might be well stated in bilateral contracts. Relation-based alliances would then suffice as an
efficient alternative to minimize negative effects of opportunism.
Actual transactions have been coordinated through specification contracts. Although the
misalignment between theory prediction and observation might give birth to frictions, minimal
problems have been reported by focus group participants and phone call interviewees. The mutual
dependence of parties and the presence of moderate/high idiosyncratic assets in both sides of the
trade might be the reasons leading to acceptable performance of specification contracts; especially
in the UP of Michigan where there are few LCLs and biomass processors (focus group interviews

29

2011). Living up to the agreement let parties enjoy rents from specific assets while behaving
opportunistically would trigger onerous losses to both trading sides. Table 1 reports the underlying
attributes of transactions, the predicted governance mechanism and the strategy actually
implemented by trading parties.

3.5. Transactions between logging firms and biomass processors
Transactions between logging firms and biomass processors are marked by an intense
difference in the degree of asset specificity. Logging firms, from one side, own highly specific
assets (i.e. chippers) for production of renewable electricity input. Specificity here is related to two
main factors. First, chippers constitute a type of investment that would not be undertaken but for
the prospect of selling large volume of biomass (focus group interview 2011). Interviewees agree
on the previous statement especially because alternative uses or alternative users are scarce.
Second, there are only three processors acquiring biomass in the UP of Michigan (one independent
power generator in the Baraga County, one manufacturing firm in the Delta County, and one
manufacturing firm in the Dickinson County), which are relatively far from each other.
The assets controlled by biomass processors in the other end of the transaction have low
degrees of specificity. The underlying reason here is associated with the small amount of biomass
each logging firm carries. Although biomass-based generators are designed to a specific function,
the low volume of biomass hauled by a single firm is unlikely to affect stable generation of
electricity and expose biomass processors to expropriation of rents (e.g. penalties from
downstream partners, lack of industrial efficiency). In other words, biomass processors incur low
costs if reallocation from one logging firm to another takes place. It is worth noting that the latter

30

argument is not valid for relationships between processors and LCLs due to operational size and
mutual dependability, as analyzed above.
The nature of operations at logging firms plays an important role to the level of uncertainty
faced by trading parties. Logging firms often procure upstream partners because long-lasting
relationships are seldom settled (except between LCLs and logging firms). That leads to unstable
transactions and relatively uncertain supply of biomass. Moreover, biomass quality tends to be
variable because of the lack of agricultural practices undertaken in SPLs’ properties and public
parks. Uncertainty of demand has been highlighted as of low importance (personal interview
2011). Not different from above, the REP policy and the interest of manufacturing firms to reduce
use of expensive inputs maintain demand for biomass at high levels. Therefore, uncertainty levels
equilibrate at low-to-moderate levels.
The necessity of processing firms to maintain stable generation of renewable electricity is
the main fact leading managers to request stable supply of biomass. Such stability only happens if
parties arrange a precise delivery schedule of activities (biomass harvest and transportation). The
development of a refined operational plan, however, is of difficult execution because of the nature
of operations at logging firms. As these firms must procure upstream partners and conduct
harvesting operations without establishing long-lasting agreements, the programmability of
activities is considerably compromised. Constrained management capability at logging firms is
also a factor favoring high complexity because it tends to influence programmability of operational
activities (personal interview 2011). Together these factors indicate that transactions are highly
complex. Finally, frequency of transactions between a given pair of logging firm and biomass
processor is moderate. Focus group participants explain that it cannot be more frequent than

31

transactions between LCLs and processing firms but it is certainly more frequent than transactions
between SPLs and logging firms (focus group interview 2011).
With a considerable difference in the degree of asset specificity between trading parties,
low-to-moderate levels of uncertainty, high complexity, and moderately frequent transactions,
Williamson’s framework indicates that a hybrid alliance is necessary to obtain transaction
efficiency. Relation-based alliance is of good use to balance risks and benefits emanating from
different degrees of asset specificity (Peterson et al. 2001). Because logging firms are highly
dependable on trades with biomass processors while the opposite is not the case, relation-based
alliance as a hybrid mode of governance (Williamson 1998) creates room for definition of mutual
objectives and safeguards against opportunism. Alliance is also functional to devise ex post control
and manage uncertainty through a set of programmable tasks.
Actual transactions between logging firms and biomass processors, nevertheless, have used
specification contracts instead. In concordance to theory predictions, focus group participants
report that such coordination mechanism has not performed efficiently. Representatives of logging
firms in particular contend that “contract specification has little meaning (...). A competing logger
comes in, undercuts the agreed price and all of a sudden the processing firm offers a new price at
its gate” (focus group interview 2011). Besides, biomass processors tend to penalize loggers who
decide to take legal action to enforce breached contracts. Penalties take the shape of lock-out
strategies and the plaintiff-logger is no longer accepted in future transactions with the defendantprocessor. Table 1 summaries the underlying attributes of transactions, the TCE prediction, and
the coordination mechanism adopted to govern transactions.
Different from the misalignment analyzed in transactions between biomass processors and
LCLs, the misalignment here culminates in hold-up problems as discussed in-depth below. It is

32

plausible to argue that the poor performance of relationships between logging firms and biomass
processors is an important factor preventing the renewable electricity industry to develop faster in
Michigan. Commentators yet suggest that the unsuccessful attempt of logging firms to supply
biomass might have generated skepticism among potential small-scale suppliers located elsewhere
in the state.
With a good understanding of where in the woody biomass value chain transactions have
struggled to perform, next section introduces a model that pushes discussion further. The private
enforcement capital model (Klein 1996) is employed to precisely explain why specification
contract is not sufficient to govern sustainable transactions between logging firms and biomass
processors. A theory-based mechanism is sequentially proposed to alleviate the existing
transaction problem.

4. Private Enforcement Capital Model (Klein 1996)

The private enforcement capital model is founded on prior substantial contributions. To
mention, the literature of hold-up problems was open for discussion when Benjamin Klein (1996)
introduced the concept of ‘self-enforcing range’ of contractual agreements and developed a
framework for minimizing the probability of hold-up potential. In this sense, two concepts and a
conditioning factor created foundations for the model of interest: (i) contract incompleteness, (ii)
asset-specificity, and (iii) changes in market conditions. For the purpose of context, the concepts
and conditioning factor are reviewed below.
Theorists and applied economists have long ago agreed that contracts are at best incomplete
(Bolton and Dewatripont 2005) for two behavioral reasons. First, a contractual relationship is the
result of interactions among people, and people’s minds are constrained to their ability to access
33

all relevant information promptly. Even if we humans were fully capable of doing so, the cost of
accessing information and mapping out all potential outcomes and related states of nature would
reach uneconomical marks. In other words, contracts are often conditional to ‘bounded rationality’
(Simon 1978). Second, a given party to trade needs not share strategic information (e.g. costs of
production, procurement strategies, and hedging strategies) with its counterparty at the moment of
contract negotiation. This latter reason defines the concept of ‘asymmetric information’ which
emerges when one side is better informed about a transaction than the other (Arrow 1963).
Together the concepts of bounded rationality and information asymmetry force any contract to be
incomplete at its best.
The concept of asset specificity, as applied above, has been understood as the degree at
which an asset can be redeployed to alternative uses and by alternative users without sacrifice of
productive value (Peterson et al. 2001). Different types of specific-assets have been identified
elsewhere (e.g. physical specific asset, dedicated asset, site specific asset) but such refinement of
the TCE literature is not necessary for the purposes of Klein’s model (1996). Finally, it is well
accepted that the occurrence of contractual hold-ups is conditional to changes in market conditions.
When changes in market conditions (e.g. unexpected decrease of demand, input scarceness) take
place, a party may have incentive to hold up its counterparty in order to expropriate quasi-rents. It
is worth mentioning that such change must not be ex ante contracted and ex post verifiable. That
is, changes in market conditions only make a hold-up threat credible if the contract does not
explicitly specify the contingency under question. If there are contracted provisions for a given
market condition and the state of nature can be verified by courts of justice, the deviator is then
subject to punishment. Thus, hold-up is unlikely to happen.

34

Taking these past contributions into consideration, the private enforcement capital model
can then be defined as follows. Let two firms, A and B, be engaged in a transaction in which
specification contract is the coordination mechanism. Also, assume that both firms own productive
assets with equal or different degrees of specificity. Under the plausible assumption of incomplete
contracts, allow decision makers at firms A and B to have expectations about future market
conditions and associated gains or losses from hold-ups. For simplicity, assume that ex post change
in market conditions is represented by a change in aggregate demand of an arbitrary good. Figure
2 illustrates a hypothetical probability distribution of the hold-up potential, 𝑓(𝐻𝑖 (𝐷)) where i = A
and B, and hold-up potential is a function of observable ex post demand. Moreover, let 𝐾𝑖 be a
threshold and for any |𝑘| > |𝐾𝑖 |, firm i will hold up its counterparty.

Figure 2: Probability of hold-up potential

Source: Adapted from Klein (1996).
Intuitively, |𝐾𝑖 | is the limiting market condition at which the contractual relationship
sustains. It represents the zero net value between gains and losses of breaching the contract. More
specifically, gains are returns a firm may have by holding up its counterparty; and losses include
35

not only the loss of future rents but also the reputation loss in the marketplace. Hence, for any
market condition 𝑘′ falling between 𝐾𝐴 and 𝐾𝐵 (i.e. self-enforcing range), it is more costly than
beneficial for either party to breach the contract so parties have incentive to live up to the
agreement. For any market condition 𝑘" falling in the outer region (𝑘" < 𝐾𝐴 or 𝑘" > 𝐾𝐵 ), firm A
has incentive to hold up B or firm B has incentive to hold up A, respectively. That is, there are
more gains than losses in breaching the contract and holding up the counterparty. Therefore, the
𝐾

probability of firm A hold up B is ∫−∞𝐴 𝑓(𝐻𝐴 (𝐷))𝑑𝐷. Similarly, the probability of firm B hold up
∞

A is ∫𝐾 𝑓(𝐻𝐵 (𝐷))𝑑𝐷.
𝐵

In the lines of the model, the probability of hold-up can be changed by shifting 𝐾𝐴 and 𝐾𝐵 .
That can be done by either increasing gains or decreasing losses at every transaction period.
Numerous studies have discussed ways to shift thresholds 𝐾𝐴 and 𝐾𝐵 (Williamson 1983, Klein
1996, Hueth et al. 1999, Gow and Swinnen 2001, Shanoyan 2011). Next section explains why the
use of credible commitments (Williamson 1983) could be used to minimize the probability of holdup problems and, along with collective action, might increase transaction efficiency for biomass
trading between loggers and processors in the UP of Michigan.

4.1. The Private Enforcement Capital Model, Applied
It is worthwhile to identify in the case the concepts of ‘asset-specificity’ and ‘contract
incompleteness’ and the conditioning factor (i.e. change in market conditions). Once these pieces
are recognized, it can be argued that logging firms are likely to hold up biomass processors if a
given set of market conditions take place. The same argument can be used for biomass processors:
managers at biomass processing firms are likely to hold up loggers if another set of market
conditions emerge.
36

Logging firms and biomass processors have decided to invest in new productive assets (i.e.
chippers and power generating equipment, respectively) as they expected that demand for
renewable electricity would sharply increase once the REP was enacted in 2008. The expectation
over demand for renewable electricity was so optimistic that new logging firms invested in
chippers and entered the market seeking positive rents through sales of biomass feedstock 8.
Although entry did not happen in the processing side due to impeditive investment costs, a large
expansion project took place in the UP of Michigan. More interestingly, parties invested in new
assets without taking under consideration potential collaboration regarding supply chain issues;
perhaps because of distrust already in place.
The intensive entry in the logging side and the expansion in the processing side ended up
altering the initial degrees of asset-specificity and unbalancing probabilities of hold-ups.
Investments made by biomass processors became less specific as new logging firms entered in the
opposite side of business. Managers at processing firms could then easily redeploy a given contract
from one logger to another without sacrificing the productive value of electricity generators.
Conversely, logging firms that had also invested in specific physical assets became dedicated firms
with few alternative uses for their chippers. As a result, investments undertaken by logging firms
were not only specific to a certain operational activity but also became dedicated to a reduced
number of biomass processors9.
With investments in place, loggers and biomass processors drafted specification contracts
for trading biomass. Focus group responses allow us to plausibly argue that contracts were drafted

8

Exit started to occur when logging firms experienced opportunism from counterparties (focus group interviews
2011). In total, eight firms entered business between 2008 and 2011 and six firms exited during the same period
(Manufacturers News Inc., 2013).
9
As of 2011, there were three biomass processors in the UP of Michigan – two manufacturing firms and one
independent power generator (EIA, 2013).

37

without either party having a good understanding of demand prospects for biomass-based
electricity (focus group interview 2011). Contractual agreements were settled based on optimistic
expectations of demand and without specifying how parties would behave (or how contracts should
perform) if negative scenarios emerged. In light of the private enforcement capital model, trading
parties ended up engaging in (incomplete) contracts in which incompleteness is represented by the
absence of specific terms regarding demand for biomass-based electricity.
Figure 3 illustrates the situation in which loggers (L) and biomass processors (B) placed
themselves. Aggregate demand (D) for renewable electricity is in the x-axis and probability of
hold-up potential, 𝑓(𝐻𝑖 (𝐷)), is in the y-axis. Ex ante expected demand for renewable electricity
is represented by 𝑘 0 .

Figure 3: Scenario preceding contract performance

Figure 3 shows that parties set the arena for potential hold-up problems. Logging firms
owned specific assets, which were exposed to expropriation of quasi-rents because the agreed
contract carried little information of how parties should perform if future demand did not follow

38

optimistic expectations. Therefore, 𝐾𝐵 represents the minimum aggregate demand for renewable
electricity that prevents a biomass processor from behaving opportunistically. Similarly, 𝐾𝐿
represents the maximum demand for renewable electricity that maintains a logger loyal to the
agreed contract.
The self-enforcing range (i.e. area below the probability density function and between 𝐾𝐵
and 𝐾𝐿 ) is relatively small as neither party requested additional specifications nor used alternative
mechanisms to enlarge or shift that range. Additionally, the probability of loggers to push hold-up
threats forward was relatively small when compared to the probability of threats that biomass
processors could impose over loggers. The first and main reason is related to the intensive entry in
the logging sector and limited entry in the processing sector. Logging firms were highly exposed
to expropriation of quasi-rents due to increased degree of asset-specificity of their investments.
Conversely, biomass processors were less exposed to hold-up threats because contracts for
biomass could be redeployed amongst numerous logging firms. Three additional reasons justify
the relatively small capability of logging firms to hold up biomass processors: (i) constrained
capacity of production; (ii) managerial independence; and (iii) lack of collective action.
The first reason suggests that the capacity of producing forest residue at a given logging
firm falls far below the quantity demanded by any average-sized biomass-based electricity
generator, which suggests that a logging firm alone has low bargaining power and low probability
to hold up its counterparty. This is true unless the demand for renewable electricity downstream
in the supply chain was exceptionally high and managers could not afford giving up any agreement
of feedstock supply. The second and third reasons imply that numerous logging firms compete to
supply biomass to a limited number of processors. As stated by a focus group participant, “logging
firms are autonomous and seldom coordinate actions collectively” (focus group interview 2011).

39

Once contracts of supply started performing, parties realized that the real demand for
renewable electricity was not as much as they expected (𝑘 ′ < 𝑘 0 ). In fact, such realization led
participants to suggest that “(loggers) spent a lot of money to gear up for something that never
showed up” (focus group interview 2011). ‘Something that never showed up’ refers to the
optimistic demand that loggers were expecting. This aspect of the case can be analyzed as the
change in market condition; the conditioning factor that turns hold-ups into credible threats. Figure
4 represents the real demand (𝑘′) such that 𝑘′ < 𝐾𝐵 . As mentioned above, a biomass processor
holds up a logger because: (i) transactions have been conducted through incomplete contracts; (ii)
the logger owns assets with high degree of idiosyncrasy; and (iii) the real demand for biomassbased electricity is below the 𝐾𝐵 threshold.

Figure 4: Scenario once contracts started performing

Under this new scenario, commentators ask why trading parties were so optimistic about
demand of renewable electricity in the first place; or what loggers failed to consider that could
have otherwise led them to better estimates of demand. As mentioned by focus group interviewees,

40

the REP and the existence of subsidies for renewable electricity motivated existing loggers (and
entrant firms) to settle bilateral contracts without considering the use of alternative enforcing
mechanisms to limit the probability of opportunistic behavior of biomass processors. In addition,
decision makers at logging firms might have failed to observe the existence of substitute
technologies that fulfill REP standards and are also eligible to receive subsidies. Windmills, for
instance, have demonstrated to be more cost-effective than biomass-based generators in Michigan.
The Michigan Public Service Commission (MPSC) reports that average contract price for windbased electricity is $80.32/MWh against $98.94/MWh for biomass-based electricity
(Quackenbush et al. 2013 p.26). Although, these prices are not conclusive about generation costs,
they are strong indicators of cost effectiveness.
These supporting numbers let us reasonably argue that distribution utilities are likely to
start procuring the most cost-effective technology, leaving the residual demand to be supplied by
less cost-effective technologies. Decision makers at logging firms should therefore have
considered the existence of substitute technologies and have drafted sale contracts based on the
expected residual demand (i.e. the expected demand for biomass-based electricity) rather than the
expected aggregate demand for renewable electricity.
An interesting conjecture is that if biomass-based electricity was the most cost-effective
technology available in Michigan, and thereby real demand was just as high as the optimistic
expectation of loggers; biomass processors would have honored contracts previously specified. A
theoretical justification behind this argument is that the demand for biomass-based electricity
would fall into the self-enforcing range and biomass processors would have little incentive to hold
up loggers. Still, loggers would be likely to threaten biomass processors if demand for biomassbased electricity overcame the upper threshold 𝐾𝐿 (i.e. 𝑘" > 𝐾𝐿 ) in picture 4. In this latter scenario

41

biomass processors would, in turn, submit to the credible threat in order to maintain stable
generation of renewable power and meet contractual specifications defined downstream with
distribution utilities.

4.2. The Use of a Credible Commitment Mechanism: a necessary but not sufficient
condition
Despite the fact that managers at logging firms could have decreased the probability of
hold-up problems to occur by noticing the existence of more cost-effective technologies, and
therefore having set their entry strategies based on the residual demand for biomass-based
electricity; the expose to opportunism could be decreased if logging firms had requested credible
commitments from biomass processors. Williamson (1983) refers to these commitments as
‘hostages’ because they force biomass processors to behave as agreed for the duration of the
contract. In his words, credible commitments create reciprocity and serve to equalize the exposure
of the parties, thereby reducing the incentive of the buyer to defect from the exchange (Williamson
1983 p. 531). The imposition of a hostage (𝜔) over biomass processors would shift thresholds 𝐾𝐵
and 𝐾𝐿 , and consequently re-set the self-enforcing range as follows in figure 5.

42

Figure 5: Potential scenario if hostages (ω) were ex ante implemented

With implementation of a credible commitment mechanism, a given biomass processor
would have lower incentive to hold up the logger because hostages decrease his gains from doing
so. Likewise, a given logger has less to lose if the biomass processor unfairly holds her up. Kreps
and Wilson (1982) and Williamson (1983, 1998) suggest that the use of hostages enhance
contractual efficiency by introducing mutual dependence between logger and biomass processor.
In the case under analysis, hostages might represent upfront payments from biomass processors to
logging firms; or even partial acquisition of the specific machinery used for collecting biomass.
Klein (1996) suggests that the ideal size of hostages occurs when the probability of hold-up
potential is equal for the logging firm and biomass processor.
However, even if a single strategist-logger had tried to request credible commitments from
the biomass processor, the problem of hold-up potential would not have been entirely managed.
The processor related in the transaction would have had little incentive to agree on paying a hostage
to the strategist-logger knowing that other, less cautious, loggers did not use the same strategy.

43

It turns out that the biomass processor could still behave opportunistically (i.e. undercut
feedstock prices) given that loggers play under almost perfect competitive conditions. A likely
equilibrium for this strategy game is as follows: the biomass processor would have rejected the
contract provision requested by the strategist-logger; instead, the processor would have signed
contracts of supply with less cautious loggers who did not request credible commitments; the
strategist-logger would not have entered into business; and the biomass processor would still
expropriate quasi-rents from less cautious loggers by holding up their contracts. Therefore, the
strategy of requesting credible commitments can be analyzed as a necessary but not sufficient
condition to improve overall transaction efficiency in this emerging industry.
A sufficient condition would only obtain if loggers behaved collectively to adopt the
strategy discussed above (i.e. request payment of hostages from the biomass processor). If loggers
meet the definition of agricultural producers under the Capper-Volstead Act (1922), then they
would be allowed to organize themselves as a cooperative, a type of hybrid mode of governance
(Williamson 1998). Loggers would need to give up some managerial independence in order to
share a common contract term regarding payment of hostages. If such collective action was in
place before individual contracts were signed between loggers and biomass processors, a given
processor would have little alternative but accept the request imposed by the cooperative.
Acquiring feedstock outside the cooperative – perhaps from new entrants – would not be optimal
because the cooperative might punish the processor by expropriating hostages (e.g. gain full
ownership over the chippers). Cooperative members would also have little incentive to trade
directly with biomass processor, especially because biomass feedstock is a homogeneous input.
It is worth to mention that collective action alone would be insufficient for improving
contractual efficiency if loggers did not agree upon a solid hostage provision. That is, if loggers

44

had formed a cooperative with the sole objective of defining a common sales price, biomass
processors would still have incentive to acquire biomass from outsiders at lower prices, and
consequently, hold up cooperative members. This observation rises against a remark (usually
addressed by the classic industrial organization school of thought) that improving bargaining
power implies better contractual outcome.
A similar argument fits if transactions between biomass processors and LCLs are taken as
example. Although parties have traded biomass using specification contract (in which credible
commitment provisions are omitted) it does not guarantee that they will continue to perform until
expiration. Transactions are performing well merely because of the mutual dependence generated
from moderate-to-high degrees of asset-specificity. However, if biomass processors could identify
other large-scale suppliers capable of delivering a stable flow of biomass hold-up problems or
other types of credible treats could still occur. Negative effects of opportunistic behavior would be
minimized only if LCLs and biomass processors agreed on a credible commitment clause. Thus,
collective action or large productive scale would imply higher transaction efficiency only if the
correct set of strategies was agreed in the first place.
In the case of loggers, higher transaction efficiency would have obtained if loggers had not
only organized themselves as a cooperative but also requested payment of hostages from the
counterparty. This result meets the criticism once expressed by Goldberg and Erickson (1987
p.371): propositions [invoking market power] have impeded research on complex exchanges by
providing superficially appealing answers that have discouraged search along more fruitful lines.

45

5. Conclusion

This paper conducts a qualitative analysis of biomass transactions in the emerging
renewable electricity industry in Michigan. Through the lenses of TCE (Williamson 1985) and
based on focus group/personal interviews, the four underlying attributes of transactions (i.e. assetspecificity, uncertainty, complexity and frequency) are examined to predict the most efficient
coordination strategy. Predictions are then compared to actual coordination mechanisms for five
biomass-related transactions. Our analysis indicates that while four sets of transactions seem to
perform efficiently, transactions between logging firms and biomass processors do not. These
trading parties face hold-up problems because specification contracts have been adopted to govern
biomass transactions while TCE predicts that hybrid strategies (i.e. relation-based alliance or
equity-based alliance) should be used instead.
We sequentially utilize Klein’s private enforcement capital model (1996) to examine why
specification contracts have exposed logging firms to hold-up problems. The generalized model is
adjusted to represent multiple contractual relations between logging firms and biomass processors.
We identify for the biomass transaction under question the concepts of idiosyncratic assets and
incomplete contracts, two fundamental factors for the occurrence of hold-ups. The mismatch
between expected aggregate demand for renewable electricity and real demand for biomass-based
electricity is interpreted as the conditioning factor trigging hold-up problems.
It is then argued that the use of a credible commitment mechanism as defined in the
literature (Williamson 1983) is a necessary but not sufficient condition to minimize probability of
hold-up potential, and thereby improve contractual efficiency. The sufficient condition would only
take place if such strategy was taken as a collective action – plausibly under a strategic alliance as

46

allowed by the Capper-Volstead Act – by loggers. If that scenario emerged, biomass processors
would have little gain from offering incentives to suppliers outside the alliance; alliance members
would not be willing to trade directly with the biomass processor; and new entrants would have
undercut prices rejected by the processor due to credible commitments made with the alliance. For
that to happen, however, loggers would need to set aside some competitive behavior and engage
in a collaborative action in order to improve transaction efficiency.
The case study of biomass trading in the UP of Michigan has also provided evidence to
support the usefulness of the private enforcement capital model (Klein 1996). Although
Williamson’s framework suffices to indicate that logging firms and biomass processors should
adopt a strategic alliance rather than specification contracts to govern transactions, it provides little
guidance of how to implement the most efficient strategy. The private enforcement capital model
(Klein 1996) complements Williamson’s framework in the sense of creating grounds for
implementing the new strategy. As one would expect, the analyses conducted through either model
indicates that hybrid coordination strategy is superior to specification contract. Results derived
from Klein’s model nevertheless push discussion further and indicate that a sufficient solution for
long-lasting relationships also requires credible commitments from the biomass processor.
Although the analyses conducted in this article do not intend to solve hold-up problems
faced by logging firms, we hope it adequately explains how parties should conduct ex ante
negotiations in order to devise stable relationships and efficient markets. Structured assessment of
transaction attributes and holistic understanding of value chain relationships are essential for the
development of solid coordination strategies. That is especially true when practitioners examine
opportunities in emerging markets as it is demonstrated to be the case of the renewable energy
industry in Michigan.

47

REFERENCES

48

REFERENCES

Arrow, Kenneth J. "Uncertainty and the Welfare Economics of Medical Care." American
Economic Review 53, no. 5 (1963): 941-973.

Bolton, Patrick, and Mathias Dewatripont. Contract Theory. Cambridge, MA. MIT Press, 2005.

Corbin, Juliet and Anselm Strauss "Grounded Theory Research: Procedures, Canons, and
Evaluative Criteria." Qualitative Sociology 13, no.1 (1990): 3-21.

Goldberg, Victor and John Erickson. "Quantity and Price Adjustment in Long-Term Contracts: A
Case Study of Petroleum Coke." Journal of Law and Economics 30, no. 2 (1987): 369-398.

Goldberg, Victor P. "Regulation and Administered Contracts." The Bell Journal of Economics 7,
no. 2 (1976): 426-48.

Gow, H.R. and J.F.M. Swinnen. "Private Enforcement Capital and Contract Enforcement in
Transition Economies." American Journal of Agricultural Economics 83, no. 3 (2001):
686-690.

Hart, Oliver, and John Moore. "Incomplete Contracts and Renegotiation." Econometrica 56, no. 4
(1988): 755-85.

Hueth, B., E. Ligon, S. Wolf and S. Wu. "Incentive Instruments in Fruit and Vegetable Contracts:
Input Control, Monitoring, Measuring, and Price Risk." Review of Agricultural Economics
21, no. 2 (1999): 374-389.

Khanna, Madhu. "Meeting the Demand for Biofuels: Implications for Land Use and Green House
Gas Emissions." Paper presented at MSU Agricultural, Food, and Resource Economics,
Department Seminar, East Lansing, 2009.

Klein, Benjamin, Robert G. Crawford, and Armen A. Alchian. "Vertical Integration, Appropriable
Rents, and the Competitive Contracting Process." Journal of Law and Economics 21, no.
2 (1978): 297-326.
49

Klein, Benjamin. "Why Hold-Ups Occur: The Self-Enforcing Range of Contractual
Relationships." Economic Inquiry 34, no. 3 (1996): 444-463.

Knight, Frank H. "Risk, Uncertainty, and Profit." Library of Economics and Liberty, 1921.
Retrieved Jun 11, 2011 from the World Wide Web: http://www.econlib.org/library/Knight
/knRUP.html

Kreps, David M. and Robert Wilson. "Reputation and Imperfect Information." Journal of
Economic Theory 27, no. 2 (1982): 253-279.

Masten, Scott E. "The Organization of Production: Evidence from the Aerospace Industry."
Journal of Law and Economics 27, no. 2 (1984): 403-17

Mahoney, Joseph T. "The Choice of Organizational Form - Vertical Financial Ownership Versus
Other Methods of Vertical Integration." Strategic Management Journal 13, no. 8 (1992):
559-84.

Peterson, H. Christopher. "An Epistemology for Agribusiness: Peers, Methods and Engagement
in the Agri-Food-Bio System." International Food and Agribusiness Management Review
14, no. 5 (2011): 11-26.

Peterson, H. Christopher, Allan Wysocki, and Steve B. Harsh. "Strategic Choice Along the
Vertical Coordination Continuum." The International Food and Agribusiness Management
Review 4, no. 2 (2001): 149-166.

Quackenbush, John D., Greg R. White and Sally A. Talberg. "Report on the Implementation of
the P.A. 295 Renewable Energy Standard and the Cost-Effectiveness of the Energy
Standards." Michigan Public Service Commission, Department of Licensing and
Regulatory Affairs, February 14, 2014
http://www.michigan.gov/documents/mpsc/pa295report_447680_7.pdf [accessed March
4, 2014].

Schelling, Thomas C. "An Essay on Bargaining." Economic Review 46 (1956): 281-306.

50

Shanoyan, Aleksan. "External Facilitation of Sustainable Market Linkages in the Agri-Food
Supply Chain: Evidence from the USDA Marketing Assistance Program in the Armenian
Dairy and Vegetable Industries." Ph.D. diss., Michigan State University, East Lansing,
2011.

Shelanski, Howard A., and Peter G. Klein. "Empirical Research in Transaction Cost Economics:
A Review and Assessment." Journal of Law, Economics, & Organization 11, no. 2 (1995):
335-61.

Simon, Herbert A. "Rationality as Process and as Product of Thought." The American Economic
Review 68, no. 2 (1978): 1-16.

Strauss, Anselm and Juliet Corbin. Basics of Qualitative Research: Grounded Theory Procedures
and Techniques. Newbury Park, CA. Sage, 1990.

Williamson, Oliver E. "The Institutions of Governance." The American Economic Review 88, no.
2 (1998): 75-79.
________. “The Mechanisms of Governance.” New York, US: Oxford University Press, 1996
________. “The Economic Institutions of Capitalism: Firms, Markets, Relational Contracting.”
New York, US: The Free Press, 1985

________. "Credible Commitments: Using Hostages to Support Exchange." American Economic
Review 73, no. 4 (1983): 519-540.

________. "Markets and Hierarchies." New York, US: The Free Press, 1975
Wooldridge, Jeffrey M. Econometric analysis of cross section and panel data, 2nd ed. Cambridge,
MA. MIT Press, 2002.
Yin, Robert K. Case Study Research: Design and Methods, 4th ed. Thousand Oaks, CA. Sage,
2009.

51

ESSAY 2: ASSESSING FARMERS’ WILLINGNESS TO PRODUCE ENERGY CROPS
IN THE U.S. MIDWEST

1. Introduction

Governments have exerted effort to develop renewable energy industries in several
countries around the globe. The public stimuli often take shape in mandates, as in the European
Union (EU) and in the United States (US). In the EU, mandatory national targets consisting of
20% share of energy from renewable sources were implemented and are due by 2020. In the US,
the Renewable Portfolio Standards (RPS) regulation – perhaps the most important measure taken
in the country to promote clean energy – defines proportion goals of renewable electricity for 30
states with multiple deadlines.
As policy unfolds academics conduct research to assess its efficiency (Menz 2005, Palmer
and Burtraw 2005, Lewis and Wiser 2007, Sanya 2009, Yin and Powers 2010, Bird et al. 2011,
Palmer et al. 2011) and to discuss successful and unsuccessful cases (Goldemberg et al. 2004, Blok
2006, Zoellner et al. 2008, Fava Neves and Conejero 2010). Some results, nevertheless, indicate
that the use of renewable sources for electric power generation has fallen short of expectations
(Roos et al. 1999).
American states provide supporting examples for the reference above. The US states of
Michigan and Wisconsin have mandated that power distribution utilities must supply endconsumers with 10% of renewable electricity (i.e. electricity based on renewable sources such as
solar, wind, geothermal, micro-hydro10, and biomass) by 2015. As of 2013, Michigan authorities

10

with capacity of less than 60 megawatts.

52

reported that renewable electricity distributed within the state has barely reached half of the RPS
mandate (MPSC 2014). Illinois and Minnesota have even higher goals. Utilities in these states are
expected to distribute 25% renewable electricity by 2025.
The disparity between real and expected participation of renewables in the electricity
portfolio of US states is associated with multiple deterring factors (Roos et al. 1999, Paulrud and
Laitila 2010). Among those factors, farmland conversion and entry barriers for feedstock
producers appear to play major roles in the establishment of sustainable relationships with
upstream supply chain players (e.g. transactions between utilities of generation and suppliers of
biomass feedstock). Previous research points out that uncertainty of biomass-based electricity
demand, asymmetric information, and idiosyncratic assets turned initial attempts to trade
renewable energy feedstock into hold-up problems with difficult solutions (Signorini et al.
forthcoming). Factors related to downstream relationships seem less concerning. Transactions
between biomass-based generating facilities and distribution utilities are likely to unfold smoothly
because state authorities recognize the inability of renewable electricity to compete in price with
electricity generated from traditional sources such as coal and petroleum. Hence, distribution
utilities are often granted some leeway to surcharge end consumers in order to acquire less pricecompetitive power and meet federal requirements11.
Contributions from two recent studies indicate that hesitation and skepticism among
potential biomass suppliers might be important underlying reasons to these deterring factors. Using
a double-hurdle modeling approach, Qualls et al. (2012) find significant effects of hesitation on
willingness to grow switchgrass and convert farmland. Using similar analytical methods, Hayden
(2013) concludes that only one third of the sampled population of landowners would be willing to

11

Residential customers in Lansing, MI have $2.50 of surcharge added to their monthly bills. Residential customers
in Milwaukee, WI might voluntarily subscribe to a renewable energy program starting at $4.50 per month.

53

rent out marginal land at $300 per acre to grow bioenergy crops. Such rates, however, far exceed
rental rates being offered for conventional field crops in the analyzed region. Skepticism has also
become an important driver of producers’ lack of interest for energy crops (Rossi and Hinrichs
2011), especially because hold-up problems experienced by some tend to spread over broader
populations. In the context of the developing biomass-based electricity industry, we borrow the
definition of hesitation from Qualls et al. (2012): hesitation is the reluctance to adopt new
production methods or crops until one sees them working for others. To draw unbiased estimates
of agricultural producers’ willingness to grow energy crops and to identify the underlying causes
of skepticism and hesitation, this study proposes an estimating approach different from those found
in the literature.
Much of the analysis of agricultural producers’ willingness to change land use and
willingness to supply biomass feedstock utilize simulation methods (Scheffran and BenDor 2009,
Bocquého and Jacquet 2010, Sherrington and Moran 2010, Griffith et al. 2012) or survey-based
choice experiments (Rahmani et al. 1996, Paulrud and Laitila 2010, Caldas et al. 2014). Other
scholars examine farmers’ intentions toward energy crops by analyzing focus group or structured
interview responses (Mattison and Norris 2007, Sherrington et al. 2008, Villamil et al. 2008).
Despite great contributions, these studies either simulate decisions using a ‘representative farm’
or estimate coefficients using data from randomly selected respondents. Our major concern is that
results drawn from these approaches produce weak approximation of reality and little effective
guidance to decision makers. That is because these results are often extrapolated to broad regions
and important factors such as regional specificities and heterogeneity of farming operations tend
to be neglected. The present study overcomes these limitations by conducting a choice experiment

54

while recognizing that regional characteristics and types of operation lead to different potentials
to convert conventional cropping systems into acres of energy crops.
In the realm of the developing bioelectricity industry in the US, the objective of this study
is to refine past attempts to assess agricultural producers’ willingness to grow energy crops.
Specific objectives are to identify locations within six Midwestern states with RPS mandates (i.e.
Illinois, Michigan, Minnesota, Missouri, Ohio, and Wisconsin) where energy crops have
competitive advantages over conventional cropping systems; and to examine agricultural
producers’ preferences for cropping system attributes conditional on location and type of farming
operation. This study uses a discrete choice experiment and incorporates agricultural producers’
preference heterogeneity into two adapted versions of the random utility maximization (RUM)
model (McFadden 1974). This study also compares the models in terms of relative performance
as suggested by Fiebig et al. (2010). We design an unlabeled choice experiment that includes
expected net margins along with other characterizing attributes of hypothetical cropping systems.
The following sections explain how regions and types of farming operations were categorized
according to energy crops substitution potential.

2. Identifying Regional Categories

Production regions are categorized by comparing predictions of net margins from a
conventional cropping system to net margins that might be obtained from growing energy crops.
As this study focuses on Midwestern states, corn-soybean rotation is assumed to be our
conventional cropping system. Such assumption is reasonable since corn-soybean rotation is the
most common cropping system in the Midwest United States. Corn and soybean together

55

correspond to 79 percent of the total area planted in the states under analysis. The state of Illinois
leads the rank with the highest share of farmland covered with corn or soybean (93 percent in
2013), followed by Ohio (82 percent), Minnesota (79 percent), Michigan and Wisconsin (70
percent), and finally Missouri (64 percent) (USDA/NASS 2014).
Because switchgrass (Panicum virgatum) is considered by some scholars the ‘model
biomass feedstock’ (McLaughlin and Kszos 2005), we base our introductory analysis in estimates
obtained elsewhere for this crop. We use public multi-level data from the United States Department
of Agriculture (USDA) (i.e. National Agricultural Statistics Service (NASS) and Economic
Research Service (ERS)) and from the Food and Agricultural Policy Research Institute
(FAPRI/MU) to estimate net margins from corn-soybean rotation. Expected net margin (ENM)
from switchgrass is based on results of highly cited research papers. Table 2 summarizes time
granularity, spatial granularity, and sources for all data.

Table 2: Summary of data
Time Granularity
Corn-Soybean Rotation
Yield
annual: 2007-2012
Price forecast
annual: 2014-2019
Farm Revenue
Operating Costs

annual: 2007-2012

Net Margin
Switchgrass
Farm Revenue
Operating Costs
Net Margin

-

Spatial Granularity

Sources

county-level
national-level

NASS QuickStats
FAPRI/MU

ERS farm resource
regions

ERS Costs and
Returns

Griffith et al. (2012)
Perrin et al. (2008)
-

Corn and soybean yields (bushels/acre) are disaggregated at county-level. We assume that
farms within a given county obtain production yields similar to those obtained in the past six years.
56

Price forecast ($/bushel) for corn and soybean is at best available for the country12. Hence,
agricultural producers are assumed to receive the prices forecasted by FAPRI/MU for the next six
years, regardless of the location. Farm revenue from corn-soybean rotation (dollars per acre) is
computed based on the assumption that revenue comes half from corn sales and half from soybean
sales. On the expenses side, operating cost (dollars per acre) is assumed to include expenditures
for seed, fertilizer, chemicals, custom operations, fuel, lubricants, electricity, repairs, and interest
on operating capital; and is available at the level of major production regions13. We also assume
that operating costs (dollars per acre) reported by USDA/ERS remain stable for the next six years.
Half of the operating costs are assumed to come from corn production and half comes from
soybean production. Finally, ENM is computed based on the data and assumptions described
above, corresponding to farm revenue minus operating cost.
Switchgrass farm revenue is assumed to be the same as the compensation offered to Eastern
Tennessee agricultural producers engaged in the University of Tennessee Biofuels Initiative
(UTBI). The UTBI contract compensates producers with an annual payment of $450 per acre
(Griffith et al. 2012). Operating cost is retrieved from Perrin et al. (2008) and is assumed to be
similar for all agricultural producers regardless of the location. The annualized cost per acre
includes seed and planting operations, fertilizer and operations, weed control, harvesting
operations, and rent. Estimates indicate that commercial-scale fields have an annualized cost of
$133 per acre. The ENM from switchgrass is then calculated to be $317 per acre for all locations.

12

Aggregated commodities prices should not be too concerning, however. The average of differences between highest
and lowest state prices (as reported by USDA/NASS for 2007-2012) is $0.58 per bushel of corn and $0.62 per bushel
of soybean. Such difference affects farm revenue in plus or minus $25 per acre.
13
See US farm resource regions link for more information:
http://webarchives.cdlib.org/sw1wp9v27r/http://ers.usda.gov/briefing/arms/resourceregions/resourceregions.htm

57

With expected county-level net margin values from corn-soybean rotation and expected
net margin from switchgrass, a comparative analysis can be undertaken. Counties are divided into
four categories: (1) counties that have ENM (over the next six years) from corn-soybean rotation
30 percent below state average; (2) those with ENM higher than 70 percent of state average but
lower than ENM from switchgrass; (3) those with ENM from corn-soybean rotation above ENM
from switchgrass but with two or more annual net margins from corn-soybean rotation below ENM
from switchgrass; and (4) counties with at most one annual net margin from corn-soybean rotation
below ENM from switchgrass. Table 3 summarizes the frequency of county categories per state
and figure 6 displays their location within states.

Table 3: Expected Net Margin from corn-soybean rotation by State and by category
(2014-2019). Frequency of county categories per state
Corn-soybean rotation
ENM by State
Illinois
374.33
Michigan
295.79
Minnesota
353.89
Missouri
172.38
Ohio
355.02
Wisconsin
299.31
Sum of counties
ENM by category

Categories (frequency)
(1)
(2)
(3)
(4)
13
20
5
64
9
35
7
10
8
22
10
38
16
85
0
0
2
13
43
28
7
32
20
6
55
207
85
146
173.65 241.01 351.53
410.52

In this study we hypothesize that agricultural producers’ willingness to dedicate farmland
to energy crops is higher in regions where switchgrass presents comparative advantage over cornsoybean rotation systems for the next six years. More precisely, agricultural producers’ preference
for energy crops is higher in counties categorized as (1) when compared to other county categories.
The underlying reason here is that producers located in 1-counties might be searching for a
cropping system with greater economic returns than corn-soybean rotation. Exceptions might
58

occur in regions where such transition has already happened, such as in Northwest Michigan (fruit
orchards), Central Missouri (livestock production), and Northeast Minnesota (timber and wood
pulp production).

Figure 6: County categories per state
WI

MN
IL

MO

MI

Legend
County category #1
County category #2
County category #3

OH

County category #4

Source: Designed by the authors.

For similar reasons, we hypothesize that agricultural producers located in 2-counties or 3counties will be more willing to grow switchgrass than producers located in 4-counties. That is,
switchgrass will have lower probability of entering counties where corn-soybean systems generate
high and stable annual revenues to growers. Our next hypothesis suggests that farmers located in
counties categorized as (2) will have higher interest in growing switchgrass than farmers located
59

in 3-counties. The reason here is that farmers exposed to low average returns from corn and
soybean might be more willing to grow switchgrass than farmers who will trade stable returns
from energy crops for slightly higher but volatile annual revenues from corn and soybean. Table
4 summarizes these null hypotheses.

Table 4: Null hypotheses: Willingness-to-grow (WTG)
Ho :
WTG1 > WTG2 > WTG3 > WTG4

3. Identifying Categories of Farming Operations

Farming operations are categorized based on agricultural producers’ responses to the 2012
census of agriculture administered by USDA/NASS. Answers to section 29 (‘Machinery and
Equipment’) of the census are used as proxy for farming capability. From these responses,
agricultural producers are grouped depending on the ownership status over machinery – selfpropelled grain combines and self-propelled forage harvesters or a set of three equipment to harvest
forage (mower/rake/baler). Agricultural producers were grouped as follows: (A) those who own
equipment to harvest grain as well as forage; (B) those who own equipment to harvest grain only;
(C) those who own equipment to harvest forage only; and (D) those who own neither type of
equipment. This categorization is only possible because the questionnaire used in this study was
administered by USDA/NASS14.
With farm operations categorized according to production capabilities, we hypothesize that
agricultural producers who need to acquire machinery to grow energy crops are less willing to

14

The agency did not disclose private information from census respondents but allowed us to sort the population of
producers as described above.

60

dedicate farmland to switchgrass. The ultimate goal here is to test whether certain characteristics
of farming operations such as ownership of machinery have significant effect on the likelihood of
growing energy crops.

4. Research Methodology

Choice experiments simulate real-world situations in order to examine how individuals –
in our case, agricultural producers – make decisions over available alternatives with varying
characteristics. Agricultural producers derive net margin expectations and make production
decisions based on past experience, peer influence, household constraints, agronomic
characteristic of cropping systems, and market information available ex ante (Ross and Signorini
2013). In our context, cropping systems can be interpreted as sets of attributes which express value
to decision makers. Previous focus group interviews with agricultural producers in Michigan (Ross
and Signorini 2013) highlight six multi-level pecuniary and non-pecuniary attributes with major
impact on how agricultural producers make planting decisions. Those attributes are: expected net
margin per acre per year, net margin variance, intensity of production practices (e.g. tillage,
fertilization, pest control, and irrigation), coordination arrangement, whether the crop under
analysis requires acquisition of machinery, and choice of neighbor farmers.
In the choice experiment, ENM takes four levels: $200/acre/year, $260/acre/year,
$320/acre/year, or $400/acre/year. These values attempt to cover the most likely range of ENM
from conventional cropping systems and switchgrass in Midwestern states. Revenue variance takes
three levels: ‘± 40%’, ‘± 25%’, or ‘± 10%’. Intensity of production practices takes two levels:
‘High’ or ‘Low’. Coordination arrangement takes three levels: ‘Cooperative’, ‘Production

61

Contract’, or ‘Spot Market’. The choice made by neighbor producers takes two levels: ‘Crop A’
or ‘Crop B’. Finally, requirement of machinery is included in the choice experiment but its levels
do not vary randomly. The levels are rather adjusted depending on the respondent’s category of
farming operation. Table 5 outlines the attributes and levels used in the experiment.

Table 5: Attributes and levels used in the choice experiment
Attributes
Revenue per acre per year

Revenue variance
Intensity of production practices
Vertical coordination
Choice of neighbor producers
Requires acquisition of specific
machinery15

Levels
$200/acre/year
$260/acre/year
$320/acre/year
$400/acre/year
± 40%
± 25%
± 10%
High
Low
Cooperative
Production Contract
Spot Market
Crop A
Crop B
Yes
No

Differently from other studies using choice experiment to examine producers’ willingness
to grow energy crops, this study takes an unlabeled approach. Without stating the crops’ name,
agricultural producers are requested to respond in a more abstract fashion, leading to skepticismfree estimates. Switchgrass and corn-soybean rotation can be identified from the attribute levels,
however. The literature in place points out that switchgrass requires less intensive production
practices when compared to corn and soybean crops (Khanna et al. 2008). FAPRI data corroborates

The levels for this attribute do not vary randomly. They rather vary in function of the respondents’ production
capabilities and levels of ‘intensity of production practices’.
15

62

with this fact and indicates that forage crops (e.g. Alfalfa – Medicago sativa) require 31% less
labor time and effort to reach average yield than corn-soybean rotation does. Hence, we use the
‘intensity of production practices’ attribute as an indicator of the cropping systems included in a
given choice scenario: a system characterized with ‘low’ intensity of production practices connotes
switchgrass and a system qualified with ‘high’ intensity of practices implicitly refers to
corn/soybean rotation.
Using this identification and the machinery ownership status of agricultural producers,
surveys are adjusted to capture realistic preferences. For example if the questionnaire is to be
answered by a producer who falls into category B (i.e. owns equipment to harvest grain only), the
crop alternative characterized by ‘low’ intensity of production practices (i.e. representing
switchgrass) should always have a ‘yes’ for ‘requires acquisition of machinery’ and a ‘no’ if the
intensity of production practices indicates ‘high’. Likewise, if the questionnaire is to be answered
by an A-producer (i.e. owns equipment to harvest grain as well as forage), there should be a ‘no’
for ‘requires acquisition of machinery’ regardless of the alternative crop.
The unlabeled experiment uses an efficient fractional factorial design, which accounts for
two-way interactions between county category dummies and cropping system attributes. The
OPTEX procedure in SAS was used to define 36 choice scenarios and divide them into four
blocks16. The blocking strategy was defined to decrease the number of scenarios presented to
agricultural producers and reduce bias due to fatigue. Hence, each survey respondent was
presented with nine different choice scenarios featuring two generic cropping systems (labeled
‘crop rotation A’ and ‘crop rotation B’) and an opt-out alternative (labeled ‘neither’). By including
the opt-out alternative, we remove the market participation assumption and recognize that

16

The experimental design obtained an optimal D-efficiency value of 97.92.

63

agricultural producers might choose not to farm if they are restricted to choose from the
alternatives presented (Adamowicz et al. 1998). Figure 7 (appendix A) provides an example of
choice scenario used in the experiment.
The data used in this study was collected from a mail survey administered through
USDA/NASS. The survey was sent to 4,216 agricultural producers who met the following criteria.
First, producers must have responded to the 2012 Census of Agriculture. This criterion implies
that the producer is listed in the USDA/NASS data bank. Second, agricultural producers must have
reported farm income of $10,000.00 or more in 2012 to be eligible as survey respondents. This
criterion limits the population of producers to professional individuals and excludes those who
grow crops for leisure. Although researchers use different techniques to obtain representative
responses from professional producers, such criterion is also found in other survey-based studies
(Jensen et al. 2007, Paulrud and Laitila 2010, Hayden 2013). Third, agricultural producers must
have reported ‘value of sales’ above $5,000.00 in section 6 (‘field crops’) and/or section 7 (‘hay
and forage crops’) of the 2012 census. This criterion seeks to include farmers who produce grains
and/or forage crops aiming economic gains while excluding those focused on other businesses.
This criterion also tries to exclude large dairy facilities, for example, that might obtain some profit
from selling leftover hay bales. Finally, the amount of grain and/or forage produced in 2012 and
reported in the Census must have been produced within the counties of interest. All these criteria
were precisely verified prior to the first mailing of the survey.
Eligible producers were randomly selected by the USDA/NASS within each county
category rather than randomly selected across the entire region under analysis. The objective here
is to obtain a balanced dataset with similar number of observations for each county category,
leading to a consistent statistical analysis of agricultural producers’ preferences while controlling

64

for regional specificities. Descriptive statistics of demographic characteristics for the survey
sample and for the population of farmers, who meet the three research criteria mentioned above,
are presented in Table 6.
In total each agricultural producer received two letters in the mail and a phone call. The
first correspondence contained a cover letter explaining the purpose of the study and a copy of the
survey with the choice experiment. The second correspondence was sent two weeks after the first
aiming to reinforce the importance of the study. It included a new copy of the cover letter and a
new copy of the survey. Finally, phone calls were made a week after the second mailing to those
participants who had not yet returned the surveys. In total 294 valid surveys were obtained, leading
to a 7% response rate17. Almost all returned surveys were filled completely, yielding a statistical
sample of 2519 observations (294 surveys X 9 choice scenarios, approximately). Survey sample
is available from the authors upon request. The dataset is distributed across county categories as
follows: 21% of the choices over cropping systems were stated by agricultural producers located
in 1-counties; 29% from producers located in county category 2; 26% were made by producers
from county category 3; and 24% from agricultural producers located in county category 4.
It is worth noting that the obtained sample appears to be a good representation of the
population. Table 6 below indicates that the summary statistics of experience, age, farm income
participation, gender, retirement status, and engagement of producers in renewable energy
enterprises are similar for sample and population. There are slight divergences for harvested acres,
however. While the sample has an overall mean of 390 acres harvested in 2012, the population
seems to have harvested 16% more land on average. Except for producers sampled in county
category 2, survey respondents appear to have harvested fewer acres than the population of their

17

The response rate achieved here is comparable to other studies. Looking at a similar research question Khanna et
al. (2014) obtained 424 usable responses from 4800 surveys, leading to 8.8% response rate.

65

corresponding counties. The distribution of harvested acres is also more disperse for the population
than it is for the sample.

Table 6: Demographic statistics
Sample

Population

Size
- county cat. 1
- county cat. 2
- county cat. 3
- county cat. 4

294 (persons)
63
84
77
70

656,917
56,852
278,816
134,056
187,193

Net Margin Revealed
- county cat. 1
- county cat. 2
- county cat. 3
- county cat. 4

229.90 (mean)
152.69
211.60
276.12
264.73

Regional Rent Price
- county cat. 1
- county cat. 2
- county cat. 3
- county cat. 4

147.93 (mean)
83.85
135.42
175.70
197.29

Acreage
- county cat. 1
- county cat. 2
- county cat. 3
- county cat. 4

390.2 ± 585.9
246.53
460.09
385.13
446.90

451.98 ±746.37
357.87
450.69
427.70
518.39

(mean ± st. dev.)

Experience (years)
- county cat. 1
- county cat. 2
- county cat. 3
- county cat. 4

29.6 ± 15.7
30.33
29.23
31.94
26.90

30.3 ±15.06
30.45
30.14
30.52
30.35

(mean ± st. dev.)

59.79 ±13.31
60.40
60.14
59.45
59.31

(mean ± st. dev.)

Age
- county cat. 1
- county cat. 2
- county cat. 3
- county cat. 4

58.4 ± 13.9
58.91
60.48
58.06
55.99

66

Obs.
2519 (n)
535
740
647
597

Table 6 (cont’d)
Sample

Population

Farm Income %
- county cat. 1
- county cat. 2
- county cat. 3
- county cat. 4

47% ± 35%
38%
48%
53%
50%

48% ±36%
40%
46%
49%
54%

Gender
- male
- female

96.6%
3.4%

96.1%
3.9%

Retired?
- yes

18%

19.6%

Obs.
(mean ± st. dev.)

4.1. The Model
The statistical model used in this article to analyze decision data relies on random utility
theory (McFadden 1974, 1981). Agricultural producers are assumed to maximize utility by
choosing from a hypothetical set of two cropping systems available to them before the planting
season. Following Lancaster (1966), system alternatives are comprised of a number of attributes
that both agricultural producers and analyst can observe. Let 𝑆𝑖𝑡 denote the set of system
alternatives (indexed j) available to agricultural producer 𝑖 in choice scenario t. Let 𝒙𝑖𝑗𝑡 be the
vector of attributes that take on various levels for each system alternative 𝑗 in choice scenario 𝑡.
Also, let 𝑈𝑖𝑗𝑡 be the utility function of agricultural producer 𝑖 from choosing cropping system 𝑗 ∈
𝑆𝑖𝑡 in choice scenario 𝑡.
The utility function is assumed to be of the following form:

𝑈𝑖𝑗𝑡 = 𝑣𝑖𝑗𝑡 + 𝜀𝑖𝑗𝑡 ,
where 𝑣𝑖𝑗𝑡 is the deterministic term and 𝜀𝑖𝑗𝑡

(1)
is the stochastic term. While the latter captures

cropping system attributes observed by agricultural producer 𝑖 but unobserved by the analyst, the
67

former contains attributes observed by both producer and analyst. The stochastic term is assumed
to be independent of 𝑣𝑖𝑗𝑡 and can be argued as agricultural producers’ heterogeneity in preferences
(Keane 1997) for unobserved characteristics of cropping systems. Regarding the deterministic
term, agricultural producer 𝑖 has preferences for net margins, risk exposure, and other nonpecuniary attributes. Let 𝑣𝑖𝑗𝑡 be linear in parameters as follows:

𝑣𝑖𝑗𝑡 = 𝜷𝒙𝑖𝑗𝑡 ,
where 𝜷 is a vector of utility weights measuring N, and 𝒙𝑖𝑗𝑡

(2)
a vector of equal length N measuring

the levels of each attribute for cropping system 𝑗. Plugging (2) into (1) we have:

𝑈𝑖𝑗𝑡 = 𝜷𝒙𝑖𝑗𝑡 + 𝜀𝑖𝑗𝑡 ,
(3)
Hence, agricultural producer 𝑖 chooses from a set of alternative cropping systems in order
to maximize utility. But because the analyst cannot observe 𝜀𝑖𝑗𝑡 , an informed assumption regarding
its distribution must be made so estimates resulting from stated decisions can be obtained. A
convenient and often made set of assumptions regarding 𝜀𝑖𝑗𝑡 is that it is distributed identically and
independently across individuals and follows a Glumbel distribution (𝜀𝑖𝑗𝑡 ~ i.i.d. Glumbel or
extreme value type I, in short). The basic multinomial logit (MNL) model (McFadden 1974) for
the probability that agricultural producer 𝑖 chooses cropping alternative 𝑐 when presented with
scenario 𝑡 can be written as:

𝑃𝑖𝑐𝑡 = 𝑃𝑟 {𝑈𝑖𝑐𝑡 = max 𝑈𝑖𝑗𝑡 } =
𝑗∈𝑆

𝑒𝑥𝑝 (𝜷𝒙𝑖𝑐𝑡 )
∑𝑗∈𝑆 𝑒𝑥𝑝 (𝜷𝒙𝑖𝑗𝑡 )

(4)
The literature in place has, however, sought to refine McFadden’s seminal model in three
main fronts: (i) addition of heterogeneity in preferences for observed attributes; (ii) inclusion of
68

preference heterogeneity for unobserved attributes (Keane 1997); and (iii) correlation of observed
attributes18. The first concern has led scholars to propose and apply the random parameters logit
(RPL) model, which is appealing to practitioners due to its ease of use (Fiebig et al. 2010). In this
case utility weights are assumed to be random variables following a pre-determined distribution.
When applied, the RPL model modifies equation 3 as follows:

𝑈𝑖𝑗𝑡 = (𝜷 + 𝜸𝑖 )𝒙𝑖𝑗𝑡 + 𝜀𝑖𝑗𝑡 ,
(5)
where 𝜷 is the vector of mean utility weights and 𝜸𝑖 is the vector of individual 𝑖-specific deviations
from the mean weight. It is worth to note that the utility weight on net margin is commonly kept
as a fixed variable rather than random (Goett et al. 2000, Morey and Rossmann 2003). Such
restriction has to do with identification of WTA estimates, calculated subsequently to model
estimation. The second concern serves as motivation for the scaled multinomial logit (S-MNL)
model, which modifies equation 3 as below:

𝑈𝑖𝑗𝑡 = 𝜷𝒙𝑖𝑗𝑡 +

𝜀𝑖𝑗𝑡
⁄𝜎𝑖 ,

(6)
where 𝜎𝑖 varies across agricultural producers and represents individual-specific scale in the
stochastic term. The scale parameter 𝜎 is commonly normalized to 1 in order to allow model
estimation. The S-MNL model takes the following form:

𝑈𝑖𝑗𝑡 = (𝜷𝜎𝑖 )𝒙𝑖𝑗𝑡 + 𝜀𝑖𝑗𝑡
(7)

Other refinements to McFadden’s original MNL model have recently been proposed such as relaxing the linear
utility assumption or estimating preference models in WTP-space. Although we recognize the importance of these
advances, the purpose of our argument is to call attention to refinements that have been tested in several empirical
studies and are now widely accepted in the literature.
18

69

The third and final concern regarding McFadden’s model is whether or not observed
attributes should be correlated (Revelt and Train 1998, Scarpa and Del Giudice 2004). While most
scholars argue that adding this flexibility tends to enhance goodness of fit, others explain that
complex models may prevent its implementation due to the rapid increase in number of parameters
(Louviere et al. 2000, Train 2009). In other words, allowing parameters to correlate freely in
complex models forces the number of parameters to increase rather fast, turning estimation
unfeasible.
Besides these three research fronts, it is worth mentioning that choice behavior specialists
have been quite active developing new models and discussing new approaches. A model that has
received recent attention is the generalized multinomial logit (G-MNL) model (Fiebig et al. 2010)
which incorporates individual deviations to utility weights – as in the RPL model – and individual
specific scale heterogeneity – as in the S-MNL model – into the same model. More specifically,
the G-MNL model combines equations 5 and 7 and is presented in equation 8.

𝑈𝑖𝑗𝑡 = [𝜷𝝈𝑖 + 𝛿𝜸𝑖 + (1 − 𝛿)𝜸𝑖 𝝈𝑖 ]𝒙𝑖𝑗𝑡 + 𝜀𝑖𝑗𝑡
(8)
Scale factor 𝜎𝑖 and the deviations to mean utility weights 𝜸𝑖 take the same descriptions
provided above. The new variable 𝛿 captures the proportions of preference heterogeneity
associated with scale factor 𝜎 or individual-specific deviation 𝛾19.
There are strong criticisms against the models presented in equations 5, 7 and 8, however.
Authors argue that the utility weight on net margin (or revenue) should be treated as random
variables, just as the utility weights related to other attributes are. The underlying reason is that
unobserved socioeconomic characteristics are likely to influence individuals to respond differently

19

For a comprehensive treatment of the G-MNL model, please refer to Fiebig et al. (2010).

70

to revenue (or expense in case of WTP estimation) (Fiebig et al. 2010). Train (2009) contends that
models with all-random coefficients can be empirically unidentified. Yet, it is also known that
WTA/WTP estimates derived from two random coefficients do not converge to identifiable
distributions. In this present study we agree with the latter set of concerns and assume net margin
attribute as non-random variables.
In a different study Hess and Rose (2012) express concern to the G-MNL model
exclusively. The authors argue that scale heterogeneity is equivalent to heterogeneity in individual
utility weights; and including both types of heterogeneity into the same model (as in equation 8)
is misleading. They explain that the relative gain in fit obtained through the G-MNL model – when
compared to the RPL model – is primarily associated with using more flexible distributions, rather
than capturing scale heterogeneity. Louviere et al. (2008) also criticize the G-MNL model. In their
empirical application, they have found that heterogeneity in preferences is better captured by SMNL than by G-MNL.
In this study two econometric models – random parameters logit (RPL) and scaled
multinomial logit (S-MNL) – are used to estimate preferences over cropping system attributes.
Both of these models have meaningful advantage over the traditional MNL model because they
recognize that subjects are heterogeneous and might have different appreciation for specific
attributes. The modeling question that remains unanswered is how to incorporate agricultural
producers’ preference heterogeneity for cropping system attributes – through individual-specific
deviations from the mean utility weight or though scale factor. Following the method proposed by
Fiebig et al. (2010), we assess model performance by examining the likelihood improvement that
is attained by including deviations from mean utility weights or individual scale factors into the
seminal MNL model proposed by McFadden’s (1974).

71

The RPL model assumes that utility weights are random parameters and follow a
predetermined distribution (Train 2009). Based on equation 5, the probability that agricultural
producer 𝑖 chooses cropping alternative 𝑐 when presented with scenario 𝑡 becomes:

𝑃𝑖𝑐𝑡 = ∫

𝑒𝑥𝑝(𝜷𝒙𝑖𝑐𝑡 )
∑𝑗∈𝑆 𝑒𝑥𝑝(𝜷𝒙𝑖𝑗𝑡 )

𝑔(𝛽)𝑑𝛽
(9)

where the distribution of random parameters 𝑔(∙) is normal and 𝜷 = 𝜷 + 𝜸𝑖 .
20

The S-MNL model assumes that utility weights are scaled proportionately across
agricultural producers. Based on equation 7, the probability of agricultural producer 𝑖 choosing
cropping system 𝑐 is:

𝑃𝑖𝑐𝑡 =

𝑒𝑥𝑝 (𝜷𝒙𝑖𝑐𝑡 )
∑𝑗∈𝑆 𝑒𝑥𝑝 (𝜷𝒙𝑖𝑗𝑡 )
(10)

2

where 𝜷 = 𝜷𝜎𝑖 , 𝜎𝑖 = 𝑒𝑥𝑝 (−𝜏 ⁄2 + 𝜏𝜔𝑖 ) and 𝜔𝑖 takes the standard normal distribution.
Correlation of coefficients was assumed absent due to the complexity of the model. Given
the objective of the study (i.e. to examine whether regional characteristics and farming capability
influence decision), two-way interaction effects bring the total number of parameters to 52 and 29
(RPL and S-MNL, respectively) as opposed to 13 in the main effects RPL and 8 in the main effects
S-MNL. Hence, allowing correlation of coefficients in the model with two-way interactions would
add over 200 parameters to be estimated which is computationally unfeasible. The next section
analyzes the results derived from these models.
Willingness to accept (WTA) values for cropping system attributes are sequentially
computed for each county category by taking the ratio of the estimated coefficient on the observed

20

Normal distribution is assumed because it provides the best fit to our data.

72

attribute to the net margin coefficient, times two due to effects coding (Lusk et al. 2003). Statistical
variability in WTA estimates are calculated using the parametric bootstrapping technique as
proposed by Krinsky and Robb (1986). For each crop attribute, WTA estimates are compared
across county categories using the complete combinatorial method proposed by Poe, Giraud, and
Loomis (2005).

5. Empirical Results

The estimates for two RPL models and two S-MNL models are reported in Tables 7 and 8,
respectively. While column 2 presents estimates for the entire region of interest (main effects
models), column 3 presents county category-specific estimates (two-way interactions models).

Table 7: Random Parameter Logit Estimates for Cropping Systems
RPL model
(partitioned by
county category)

Variable

RPL Model

Net margin per acre per year
- county cat.#1
- county cat.#2
- county cat.#3
- county cat.#4

1.390 (0.074)**

Variance of net margin
- county cat.#1
- county cat.#2
- county cat.#3
- county cat.#4

-1.546 (0.325)**

Does not require acquisition
- county cat.#1
- county cat.#2
- county cat.#3
- county cat.#4

0.714 (0.097)**

73

Standard
Deviations

1.279 (0.174)**
1.249 (0.120)**
1.495 (0.142)**
1.586 (0.160)**

-

-2.41 (0.837)**
-0.632 (0.551)
-1.972 (0.665)**
-1.894 (0.715)**

1.901 (0.567)**
2.623 (1.244)*
0.885 (1.229)
2.475 (0.958)**
2.569 (0.939)**

1.237 (0.212)**
0.375 (0.135)**
0.634 (0.180)**
0.772 (0.238)**

0.713 (0.126)**
0.091 (0.335)
0.508 (0.177)**
0.672 (0.348)
0.935 (0.248)**

Table 7 (cont’d)
RPL model
(partitioned by
county category)

Variable

RPL Model

Low intensity of practices
- county cat.#1
- county cat.#2
- county cat.#3
- county cat.#4

0.540 (0.060)**

Cooperative
- county cat.#1
- county cat.#2
- county cat.#3
- county cat.#4

-0.189 (0.063)**

Specification Contract
- county cat.#1
- county cat.#2
- county cat.#3
- county cat.#4

-0.248 (0.056)**

Opt out
- county cat.#1
- county cat.#2
- county cat.#3
- county cat.#4

1.858 (0.325)**

Log likelihood function
AIC

-1915.63
3857.3

Standard
Deviations

0.855 (0.149)**
0.510 (0.091)**
0.471 (0.117)**
0.428 (0.142)**

0.402 (0.101)**
0.390 (0.219)
0.179 (0.192)
0.509 (0.171)**
0.435 (0.176)*

-0.245 (0.173)
-0.187 (0.1)
-0.068 (0.109)
-0.264 (0.153)

0.511 (0.085)**
0.813 (0.206)**
0.353 (0.154)*
0.127 (0.389)
0.752 (0.176)**

-0.321 (0.146)*
-0.171 (0.096)
-0.134 (0.098)
-0.415 (0.125)**

0.457 (0.08)**
0.607 (0.178)**
0.451 (0.134)**
0.242 (0.172)
0.519 (0.169)**

2.152 (0.815)**
2.060 (0.538)**
1.858 (0.546)**
1.689 (0.607)**

3.442 (0.28)**
5.906 (1.096)**
3.282 (0.464)**
2.668 (0.428)**
2.051 (0.44)**

-1878.10
3860.2

Notes: The models were estimated using Nlogit 5.0, with Halton draws and 800 replications for simulated
probability. Standard errors are reported in parentheses.
* and ** indicate statistical significance at 5% and 1% level respectively.

Table 8: Scaled Multinomial Logit Estimates for Cropping Systems
Variable

S-MNL Model

Net margin per acre per year
- county cat.#1
- county cat.#2
- county cat.#3
- county cat.#4

1.441 (0.221)**

S-MNL model
(partitioned by
county category)
1.038 (0.239)**
1.356 (0.269)**
1.953 (0.473)**
1.849 (0.509)**

74

Table 8 (cont’d)
Variable

S-MNL Model

Variance of net margin
- county cat.#1
- county cat.#2
- county cat.#3
- county cat.#4

-1.546 (0.350)**

Does not require acquisition
- county cat.#1
- county cat.#2
- county cat.#3
- county cat.#4

0.681 (0.117)**

Low intensity of practices
- county cat.#1
- county cat.#2
- county cat.#3
- county cat.#4

0.592 (0.099)**

Cooperative
- county cat.#1
- county cat.#2
- county cat.#3
- county cat.#4

-0.173 (0.068)*

Specification Contract
- county cat.#1
- county cat.#2
- county cat.#3
- county cat.#4

-0.176 (0.056)**

Opt out
- county cat.#1
- county cat.#2
- county cat.#3
- county cat.#4

2.302 (0.448)**

𝜏
𝜎
Log likelihood function
AIC

S-MNL model
(partitioned by
county category)
-2.474 (0.843)**
-0.388 (0.462)
-1.842 (0.665)**
-0.994 (0.805)
0.899 (0.218)**
0.391 (0.131)**
0.687 (0.218)**
1.078 (0.366)**

0.697 (0.164)**
0.538 (0.135)**
0.630 (0.182)**
0.648 (0.169)**

-0.096 (0.131)
-0.256 (0.102)*
-0.073 (0.127)
-0.064 (0.141)

-0.048 (0.110)
-0.302 (0.100)**
-0.132 (0.116)
-0.299 (0.122)*

1.920 (0.619)**
2.727 (0.627)**
3.371 (0.975)**
1.958 (0.700)**

-2245.00
4506.0

1.095 (0.153)**
0.97613 (1.314)
-2213.89
4485.8

Notes: The models were estimated using Nlogit 5.0, with Halton draws and 800 replications for simulated
probability. Standard errors are reported in parentheses.
* and ** indicate statistical significance at 5% and 1% level respectively.

75

The main effects RPL model shows that all attributes included in the experiment influence
utility and drive agricultural producers’ preferences for crop attributes. As expected, the estimates
on net margin, absence of machinery acquisition, and low intensity of production practices are
valuable and increase agricultural producers’ utility; variance of annual net margins and use of
hybrid governance strategies (i.e. specification contract or cooperative), on the other hand, tend to
decrease utility. The model indicates, however, that there is strong heterogeneity in agricultural
producers’ preferences for all experimented cropping system attributes. Although the main effects
S-MNL model obtains similar estimates, deviations from mean utility weights seem to
accommodate preference heterogeneity far better than the scale factor. Departing from the basic
MNL model simple calculations indicates that inclusion of individual deviations improves
goodness of fit by 18.8% while inclusion of scale factors improves fit by 4.8%.
Two-way interactions refine the initial model and allow one to examine whether regional
specificities capture heterogeneity in preferences. Results derived from the RPL model indicate
that some heterogeneity is indeed related to county categories. Variance of net margins, for
example, is less of a concern for agricultural producers located in county category 2. Agricultural
producers located in county categories 1 and 3 have homogeneous preference for cropping systems
that do not require acquisition of machinery. Cropping systems that require low intensity of
production practices are homogeneously preferred to producers located in county categories 1 and
2. While producers located in 3-counties seem to be indifferent of the coordination arrangement
used to govern transactions with buyers, producers from 2-counties have heterogeneous
preferences without a typical preference.
Results derived from the S-MNL model are similar. Variance of net margins appears not
to influence decisions of agricultural producers located in county categories 2 and 4. Requirement

76

of machinery acquisition and intensity of production practices necessary to obtain average yields
influence planting decisions in all county categories. Producers located in 1-counties and 3counties seem to derive little value from the type of coordination arrangement in place. Producers
located in county category 4 are, on one hand, indifferent between trading via spot markets or
cooperatives; but on the other, are less willing to accept a hypothetical engagement offer if the
related coordination arrangement is specification contract. County category 2 farmers are less
willing to adopt a cropping system if the associated coordination arrangement is different from
spot market.
It is worth noting that the inclusion of preference heterogeneity via individual deviations
from mean utility weights improves model performance faster than via scale factors. Taking the
MNL model as base for comparison, RPL with interactions improves fit by 18.63% while S-MNL
improves fit by 4.09%.
In any case interpreting the magnitudes of coefficients is discouraged because only relative
parameter values matter (Scarpa and Del Giudice 2004). The conventional alternative in these
circumstances is to estimate agricultural producers’ willingness to accept (WTA) based on model
estimates. Due to the substantial difference in model performance, Table 9 reports mean WTA
estimates, 95% confidence intervals, and statistical evidence of differences amongst county
categories derived from the RPL model only.

77

Table 9: Mean Willingness-to-accept (WTA) and 95% confidence intervals
Variable
Variance of net margin
- county cat. 1
- county cat. 2
- county cat. 3
- county cat. 4

Mean WTA [95% CI]

*1

-222.54 [-313.87, -133.79]

Does not require acquisition
- county cat. 1
- county cat. 2
- county cat. 3
- county cat. 4

102.92 [75.35, 131.07]

Low intensity of practices
- county cat. 1
- county cat. 2
- county cat. 3
- county cat. 4

77.85 [61.32, 95.06]

Cooperative
- county cat. 1
- county cat. 2
- county cat. 3
- county cat. 4

-27.18 [-44.68, -9.38]

Specification Contract
- county cat. 1
- county cat. 2
- county cat. 3
- county cat. 4

-35.69 [-51.70, -19.89]

-380.19 [-659.01, -118.54]
-101.20 [-278.82, 73.82]
-263.74 [-438.57, -93.58]
-239.34 [-410.67, -65.82]






196.00 [132.21, 273.15]
60.83 [17.46, 106.86]
85.43 [38.94, 135.52]
97.85 [39.67, 160.54]






135.06 [90.83, 186.89]
82.01 [54.15, 113.05]
63.02 [33.02, 93.63]
54.12 [18.98, 91.65]






-38.66 [-91.94, 14.99]
-30.06 [-61.11, 1.27]
-9.00 [-37.33, 19.99]
-33.24 [-70.98, 4.94]






-50.57 [-97.79, -5.70]
-27.65 [-58.06, 3.03]
-17.87 [-43.72, 7.76]
-52.34 [-83.22, -21.98]






Notes: Different symbols in the final column indicate statistical difference at 5% confidence level within each attribute.
*1: Complete combinatorial test results.

The estimates of WTA for variance of net margins differ across county categories.
Agricultural producers located in county category 1 show the highest aversion to variable annual
returns. These producers would be willing to adopt a cropping system with high variance of returns
if the expected net margin was on average $380.19 higher than the average net margin of a
cropping system characterized by low variance of returns, maintaining all other attributes fixed.
This result might be associated with revealed net margins (from the survey). Producers located in
county category 1 reported the thinnest net margins ($152.69/acre/year – Table 5), which is 33.6%

78

lower than the average revealed net margin ($229.90/acre/year). Therefore, it is reasonable that 1county producers demand such high returns because any unexpected reduction of net margins
could lead to desertion or bankruptcy of farms in extreme cases. Producers located in county
category 2 present the lowest aversion to variable net margins. These producers would adopt a
cropping system with variable returns if its expected net margin was $101.20/acre/year higher than
returns from a crop with low variance of net margins, ceteris paribus. For comparison, producers
located in county categories 3 and 4 would need additional $263.74/acre/year and
$239.34/acre/year respectively to adopt a cropping system with high variance of returns, keeping
other attributes equal. These results should be used with caution, however, since the mean
estimates of WTA for net margin variance are not statistically different (Table 9).
The absence of specific investments in machinery is preferable to most producers,
regardless of the location. Estimates of WTA for investment requirements computed from
coefficients of the two-way interactions RPL model presented on Table 6 indicate that producers
would be willing to accept $102.92/acre/year less, on average, to adopt a cropping system that
does not require acquisition of new machinery, ceteris paribus. There are differences across county
categories, however. Producers from 1-counties would be willing to give up $196/acre/year to
adopt a crop or set of crops that do not require investment in machinery. Producers located in
county categories 2, 3, and 4 would request additional $60.83, $85.43, and $97.85/acre/year
respectively to engage in a given cropping system that does require investment in machinery,
leaving other attributes unaltered.
It is interesting to observe that economies of scale seem to correlate with preference for
investments in machinery. While producers located in county category 2 operate the largest
production fields in the sample (460 acres on average – Table 6), they also require the lowest ‘fee’

79

to engage in a cropping system that requires acquisition of machinery. On the other hand,
producers located in 1-counties have the highest WTA estimates because of the relative small size
of production fields (246.53 acres on average – Table 6). In addition to small production fields,
agricultural producers from 1-counties also demonstrate little ability to afford investments due to
the thinnest revealed net margins, leading to the highest WTA estimate. In fact, the WTA estimate
for producers located in 1-counties is statistically different at 5% confidence level from estimates
computed for producers located elsewhere.
Producers expressed overall preference for crops that require low intensity of production
practices. The mean WTA for a low intensity cropping system is $77.85/acre/year. Looking into
specific counties, agricultural producers from 1-counties require $135.06/acre/year to engage in a
high intensity cropping system. Producers located in counties 2, 3, and 4 would demand additional
$82, $63, and $54.13/acre/year to cultivate high intensity crops. An underlying factor leading to
this difference across county categories might be farm income participation in the total household
income. Because producers from 1-counties are less dependent on farm income and consequently
more dependent on off-farm returns (farm income participation of 38% – Table 6), crops with
moderate production practices give them more flexibility to seek full-time positions outside the
farm. Producers located in 3-counties and 4-counties, on the other hand, are more attached to farm
work (average farm income participation of 53% and 50% respectively – Table 6), making them
accept a considerably lower ‘fee’ for a high intensity crop (less than 50% the amount producers
from 1-counties would be willing to accept).
When it comes to coordination arrangements, agricultural producers indicate strong
preference for spot markets. Regardless of the location, producers would be willing to accept an
additional of $27.18/acre/year to trade via cooperatives and an additional of $35.69 to trade via

80

specification contracts. Producers from county category 3 are the least averse to trading through
hybrid governances (i.e. specification contracts or cooperatives). County-3 farmers would be
willing to join a cooperative if the expected net margin was on average $9/acre/year higher than
trading via spot markets; or willing to sign a specification contract if the expected net margin was
$17.87/acre/year more than trading through spot markets. Producers from county category 1 are
the most averse to joining a cooperative while producers from county category 4 are the most
averse to signing production contracts. The former group would require an additional net margin
of $38.66/acre/year to become a cooperative member and the latter group would demand
$52.34/acre/year net to sign a written agreement of supply.
Stated preferences for crop attributes indicate that agricultural producers behave as we
initially expected in regards to net margins. The summary statistics presented in Table 10
(appendix B) shows that the probability of a crop being chosen increases with net margin, over all
counties as well as for each separate county category. Preferences for high net margins, however,
seem to increase in county categories as the trend line approximation indicates. Agricultural
producers located in 3-counties and 4-counties, for example, show stronger preference for high net
margins when compared to agricultural producers located elsewhere. This result makes sense in
light of regional rent prices and revealed net margins. Producers located in 3-counties and 4counties operate in regions where farmland rent prices are the second and first highest across all
county categories ($175.70 and $197.29, respectively – Table 6). Producers located in these
counties also have the highest and second highest revealed net margins, suggesting that they are
used to higher compensations when compared to producers located in 1-counties and 2-counties.
Finally, hypothetical choices made by peers seem to influence the decisions of producers
located in 1-counties differently from producers located elsewhere. While producers located in 1-

81

counties tend to show competitive attitude and adopt cropping systems that differ from neighbor
producers, producers from other counties are likely to adopt the same cropping system, evidencing
collaboration. This result comes hand and hand to the fact that cooperative leads to little value loss
for producers located in 3-counties (additional $9/acre/year), where producers show collaborative
posture. Conversely, producers tend to show competitive behavior in 1-counties (Table 10) where
cropping systems with cooperative coordination arrangement would be accepted for extra
$38.66/acre/year, ceteris paribus.

6. Implications for Energy Crops

Agricultural producers located in all county categories show considerable preference for
low intensity cropping systems; and as expected, WTA decreases from 1-counties to 4-counties.
This result can translate into agricultural producers’ willingness to grow energy crops given that
switchgrass resembles a low intensity crop while corn/soybean resemble high intensity ones.
Therefore, our original hypotheses hold in light of our empirical results – willingness to grow
energy crops (low intensity crops) is greater than willingness to grow conventional row crops (high
intensity crops) and willingness to grow decreases orderly from county category 1 to county
category 4.
These results are promising for extension initiatives wishing to incentivize energy crop
production. As producers become aware of agronomic characteristics of switchgrass (i.e. it
requires 31% less labor time and effort to obtain average yield than conventional crops),
appreciation for low intensity of agricultural practices is likely to broaden, and the low intensity
characteristic to become a source of differentiation leading to opportunities for switchgrass

82

expansion and crop substitution in Midwestern states. Expansion of energy crops might unfold
relatively faster in county category 1 than in other county categories since WTA for low intensity
crops is statistically higher in this county category. It means that well-aligned extension initiatives
can motivate farmland conversion into switchgrass faster in 1-counties than in other county
categories. Nevertheless, all county categories have presented opportunities for crop substitution
and switchgrass expansion.
Producer WTA for machinery acquisition requirement also follows our original hypothesis.
If a given cropping system does not require acquisition of machinery, producers would be more
willing to adopt that system. This result is of great importance for bioelectricity investors because
building a biomass processing plant in regions where few producers own forage harvester or forage
mower/rake/baler (or have expensive access to them) would make switchgrass lose value
associated with its low intensity of production practices. A better strategy would be to first enter
regions where most producers have productive capability for cultivating forage crops, which can
be adapted for switchgrass and production of biomass. In fact, targeting producers who own
harvesting capabilities in addition to educational sessions would boost crop substitution potential
and expansion of energy crops across the Mid-West United States.
That is especially the case if switchgrass enters commercial farms using a contract
specification similar to that used in the University of Tennessee Biofuels Initiative (UTBI). As
presented above, the UTBI contract has a net margin of $317/acre/year, which overcomes average
net margin revealed for all county categories. A combined entry strategy with extension sessions
targeting agricultural producers who own desirable farming capability and net margin of
$317/acre/year should generate enough value to producers so they move away from the preferred
coordination arrangement (i.e. spot market) and sign biomass production contracts.

83

County category 2 in particular shows great opportunities for emergence of the energy crop
industry in Midwestern states if such entry strategy is put forward. Differently from other regions,
the UTBI net margin of $317/acre/year appears to be large enough to offset revealed net margins
in that county category as well as the negative values associated with machinery acquisition and
engagement in specification contract. More specifically, the UTBI net margin overcomes revealed
net margin in county category 2 by $105.40/acre/year, which can be conjectured as sufficient to
cover machinery acquisition – valued at $60.83/acre/year on average – as well as engagement of
producers in specification contracts – averaged at $27.65/acre/year. Nevertheless, utility weights
obtained from the RPL model with two-way interactions indicate that producers in 2-counties have
heterogeneous preferences for machinery acquisition and coordination arrangements; which
emphasizes the importance of extension sessions in an effective entry strategy.
These implications must be used with discretion because WTA estimates rely on the notion
of ceteris paribus. In other words, WTA estimates should be interpreted while maintaining other
factors constant since one cannot guarantee that the effect of two or more attributes combined is
equal to the sum of the individual effects. Additionally, the revealed net margin value of
$211.60/acre/year for producers from 2-counties is averaged across survey respondents, which
does not take dispersion or variance into account. We recognize these points as limitations to our
approach but our results are the first of its kind and can be used to motivate expansion of energy
crops in the Mid-west United States. If past empirical studies have identified hesitation and
skepticism (Rossi and Hinrichs 2011, Qualls et al. 2012, Hayden 2013) from agricultural producers
in regards to energy crops, this article shows that producers are interested in growing crops with
key characteristics found in switchgrass. A good extension initiative could benefit expansion of

84

energy crops because it would not only inform producers that switchgrass possesses characteristics
they value, but it would also minimize the negative effects of hesitation and skepticism.

7. Conclusions

Despite government efforts to develop renewable electricity industries, hesitation and
skepticism of decision makers in upstream links of the supply chain have been identified as
deterring factors for expansion and farmland conversion. This article utilizes an innovative choice
experiment to estimate willingness to grow energy crops. Based on choice data collected
throughout Midwestern states with RPS mandates in place, results indicate that agricultural
producers value (i) cropping systems that require low intensity of production practices to obtain
average yield; (ii) absence of required investments in machinery; (iii) low variance of net margins;
and (iv) coordinating transactions through spot markets.
While recognizing that regional specificities cause preference for cropping system
attributes to vary, results point out that producers located in places where conventional cropping
systems (e.g. corn-soybean rotation) are less competitive to switchgrass tend to prefer crop
characteristics found in switchgrass. In other words, low intensity of production practices have
diminishing WTA as one moves from county category 1 to county category 4.
Farming capabilities also influence preference for cropping system attributes. As one
would plausibly suggest, producers value crops that do not require investment in machinery. The
present study refines this notion and estimates magnitudes for such preference. Results indicate
that absence of investment not only places first in the rank of preferable attributes but also identify
locations across the Mid-west where producers would be less concerned to purchase machinery

85

for cultivating a desirable crop. Producers located in county category 2 would be considerably less
concerned with acquiring equipment whereas producers in county category 1 would be highly
concerned.
Combining these results and taking into consideration agricultural producers’ demographic
characteristics, we show that $317/acre/year (same as the UTBI net margin) might suffice to
motivate farmland conversion and adoption of switchgrass in county category 2 – regardless of
whether investments in machinery are necessary – or in county categories 1, 3, and 4 if investments
are unnecessary. These would hold even if specification contracts must be agreed between
producers and biomass processor. However, it is important to take these interpretations with
caution given the notion of ceteris paribus associated with model estimates and derived WTA
calculations. Further refinement of this study to allow precise interpretation of WTA estimates is
strongly encouraged.
Finally, this study provides evidence that modeling preference heterogeneity through
individual deviations to mean utility weights reaches superior performance than modeling
individual scale effects. Results show that log likelihood improves about 18% when RPL model is
utilized versus 4% improvement when S-MNL model is used, regardless of whether county
categories are specified.
This article leaves an important question unattended. Although it has been observed that
producers from county category 2 and 3 do not have homogeneous preference for coordination
arrangements while producers from counties 1 and 4 do, this study does not address how contract
provision or cooperative membership should be drafted. It might be important to assess
agricultural producers’ preferences for provisions or cooperative membership rules as these

86

points may have substantial influence on producers’ cropping system adoption. Essay 3 of this
dissertation tackles this question in a comprehensive manner.

87

APPENDICES

88

APPENDIX A Sample choice scenario used in the unlabeled experiment

Figure 7: Sample choice scenario used in the unlabeled experiment

Choice Scenario #1 of 9: Which crop rotation would you prefer?
Crop

Crop

System Attributes

Neither
Rotation A

Rotation B

Net margin/acre/year

$260

$400

Net margin variance

± 10% (± $40)

± 40% (± $160)

Requires acquisition of machinery

No

Yes

Intensity of production practices*

Low

High

Marketing arrangement

Cooperative

Spot Market

My neighbors are choosing

I would not farm if
A and B were the
only alternatives
available

-------- Crop Rotation B --------

I would grow:
* cropping systems with LOW intensity of production practices require 31% less labor time and effort to obtain average
yield than cropping systems with HIGH intensity of production practices.

89

APPENDIX B Summary statistics – aggregate counties and individual county categories
Table 10: Summary statistics - all counties and separate county categories
Prob(Net Margin)
All counties
Net Margins Abs. Obs. Freq.
0.40
200
288
0.11
0.11
0.30
260
357
0.14
0.14
0.20
320
582
0.23
0.23
0.10
400
779
0.31
0.31
0.00
Opt-out
513
0.20
Total (n)
2519
Weighted average of margins: $257.35
County Category #1
Net Margins Abs. Obs. Freq.
0.40
200
56
0.10
0.10
0.30
260
86
0.16
0.16
0.20
320
95
0.18
0.18
0.10
400
137
0.26
0.26
0.00
Opt-out
161
0.30
Total (n)
535
Weighted average of margins: $221.98
County Category #2
Net Margins Abs. Obs. Freq.
0.40
200
81
0.11
0.11
0.30
260
103
0.14
0.14
0.20
320
172
0.23
0.23
0.10
400
212
0.29
0.29
0.00
Opt-out
172
0.23
Total (n)
740
Weighted average of margins: $247.05
County Category #3
Net Margins Abs. Obs. Freq.
0.40
200
72
0.11
0.11
0.30
260
84
0.13
0.13
0.20
320
160
0.25
0.25
0.10
400
217
0.34
0.34
0.00
Opt-out
114
0.18
Total (n)
647
Weighted average of margins: $269.30
County Category #4
Net Margins Abs. Obs. Freq.
0.40
200
79
0.13
0.13
0.30
260
84
0.14
0.14
0.20
320
155
0.26
0.26
0.10
400
213
0.36
0.36
0.00
Opt-out
66
0.11
Total (n)
597
Weighted average of margins: $288.84

90

Slope = 0.0674

200

260

320

400

Opt-out

320

400

Opt-out

320

400

Opt-out

320

400

Opt-out

320

400

Opt-out

Slope = 0.0471

200

260

Slope = 0.0624

200

260

Slope = 0.079

200

260

Slope = 0.0792

200

260

Table 10 (cont’d)
Prob(Variance of Net Margins)
All counties
Variance Abs. Obs. Freq.
±10%
737
0.29
±25%
640
0.25
±40%
629
0.25
Opt-out
513
0.20
Total (n)
2519
County Category #1
Variance Abs. Obs. Freq.
±10%
149
0.28
±25%
112
0.21
±40%
113
0.21
Opt-out
161
0.30
Total (n)
535
County Category #2
Variance Abs. Obs. Freq.
±10%
202
0.27
±25%
178
0.24
±40%
188
0.25
Opt-out
172
0.23
Total (n)
740
County Category #3
Variance Abs. Obs. Freq.
±10%
198
0.31
±25%
170
0.26
±40%
165
0.26
Opt-out
114
0.18
Total (n)
647
County Category #4
Variance Abs. Obs. Freq.
±10%
188
0.31
±25%
180
0.30
±40%
163
0.27
Opt-out
66
0.11
Total (n)
597

0.40

Slope = -0.0214
0.30 0.29

0.20 0.25
0.10 0.25

0.00
±10%

±25%

±40%

Opt-out

±40%

Opt-out

±40%

Opt-out

±40%

Opt-out

±40%

Opt-out

0.40

Slope = -0.0336
0.300.28
0.200.21
0.100.21
0.00
±10%

0.40
0.300.27

±25%

Slope = -0.0095

0.200.24
0.100.25
0.00
±10%

0.40
0.300.31

±25%

Slope = -0.0255

0.200.26
0.100.26
0.00
±10%

0.40

0.300.31

±25%

Slope = -0.0209

0.200.30

0.100.27
0.00

±10%

91

±25%

Table 10 (cont’d)
Prob(Requires Acquisition of Machinery)
All counties
Acquisition Req.
Abs. Obs. Freq.
Acq. Requested
833
0.33
Acq. not requested
1173
0.47
Opt-out
513
0.20
2519

County Category #1
Acquisition Req.
Abs. Obs. Freq.
Acq. Requested
128
0.24
Acq. not requested
246
0.46
Opt-out
161
0.30
535

County Category #2
Acquisition Req.
Abs. Obs. Freq.
Acq. Requested
250
0.34
Acq. not requested
318
0.43
Opt-out
172
0.23
740

County Category #3
Acquisition Req.
Abs. Obs. Freq.
Acq. Requested
200
0.31
Acq. not requested
333
0.51
Opt-out
114
0.18
647

County Category#4
Acquisition Req.
Abs. Obs. Freq.
Acq. Requested
255
0.43
Acq. not requested
276
0.46
Opt-out
66
0.11
597

0.60
0.40

0.13

0.20
0.00

Acq.
Requested

Acq. not
requested

Opt-out

Acq.
Requested

Acq. not
requested

Opt-out

Acq.
Requested

Acq. not
requested

Opt-out

Acq.
Requested

Acq. not
requested

Opt-out

Acq.
Requested

Acq. not
requested

Opt-out

0.60
0.40

0.22

0.20
0.00

0.60
0.40

0.09

0.20
0.00

0.60
0.40

0.21

0.20
0.00

0.60
0.40
0.200.04
0.00

92

Table 10 (cont’d)
Prob(Intensity of Production Practices)
All counties
0.60
Intensity Abs. Obs. Freq.
High
872
0.35
0.40
Low
1134
0.45
0.10
0.20
Opt-out
513
0.20
Total (n)
2519
0.00

County Category #1
Intensity Abs. Obs. Freq.
High
144
0.27
Low
230
0.43
Opt-out
161
0.30
Total (n)
535

County Category #2
Intensity Abs. Obs. Freq.
High
243
0.33
Low
325
0.44
Opt-out
172
0.23
Total (n)
740

County Category #3
Intensity Abs. Obs. Freq.
High
238
0.37
Low
295
0.46
Opt-out
114
0.18
Total (n)
647

County Category #4
Intensity Abs. Obs. Freq.
High
247
0.41
Low
284
0.48
Opt-out
66
0.11
Total (n)
597

High

Low

Opt-out

High

Low

Opt-out

High

Low

Opt-out

High

Low

Opt-out

High

Low

Opt-out

0.60
0.40

0.16

0.20
0.00

0.60
0.40

0.11

0.20
0.00

0.60
0.40

0.09

0.20
0.00

0.60
0.40

0.06

0.20
0.00

93

Table 10 (cont’d)
Prob(Vertical Coordination)
All counties
Coord. Strategy Abs. Obs. Freq.
Spot
678
0.27
Contr
632
0.25
Coop
696
0.28
Opt-out
513
0.20
Total (n)
2519
County Category #1
Coord. Strategy Abs. Obs. Freq.
Spot
130
0.24
Contr
114
0.21
Coop
130
0.24
Opt-out
161
0.30
Total (n)
535
County Category #2
Coord. Strategy Abs. Obs. Freq.
Spot
190
0.26
Contr
190
0.26
Coop
188
0.25
Opt-out
172
0.23
Total (n)
740
County Category #3
Coord. Strategy Abs. Obs. Freq.
Spot
173
0.27
Contr
171
0.26
Coop
189
0.29
Opt-out
114
0.18
Total (n)
647
County Category #4
Coord. Strategy Abs. Obs. Freq.
Spot
185
0.31
Contr
157
0.26
Coop
189
0.32
Opt-out
66
0.11
Total (n)
597

0.30
0.20
0.10
0.00
Spot

Contr

Coop

Opt-out

Spot

Contr

Coop

Opt-out

Spot

Contr

Coop

Opt-out

Spot

Contr

Coop

Opt-out

Spot

Contr

Coop

Opt-out

0.30
0.20
0.10

0.00

0.30
0.20
0.10
0.00

0.30
0.20

0.10
0.00

0.30
0.20
0.10

0.00

94

Table 10 (cont’d)
Prob(Neighbor Choice)
All counties
Neighbor choice Abs. Obs. Freq.
Different
978
0.39
Equal
1028
0.41
Opt-out
513
0.20
Total (n)
2519

County Category #1
Neighbor choice Abs. Obs. Freq.
Different
201
0.38
Equal
173
0.32
Opt-out
161
0.30
Total (n)
535

County Category #2
Neighbor choice Abs. Obs. Freq.
Different
263
0.36
Equal
305
0.41
Opt-out
172
0.23
Total (n)
740

County Category #3
Neighbor choice Abs. Obs. Freq.
Different
254
0.39
Equal
279
0.43
Opt-out
114
0.18
Total (n)
647

County Category #4
Neighbor choice Abs. Obs. Freq.
Different
260
0.44
Equal
271
0.45
Opt-out
66
0.11
Total (n)
597

0.50
0.40

0.30

0.02
0.20
0.10
0.00
Different

Equal

Opt-out

Different

Equal

Opt-out

Different

Equal

Opt-out

Different

Equal

Opt-out

Different

Equal

Opt-out

0.50
0.40
0.30
-0.05
0.20
0.10
0.00

0.50
0.40
0.30
0.06
0.20
0.10
0.00

0.50
0.40
0.30
0.04
0.20
0.10
0.00

0.50
0.40
0.30
0.02
0.20
0.10
0.00

95

REFERENCES

96

REFERENCES

Adamowicz, Wiktor, Peter Boxall, Michael Williams, and Jordan Louviere. "Stated preference
approaches for measuring passive use values: Choice experiments and contingent."
American Journal of Agricultural Economics no. 80 no. 1 (1998):64.

Bird, Lori, Caroline Chapman, Jeff Logan, Jenny Sumner, and Walter Short. "Evaluating
Renewable Portfolio Standards and Carbon Cap Scenarios in the U.S. Electric Sector."
Energy Policy 39, no. 5 (2011): 2573-85.

Blok, Kornelis. "Renewable Energy Policies in the European Union." Energy Policy 34, no. 3
(2006): 251-55.

Bocquého, G., and F. Jacquet. "The Adoption of Switchgrass and Miscanthus by Farmers: Impact
of Liquidity Constraints and Risk Preferences." Energy Policy 38, no. 5 (2010): 2598-607.

Caldas, Marcellus, Jason Bergtold, Jeffrey Peterson, Russell Graves, Dietrich Earnhart, Sheng
Gong, Brian Lauer and J. Christopher Brown. "Factors affecting farmers' willingness to
grow alternative biofuel feedstocks across Kansas", Biomass and Bioenergy 66, no. 0
(2014), 223-231.

Fava Neves, Marcos, and Marco Antonio Conejero. Estratégias para a Cana no Brasil: um Negócio
Classe Mundial. [Portuguese] São Paulo, Brazil: Editora Atlas, 2010.

Fiebig, Denzil, Michael Keane, Jordan Louviere, and Nada Wasi. "The Generalized Multinomial
Logit Model: Accounting for Scale and Coefficient Heterogeneity", Marketing Science 29,
no.3 (2010): 393-421.

Goett, Andrew, Kenneth Train, and Kathleen Hudson. "Customers' Choice Among Retail Energy
Suppliers: The Willingness-to-pay for Service Attributes", Energy Journal 21, no. 4
(2000): 1.

Goldemberg, José, Suani Teixeira Coelho, Plinio Mário Nastari, and Oswaldo Lucon. "Ethanol
Learning Curve—the Brazilian Experience." Biomass and Bioenergy 26, no. 3 (2004): 30104.

97

Griffith, Andrew, James Larson, Burton English, and Dan L. McLemore "Analysis of contracting
alternatives for switchgrass as a production alternative on an East Tennessee beef and crop
farm" AgBioForum, 15 no. 2 (2012): 206-216.

Hayden, Noel J., "Landowner Willingness to Supply Marginal Land for Bioenergy Production: A
Case Study from Michigan." M.S. Thesis, Agricultural Economics, Michigan State
University, 108 pp, 2013.

Hess, Stephane. and John Rose. "Can scale and coefficient heterogeneity be separated in random
coefficients models?", Transportation 39, no. 6 (2012): 1225-1239.

Jensen, Kimberly, Christopher D. Clark, Pamela Ellis, Burton English, Jamey Menard, Marie
Walsh, and Daniel de la Torre Ugarte. "Farmer Willingness to Grow Switchgrass for
Energy Production." Biomass and Bioenergy 31, no. 11–12 (2007): 773-81.

Keane, Michael. "Modeling Heterogeneity and State Dependence in Consumer Choice Behavior",
Journal of Business & Economic Statistics 15, no. 3 (1997): 310-327.

Khanna, Madhu, Basanta Dhungana, and John Clifton-Brown. "Costs of Producing Miscanthus
and Switchgrass for Bioenergy in Illinois." Biomass and Bioenergy 32, no. 6 (2008): 48293.

Klein, Benjamin. "Why Hold-Ups Occur: The Self-Enforcing Range of Contractual
Relationships." Economic Inquiry 34, no. 3 (1996): 444-63.

Krinsky, Itzhak and Leslie A. Robb. "On Approximating the Statistical Properties of Elasticities",
Review of Economics & Statistics 68 no. 4 (1986), 715.

Lancaster, Kelvin J. "A New Approach to Consumer Theory." Journal of Political Economy 74,
no. 2 (1966): 132.

Lewis, Joanna I., Ryan H. Wiser "Fostering a Renewable Energy Technology Industry: An
International Comparison of Wind Industry Policy Support Mechanisms." Energy Policy
35, no. 3 (2007): 1844-57.

98

Louviere, Jordan J., David A. Hensher, Joffre D. Swait. "Stated Choice Methods: Analysis and
Applications" Cambridge, UK: University Press, 2000

Louviere, Jordan J., Deborah Street, Leonie Burgess, Nada Wasi, Towhidul Islam, and Anthony
Marley. "Modeling the choices of individual decision-makers by combining efficient
choice experiment designs with extra preference information", Journal of Choice
Modelling 1, no. 1 (2008): 128-164.

Lusk, Jayson L., Jutta Roosen, and John A. Fox. "Demand for Beef from Cattle Administered
Growth Hormones or Fed Genetically Modified Corn: A Comparison of Consumers in
France, Germany, the United Kingdom, and the United States." American Journal of
Agricultural Economics 85, no. 1 (2003): 16-29.

Mattison, Elizabeth H. A., Ken Norris. "Intentions of Uk Farmers toward Biofuel Crop
Production:  Implications for Policy Targets and Land Use Change." Environmental
Science & Technology 41, no. 16 (2007): 5589-94.

McFadden, Daniel L. "Conditional Logit Analysis of Qualitative Choice Behavior" In: Frontiers
in Econometrics, Zarembka, P. (ed). Academic Press, (1974) 105-142

McFadden, Daniel L. "Structural Discrete Probability Models from Theories of Choice." In:
Structural Analysis of Discrete Data and Econometric Applications, Manski, C.F.,
McFadden, D.L. (eds). The MIT Press (1981): 198-272.

McLaughlin, Samuel B. and L. Adams Kszos. "Development of switchgrass (Panicum virgatum)
as a bioenergy feedstock in the United States", Biomass and Bioenergy 28, no. 6 (2005):
515-535.

Menz, Fredric C. "Green Electricity Policies in the United States: Case Study." Energy Policy 33,
no. 18 (2005): 2398-410.

Morey, Edward and K. Greer Rossmann. "Using Stated-Preference Questions to Investigate
Variations in Willingness to Pay for Preserving Marble Monuments: Classic
Heterogeneity, Random Parameters, and Mixture Models", Journal of Cultural Economics
27 no. 3-4 (2003): 215-229.

99

Palmer, Karen, David Burtraw. "Cost-Effectiveness of Renewable Electricity Policies." Energy
Economics 27, no. 6 (2005): 873-94.

Palmer, Karen, Anthony Paul, Matt Woerman, and Daniel C. Steinberg. "Federal Policies for
Renewable Electricity: Impacts and Interactions." Energy Policy 39, no. 7 (2011): 397591.
Paulrud, Susanne and Thomas Laitila. "Farmers’ Attitudes About Growing Energy Crops: A
Choice Experiment Approach." Biomass and Bioenergy 34, no. 12 (2010): 1770-79.

Perrin, Richard, Kenneth Vogel, Marty Schmer, and Rob Mitchell. "Farm-Scale Production Cost
of Switchgrass for Biomass." Bioenergy Research 1, no. 1 (2008): 91-97.

Poe, Gregory. L., Kelly Giraud, and John B. Loomis. "Computational Methods for Measuring the
Difference of Empirical Distributions", American Journal of Agricultural Economics 87,
no. 2 (2005): 353-365.

Qualls, D. Joshua, Kimberly L. Jensen, Christopher D. Clark, Burton English, James A. Larson,
and Steven Yen. "Analysis of Factors Affecting Willingness to Produce Switchgrass in the
Southeastern United States." Biomass and Bioenergy 39, no. 0 (2012): 159-67.

Rahmani, M., A.W. Hodges, J.A. Stricker "Potential Producers and Their Attitudes Toward
Adoption of Biomass Crops in Central Florida" Prepared for presentation at 7th National
BioEnergy Conference, 1996.

Revelt, David, and Kenneth Train. "Mixed Logit with Repeated Choices: Households' Choices of
Appliance Efficiency Level", Review of Economics and Statistics 80, no. 4 (1998): 647657.

Roos, Anders, Robin L. Graham, Bo Hektor, and Christian Rakos. "Critical Factors to Bioenergy
Implementation." Biomass and Bioenergy 17, no. 2 (1999): 113-26.

Ross, Robert. B., and Guilherme Signorini "Contract Design for Biomass Procurement." Focus
Group Report. Michigan State University, East Lansing, MI. (08/15/2013).

100

Rossi, Alissa M., and C. Clare Hinrichs. "Hope and skepticism: Farmer and local community views
on the socio-economic benefits of agricultural bioenergy." Biomass and Bioenergy 35, no.
4 (2011):1418-28.

Scarpa, Riccardo and Teresa Del Giudice. "Market Segmentation via Mixed Logit: Extra Virgin
Olive Oil in Urban Italy", Journal of Agricultural and Food Industrial Organization 2, no.
1-18 (2004).

Scarpa, Riccardo, Mara Thiene, and Kenneth Train. "Utility in Willingness to Pay Space: A Tool
to Address Confounding Random Scale Effects in Destination Choice to the Alps",
American Journal of Agricultural Economics 90, no. 4 (2008): 994-1010.

Sanya, Carley. "State Renewable Energy Electricity Policies: An Empirical Evaluation of
Effectiveness." Energy Policy 37, no. 8 (2009): 3071-81.
Scheffran, Jurgen, Todd BenDor. “Bioenergy and Land Use: a Spatial-agent dynamic model of
energy crop production in Illinois”. International Journal of Environment and Pollution
39, no. 1 (2009):4-27.

Sherrington, Chris, Justin Bartley, and Dominic Moran. "Farm-Level Constraints on the Domestic
Supply of Perennial Energy Crops in the UK." Energy Policy 36, no. 7 (2008): 2504-12.

Sherrington, Chris, and Dominic Moran. "Modelling Farmer Uptake of Perennial Energy Crops in
the UK." Energy Policy 38, no. 7 (2010): 3567-78.

Signorini, Guilherme, R. Brent Ross, William Knudson, Vivek Pandey, and H. Christopher
Peterson. "Assessing Contract Efficiency for Biomass: A Case Study of the Upper
Peninsula of Michigan." Working Paper (forthcoming).

Train, Kenneth E. "Discrete Choice Methods with Simulation". Cambridge, US: Cambridge
University Press, 2009.
Villamil, María B., Anne Heinze Silvis, and Germán A. Bollero. "Potential Miscanthus’ Adoption
in Illinois: Information Needs and Preferred Information Channels." Biomass and
Bioenergy 32, no. 12 (2008): 1338-48.

101

Yin, Haitao, and Nicholas Powers. "Do State Renewable Portfolio Standards Promote in-State
Renewable Generation?." Energy Policy 38, no. 2 (2010): 1140-49.

Zoellner, Jan, Petra Schweizer-Ries, and Christin Wemheuer. "Public Acceptance of Renewable
Energies: Results from Case Studies in Germany." Energy Policy 36, no. 11 (2008): 413641.

102

ESSAY 3: DOES THEORY PREDICT ADOPTION OF PRODUCTION CONTRACTS?
EVIDENCE FROM THE DEVELOPING BIOENERGY INDUSTRY IN U.S. MIDWEST

1. Introduction

Economic theory has developed two distinct routes to deal with issues of market structure
and behavior of firms. While the first – recognized as industrial organization – relies on game
theory and often justifies firm behavior using arguments of incentive (Tirole 1988); the second –
institutional economics – provides justification based on arguments of friction costs, externalities,
and opportunism (Williamson 1985). In spite of such distinction, the concepts of information
asymmetry, uncertainty and idiosyncratic assets seem to have inspired theorists dedicated to either
route, whose contributions originated the rich literature of incomplete contracts.
A central issue in the incomplete contracts literature is the hold-up problem (Williamson
1975, Goldberg 1976, Hart and Moore 1988). From the industrial organization perspective, these
problems emerge because observable-but-unverifiable information exists (Bolton and Dewatripont
2005). Information is said to be observable but not verifiable when trading parties can see a given
contingency but outsiders (e.g. judges and public jury) cannot (Hart 1995). An institutional
economics theorist would conversely argue that hold-ups occur because one party behaves
opportunistically to extract quasi-rents from the counterparty when favorable unanticipated market
conditions arise.
Occurrence of hold-up problems lead to expropriation of quasi-rents on specific
investments made to conduct a transaction. It is especially concerning in emerging industries
where uncertainty of future events tends to be high and investments carry high degrees of

103

specificity (i.e. few redeployment options with major losses of productive value). It also concerns
emerging industries because the occurrence of hold-ups might hinder future entrepreneurship
endeavors in the region or elsewhere.
Regardless of the theoretical interpretation, both routes of economic thought have arrived
at solutions to the problem at hand. From the empirical point of view, however, the questions that
remain unanswered are whether the offered solutions hold in real-world settings; and if so, which
one is preferred by economic agents. Industrial organization theory shows that a ‘specific
performance contract’ with a default option and a penalty for delayed trade suffices to solve holdup problems and results in the first-best outcome (Aghion et al. 1994). In a similar fashion,
Nöldeke and Schmidt (1995) demonstrate that an ‘option contract’ overcomes hold-ups, leading
to efficient idiosyncratic investments and Pareto outcomes. Institutional economics theory argues
that the strategic allocation of private enforcement capital minimizes the probability that hold-up
problems will occur (Klein 1996). Although the latter contribution does not formally show whether
a first-best outcome is obtainable, the author uses multiple real-world case studies to present
efficiency enhancing properties. In other words, the use of sufficiently large hostages (Williamson
1983) deters opportunism and increases expected returns to both parties.
When it comes to testing and implementing these solutions applied economists often face
a major limitation: access to field data on contractual relationships and strategic alliances. Even
with theory shedding light on hold-up issues, predictions of how parties to trade should set strategic
partnerships can seldom be tested empirically due to the limited availability of field data. That is
not surprising because firms spend resources writing agreements to enhance supply chain
efficiencies, which consequently carries strategic information. Under these circumstances firms
are often opposed to sharing information that might expose their competitive advantages, even if

104

the data are only used for academic purposes (Just and Wu 2009). Experimental economics,
nevertheless, provides an alternative approach to overcome the data problem. Generating data
through experiments allows applied economists to test empirical data against theory-based
equilibria. More precisely, analyzing how decision makers state their likely behavior when
hypothetical market situations are posed to them helps us understand the robustness of theoretical
predictions.
The main purpose of this article is to examine what contract designs key economic agents
– agricultural producers in this study – involved in an emerging industry prefer when hypothetical
market situations are presented. Using the nascent bioelectricity industry in the United States as a
case study, this article examines whether theoretical solutions to hold-ups are obtained when
decision makers are more or less vulnerable to hold-up problems. Supporting objectives are (i) to
inform entrepreneurs of what provisions might cause greater engagement of agricultural producers
in the bioelectricity industry; and (ii) to inform extension branches of land grant universities about
information barriers that might prevent producers from reaching transaction efficiency.
The developing renewable energy industry provides a fruitful economic environment to
test whether theory-based solutions to hold-up problems are available. As previously identified,
hold-ups constitute a potential problem in transactions between suppliers of biomass feedstock and
generation utilities, especially because idiosyncrasy tends to characterize assets on both trading
sides and agreements are unlikely to carry provisions to all future contingencies (Signorini et al.
forthcoming). In light of these threats, we test whether the contract solutions discussed by theorists
(Aghion et al. 1994, Klein 1996) emerge from agricultural producers’ stated preferences when
facing hypothetical market situations in the bioelectricity industry.

105

This study uses a discrete choice experiment and incorporates agent heterogeneity into
three adapted versions of the random utility maximization (RUM) model (McFadden 1974, 1981).
This study also compares the empirical models in terms of relative performance as described by
Fiebig et al. (2010). In light of contract theory, we design a choice experiment that includes
multiple contract designs and two hypothetical market conditions.
The experiment is conducted with agricultural producers located in six Midwestern states
(i.e. Illinois, Michigan, Minnesota, Missouri, Ohio, and Wisconsin) where the development of
renewable electricity industries has been pushed by public policies – Renewable Portfolio Standard
(RPS) mandate and other state programs21. We then examine agricultural producers’ willingness
to accept (WTA) various contract attributes and contractual designs under two hypothetical market
conditions.
The remainder of this article is organized as follows: section 2 reviews three theoretical
solutions to hold-up problems; section 3 presents the research methodology; section 4 analyzes
empirical results; section 5 discusses implications for bioenergy initiatives; and section 6
concludes.

2. Theoretical Framework
2.1. Industrial Organization Approach to Hold-up Problems
The industrial organization literature suggests that hold-up problems might be overcome
with either ‘specific performance contracts’ (Aghion et al. 1994) or ‘option contracts’ (Nöldeke
and Schmidt 1995). In order to examine these potential solutions, Aghion et al. (1994) examine
three variations of the following game:

21

In Michigan, the Renewable Energy Plan (REP) enacted in 2008 by the Michigan Public Service Commission
(MPSC) provides an example for state-level programs.

106



in 𝑡0 trading parties agree on initial price 𝑝0 , quantity 𝑞0 (called default option), and
allocation of bargaining power 𝛼, which might or might not be contingent on
announcements about the state of nature in period 𝑡3 ; buyer and seller invest in relationship
specific assets (𝑗 and 𝑖, respectively)



in period 𝑡1 state of nature 𝜃 is realized



in 𝑡2 trading parties observe 𝜃, 𝑗, and 𝑖



in 𝑡3 both parties announce a “state” (𝑖 ′ , 𝑗 ′ and 𝜃 ′ ) and renegotiate price and quantity if
necessary; trade occurs.
Aghion et al. (1994) consider three scenarios: (i) parties are risk-neutral and investments

are present, (ii) parties are risk-averse and investments are absent, and (iii) parties are risk-averse
and investments are present. They show that two contractual instruments are necessary and
sufficient to reach Pareto efficiency. The first instrument is the adequate choice of the default
option (𝑝0 , 𝑞0 ), which applies whenever ex post renegotiation fails (in 𝑡3 ). The second is the
allocation of all bargaining power to one party by making the default option sufficiently attractive
to the other party for all 𝑗, 𝑖, and 𝜃. For instance, allocating full bargaining power to the seller can
be achieved by penalizing the buyer if trade is delayed.
Renegotiation in period 𝑡3 ensures ex post efficiency because information symmetry
prevails (i.e. parties have made investments 𝑗 and 𝑖 and state of nature 𝜃 has realized). At this stage
renegotiation follows the bargaining model with discounting in which either party can unilaterally
impose a pre-specified trade (the default option) as in Binmore et al. (1986). Besides, monetary
penalties take place if no trade occurs until a pre-specified deadline. Aghion et al. (1994) add that

107

a given party has little incentive to deviate and announce a “state” different from the realized. They
show that a simple contract in which price, quantity, and allocation of bargaining power are not
contingent on announcements suffices to implement Pareto efficiency when parties are riskneutral. Conversely, a complex contract with a revelation mechanism (Maskin 1999) is necessary
to implement the first-best solution when parties are risk-averse, either in the presence or in
absence of specific investments.
Nöldeke and Schmidt (1995) use a different approach to reach the same conclusions as
Aghion et al. (1994). The authors demonstrate that every option contract results in the allocation
of all bargaining power to the buyer, meeting the second condition proposed by Aghion et al.
(1994). The first condition is obtained by appropriately choosing the option price 𝑘 = 𝑝1 − 𝑝0 ,
where 𝑝0 is the no-trade payment and 𝑝1 is the trade price to be paid by the buyer if seller supplies
quantity 𝑞. The game unfolds as follows:


parties sign the initial option contract (𝑝0 , 𝑝1 , 𝑞) in period 𝑡0



buyer and seller make investment decisions and state of nature is realized in period 𝑡1 ;
contract can be renegotiated at the end of period 𝑡1 once parties observe investment
decisions and state of nature



in period 𝑡2 buyer and seller decide whether to present renegotiation offers to the court in
order to adjust the initial option contract; trade occurs.
The authors demonstrate that renegotiation at the end of period 𝑡1 happens only if the

private decision of the seller is not socially optimal. If socially optimal, both parties are better off
sticking to the original agreement. If not socially optimal, the buyer has all the bargaining power
in the renegotiation game (i.e. any offer made by the seller can be withheld in period 𝑡2 ). Thus, the
buyer makes a new offer only if it is under his best interest to do so. The first scenario in which it

108

is profitable for the buyer to make a new offer is when 𝑝1 − 𝑝0 < 𝑐(𝜃, 𝑖) < 𝑣(𝜃, 𝑗), where 𝑐(∙,∙) is
the seller’s production cost and 𝑣(∙,∙) is the buyer’s value for the good. In this case, the buyer
maintains the no-trade price at 𝑝0 and raises the trade payment to 𝑝̃1 = 𝑝0 + 𝑐 + 𝜀 (𝜀 → 0 in
equilibrium) in order to make the seller slightly better off under the new contract. The second
scenario is when 𝑝1 − 𝑝0 > 𝑐(𝜃, 𝑖) > 𝑣(𝜃, 𝑗). In this case, trade is inefficient and the buyer must
offer a new set of prices to prevent the seller from delivering the good. Hence the seller increases
the no-trade price to 𝑝̃0 = 𝑝1 − 𝑐 + 𝜀, with 𝜀 converging to zero in equilibrium. Following that the
renegotiation game is ex post efficient, Nöldeke and Schmidt (1995) also show that there exists an
option contract (𝑝0 ,𝑘) that implements efficient investment for both parties. This final proof leads
to the conclusion that option contracts obtain Pareto efficiency. The downside of this model in
comparison to Aghion et al. (1994) is that it reaches Pareto efficiency under risk-neutrality only.
Although valid in a limited setting, the model of Nöldeke and Schmidt (1995) also provides a
solution to hold-up problems.

2.2. Institutional Economic Approach to Hold-up Problems
Under a different set of lenses, Klein (1996) proposes a mechanism to alleviate hold-up
problems. Without formally proving whether Pareto efficiency is obtained, Klein (1996) argues
that credible commitments or hostages (Williamson 1983) can be used to maximize the selfenforcing range of contractual relationships. Commitments might represent upfront payments from
party A to party B or co-investment in specific assets utilized to conduct B’s activity – more
generally defined as party A’s private enforcement capital. When well implemented, credible
commitments expand the self-enforcing range and equalize exposure of parties to unanticipated

109

events. In other words, it broadens the extent to which market conditions can change without
making potential hold-up gains larger than returns associated with honoring the agreement.
Starting from the fact that idiosyncratic investments might characterize both sides of a
relationship and that parties can seldom draft contingencies for all future states of nature, the author
suggests that a sufficiently large credible commitment leads parties to behave as expected for the
duration of the contract. Using empirical case studies to explain the usefulness of concepts (e.g.
self-enforcing range and private enforcement capital), Klein (1996) argues that the trading party
making the credible commitment, say the buyer, has little incentive to hold up its counterparty
because his private enforcement capital would be lost otherwise. In the opposite side of the
transaction, the seller is better off seeking future benefits rather than acting with guile for one-shot
gain plus zero returns thereafter (due to reputation loss). The use of credible commitments along
with a repayment mechanism of hostages increases transaction efficiency and minimizes the
probability of hold-up potential.
The literature of incomplete contracts has introduced several models illustrating that holdup problems can be manageable. From the empirical perspective, however, it is still important to
evaluate whether any of the proposed solutions is implementable in reality.

3. Research Methodology

This article conducts a choice experiment to test whether agricultural producers’
preferences over various contract provisions resemble contract designs capable of overcoming
hold-up problems. Using Lancaster’s (1996) approach to utility maximization as point of
departure, this study assumes that agricultural producers derive utility from various provisions in

110

contracts for trading biomass feedstock with electricity generation utilities. The experiment
attempts to reproduce decision situations to examine how individuals trade off provisions and
derive value from contract designs under different market conditions.
In the context of a stated preference experiment, we introduce two hypothetical future
market conditions for the developing bioelectricity industry. These treatments illustrate potential
market conditions such that the expected probability of hold-up problems differs. Namely, the first
market condition is characterized by high hold-up potential: entry into the biomass production
business implies investing in idiosyncratic assets, crop yield variability is high across seasons, and
producers may only trade with a single processor (i.e. little or no commercial alternative for the
produced feedstock). Comparatively, the second market condition is characterized by
comparatively low potential for hold-up: two processing plants compete for biomass supply, yield
expectations are stable across seasons, and redeployment of investments sacrifices little productive
value. Table 11 below summarizes the market conditions used in the choice experiment.

Table 11: Hypothetical market conditions
Market
conditions

MC#1

MC#2

Description
There is an electricity generating firm in your region willing to sign a production
contract for biomass feedstock. Price for biomass-based electricity between the
generating firm and distribution utilities is volatile. Agricultural operations
require acquisition of an expensive piece of machinery, which might be
redeployed in a future for its salvage value. Switchgrass yield is as variable as
corn/soybeans yield over the seasons.
There are two electricity generators willing to sign a production contract for
biomass feedstock in your region. There is also a well-established cash market
for biomass (i.e. fiber pellets manufacturing company), however the expected
returns from this market are lower than the expected returns from the electricity
market. Price for biomass-based electricity is relatively stable. Agricultural
operations require acquisition of an inexpensive piece of machinery, which might
be redeployed in a future period without loss of productive value. Variability of
switchgrass yield is low in comparison to corn/soybeans yield variation over
seasons.
111

Market conditions are intentionally designed to hold different levels of hold-up potential.
Condition one creates high hold-up potential for the sole biomass processing firm to expropriate
quasi-rents from feedstock producers in case of changes in market conditions. Without a secondbest market and little alternative use for productive assets, the only option available for the engaged
feedstock producer to control for hold-up potential is to sign a complex contract ex ante. Condition
two generates little chances for the processor to hold-up biomass producers. If the processor
behaves with guile, producers might trade biomass with alternative buyers (e.g. a fiber pellets
manufacturing company, for instance); or even exit the market by liquidating nonspecific
productive assets.
As part of the choice experiment setup, a menu of contracts with different capabilities to
deal with hold-up issues is offered to agricultural producers. The objective here is to create a quasicontinuum of contracts such that exposure to hold-ups varies as transactions unfold. Two fourlevel attributes and one two-level attribute were selected to characterize contracts: compensation
structure, base payment, and contract length. The levels used for compensation structure are
borrowed from the literature of hold-up problems and from habitual practice in agribusiness
settings. Base payment is included as a contract term because it might influence agricultural
producers engagement decision as well as their willingness to bear more or less exposure to holdup potential. Length characterizes contract designs because it tends to affect agricultural
producers’ utility expectations, especially in light of varying degrees of hold-up vulnerability
(Crocker and Masten 1988).
Compensation structure takes four levels. ‘Structure A’ represents the specific performance
contract as presented by Aghion et al. (1994), ‘Structure B’ represents the private enforcement
capital contract proposed by Klein (1996), ‘Structure C’ represents a share-cost contract, and

112

‘Structure D’ represents an acreage-based contract. Structures C and D were added to the CE
because of their frequent use to govern transactions in agribusiness settings.
The base payment term also has four levels: ‘$106/ton’, ‘$125/ton’, ‘$145/ton’, or
‘$170/ton’. These values assume an annualized yield of 2.83Mg/acre (7Mg/ha) as obtained in
agronomic tests conducted in multiple experimental stations by Perrin et al. (2008) and are
equivalent to annual farm revenues from switchgrass of ‘$330/acre/year’, ‘$390/acre/year’,
‘$450/acre/year’, and ‘$530/acre/year’, respectively. Based on the operating costs for switchgrass
obtained by Perrin et al. (2008) and Chen et al. (2014), these farm revenues define an equivalent
range of gross returns (i.e. sales minus operating costs) as found from conventional cropping
systems in Midwestern states. In other words, we assume that the hypothetical menu of contracts
offered to producers is competitive to returns from a corn-soybean rotation system, the most
common agriculture practice in the Midwest. Finally, contract length has two levels: ‘five years’,
or ‘ten years’. It is worth noting that the attributes and levels detailed above were previously
discussed with Michigan agricultural producers in focus group interviews (Ross and Signorini
2013). Table 12 provides further details about contract terms and levels.

113

Table 12: Attributes, levels and hypotheses tested in the choice experiment
Contract Terms

Levels
- Structure A
(specific performance)

Compensation
structure

- Structure B
(private enforcement capital)

- Structure C
(share-cost)
- Structure D
(acreage-based)

Description () / Expected Effect on stated preferences ()
 Parties agree ex ante on price and quantity to be delivered
(p0, q0) and renegotiate terms (p1, q1) after realization of state of
nature and investment levels. Producer invests in productive
assets if necessary. Biomass processor is penalized if trade is
delayed.
 Parties agree ex ante on price and quantity to be delivered.
Biomass processor and producer co-invest in productive assets
if necessary. Biomass processor loses ownership of assets in
case of contract breach due to hold-up.
 Parties agree ex ante on price. Biomass processor and
producer share establishment costs equally. Producer invests in
productive assets if necessary.
 Parties agree ex ante on price per acre farmed. Producer
invests in productive assets if necessary.
 Preference for complex contracts increases as market
situations are more vulnerable to hold-ups.

Base payment

- $106/ton ($330/acre)
- $125/ton ($390/acre)
- $145/ton ($450/acre)
- $170/ton ($530/acre)

 Higher base payments induce decision makers to bear more
exposure to hold-up problems.

Length of contract

- Five years
- Ten years

 Short-term contracts are preferable in situations of high
variability of yield and prices.

114

3.1. Hypotheses
Economic theory predicts agricultural producers to recognize that some market conditions
are more susceptible to hold-ups than others, and choose contracts accordingly. In concordance
with theory, this study hypothesizes that producers choose contract structures that are better
equipped to deal with hold-up problems when such problems are more likely to occur as in market
condition one. Conversely, we hypothesize that producers choose less complex structures when
market condition two (i.e. hold-up problems less likely to occur) characterizes the business
environment. We also hypothesize, however, that bounded rationality (Simon 1978) and traditions
might lead producers to make inefficient contract choices. Figure 8 below illustrates these
hypotheses. Light grey dots depict efficient decisions by agricultural producers and dark grey dots
exemplify inefficient choices. While mainstream economic theory suggests that fully rational
individuals’ preferences move from less to more complex contract structures as hold-up potential
increases in market conditions (light grey dots), unorthodox theory proposes that bounded rational
individuals might prefer contract structures according to previous experiences and traditions (dark
grey dots). Finally, this study hypothesizes that short-term contracts are preferred when market
conditions set the stage for hold-up problems to happen. We discuss each hypothesis in depth in
the next section. Hypotheses are summarized in table 12.

115

Figure 8: Primary hypothesis, illustrated

3.2. Survey design, data collection, and data
The experiment uses an optimal design and accounts for two-way interactions between
contract design attributes and market condition dummies. The OPTEX procedure in SAS was used
to identify 24 choice scenarios and divide them into eight blocks of three scenarios each22. The
blocking strategy was used to decrease the number of scenarios presented to agricultural producers
and to reduce statistical bias due to fatigue (Tonsor et al. 2005). Sequentially, blocks were treated
with both market condition one and market condition two such that each respondent faced three
choice scenarios under market condition one (high hold-up potential) and three choice scenarios
under market condition two (low hold-up potential). The order of market condition appearance
was randomized across survey versions to control for potential bias. All choice scenarios featured

22

The experimental design obtained an optimal D-efficiency value of 95.29.

116

two contract alternatives and an opt-out alternative. By including the opt-out alternative, we
remove the market participation assumption and recognize that agricultural producers might
choose not to engage in a production contract for biomass feedstock (Adamowicz et al. 1998).
Figures 9 and 10 in the appendix provide examples for the choice scenarios under information
treatment one and under information treatment two, respectively.
The data used in this study comes from agricultural producers’ responses to a mail survey
administered through USDA/NASS. The survey was sent out to 4,176 agricultural producers with
operations located in Illinois, Michigan, Minnesota, Missouri, Ohio, or Wisconsin. Prior to the
mailing, producers were screened and only those who meet the following three criteria were
considered eligible:
1. Producers must have responded to the 2012 Census of Agriculture. This criterion implies
that the producer is listed in the USDA/NASS data bank.
2. Agricultural producers must have reported farm income of $10,000.00 or more in 2012 to
be eligible as survey respondents. This criterion limits the population of producers to
professional individuals and excludes those who grow crops for leisure. Such criterion is
widely found in survey-based studies (Jensen et al. 2007, Paulrud and Laitila 2010, Hayden
2013).
3. Producers must have reported ‘value of sales’ above $5,000.00 in section 6 (‘field crops’)
and/or section 7 (‘hay and forage crops’) of the 2012 census. This criterion seeks to include
in the population farmers who produce grains and/or forage crops and seek economic gains
from those activities. It attempts to include diversified farms with multiple business units
and exclude non-diversified farms unless the single operation is related to field crops or
forage crops.

117

Eligible producers were randomly selected by USDA/NASS within each state. The
objective here was to obtain a balanced dataset, allowing consistent statistical analysis of
agricultural producers’ preferences and representation across the Midwest. Sampled producers
received two letters in the mail and a phone call. The first correspondence contained a cover letter
explaining the purpose of the study and a copy of the survey containing the choice experiment as
well as demographic questions for characterization of the sample of producers. The second
correspondence was sent two weeks after the first mailing with the aim to reinforce the importance
of the study. It included copies of the cover letter and the survey that were sent in the first mailing.
Finally, phone calls were made a week after the second mailing to those participants who had not
yet returned the surveys. In total 439 valid surveys were obtained, leading to a 10.5% response
rate. Almost all returned surveys were filled completely, yielding a statistical sample of 2,626
observations (439 X 6 choice scenarios, approximately) – 1,312 for information treatment one and
1,314 for information treatment two23. Survey sample is available from the authors upon request.
The dataset is distributed across states as follows: 14.75% of the choices over contract
designs were stated by agricultural producers from Illinois; 22.77% from Michigan; 14.55% from
Minnesota; 15.16% from Missouri; 14.47% from Ohio; and 18.32% from Wisconsin. Descriptive
statistics of demographic characteristics for the survey sample and population of farmers who meet
the research criteria are presented in Table 13.

23

The response rate achieved here is comparable to other studies. Khanna et al. (2014) obtained 424 usable responses
from 4800 surveys, leading to 8.8% response rate.

118

Table 13: Demographic statistics
Sample

Population

Operation size
- Illinois
- Michigan
- Minnesota
- Missouri
- Ohio
- Wisconsin

439 (persons)
63
98
65
68
64
81

656,917
109,123
76,095
106,537
138,339
115,743
111,080

Acreage
- Illinois
- Michigan
- Minnesota
- Missouri
- Ohio
- Wisconsin

438.6 ±694.8
512.50
397.40
531.55
280.04
522.88
423.19

451.98 ±746.37
617.73
419.52
533.85
477.12
355.81
318.34

(mean ± st. dev.)

Experience (years)
- Illinois
- Michigan
- Minnesota
- Missouri
- Ohio
- Wisconsin

30.16 ±14
29.13
31.88
30.89
29.94
29.25
29.19

30.3 ±15.06
31.27
31.18
29.69
30.25
30.10
29.75

(mean ± st. dev.)

Age

60.07 ±12.35
58.63
61.77
58.92
61.40
58.34
60.32

59.79 ±13.31
60.16
60.62
58.08
61.19
59.86
59.07

(mean ± st. dev.)

48% ± 36%
51%
39%
61%
39%
47%
53%

48% ±36%
52%
44%
56%
44%
41%
52%

(mean ± st. dev.)

- Illinois
- Michigan
- Minnesota
- Missouri
- Ohio
- Wisconsin
Farm Income Participat.
- Illinois
- Michigan
- Minnesota
- Missouri
- Ohio
- Wisconsin

119

Obs.
2626 (n)
387
598
382
398
380
481

Table 13 (cont’d)
Sample

Population

Gender
- male
- female

97%
3%

96%
3.9%

Retired?
- yes
- no

20%
80%

19.6%
80.4%

Renewable Energy
Enterprise?
- yes

4%

2.9%

Familiarity to contracts
- Never heard
- Heard but never used
- Used
- No response

Structure A
0.59
0.27
0.05
0.10

Structure B
0.66
0.24
0.00
0.10

Structure C
0.52
0.33
0.03
0.12

Structure D
0.50
0.30
0.10
0.10

It is worth noting that the obtained sample appears to be a good representation of the
population. Table 13 above indicates that the summary statistics of experience, age, farm income
participation, gender, retirement status, and engagement of producers in renewable energy
enterprises are similar for sample and population. There are slight divergences for harvested acres,
however. Although sample and population present similar overall mean for harvested acres,
sampled farmers from Illinois and Missouri seem to have harvested 21% and 70% less than the
population of their corresponding states. Sampled farmers from Ohio and Wisconsin, on the other
hand, harvested 32% and 25% more than their respective state populations. These differences in
acreage must be considered prior to extrapolating research findings to the entire Midwest region.
Table 13 also reports the level of familiarity with contract structures expressed by survey
respondents. The contract structures included in the choice experiment are unknown to at least half
of the individuals. Acreage-based contract (structure D) is the most familiar with 40% of the

120

respondents having either heard or used it before. Share-cost contract (structure C) comes second
with 36% of familiarity. Specific performance contract (structure A) reaches 32% of familiarity
and comes third. Finally, private enforcement capital (structure B) is the most obscure contract
type with 24% of the respondents having heard about it but no past users.

3.3. The Model
The three empirical models used to analyze agricultural producers’ preference data are
based on the random utility maximization (RUM) model (McFadden 1974, 1981). The McFadden
model assumes that agricultural producers maximize utility by choosing from a set of finite
production contracts when presented with hypothetical decision scenarios. In mathematical
notation, let 𝒁𝑠 be the vector of 𝑘 quantifiable attributes in contract design s and 𝒀𝑡 be a vector of
market condition characteristics in decision scenario 𝑡. Let utility to agricultural producer 𝑖 from
contract design 𝑠 in market scenario 𝑡 be given by the following indirect utility function:
𝑉𝑖𝑠𝑡 = 𝑣𝑖𝑠𝑡 (𝒁𝑠 𝒀𝑡 ) + 𝜀𝑖𝑠𝑡
(1)
The first element in the right hand side of equation (1) is a deterministic term representing
the set of information available to both decision makers – agricultural producers in this case – and
the analyst. The second element is only observable by agricultural producers, and from the
analyst’s perspective it is equivalent to a stochastic term. Deterministic and stochastic terms are
assumed independent where the latter can be argued as decision makers’ heterogeneity in
preferences (Keane 1997).
Vector 𝒁𝑠 in the deterministic term includes the following 𝑘 attributes: (i) compensation
structure, (ii) base payment, and (iii) length of contract – with randomized levels within contract

121

designs. Vector 𝒀𝑡 includes characteristics of hypothetical market conditions such as (i) number
of processing plants in the market, (ii) degree of asset specificity, (iii) expected price variability
for biomass feedstock, and (iv) expected yield variation. We use a single dummy variable to
capture variation in these attributes across the two market conditions24. It is worth stressing that
the dummy variable associated with market characteristics is interacted with the vector of
contract attributes in the empirical model. The reason for such interaction is to examine whether
market conditions and socio-economic characteristics explain preference heterogeneity for
contract attributes.
The model also assumes that the deterministic term 𝑣𝑖𝑠𝑘𝑡 (𝒁𝑠𝑘 𝒀𝑡 ) is linear in parameters:
𝑣𝑖𝑠𝑡 (𝒁𝑠 𝒀𝑡 ) = 𝜷𝒁𝑠 𝒀𝑡
(2)
where 𝜷 is a vector of utility weights for interactions between market conditions and contract
attributes. Plugging (2) into (1) we have:
𝑉𝑖𝑠𝑡 = 𝜷𝒁𝑠 𝒀𝑡 + 𝜀𝑖𝑠𝑡
(3)
Therefore, we assume that agricultural producer 𝑖 chooses the contract alternative 𝑠 if
𝑉𝑖𝑠∗𝑡 > 𝑉𝑖𝑠𝑡 ∀𝑠 ∗ ≠ 𝑠. Consequently, the probability of producer 𝑖 choosing contract design 𝑠 is
given by:
𝑃𝑟𝑖𝑠∗ 𝑡 = 𝑃𝑟(𝑉𝑖𝑠∗𝑘𝑡 + 𝜀𝑖𝑠∗𝑡 > 𝑉𝑖𝑠𝑡 + 𝜀𝑖𝑠𝑡 ; ∀𝑠 ∗ ≠ 𝑠)
(4)

24

This approach does not harm interpretation of CE results. As the characterizing levels of market conditions are
nonrandom across decision scenarios, the use of a dummy variable is appropriate and measures whether agricultural
producers’ preferences differ between market conditions one and two.

122

However, because the analyst cannot observe 𝜀𝑖𝑠𝑡 , an informed assumption regarding its
distribution must be made so estimates resulting from stated decisions can be obtained. A
convenient and often made set of assumptions regarding 𝜀𝑖𝑠𝑡 is that it is distributed identically and
independently across individuals and follows a Glumbel distribution (𝜀𝑖𝑠𝑡 ~ i.i.d. Glumbel or
extreme value type I, in short). Thus, the basic multinomial logit (MNL) model (McFadden 1974)
for the probability that agricultural producer 𝑖 chooses contract design 𝑠 ∗ when presented with
decision scenario 𝑡 can be written as:
𝑃𝑟𝑖𝑠∗𝑘𝑡 =

𝑒𝑥𝑝 [𝜶𝒁𝑠∗ 𝒀𝑡 ]
∑𝑠≠𝑠∗ 𝑒𝑥𝑝 [𝜶𝒁𝑠 𝒀𝑡 ]
(5)

3.4. Refinements of the Multinomial Logit Model
The recent discrete choice literature has suggested three main refinements to the McFadden
model: (i) addition of heterogeneity in preferences for observed attributes; (ii) inclusion of
preference heterogeneity for unobserved attributes (Keane 1997); and (iii) addition of both types
of heterogeneity into a model capable of estimating parameters in WTA-space. The first suggestion
has led scholars to propose and apply the random parameters logit (RPL) model, which is appealing
to practitioners due to its ease of use (Fiebig et al. 2010). In this case, utility weights are assumed
to be random variables following a pre-determined distribution. When applied, the RPL model
modifies equation 3 as follows:
𝑉𝑖𝑠𝑡 = (𝜷 + 𝜸𝑖 )𝒁𝑠 𝒀𝑡 + 𝜀𝑖𝑠𝑡 ,
(6)
where 𝜷 become vector of mean utility weights on observed attributes, and 𝜸𝑖 is vector of
individual 𝑖-specific deviations from the mean weight.

123

The second refinement serves as motivation for the scaled multinomial logit (S-MNL)
model, which modifies equation 3 as below:
𝑉𝑖𝑠𝑡 = 𝜷𝒁𝑠 𝒀𝑡 +

𝜀𝑖𝑠𝑡
⁄𝜎𝑖 ,
(7)

where 𝜎𝑖 is individual-specific and represents different degrees of unobserved heterogeneity for
individual respondents across multiple decision observations (Fifer et al. 2014). The scale
parameter 𝜎 is commonly normalized to one in order to allow model estimation (Fiebig et al.
2010). The S-MNL model can be re-written in the following form:
𝑉𝑖𝑠𝑡 = (𝜷𝜎𝑖 )𝒁𝑠 𝒀𝑡 + 𝜀𝑖𝑠𝑡
(8)
The third refinement to the MNL model has received recent attention in the literature due
to its flexibility and capability of allowing estimation in WTA-space. Incorporating both types of
preference heterogeneity into the same model – deviations from mean utility weights and 𝑖-specific
scale heterogeneity – has led scholar to the generalized multinomial logit (G-MNL) model (Fiebig
et al. 2010). More specifically, the G-MNL model combines equations 6 and 8 and is presented in
equation 9.
𝑉𝑖𝑠𝑡 = [𝜷𝜎𝑖 + 𝛿𝜸𝑖 + (1 − 𝛿)𝜸𝑖 𝜎𝑖 ]𝒁𝑠 𝒀𝑡 + 𝜀𝑖𝑠𝑡
(9)
Scale factor 𝜎𝑖 and the deviations to mean utility weights 𝜸𝑖 take the same descriptions
provided above. The new variable 𝛿 captures the proportions of preference heterogeneity
associated with individual scale factor 𝜎𝑖 or individual-specific deviation 𝛾𝑖 25.

25

For a comprehensive treatment of the G-MNL model, please refer to Fiebig et al. (2010).

124

In addition to these major refinements, choice behavior specialists have been active in
numerous discussion fronts regarding the RPL, S-MNL, and G-MNL models. To mention a few,
authors have discussed whether or not observed attributes should be correlated (Revelt and Train
1998, Scarpa and Del Giudice 2004). While most scholars agree that adding this flexibility tends
to enhance goodness of fit, others explain that complex models may prevent its implementation
due to the rapid increase in number of parameters (Louviere et al. 2000, Train 2009). In other
words, allowing parameters to correlate freely in complex models forces the number of parameters
to increase rather fast, making estimation computationally infeasible.
Authors have also argued that utility weights on pecuniary parameters should be treated as
random variables, just as the weights related to other attributes are in RPL and G-MNL models.
The underlying reason is that unobserved characteristics are likely to influence individuals to
respond differently to revenue (or to expense in case of willingness-to-pay estimation) (Fiebig et
al. 2010). Train (2009) contends that models with all random coefficients can be empirically
problematic as WTA/WTP values derived from two random coefficients do not converge to
identifiable distributions.
In a different study Hess and Rose (2012) express concerns about the G-MNL model
exclusively. The authors argue that scale heterogeneity is equivalent to heterogeneity in individual
utility weights, and including both types of heterogeneity into the same model (as in equation 8)
is misleading. They explain that the relative gain in fit obtained through the G-MNL model – when
compared to the RPL model – is primarily associated with using more flexible distributions, rather
than capturing scale heterogeneity. Louviere et al. (2008) also criticize the G-MNL model, as they
have found that heterogeneity in preferences is better captured by S-MNL than by G-MNL.

125

3.5. The Empirical Models
In this study, three empirical models – RPL, S-MNL, and G-MNL – are used to estimate
WTA for contract design attributes. Either model has meaningful advantages over the traditional
MNL model because they recognize that subjects are heterogeneous and might have different
appreciation for specific attributes. The modeling question that remains unanswered is of how to
incorporate agricultural producers’ preference heterogeneity – through deviations from mean
utility weights, through individual-specific scalar, or incorporating both types at once. Following
the method proposed by Fiebig et al. (2010), we assess model performance by examining the
likelihood improvement that is attained by including preference heterogeneity into the MNL
model.
The estimation function derived from the RPL model – equation 5 above – assumes that
utility weights on non-pecuniary attributes are random parameters and follow a normal
distribution. The utility weight on base payments is assumed to be nonrandom to allow
identification of WTA estimates (Goett et al. 2000, Morey and Rossmann 2003). Hence, the
probability that agricultural producer 𝑖 chooses contract design 𝑠 ∗ when presented with scenario 𝑡
becomes:
𝑃𝑟𝑖𝑠∗𝑡 = ∫

𝑒𝑥𝑝[𝜷𝒁𝑠∗ 𝒀𝑡 ]
𝑔(𝛽)𝑑𝛽
∑𝑠∈𝑆 𝑒𝑥𝑝[𝜷𝒁𝑠 𝒀𝑡 ]
(10)

where the distribution of random parameters 𝑔(∙) is normal; 𝜷 = 𝜷 + 𝜸𝒊 26.

26

Normal distribution is assumed because it provides the best fit to our data.

126

The S-MNL model assumes that utility weights are scaled proportionately across
agricultural producers. Based on equation 7, the probability of agricultural producer 𝑖 choosing
contract design 𝑠 ∗ in the decision scenario 𝑡 is:
𝑃𝑟𝑖𝑠∗𝑡 =

𝑒𝑥𝑝 [𝜷𝒁𝑠∗ 𝒀𝑡 ]
∑𝑠≠𝑠∗ 𝑒𝑥𝑝 [𝜶𝒁𝑠 𝒀𝑡 ]
(11)
2

where 𝜶 = 𝜶𝜎𝑖 ; 𝜷 = 𝜷𝜎𝑖 ; 𝜎𝑖 = 𝑒𝑥𝑝 (−𝜏 ⁄2 + 𝜏𝜔𝑖 ); and 𝜔𝑖 takes the standard normal
distribution.
The G-MNL model incorporates scaled heterogeneity in preferences as well as deviation
from mean utility weights. Derived from equation 8, the probability of agricultural producer 𝑖
choosing contract design 𝑠 ∗ in the decision scenario 𝑡 becomes:
𝑃𝑟𝑖𝑠∗𝑡 = ∫

𝑒𝑥𝑝[𝜷𝒁𝑠∗ 𝒀𝑡 ]
𝑔(𝛽)𝑑𝛽
∑𝑠∈𝑆 𝑒𝑥𝑝[𝜷𝒁𝑠 𝒀𝑡 ]
(12)

where 𝜷 = 𝜷𝜎𝑖 + 𝛿𝜸𝑖 + (1 − 𝛿)𝜸𝑖 𝜎𝑖 ; and 𝛿 is a scalar to be estimated. Similar to the RPL model,
coefficients on non-pecuniary attributes are assumed to be normally distributed while coefficients
on base payments – interacted with dummy variables for market condition one and market
condition two – are fixed to allow identification of WTA values. Although choice behavior
specialists are enthusiastic with the possibility of estimating preferences in WTA-space, the
estimates we derive from the G-MNL model are in the usual preference-space. Because the
primary objective of this study is to determine whether market conditions (high vs. low holdup
potential) influence preference for industrial organization-related contracts or institutional

127

economics-related contracts, the models contain two interacted parameters for the pecuniary
attribute, which cannot be modeled at once for direct estimation of WTA values27.
Correlation of coefficients is assumed absent due to the complexity of the model. Given
that the model considers two-way interaction effects between market conditions and contract
design attributes and between demographic characteristics and contract attributes, allowing
correlation of coefficients would add 60 plus parameters to be estimated which is computationally
infeasible.
Willingness to accept values for contract design attributes are sequentially computed for
each hypothetical market condition. Following the literature, WTA values are the ratio of the
estimated coefficient on the attribute of interest to the coefficient on base payment, times two due
to our use of effects coding (Lusk et al. 2003). Statistical variability in WTA estimates are
calculated using the parametric bootstrapping technique as proposed by Krinsky and Robb (1986).
For each contract design attribute, WTA estimates are compared across market conditions using
the complete combinatorial method proposed by Poe, Giraud, and Loomis (2005).

4. Empirical Results

Estimates for two RPL models are reported in Table 14. The left hand side column presents
estimates for the main effects model while the right hand side shows estimates for the model with
interactions between contract attributes and market condition dummies. Estimates from the main
effects model show that all attributes included in the experiment influence decision over contract

27

Different strategies were attempted to overcome this limitation. Separating the dataset into two sub-sets - one for
each market condition – did not yield a solution. The models do not perform and fail to return robust estimates when
sub-sets of the original dataset are utilized.

128

designs. Base payment is a valuable attribute and tends to increase agricultural producers’ utility.
Contract structures that resemble specific performance contracts (structure A), private enforcement
capital contracts (structure B), or share-cost contracts (structure C) tend to influence agricultural
producers’ utility negatively when compared to acreage-based contracts (structure D). Finally,
long-term contracts (i.e. 10 years long) are less preferred than short-term contracts (i.e. 5 years
long). The main effects model also indicates that there is strong heterogeneity in producers’
preferences for non-pecuniary attributes, as shown by the statistically significant estimates of
‘standard deviation’ next to parameter estimates.
Results from the RPL model with two-way interactions between contract attributes and
market conditions capture some of the observed heterogeneity. Coefficients on utility weights are
significant at 1% confidence level except for share-cost contract structures under high levels of
vulnerability. With exception made to share-cost contract structures, non-pecuniary attributes
continue to show statistically significant heterogeneity in preferences. Mean utility weights,
nevertheless, lead to interesting interpretations. When the contract engagement decision is
evaluated in market condition one (high hold-up potential), base payment appears to be slightly
less important than it is under conditions of low hold-up potential. Acreage-based contract
structure attains agricultural producers’ preference regardless of the market condition because of
the negative coefficients in alternative contract structures. Coefficients on alternative contract
structures also indicate that decision makers are less reluctant to adopt complex contracts if the
decision is made under high hold-up potential, and more reluctant if hold-ups do not constitute
credible threats. More specifically, parameter estimates reported in Table 14 indicate that mean
utility weights on structure A and structure B under high hold-up potential are systematically larger
than coefficients on the same structures but under low hold-up potential, suggesting that preference

129

for complex contracts increases in probability of hold-up problems to occur. Agricultural
producers also seem indifferent to the possibility of signing share-cost contracts with biomass
suppliers when market conditions set the stage for high hold-up potential. Finally, coefficients on
contract length are similar and indicate preference for short-term contracts regardless of the market
condition.
Despite the clear preference for acreage-based contract structures, these initial results
suggest that decision makers weigh market conditions against financial gains and seek appropriate
mechanisms to engage in transactions. Results also indicate that one cannot reject the hypothesis
that having the possibility of signing a share-cost contract does not influence agricultural
producers’ preference under market conditions with high hold-up potential, given that its
parameter estimate is not significant when high hold-up potential characterizes the market. Under
low hold-up potential, however, share-cost structures seems to influence preference. Other
parameter estimates provide insights on preferred structures if high hold-up potential characterizes
the market condition. Conditional on high potential, producers would give similar consideration
to specific performance contract (structure A) or private enforcement capital contract (structure B)
to deter opportunism. Statistical difference in parameter magnitude is discouraged at this point
(Scarpa and Del Giudice 2004). Statistical difference in preference for complex contracts is
discussed below once WTA values are computed.

130

Table 14: Parameter estimates from RPL models (main effects vs. two-way interactions)
Variable
Base Payment
Low hold-up potential
High hold-up potential
Contract Structure
- Specific Performance
Low hold-up potential
High hold-up potential

RPL model

Standard
Deviation

RPL model
(partitioned by
market condition)

Standard
Deviation

0.032 (0.002)**
0.033 (0.004)**
0.029 (0.003)**
-0.435 (0.071)**

- Private Enforc. Capital
Low hold-up potential
High hold-up potential

-0.524 (0.074)**

- Share-Cost
Low hold-up potential
High hold-up potential

-0.226 (0.061)**

Long-term length
Low hold-up potential
High hold-up potential

-0.429 (0.064)**

Opt Out
Low hold-up potential
High hold-up potential

4.785 (0.427)**

Log likelihood function
AIC

-2275.59
4116.2

0.626 (0.099)**
-0.490 (0.098)**
-0.395 (0.104)**

0.524 (0.216)*
0.762 (0.191)**

-0.631 (0.103)**
-0.389 (0.105)**

0.672 (0.174)**
1.002 (0.172)**

-0.306 (0.087)**
-0.139 (0.083)

0.107 (0.567)
0.103 (0.343)

-0.440 (0.088)**
-0.465 (0.870)**

0.795 (0.159)**
0.529 (0.201)**

4.839 (0.628)**
4.297 (0.586)**

4.116 (0407)**
4.251 (0.423)**

0.820 (0.093)**

0.223 (0.171)

0.715 (0.083)**

4.246 (0.305)**

-2269.45
4582.9

Notes: The models were estimated using Nlogit 5.0, with Halton draws and 1000 replications for simulated probability.
Standard errors are reported in parentheses.
* and ** indicate statistical significance at 5% and 1% level respectively.
A likelihood-ratio (LR) test was performed. Preference data fit both models similarly. We cannot reject the main effects
model in favor of the two-way interactions model.

131

Both S-MNL and G-MNL models with two-way interactions show similar relations among
coefficients (Table 15). Base payment under high hold-up potential influence utility at a lesser
degree than base payment under low hold-up potential. Agricultural producers prefer acreagebased contract structures to any other structure – equipped to deal with hold-ups or not – in either
market condition. In contrast to the RPL model, producers appear to consider the share-cost
structure a feasible alternative under high hold-up potential. Looking specifically at complex
contracts (i.e. structure A vs. structure B) under high potential, S-MNL and G-MNL offer opposite
results. While S-MNL suggests that producers prefer private enforcement capital contracts to
specific performance contracts, G-MNL shows that producers would be more reluctant to accept
the former and more willing to accept the latter. In any case, both models do not reject the
hypothesis that producers’ preference for complex contracts increases when probability of holdups increases (i.e. coefficients on structure A and structure B increase as market conditions change
from low to high hold-up potential). Finally, coefficients on contract length confirm that producers
prefer shot-term contracts to long-term contracts across all market conditions. Parameter estimates
from the S-MNL and G-MNL models are reported in table 15.

132

Table 15: Parameter estimates from S-MNL and G-MNL models with two-way
interactions
S-MNL model
(partitioned by
market conditions)

G-MNL model
(partitioned by
market conditions)

Base Payment
Low hold-up potential
High hold-up potential

0.020 (0.002)**
0.017 (0.002)**

0.033 (0.003)**
0.031 (0.003)**

Contract Structure
- Specific Performance
Low hold-up potential
High hold-up potential

-0.322 (0.056)**
-0.232 (0.056)**

-0.852 (0.219)**
-0.724 (0.224)**

0.814 (0.296)**
1.054 (0.307)**

- Private Enforc. Capital
Low hold-up potential
High hold-up potential

-0.385 (0.055)**
-0.199 (0.055)**

-1.271 (0.295)**
-0.826 (0.246)**

0.834 (0.281)**
1.305 (0.342)**

- Share-cost
Low hold-up potential
High hold-up potential

-0.208 (0.059)**
-0.119 (0.059)*

-0.557 (0.189)**
-0.351 (0.164)*

0.158 (0.718)
0.003 (4.125)

Long-term length
Low hold-up potential
High hold-up potential

-0.256 (0.045)**
-0.249 (0.044)**

-0.725 (0.190)**
-0.832 (0.192)**

0.979 (0.278)**
0.582 (0.326)

Opt Out
Low hold-up potential
High hold-up potential

3.443 (0.282)**
2.885 (0.275)**

7.465 (1.201)**
7.156 (1.105)**

1.195 (2.124)
1.036 (3.274)

𝜏
𝜎𝑖
Log likelihood function
AIC

0 (0.221)
1 (0.005)**
-2701.29
5428.6

0.971 (0.111)**
0.988 (1.158)
-2260.28
4568.6

Variable

Standard
Deviation

Notes: The models were estimated using Nlogit 5.0, with Halton draws and 1000 replications for
simulated probability. Standard errors are reported in parentheses.
* and ** indicate statistical significance at 5% and 1% level respectively.

Despite robust findings across the three empirical models, it is important to assess model
performance when decision makers exhibit preference heterogeneity. Fiebig et al. (2010) suggest
that model performance can be assessed by looking at the percentage log-likelihood improvement
in going from the MNL model – where preference homogeneity is assumed – to either RPL, SMNL, or G-MNL. Including heterogeneity via deviations from mean utility weights improves

133

model performance faster than via individual scale heterogeneity. Taking the MNL model with
two-way interactions as base for comparison, RPL improves fit by 16.01% while S-MNL improves
fit by 0.03%. Allowing even greater flexibility and modeling both types of heterogeneity – as in
the G-MNL – leads to a 16.35% performance improvement. Although results indicate that G-MNL
outperforms the other modeling strategies, it is important to consider the argument offered by Hess
and Rose (2012) regarding the use of both deviation from mean utility weights and individual
scalar at once. As mentioned above the authors explain that G-MNL tends to perform better due
to its increased flexibility as opposed to its capability of capturing different types of heterogeneity
in preferences. Empirical results obtained here seem to corroborate with Hess and Rose (2012)
given minimal performance gain obtained when individual-specific deviation is modeled alone.
In any case interpreting the magnitudes of parameter estimates is discouraged because only
relative parameter values matter (Scarpa and Del Giudice 2004). The conventional alternative in
this circumstance is to estimate agricultural producers’ WTA based on parameter estimates. Tables
16 and 17 report mean WTA values, 95% confidence intervals, and statistical evidence for
difference in preferences under market conditions one and two (high hold-up potential and low
hold-up potential, respectively).

134

Table 16: Mean Willingness-to-accept (WTA) and 95% confidence intervals
Variable
Contract Structure
- Specific Performance
Low hold-up potential
High hold-up potential

RPL
main effects

*1

-27.01 [-37.42, -17.84]



- Private Enforc. Capital
Low hold-up potential
High hold-up potential

-32.48 [-43.78, -22.78]

- Share-cost
Low hold-up potential
High hold-up potential

-14.10 [-22.57, -6.53]

Long-term length
Low hold-up potential
High hold-up potential

-26.59 [-35.45, -18.65]

RPL
two-way interactions

*1

-29.41 [-42.42, -17.86]
-26.96 [-42.70, -12.80]




-37.84 [-50.65, -26.17]
-26.53 [-42.96, -12.55]




-18.44 [-29.95, -8.35]
-9.60 [ -21.73, 1.77]




-31.63 [-44,48, -20.32,]
-26.43 [-37.18, -16.46,]








Notes: Confidence intervals (presented in brackets) were simulated using 1000 Krinsky-Robb
replications.
Different symbols indicate statistical difference at 5% confidence level across contract structures (for maineffects RPL) or within attribute (for two-way interactions RPL).
*1: Complete combinatorial test results.

Producers expressed overall preference for acreage-based contracts. Values of WTA for
contract structures computed from coefficients of the main effect RPL model presented in Table 6
below indicate that producers would require on average an additional $27.01/ton to sign a specific
performance contract. Alternatively, producers would require $32.48/ton extra to engage in a
private enforcement capital contract or an additional $14.10/ton to enter in a share-cost contract
agreement when compared to an acreage-based contract, ceteris paribus. In fact, the WTA estimate
for share-cost structure is statistically different at 5% level from estimates computed for complex
contract structures. Computations report that WTA values for specific performance structure and
private enforcement capital structure are not statistically different at 5% of confidence level when
market conditions are not formally modeled.

135

This result seems well aligned with producers’ revealed familiarity to contract types (Table
3), suggesting that bounded rationality (Simon 1978) and habitual practices play important roles
in agricultural producers’ business decisions. Acreage-based contract was reported as the most
familiar contract: 40% of survey respondents have either heard or used this type of contract.
Producers have also shown good familiarity to share-cost contracts. Out of 439 subjects, 36% have
heard or used this contract structure. When it comes to complex contracts, however, experience
with specific performance contract and private enforcement capital contract falls to 32% and 24%,
respectively.
Values of WTA differ across market conditions when interactions are modeled. Although
producers prefer the contract type that is mostly familiar to them (i.e. acreage-based contract),
WTA values indicate increasing acceptance for complex contracts as hold-up potential increases.
For specific performance structure, agricultural producers would require $26.96/ton under high
hold-up potential and $29.41/ton under low hold-up potential. For private enforcement capital
structure, producers would demand $26.53/ton to sign it under high hold-up potential versus
$37.84/ton under low hold-up potential. This suggests that producers do understand the importance
of using contracts better equipped to deal with hold-up problems when such problems are more
likely to happen. The latter result combined with the smaller coefficient on base payment under
high hold-up potential (versus low hold-up potential – table 5) leads to interesting interpretations.
Survey respondents have not only shown that financial gains have smaller influence on utility
when market conditions set the stage for hold-ups, but have also shown that complex contracts are
comparatively more attractive under high hold-up potential than they are under low hold-up
potential.

136

Nevertheless, statistical results show that agricultural producers perceive specific
performance structure and private enforcement capital structure similarly under high hold-up
potential, given that WTA values for these structures are not statistically different. The same
comment is valid under low hold-up potential. Willingness to accept for specific performance
structure and WTA for private enforcement capital structure are not statistically different at 5% of
confidence level.
For the sake of completeness, producers request an additional of $9.60/ton to sign sharecost contract under high hold-up potential and $18.44/ton to sign the same contract structure under
low probability of hold-ups to occur. It is worth noting that the 95% confidence interval for the
mean WTA of the share-cost contract under hold-up potential includes zero, so the result is not
statistically significant.
Regarding contract length, agricultural producers prefer short-term contracts regardless of
the market condition. Producers would require $26.59/ton on average to accept a 10-year contract
rather than a 5-year. Conditional on hold-up potential, producers would demand an additional
$26.43/ton to engage in a 10-year contract under high hold-up potential and $31.63/ton to sign a
10-year contract under low hold-up potential. While these magnitudes are consistent with theory
(Crocker and Masten 1988), which indicates that decision makers prefer short-term contracts to
protect themselves against high levels of uncertainty; our results should be used with discretion
because coefficients are not statistically different under low and high hold-up potential.

137

Table 17: Mean Willingness-to-accept (WTA) and 95% confidence intervals
Variable

S-MNL
Two-way interactions

*1

G-MNL
two-way interactions

*1

Contract Structure
- Specific Performance
Low hold-up potential
High hold-up potential

-31.69 [-45.41, -19.82]
-27.27 [-42.86, -13.45]




-51.66 [-79.41, -26.33]
-46.52 [-76.73, -18.07]




- Private Enforc. Capital
Low hold-up potential
High hold-up potential

-37.92 [-52.24, -25.80]
-23.50 [-39.02, -10.01]




-76.94 [-114.51, -42.31]
-53.47 [-88.78, -21.36]




- Share-cost
Low hold-up potential
High hold-up potential

-20.48 [-33.82, -8.51]
-14.10 [-29.39, -0.54]




-33.77 [-56.94, -11.94]
-22.80 [-45.04, -2.16]




Long-term length
Low hold-up potential
High hold-up potential

-25.20 [-34.94, -16.51]
-29.27 [-41.33, -18.68]




-44.03 [-69.15, -21.22]
-53.61 [-81.44, -28.23]




Notes: Confidence intervals (presented in brackets) were simulated using 1000 Krinsky-Robb
replications.
Different symbols indicate statistical difference at 5% confidence level within each attribute.
*1: Complete combinatorial test results.

Table 17 presents the mean estimates of agricultural producers WTA for contract attributes
derived from S-MNL and G-MNL coefficients as well as the 95% confidence intervals around
those means. The interpretation and discussion that follows consider results obtained from the GMNL model only because of the relatively low modeling performance of S-MNL.
Results indicate that WTA values calculated from G-MNL estimates are dramatically
larger than values calculated from RPL, despite similar performance indicators for the underlying
models. Using G-MNL estimates, specific performance structure is consistently preferable to
private enforcement capital in either market condition. It would take $46.52/ton to drive preference
of producers to specific performance structure and $53.47/ton to stimulate preference for private
enforcement capital when high hold-up potential characterizes the marketplace. Under low holdup potential, agricultural producers would settle for a specific performance structure if $51.66/ton
was added to the base payment of an acreage-based contract, or private enforcement capital

138

structure if $76.94/ton was offered in addition to the base payment of an acreage-based structure.
The complete combinatorial test (Poe et al. 2005) indicates, nevertheless that WTA values are not
statistically different from each other at 5% level. Test were performed to compare specific
performance under high hold-up potential versus low hold-up potential; private enforcement
capital under high hold-up potential versus low hold-up potential; specific performance versus
private enforcement capital under high hold-up potential; and specific performance versus private
enforcement capital under low hold-up potential.
Estimate magnitudes, nevertheless, suggest that our original hypothesis about preferences
for complex contracts is correct. Preference for complex contracts increases when the probability
of hold-up problems to happen also increases. In concordance to the discussion conducted above
(based on RPL coefficients), decision makers seem to understand the importance of adopting
complex contracts under high potential to hold-ups, despite the overall preference for acreagebased contract structures. On average, producers would give up 10% base payment if market
conditions changed from low to high potential and the contract had a specific performance
structure. Similarly, decision makers would accept 30.5% less to continue to use a contract derived
from the institutional economics literature – private enforcement capital contract – if the
probability of hold-up occurrence increased.
Other main results obtained from the RPL model are supported here. Acreage-based
contract structure is preferable over any alternative structure, followed by share-cost contracts.
Values of WTA for share-cost structure are statistically significant at 5% confidence level or less.
The share-cost structure is preferred if producers are offered $22.80/ton on average under high
hold-up potential, ceteris paribus. Producers would also accept a share-cost contract under low
probability of hold-ups to happen if base payment was $33.77/ton higher than acreage-based

139

contract. The complete combinatorial test points out that there is no statistical difference of WTA
for share-cost under low hold-up versus high hold-up potential; no statistical difference between
share-cost structure and complex structures when probability of hold-ups is low; and mixed results
under high hold-up potential. WTA for share-cost structure is statistically different at 5% level
from private enforcement capital structure under high hold-up potential; but it is statistically
indifferent from specific performance structure also under high-hold-up potential.
The complete combinatorial test also indicates that estimates of WTA for contract length
are not statistically different between market conditions with high hold-up potential or low holdup potential. Producers do prefer short-term contracts regardless of the level of vulnerability to
hold-ups. Agricultural producers would sign a 10-year contract under low potential if $44.03/ton
on average were added to the amount offered in a short-term contract. Producers would finally
require $53.61/ton extra to engage in a long-term contract under high potential of hold-ups.
The hypothesis that producers prefer short-term contracts under high hold-up potential is
also confirmed here. Another interesting interpretation is that values of WTA for contract length
are increasing in probability of hold-ups, meaning that aversion to long-term contracts increases
as market conditions lean towards occurrence of hold-ups. This latter finding also confirms the
literature of contract duration (Crocker and Masten 1988).
In light of these results, it is important to emphasize that theoretical predictions regarding
contract adoption are partially satisfied at best. Decision makers seem to strongly rely on previous
experience and traditions, leading to preferences that might not be appropriate to deal with future
transaction problems. Acreage-based structure emerges not only as the most familiar contract type
but also as the preferred structure regardless of the hold-up potential. Theoretical predictions are
marginally supported when results indicate that WTA for complex contracts increases in exposure

140

to hold-up problems. Results also support the theory that short-term contracts are preferred when
high levels of uncertainty characterize the market.

5. Implications for Bioelectricity Initiatives

This study illustrates that Midwest U.S. agricultural producers prefer an acreage-based
contract structure and short-term contracts to facilitate transactions in the bioelectricity market.
Empirical findings derived from RPL estimates are consistent with our original hypothesis and
indicate that preference for complex contracts (i.e. specific performance contracts and private
enforcement capital contracts) increases in probability of hold-up occurrence, although not
statistically significant at 5% confidence level. Results from G-MNL estimates further support
these findings and provide deeper insights. Although not statistically significant, specific
performance structure seems preferable to private enforcement capital structure. The estimated
WTA for long-term contracts derived from G-MNL coefficients also suggest that aversion to longterm contracts increases from low to high hold-up potential; although WTA values are not
statistically different at 5% confidence level.
These results have fundamental implications to the emerging biomass supply chain.
Specifically, Midwestern agricultural producers have shown propensity to accept acreage-based
contracts, a contract type that has been implemented elsewhere for trading biomass feedstock.
Through the University of Tennessee Biofuels Initiative (UTBI), agricultural producers are offered
acreage-based contracts to supply biomass to an experimental biorefinery located in Vonore, TN.
Producers engaged in this initiative receive an annual payment of $450/acre and face minimal
exposure to risk (e.g. price risk, yield risk, and cost of production risk due to provisions concerning

141

volatility of diesel prices). University of Tennessee (UT) extension services provide free
switchgrass seeds and technical training to engaged producers (UT Contract 2009).
The propensity to accept acreage-based contracts as demonstrated by Midwestern
producers may not lead to the development of a sustainable supply chain for biomass, however.
As economic theory suggests, more complex contract structures are needed to sustain exchange
relationships when market conditions create opportunities for hold-up behavior (Hueth et al. 1999,
Gow and Swinnen 2001, Shanoyan 2011). Therefore, although acceptance of acreage-based
contracts might be of assistance to train producers in the short-run and raise awareness about
environmental benefits of switchgrass, this structure is unlikely to lead to a sustainable biomass
supply chain without intermediation of a facilitator. In Tennessee, the successful use of acreagebased contract has been supported by the constant presence of an impartial facilitator – role played
by UT extension (Griffith et al. 2012). The use of this type of contract structure, however, has
failed in the absence of UT support. The DuPont-Danisco joint venture (JV), a Tennessee
cellulosic ethanol producer, has recently announced that a “stable biomass feedstock supply
remains [for the last five years] as the biggest challenge to push renewable energy generation
further”28. East Tennessee producers have indeed acquired experience with switchgrass crop,
promoted through the UTBI engagement, but hesitation to sign supply contracts directly with
biomass supplier persists.
From the perspective of this study, hesitation towards energy crops demonstrated by
producers is primarily aligned with two factors: (i) transaction uncertainties rather than lack of
technical knowledge about switchgrass, and (ii) familiarity or knowledge of different contract

28

DuPont Danisco Cellulosic Ethanol LLC is a 50/50 JV responsible for the demonstration biorefinery located in
Vonore, TN. The JV partnership was forced to review the original plan of turning the facility into a commercial plant
given the poor supply chain development (BiofuelsDigest, Nov 2013).

142

types. Anticipating these challenges, this study estimates agricultural producers’ WTA for a broad
array of contract structures. Results from RPL estimates indicate that producers would adopt
contracts capable of safeguarding against hold-up problems for as low as $26.53/ton/year on
average ($83/acre per year, assuming average yield of 3.12 tons/acre as in Perrin et al. 2008) added
to the acreage-based contract value, conditional on hold-ups being credible threats. Results from
G-MNL estimates suggest a larger ‘premium’. Midwestern producers would govern transactions
via specific performance contracts under high exposure to hold-up problems if the offered payment
was $46.52/ton/year ($145/acre/year) on average higher than acreage-based contract payment.
It is worthwhile to note that WTA estimates computed in this study show strong correlation
with familiarity to contract structures, which is a promising result for extension initiatives wishing
to incentivize the establishment of stable supply chain relationships for biomass. Our interpretation
is that Midwestern producers might feel increasingly comfortable trading through complex
contracts (without direct assistance from facilitators) as information barriers regarding complex
contract provisions and their effectiveness in case of breach are overcome. To put it in different
words, complex contracts should become more acceptable across agricultural producers as
extension initiatives lower information barriers.
In that sense, an entry strategy combining forces from entrepreneurs performing
investments and extension branches conducting hands-off training with agricultural producers
would reduce hesitation among producers. Taking the Tennessee experience as an example,
extension activities should tackle agronomic aspects of switchgrass as well as feasible contract
designs to govern long-term bilateral transactions. This study focuses on specification contracts
and shows that producers seem to understand the importance of using either specific performance
contracts or private enforcement capital contracts when opportunism from counterparties is a

143

credible threat. If effective training sessions are put in place and producers gain familiarity with
complex contract structures, WTA values are likely to decrease, bringing the biomass-based
energy sector a step closer to its sustainable development. It is crucial that producers become
comfortable with specific performance contracts or private enforcement capital contracts so they
move away from the preferred contract structure (i.e. acreage-based structure), especially when
market conditions set the stage for hold-up problems to happen.

6. Conclusions

This essay seeks to estimate agricultural producers’ preference for specification contract
characteristics and compare the findings with theory-driven predictions. Taking the developing
bioelectricity sector as a case, this article utilizes stated choice data on preference for contract
provisions collected throughout Midwestern states with RPS mandates in place. The specific
objective is to study whether overall preference for contract attributes vary conditional on different
levels of hold-up potential. The paper tests whether preferences for contracts better equipped with
provisions against opportunistic behavior prevail when the probability of hold-up problems to
occur is high. If so, the paper examines whether structures resembling specific performance
contracts (derived from the industrial organization literature) or private enforcement capital
contracts (derived from the institutional economics literature) are preferable to decision makers.
Moreover, empirical results from three models – RPL, S-MNL, and G-MNL – were generated and
discussed alongside to contribute to the choice experiment literature.
Results from all three empirical models indicate that producers would prefer acreage-based
contract structures to any alternative regardless of the market condition and exposure to hold-ups.

144

Share-cost contract structure comes second in the preference rank independent of the probability
of hold-up problems to happen. On average, a minimum premium of $9.60/ton would stimulate a
change from acreage-based contracts to share-cost contracts under high hold-up potential. Under
low probability of hold-up occurrence, producers would require a premium of $18.44/ton to adopt
a share-cost contract.
Although all three models similarly indicate that complex structures are not preferred to
share-cost contracts or acreage-based contracts in any market circumstance, empirical results are
mixed when it comes cross-comparing WTA for these structures. While RPL suggests that
producers prefer private enforcement capital structure ($26.56/ton premium on average), G-MNL
indicate that specific performance structure dominates under high probability of hold-ups
($46.52/ton on average). Under market conditions with low hold-up potential, specific
performance structure dominates for all three empirical models (minimum premium of
$29.41/ton). The complete combinatorial test (Poe et al. 2005), however, rejects the hypothesis
that there is statistical difference in preferences for complex contract structures in either market
condition at 5% confidence level.
Parameter estimates also point out that producers prefer short-term contracts (5 years long)
to long-term contracts (10 years long) in ether market condition. Sequentially to model estimation
and WTA calculation, complete combinatorial tests were conducted and results cannot reject the
hypothesis of statistical similarity between WTA for long-term contracts across market conditions.
Results seem to partially satisfy theoretical predictions. Although decision makers heavily
rely on traditions to make decisions over contract adoption – generating evidence that plays against
predictions driven by mainstream economic theories – WTA estimates point out that complex
contracts are more likely to be adopted when they are most needed. Results also corroborate with

145

other studies on contract longevity. Decision makers are not only averse to long-term contracts but
such aversion tends to rise when market conditions sets the stage for hold-up problems to occur.
Taking into account the latest developments in Tennessee – perhaps the vanguard state in
biomass-based energy generation – this article offers valuable implications to bioelectricity in the
Midwest U.S. Transactions between potential Midwest switchgrass growers and biomass
processors might unfold via acreage-based contracts in presence of a facilitator (e.g. extension
branches of land grant universities). The same specification contract, however, is unlikely to
stimulate stable relationships, especially if biomass transactions are conducted without an
intermediator. A better set of strategies would be to motivate producers to trade under contracts
capable to mitigating opportunism or unexpected breach. Although Midwest producers are
somehow averse to complex contracts in either market condition, this article shows that producers’
WTA for specific performance contract structure or private enforcement capital contract structure
consistently increase in probability of hold-ups (similar for all three empirical models). An entry
strategy of biomass-based electricity generators in the Midwest might reach favorable outcomes if
combined with hands-off support from extension branches of land grant universities. Without
pushing a particular contract structure, educational sessions with agricultural producers ought to
demonstrate why complex contracts are effective in market conditions with high hold-up potential.
From the results obtained here, such strategy is likely to be effective given the strong correlation
between WTA values and revealed familiarity with contract structures. As producers become
familiarized with specific performance structure and private enforcement capital structure, it is
likely that their WTA for contract structures capable of sustaining relationships in the emerging
bioelectricity industry increases.

146

This article also provides evidence that modeling preference heterogeneity through
deviations from mean utility weights performs better than modeling it through individual scale
effects. Results show that log-likelihood improves 16.01% when RPL model is utilized versus
0.03% improvement when S-MNL model is used, taking the seminal MNL model as base for
comparison. Performance improvement reaches 16.35% when both variations of heterogeneity are
incorporated into the G-MNL model.
It is important to interpret our model results with caution, however. First, none of the
empirical models used here allowed free correlation across parameters due to the rapid loss in
degrees of freedom. That can be interpreted as a weakness of the study given that the choice
experiment literature in place would recommend free correlation across attributes. Second, this
article is limited to preference-space estimation given our research objective of comparing
coefficients and WTA values across two market conditions – leading to two pecuniary attributes.
Our attempts to estimate in WTA-space failed to return robust parameter estimates. Refinements
of this study to overcome these two main limitations and others are strongly encouraged.

147

APPENDIX

148

Example of choice scenarios

Figure 9: Example of a choice scenario under information treatment 1

Assume that YOU are making the decision of what contract to sign under these market conditions:
There is an electricity generating firm in your region willing to sign a production contract for
biomass feedstock. Price for biomass-based electricity between the generating firm and distribution
utilities is volatile. Agricultural operations require acquisition of an expensive piece of machinery, which
might be redeployed in a future for its salvage value. Switchgrass yield is as variable as corn/soybeans
yield over the seasons.
Choice Scenario #3 of 6: Which switchgrass contract would you prefer?
Attributes

Contract 1

Contract 2

Structure C

Structure A

- Parties agree on price
- Grower and biomass processor share
establishment costs of crop equally
- Grower invests in machinery

- Parties agree on price and quantity to be
delivered
- Renegotiation is possible before
harvesting season
- Grower invests in machinery
- Processor is penalized if trade is delayed

5 years

5 year

$145/ton/year

$106/ton/year

Compensation
structure

Length of contract
Base payment
Which contract would you prefer?

149

Neither

I would negotiate
different terms
with the biomass
processor.

Figure 10: Example of a choice scenario under information treatment 2

Assume that YOU are making the decision of what contract to sign under these market conditions:
There are two electricity generators willing to sign a production contract for biomass feedstock in
your region. There is also a well established cash market for biomass (i.e. fiber pellets manufacturing
company), however the expected returns from this market are lower than the expected returns from the
electricity market. Price for biomass-based electricity is relatively stable. Agricultural operations require
acquisition of an inexpensive piece of machinery, which might be redeployed in a future period without
loss of productive value. Variability of switchgrass yield is low in comparison to corn/soybeans yield
variation over seasons.
Choice Scenario #4 of 6: Which switchgrass contract would you prefer?
Attributes
Compensation
structure

Contract 1

Contract 2

Structure D

Structure B

- Parties agree on price per acre
farmed
- Grower invests in machinery

Length of contract
Base payment

- Parties agree on price and quantity to be
delivered
- Grower and biomass processor co-invest
in machinery
- Processor loses its ownership share in
case of contract breach

10 years

5 year

$170/ton/year

$125/ton/year

Which contract would you prefer?

150

Neither

I would negotiate
different terms
with the biomass
processor.

REFERENCES

151

REFERENCES

Adamowicz, Wiktor, Peter Boxall, Michael Williams, and Jordan Louviere. "Stated preference
approaches for measuring passive use values: Choice experiments and contingent."
American Journal of Agricultural Economics no. 80 no. 1 (1998):64.

Aghion, Philippe, Mathias Dewatripont, and Patrick Rey. "Renegotiation Design with Unverifiable
Information." Econometrica 62, no. 2 (1994): 257-82.

Binmore, Ken, Ariel Rubinstein, and Asher Wolinsky. "The Nash Bargaining Solution in
Economic Modelling." RAND Journal of Economics 17, no. 2 (1986): 176-88.
Bolton, Patrick, and Mathias Dewatripont “Contract Theory” Cambridge, US: The MIT Press,
2005

Chen, Xiaoguang, Haixiao Huang, Madhu Khanna, and Hayri Önal. "Alternative transportation
fuel standards: Welfare effects and climate benefits." Journal of Environmental Economics
and Management 67, no. 3 (2014):241-57.

Crocker, Keith J., and Scott E. Masten. "Mitigating Contractual Hazards: Unilateral Options and
Contract Length." RAND Journal of Economics 19, no. 3 (1988): 327-43.

Fiebig, Denzil G., Michael P. Keane, Jordan Louviere, and Nada Wasi. "The Generalized
Multinomial Logit Model: Accounting for Scale and Coefficient Heterogeneity."
Marketing Science 29, no. 3 (2010):393-421.

Fifer, Simon, John Rose, and Stephen Greaves. "Hypothetical bias in Stated Choice Experiments:
Is it a problem? And if so, how do we deal with it?" Transportation Research Part A:
Policy and Practice 61, no. 0 (2014):164-177.

Goett, Andrew A., Kenneth Train, and Kathleen Hudson. "Customers' Choice Among Retail
Energy Suppliers: The Willingness-to-pay for Service Attributes", Energy Journal 21, no.
4 (2000): 1.

152

Goldberg, Victor P. "Regulation and Administered Contracts." The Bell Journal of Economics 7,
no. 2 (1976): 426-48.

Gow, Hamish R. and Johan M. Swinnen. Private Enforcement Capital and Contract Enforcement
in Transition Economies. American Journal of Agricultural Economics 83, no. 3 (2001):
686-690.

Hart, Oliver. Firms, Contracts, and Financial Structure. Oxford: Oxford University Press, 1995

Hart, Oliver, and John Moore. "Incomplete Contracts and Renegotiation." Econometrica 56, no. 4
(1988): 755-85.

Hayden, Noel J., "Landowner Willingness to Supply Marginal Land for Bioenergy Production: A
Case Study from Michigan." M.S. Thesis, Agricultural Economics, Michigan State
University, 108 pp, 2013.

Hueth, Brent, Ethan Ligon, Steven Wolf and Steven Wu. "Incentive Instruments in Fruit and
Vegetable Contracts: Input Control, Monitoring, Measuring, and Price Risk." Review of
Agricultural Economics 21, no. 2 (1999): 374-389.

Jensen, Kimberly, Christopher D. Clark, Pamela Ellis, Burton English, Jamey Menard, Marie
Walsh, and Daniel de la Torre Ugarte. "Farmer Willingness to Grow Switchgrass for
Energy Production." Biomass and Bioenergy 31, no. 11–12 (2007): 773-81.

Just, David R., and Steven Y. Wu. "Experimental Economics and the Economics of Contracts."
American Journal of Agricultural Economics 91, no. 5 (2009): 1382-88.

Klein, Benjamin. "Why Hold-Ups Occur: The Self-Enforcing Range of Contractual
Relationships.". Economic Inquiry 34, no. 3 (1996): 444-63.

Krinsky, Itzhak and Leslie A. Robb. "On Approximating the Statistical Properties of Elasticities",
Review of Economics & Statistics 68 no. 4 (1986), 715.

Louviere, Jordan J., David A. Hensher, Joffre D. Swait. "Stated Choice Methods: Analysis and
Applications" Cambridge, UK: University Press, 2000

153

Louviere, Jordan J., Deborah Street, Leonie Burgess, Nada Wasi, Towhidul Islam, and Anthony
A. J. Marley. "Modeling the choices of individual decision-makers by combining efficient
choice experiment designs with extra preference information." Journal of Choice
Modelling 1, no. 1 (2008):128-164.

Lusk, Jayson L., Jutta Roosen, and John A. Fox. "Demand for Beef from Cattle Administered
Growth Hormones or Fed Genetically Modified Corn: A Comparison of Consumers in
France, Germany, the United Kingdom, and the United States." American Journal of
Agricultural Economics 85, no. 1 (2003): 16-29.

Mahoney, Joseph T. "The Choice of Organizational Form - Vertical Financial Ownership Versus
Other Methods of Vertical Integration." Strategic Management Journal 13, no. 8 (1992):
559-84.

Maskin, Eric. "Nash Equilibrium and Welfare Optimality." The Review of Economic Studies 66,
no. 1 (1999): 23-38.

Masten, Scott E. "Case Studies in Contracting and Organization" New York, US: Oxford
University Press, 1996

McFadden, Daniel L. "Conditional Logit Analysis of Qualitative Choice Behavior" In: Frontiers
in Econometrics, Zarembka, P. (ed). Academic Press (1974): 105-142

McFadden, Daniel L. "Structural Discrete Probability Models from Theories of Choice." In:
Structural Analysis of Discrete Data and Econometric Applications, Manski, C.F.,
McFadden, D.L. (eds). The MIT Press (1981): 198-272.

Morey, Edward and K. Greer Rossmann. "Using Stated-Preference Questions to Investigate
Variations in Willingness to Pay for Preserving Marble Monuments: Classic
Heterogeneity, Random Parameters, and Mixture Models", Journal of Cultural Economics
27, no. 3-4 (2003): 215-229.

Nöldeke, George, and Klaus M. Schmidt. "Option Contracts and Renegotiation: A Solution to the
Hold-up Problem." RAND Journal of Economics 26, no. 2 (1995): 163-79.
Paulrud, Susanne, and Thomas Laitila. "Farmers’ Attitudes About Growing Energy Crops: A
Choice Experiment Approach." Biomass and Bioenergy 34, no. 12 (2010): 1770-79.

154

Poe, Gregory. L., Kelly Giraud, and John B. Loomis. "Computational Methods for Measuring the
Difference of Empirical Distributions", American Journal of Agricultural Economics 87,
no. 2 (2005): 353-365.

Ross, R. Brent, Guilherme Signorini. "Contract Design for Biomass Procurement." Focus Group
Report. Michigan State University, East Lansing, MI. (08/15/2013).

Shanoyan, Aleksan. External Facilitation of Sustainable Market Linkages in the Agri-Food Supply
Chain: Evidence from the USDA Marketing Assistance Program in the Armenian Dairy
and Vegetable Industries. Ph.D. diss., Michigan State University, East Lansing, 2011.

Signorini, Guilherme, R. Brent Ross, William Knudson, Vivek Pandey, and H. Christopher
Peterson. "Assessing Contract Efficiency for Biomass: A Case Study of the Upper
Peninsula of Michigan." Working Paper (forthcoming).

Simon, Herbert A. "Rationality as Process and as Product of Thought." The American Economic
Review 68, no. 2 (1978): 1-16.
Tirole, Jean “The Theory of Industrial Organization” Cambridge, US: The MIT Press, 1988

Tonsor, Glynn T., Ted C. Schroeder, John A. Fox, and Arlo Biere. "European Preferences for Beef
Steak Attributes." Journal of Agricultural and Resource Economics 30, no. 2 (2005): 36780.

Train, Kenneth "Discrete Choice Methods with Simulation" Cambridge, US: Cambridge
University Press, 2009
Williamson, Oliver E. “The Economic Institutions of Capitalism: Firms, Markets, Relational
Contracting.” New York, US: The Free Press, 1985

______. "Credible Commitments - Using Hostages to Support Exchange." American Economic
Review 73, no. 4 (1983): 519-40.
______. “Markets and Hierarchies.” New York, US: The Free Press, 1975

155

CONCLUSION

This dissertation constitutes a comprehensive attempt to diagnose deterring factors and
propose sustainable relationship mechanisms for the renewable energy supply chain in the U.S.
Midwest. It aims at identifying information barriers preventing agricultural producers from
engaging in stable transactions with biomass processors and at proposing solutions derived from
stated preference data, gathered from producers.
The literature on willingness to convert farmland and produce energy crops has previously
identified hesitation and skepticism as two major barriers to conversion (Hayden 2013, Qualls et
al. 2012, Rossi and Hinrichs 2011, Kelsey and Franke 2009). Hesitation expressed by agricultural
producers has often been associated with the lack of technical or agronomic knowledge about
energy crops and can be defined as ‘the reluctance to adopt new production methods or crops until
one sees them working for others’ (Qualls et al. 2012). Essay 1 of this dissertation diagnoses an
additional underlying factor that is also leading to agricultural producers’ hesitation: transaction
problems resulting from inappropriate incentive structures utilized in early attempts to govern
biomass transactions.
Using a case study methodology (Yin 2009), essay 1 examines biomass transactions in the
Upper Peninsula of Michigan. Five transactions were evaluated through the lenses of transaction
cost economics (TCE) theory (Williamson 1985; 1996). Out of the five transactions, the exchange
between logging firms and a biomass processor has failed to perform, leading parties to experience
hold-up problems. Sequentially, the private enforcement capital model (Klein 1996) served as a
framework to inspect the underlying causes of the problem. The exchange based on an incomplete
contract (Hart and Moore 1988) and in presence of specific assets (Peterson et al. 2001) created

156

high hold-up potential, which turned into a real problem as changes in market conditions set the
stage for the biomass processor to behave opportunistically over logging firms. Essay 1, therefore,
suggests that the inability of trading parties to design effective coordination strategies can be
considered as a factor causing hesitation across broad populations of agricultural producers.
Combining the results obtained in essay 1 with the argument of agronomic information
barriers regarding switchgrass, essays 2 and 3 examine stated preference data to elaborate effective
and implementable mechanisms to govern exchange of biomass feedstock. Essay 2 estimates how
agricultural producers value general attributes of cropping systems while recognizing regional
particularities and differences in farming capabilities. Based on the random utility maximization
(RUM) model (McFadden 1974, 1981), the essay implements and compares estimates from two
empirical models – Random Parameters Logit (RPL) and Scaled Multinomial Logit (S-MNL).
Results indicate that agricultural producers value (i) cropping systems that require low intensity of
production practices to obtain average yield, and (ii) low variance of net margins – characteristics
that can be found in switchgrass but seldom in conventional cropping systems (e.g. corn-soybean
rotation systems). Results also point out that willingness to accept (WTA) for switchgrass
characteristics increases as one moves away from counties where conventional cropping systems
are competitive to switchgrass. Farming capability also influences preference for cropping system
attributes, but WTA for switchgrass characteristics does not decrease linearly in county
competitiveness. Instead, producers located in county category 2 (counties where expected net
margin from conventional crops is at least 70 percent of state average but lower than expected net
margin from switchgrass) would be less reluctant to make specific investments for growing
switchgrass than producers located elsewhere.

157

Having identified regions where energy crops have relatively high acceptance and
population segments that are more willing to grow switchgrass, Essay 3 examines how agricultural
producers would prefer to draft contract provisions for trading biomass feedstock. The specific
objectives are to evaluate (i) whether producers distinguish market conditions with high hold-up
potential from market conditions with low hold-up potential, and (ii) whether producers choose
contracts better equipped to deal with hold-up problems when those problems are credible threats
– as they were discussed in Essay 1. If so, the paper examines whether structures resembling
specific performance contracts (derived from the industrial organization literature) or private
enforcement capital contracts (derived from the institutional economics literature) are more in line
with agricultural producers’ preference. The RUM model (McFadden 1974, 1981) is also utilized
in Essay 3 and parameter estimates from three empirical models are compared – RPL, S-MNL,
and Generalized Multinomial Logit (G-MNL). Results indicate that acreage-based contracts are
preferable in any market circumstance. Share-cost contract comes second in the preference rank
and is partly dependent on market conditions. Agricultural producers also show increasing
acceptance for either complex contract structure as market conditions change from low to high
hold-up potential. Producers also seem to rely on traditions to make decisions over contract
adoption as preferences for contract structures are closely related to revealed familiarity to contract
types. Parameter estimates also point out that producers prefer short-term contracts (5 years long)
to long-term contracts (10 years long) in ether market condition.
Essay 3 contends, however, that acreage-based contracts might govern transactions in the
presence of a third-party external facilitator, but is unlikely to sustain stable relationships without
an intermediary. The argument comes from the rich literature on external facilitation to promote
supply chain linkages (Glover and Kusterer 1990, Porter and Phillips-Howard 1997, Shepherd

158

2007). A better set of strategies would be to motivate producers to trade under contracts capable
to mitigating opportunism or unexpected breach when such problems constitute credible threats.
An entry strategy of biomass-based electricity generators in the Midwest might reach favorable
outcomes if combined with hands-off support from extension branches of land grant universities.
Without pushing a particular contract structure, educational sessions with agricultural producers
ought to demonstrate why complex contracts are effective in market conditions with high hold-up
potential.
Essays 2 and 3 also contribute to the literature of discrete choice by comparing model
performance as suggested in Fiebig et al. (2010). The preference data used in essay 2 fit the RPL
model better than the S-MNL model. In essay 3, allowing preference heterogeneity via deviation
from mean utility weights and individual scale – as in the G-MNL model – was superior to either
modelling strategy alone.
Even though the findings of this dissertation suggest a path for the development of the
renewable electricity industry in the U.S. Midwest, interpretation of WTA estimates must be
conducted with caution. In essay 2, mean estimates of WTA for net margin variance are not
statistically different, for example. In essay 3, WTA for specific performance contracts or WTA
for private enforcement capital contracts are not statistically different at 5% of confidence. In either
essay the statistically significance of standard deviations from mean utility weights (in RPL models
or in the G-MNL model) suggests that the mean WTA estimates may not be statistically significant
for the entire sample. Further refinements to allow precise interpretation of WTA estimates are
strongly encouraged. Finally, our attempts to estimate in WTA-space failed to return robust
parameter estimates. Given the growing importance of this latter modeling strategy, refinements
of essays 2 and 3 to overcome this limitation are strongly encouraged.

159

REFERENCES

160

REFERENCES

Fiebig, Denzil G., Michael P. Keane, Jordan Louviere, and Nada Wasi. "The Generalized
Multinomial Logit Model: Accounting for Scale and Coefficient Heterogeneity."
Marketing Science 29, no. 3 (2010):393-421.

Glover, David, and Ken Kusterer. Small farmers, big business: contract farming and rural
development. Macmillan Press Ltd., 1990.

Hayden, Noel J., "Landowner Willingness to Supply Marginal Land for Bioenergy Production: A
Case Study from Michigan." M.S. Thesis, Agricultural Economics, Michigan State
University, 108 pp, 2013.

Hart, Oliver, and John Moore. "Incomplete Contracts and Renegotiation." Econometrica 56, no. 4
(1988): 755-85.

Kelsey, Kathleen D., and Tanya C. Franke. "The Producers' Stake in the Bioeconomy: A Survey
of Oklahoma Producers' Knowledge and Willingness to Grow Dedicated Biofuel Crops."
Journal of Extension 47, no. 1 (2009).

Klein, Benjamin. "Why Hold-Ups Occur: The Self-Enforcing Range of Contractual
Relationships." Economic Inquiry 34, no. 3 (1996): 444-63.

McFadden, Daniel L. "Conditional Logit Analysis of Qualitative Choice Behavior" In: Frontiers
in Econometrics, Zarembka, P. (ed). Academic Press, 1974; 105-142

McFadden, Daniel L. "Structural Discrete Probability Models from Theories of Choice." In:
Structural Analysis of Discrete Data and Econometric Applications, Manski, C.F.,
McFadden, D.L. (eds). The MIT Press, 1981; 198-272.

Peterson, H. Christopher, Allan Wysocki and Steven B. Harsh. "Strategic Choice Along the
Vertical Coordination Continuum." The International Food and Agribusiness Management
Review 4, no. 2 (2001): 149-166.

161

Porter, Gina, and Kevin Phillips‐Howard. Comparing contracts: an evaluation of contract farming
schemes in Africa. World Development 25, no. 2 (1997): 227-238.

Qualls, D. Joshua, Kimberly L. Jensen, Christopher D. Clark, Burton C. English, James A. Larson,
and Steven T. Yen. "Analysis of Factors Affecting Willingness to Produce Switchgrass in
the Southeastern United States." Biomass and Bioenergy 39, no. 0 (2012): 159-67.

Rossi, Alissa M., and C. Clare Hinrichs. "Hope and skepticism: Farmer and local community views
on the socio-economic benefits of agricultural bioenergy." Biomass and Bioenergy 35, no.
4 (2011):1418-28.

Shepherd, A. W. Approaches to linking producers to markets. A review of experiences to date.
Agricultural Management, Marketing and Finance Occasional Paper (FAO), 2007
Williamson, Oliver E. “The Economic Institutions of Capitalism: Firms, Markets, Relational
Contracting.” New York, US: The Free Press, 1985
———. “The Mechanisms of Governance.” New York, Oxford University Press, 1996
Yin, R.K. “Case Study Research: Design and Methods” Thousand Oaks, CA: Sage, 2009

162