Professional Documents
Culture Documents
Risk Management and The Environment: Agriculture in Perspective
Risk Management and The Environment: Agriculture in Perspective
Edited by
Bruce A. Babcock
Center for Agricultural and Rural Development (CARD),
Iowa State University, U.S.A.
Robert W. Fraser
Imperial College,
Wye, u.K.
and
Joseph N. Lekakis
Department of Economics,
University of Crete, Greece
"
~.
Tables vii
Figures ix
Contributors x
Preface xii
Introduction
Risk Management and the Environment 1
in Agriculture: A Key Policy Theme
Bruce A. Babcock, Robert W. Fraser, Joseph N. Lekakis
Part I Theory
Chapter
Conclusion
Bruce A. Babcock, Robert W. Fraser, Joseph N. Lekakis 196
2. 1: Parameter Values 37
Richard M. Adams
Professor, Department of Agricultural and Resource Economics,
Oregon State University, USA
Bruce A. Babcock
Professor of Agricultural Economics
and Director, Center for Agricultural and Rural Development (CARD),
Iowa State University, USA
Robert G. Chambers
Professor, Department of Agricultural and Resource Economics,
University of Maryland, USA
Robert W. Fraser
Professor of Agricultural Economics, Imperial College, Wye, UK,
and Adjunct Professor of Agricultural and Resource Economics,
University of Western Australia, AUSTRALIA.
Barry K. Goodwin
Professor, Andersons Endowed Chair of Agricultural Policy,
Trade, and Marketing, Department of Agricultural, Environmental, and
Development Economics, Ohio State University, USA.
David A. Hennessy
Professor, Department of Economics, Iowa State University, USA
Greg Hertzler
Senior Lecturer in Agricultural and Resource Economics
and Associate Dean of Agriculture and Animal Science
Faculty of Natural and Agricultural Science
The University of Western Australia, AUSTRALIA
Terrance Hurley
Assistant Professor, Department of Applied Economics
University of Minnesota, USA
Contributors Xl
Ross Kingwell
Visiting senior lecturer, University of Western Australia
and Senior adviser, Western Australian Department of Agriculture,
AUSTRALIA
Joseph N. Lekakis
Professor of Resource & Environmental Economics,
Department of Economics, University of Crete, GREECE
Paul D. Mitchell
Assistant Professor, Department of Agricultural Economics,
Texas A&M University, USA.
David J. Pannell
Associate Professor and Principal Research Fellow
School of Agricultural and Resource Economics,
University of Western Australia, AUSTRALIA
John Quiggin
Senior Fellow, Australian Research Council
School of Economics, Faculty of Economics and Commerce
Australian National University, AUSTRALIA
HiJd Rygnestad
Managing Director, The Rygnestad Group,
Sevanna Park T-7C, Ithaca, NY 14850, USA
Vincent H. Smith
Professor of Economics, Department of Agricultural Economics
and Economics, and Co-Director, Agricultural Marketing Policy Center,
Montana State University, USA.
JunJie J. Wu
Associated Professor, Department of Agricultural and Resource
Economics, Oregon State University, USA.
Preface
The US Model
During the last decade, US farmers have been using various price risk-
management methods including storage, forward cash contracts, selling
futures, buying futures, and private and public insurance programs.
Storage on the farm allows producers to sell their output at different
points throughout the year in an effort to secure an average price. The
risky part of this tool is that farmers may pass up good pricing opportuni-
ties. Forward cash contracts allow farmers to forward price with their lo-
cal elevator anytime during the growing season and deliver later. The
major problem with this method would be failure of the farmer to deliver
due to crop failure.
In addition to these options, the US Department of Agriculture and pri-
vate insurance companies have developed three major programs, In-
come Protection (IP), Crop Revenue Coverage (CRC), and Revenue As-
surance (RA), which are there to assist the farmer's own risk manage-
ment options (Harwood, Goble, and Perry, 1997). All three products offer
an annual revenue guarantee based on producer planted area, expected
2 Bruce A. Babcock, Robert W. Fraser, Joseph N. Lekakis
yield, and sign-up time futures price for harvest-time delivery. If the har-
vest-time price multiplied by the actual yield in that year falls below the
guarantee, the producer receives a compensatory amount of money.
CRC and RA both allow a farmer to replace yield shortfalls at current
market prices. If a farmer's yield falls short and the projected price is
lower that the actual harvest price, he is compensated for the yield loss
at the harvest price in order to allow him to purchase 'replacement' crop
quantities in the market. In addition, RA allows farmers flexibility by al-
lowing them to insure all theirn crops together in a "whole-farm" unit, or
to insure the crops separately.
The EU Experience
Book Contents
The book is structured into two parts: a larger theoretical part containing
a selection of contributions to the general nature of the issues which
arise in agriculture in the context of risk management and the environ-
ment; and a smaller set of case studies which illustrates the main themes
of the theoretical papers in particular situations. We think that this struc-
ture facilitates an appreciation of the generic nature of many of the
problems associated with considering risk management and the envi-
ronment in agriculture, while at the same time showing the application of
theoretical approaches in a suitable range of cases.
More specifically, the theoretical section opens with an examination by
John Quiggin and Robert Chambers of "The state-contingent approach to
modeling environmental risk management". The aim of their chapter is to
describe the state-contingent approach to modeling uncertainty, and to
assess its implications for the analysis of problems involving agricultural
production and environmental risk. In so doing, they argue that there is a
very broad field of potential applications of the state-contingent frame-
work to problems involving environmental risk in agriculture, and that the
state-contingent approach is more amenable to the analysis of uncer-
tainty than other approaches. The second in this pair of chapters exam-
ining environmental risk is Greg Hertzler's study of "The precautionary
principle in practice: how to write a call option on the environment". This
study examines options and irreversibilities in environmental decision-
making. It aims to develop option-pricing formulas and show how much
society will invest in order to avoid undesirable outcomes. The approach
taken is to adapt dynamic investment theory from the finance literature -
an adaptation that requires considerable development as there are many
assumptions made in finance, which do not apply in an environmental
context. The chapter uses examples such as water rights to illustrate its
findings.
The second pair of chapters looks at theoretical issues associated with
the adoption of environmentally sensitive farming practices. Chapter 3,
written by Paul Mitchell and David Hennessy, studies "Factors determin-
ing best management practice adoption incentives and the impact of
green insurance". It develops a framework that allows the factors deter-
Introduction: Risk Management in Agriculture 5
References
Babcock, B. A. and D. A. Hennessy (1996) 'Input Demand under Yield and Reve-
nue I nsurance' American Journal of Agricultural Economics, 78: 416-427.
Goodwin, B. K. and V. H. Smith (1997) Crop Insurance, Disaster Relief, and En-
dogenous Soil Erosion' North Carolina State University (unpublished paper).
Harwood, J., D. Heifner, K. Goble, and J. Perry (1997) 'Alternatives for Producer
Risk Management' in USDA, Agricultural Outlook Forum '97 Proceedings,
Washington, DC, February.
Horowitz, J. K. and E. Lichtenberg (1993) 'Insurance, Moral Hazard, and Chemi-
cal Use in Ahriculture, American Journal of Agricultural Economics, 75: 926-
935.
Quiggin, J. (1991) 'Comparative Statics of Rank Dependent Expected Utility The-
ory,' Journal of Risk and Uncertainty, 4: 339-350.
Quiggin, J. G. Karagiannis, and J. Stanton (1993) 'Crop Insurance and Crop
Production: An Empirical Study of Moral Hazard and Adverse Selection,' Aus-
tralian Journal of Agricultural Economics, 37: 95-113.
Smith, V. H. and B. K. Goodwin (1996) 'Crop Insurance, Moral Hazard, and Agri-
cultural Chemical Use' American Journal of Agricultural Economics, 78: 428-
438.
Part I
Theory
1
The State-contingent Approach to
Modeling Environmental Risk Management
Intuitively, X (z.e), which we typically refer to, as the input set, is eve-
rything on or above an isoquant for the state-contingent technology.
14 John Quiggin and Robert C. Chambers
1.1.3. Preferences
mN
In general, the preferences of producers depend on the !nputs xE,,~ +
that they commit and the state-contingent income y E D1: that they re-
ceive. Thus. the most general representation of preferences is a mapping
V: D1: x D1~ -- D1. Two special cases are of particular interest. The
separable effort specification
V (y,x) = W(y)-g(x)
V (y,x) = W(y-w· x)
The net returns objective function may also be represented by the cer-
tainty equivalent, that is, the smallest certain income which leaves the
individual as well off as y. More formally,
For an individual risk-averse with respect to 7r, we now define the ab-
solute risk premium:
= 8(W(y), E 7r [yj1),
Definition 1: W displays constant absolute risk aversion (CARA) if, for all
y, t, r(y + t1) = r(y).
Definition 2: W displays constant relative risk aversion (CRRA) if. for all
y, t, u(ty) = u(y) .
Extension of these definitions to more general concepts, such as de-
clining absolute risk-aversion, is discussed by Quiggin and Chambers
(2001 ).
= nw>o{x:w·x<!:c(w,z,e)}=X(z).
eC(z)=sup{e:C(el S sC(z)}.
v = E [z - yJ + e (y - w· x) - m (e)
that is, the sum of the expected net tax payment by the producer, the
producer's certainty-equivalent income and m(e) , a monetary measure
of the environmental damage e. The planner may be required to satisfy
some reservation utility constraint yielding the producer at least some
minimum level of certainty-equivalent income.
The producer is assumed to have full information regarding the state-
contingent private and environmental output vector (z. e) and to be able
to observe the state of nature, s that is actually realized. By contract, the
planner is assumed to have access only to some subset of this informa-
tion. With varying assumptions about the planner information set, a range
of different policy problems can be formulated. Table 1.1 that follows was
adapted from Chambers and Quiggin (2001) illustrates a range of possi-
ble cases. The possibilities include not only hidden-information problems
but also moral hazard problems (commonly seen as a separate analytical
category involving 'hidden action'). The various cases have been ana-
lyzed by Chambers and Quiggin (1996, 2000, 2001) and Quiggin and
Chambers (1998, 2001).
Observables
State of Private Environmental Class of Problem
nature output output
yes yes yes full information
yes yes no externality with insurance
yes no yes equivalent to full information
yes no no nonpoint pollution
no yes yes point-source pollution with moral hazard
no yes no externality with moral hazard
no no yes externality
no no no private optimum
In all of these problems, the crucial interest lies in the relationship be-
tween the environmental externality, reflected in the exclusion of e from
The State-contingent Approach 21
the producer's private cost function, and the possibility for insurance
arising from interactions between the risk-averse producer and the risk-
neutral planner.
The full-information problem, listed in the first row of Table 1.1, is a
natural benchmark for the analysis. In this case, the producer chooses
the socially optimal output vector (z,e) which maximizes
z = f(x,s) ,
References
Arrow. K. (1953), Le role des valeurs boursiers pour la, repartition la meillure des
risques, Cahiers du Seminair d'Economie CNRS, Paris.
Chambers, R. G. and J. Quiggin (1996), 'Nonpoint pollution control as a multi-
task principal-agent problem', Journal of Public Economics 59(1): 95 116.
Chambers, R. G. and J. Quiggin (1997), 'Separation and hedging results with
state-contingent production', Economica 64(254): 187 209.
Chambers, R. G. and J. Quiggin (1998), 'Cost functions and duality for stochastic
technologies' American Journal of Agricultural Economics 80(2): 28895.
Chambers, R. G. and J. Quiggin (2000), Uncertainty, Production. Choice and
Agency: The State-Contingent Approach, Cambridge University Press, New
York.
Chambers, R. G. and J. Quiggin (2001), 'Production externalities and multitask
principal-agent problems', Working Paper, Australian National University. Can-
berra.
Debreu, G. (1952). 'A social equilibrium existence theorem', Proceedings of the
National Academy of Sciences 38: 886 93.
Grossman. S. and O. Hart (1983), 'An analysis of principal-agent problems',
Econometrica 51: 7-46.
Luenberger, D. G. (1992), 'Benefit functions and duality'. Journal of Mathematical
Economics 21 , 461-481.
Malmquist. S. (1953), 'Index numbers and indifference surfaces', Trabajos de
Estotistica 4: 209-242.
McFadden, D. (1978), 'Cost, revenue, and profit functions', in Fuss, M. and
McFadden, D. (eds.), Production Economics: A Dual Approach to Theory and
Applications, North Holland, Amsterdam.
Mirrlees, J. (1974), 'Notes on welfare economics, information and uncertainty', in
Balch. M., McFadden, D. and Wu. S. (ed.), Essays on Economic Behaviour un-
der Uncertainty, North-Holland. Amsterdam, pp. 243-61.
Quiggin, J. (2002), 'Risk and self-protection: a state-contingent view ',Journal of Risk
and Uncertainty, forthcoming.
Quiggin, J. and R. G. Chambers (1998), 'A state-contingent production approach
to principal-agent problems with an application to point-source pollution con-
trol', Journal of Public Economics 70: 441-472.
Quiggin, J. and R.G. Chambers (2000), 'Increasing and decreasing risk aversion
for generalized preferences'. Working Paper, Australian National University.
Canberra .
28 John Quiggin and Robert C. Chambers
Quiggin, J. and R.G. Chambers (2001), The firm under uncertainty with general risk-
averse preferences: a state-contingent approach, Journal of Risk and Uncertainty,
22(1): 5-20.
Savage. L. J. (1954), Foundations of statistics, Wiley, New York.
Shephard, R. W. (1953). Cost and Production Functions, Princeton University
Press, Princeton, New Jersey.
Shogren, J. and T. Crocker (1999), 'Risk and its consequences', Journal of Envi-
ronmental Economics and Management, 37(1): 44-51.
2
The Precautionary Principle in Practice:
How to Write a Call Option
on the Environment
Greg Hertzler
Over the years, there have been three major approaches to making deci-
sions under risk. The first, static expected utility theory, is perhaps the
most popular, used by many applied economists. The second is dynamic
investment theory, used by financial economists. The third is difficult to
name. It is the approach we as natural resource and environmental
economists use. It seems to be a mixture of philosophy, common sense,
and confusion. We still haven't sorted out what we mean by the term
"option value." To avoid confusion in this chapter, I will try to be precise
in defining how risk affects decisions. This will, hopefully, lead to more
precise results about how to value our options for preserving the envi-
ronment.
We often talk about the "risk preferences" of decision makers. This
leads to confusion and misinterpretation because, strictly speaking, pref-
erences about risk aren't in our models. We include preferences about
consumption, wealth and time, but not about risk. Risk affects decisions
whenever the world is non linear. One source of non-linearity is the pref-
erence or utility functions in our models. There are many other sources,
however. Behavior under risk can be explained by the non-linearity that
is affected by risk. For the purposes of this discussion, non-linearities will
fall into the following categories:
• Functions: consumption preferences, wealth preferences, time
preferences, production and cost functions, endogenous prices;
• Probability distributions: non-normal distributions, co-variances;
• Asymmetries: infeasibilities, options, irreversibilities.
30 Greg Hertzler
formulas for pricing options are different and give different answers. The
examples in this chapter are small ones and there are bigger problems
yet to study. This chapter concludes with suggestions about how this
might be done.
where, A.(W)_~'R(W)=_a2J/aw2.
oW' oj/oW
The first optimality condition is for consumption and the second is for
production and investment decisions. The marginal utility of consumption
is normalized by the expected marginal utility of wealth, A. The terms
containing R are marginal risk premiums. To simplify notation, R is de-
fined as the coefficient of absolute risk aversion. It measures the curva-
ture of expected social welfare with respect to current wealth. It is distin-
guished from an Arrow-Pratt coefficient of risk aversion that would meas-
ure the curvature with respect to terminal wealth. Marginal utility of
wealth and the risk aversion coefficient encapsulate all of the information
about the future. If their current values can be measured, optimal deci-
sions in a single period are also dynamically optimal.
resource stocks, S, are valued at the commodity price for the economy
with a return on investment above the risk-free rate of Y(O'y - O'W). Stocks
will depreciate at the rate O'S, with a depreciation cost of yO'S. Exploring
for resource stocks and maintaining the stocks will cost 5 dollars per unit.
Some risk management decisions are made once in a year. These in-
clude production contracts and forward contracts. Production contracts,
K, are not for production directly but, rather, for the environmental vari-
able, X. If the environmental variable is less than a contract level, X,
production contracts will be profitable. Production contracts are an artifi-
cial construct that may, but probably won't exist. They are included in the
model as a way to design options on the environment. Forward contracts,
H, are profitable if the expected price at the end of the decision interval,
y(1 + O'y), is less than the contract price, h, at which commodities must
be sold. Commodity futures are traded continuously and there is no fixed
contract price that remains in force throughout the life of the contract.
Trading in commodity futures will be profitable if the change in the fu-
tures price, 0" is negative.
Also from equation (2.12) in the Appendix, the error terms for the
change in wealth are:
The first line of this equation includes price risks and the second line
includes quantity risks. Variances and co variances are found by squar-
ing the errors.
The first line includes variances for price risks; the second line includes
variances for quantity risks; and the third and fourth lines contain co
variances.
Resource Stocks
y (Oy - Ow - os) -s -R [y2(y +S -H + (amjay) M)O':
(2.5)
-yf (F - (anjaf)N)O'yO'twyt + y2y O'yO'yWyy ] = o.
Production Contracts
X -x -R[(K -(aejaX)E)X 2 0'; -ylXO'YO'XCVr.¥] =0. (2.6)
Forward Contracts
y (1 + Oy) - h - R [y2 (y + S - H + (amjay) M)O':
(2.7)
-yf(F - (anjat)N)O'YO'twyt + y2yO'py{Uyy] =0.
Futures Contracts
tOf -R [yf(Y +S-H+(am/ay)M)crYO'fWyf -f2(F -(an/at)N)o?] =0. (2.8)
Options on Futures
n( on -4v) - R [yf(Y +S - H+( ami ay) M) O'yo;Wyf -t2(F - (en/ at)N) o?]( en/ at) =0. (2.10)
Although the optimality conditions seem complex, they are easy to in-
terpret. Terms beginning with R are marginal risk premiums. If prefer-
ences are linear, agents are risk neutral in the traditional sense, the mar-
ginal risk premiums are zero and the optimality conditions collapse to
profit maximizing behavior. If agents are averse to risk they will choose
an optimal portfolio that balances profit and risk.
In the finance literature, lack of profitable arbitrage and continuous
trading are assumed. The portfolio becomes riskless and the prices of
options are independent of risk preferences (Jarrow, 1999; Merton, 1990,
p. 281, Hull, 1997, p. 539). For a society making decisions about re-
The Precautionary Principle in Practice 37
The marginal risk premiums are eliminated and the expected loss or
gain from holding a futures contract equals the expected capital gains
from owning options on futures. Options incorporate the same informa-
tion as futures prices and hedge against the same risks.
Substituting for the expected change in the price of an option on fu-
tures, n8n , from equation (2.3) gives a partial differential equation defin-
ing how the price of an option evolves beginning from the current time.
This is combined with boundary conditions that an option must satisfy at
the time of maturity.
38 Greg Hertzler
a? 11 iJ 2n 2 2
-+ 12-2 f at -now =0;
a if
n (r, f) -If
- - i; f >• i
0; f sf.
The boundary conditions are for a European call option that can be exer-
cised only at the time of !!laturity, r, but not before. If the futures price, f,
exceeds exercise price, f, the option is "in the money" and will be exer-
cised; otherwise it is worthless. With the boundary conditions, the solution
becomes a cumulative probability, in this case the Black-Scholes formula
for pricing options on futures (Merton, 1990, p. 347; Hull, 1997 p. 277).
Figure 2.1 shows the prices of a call option on futures as a function of
the futures price and time. The exercise price is set to equal the agent's
expected futures price, f(1 + 0,), or $172/ton and the options prices are
the amounts the agent would pay to avoid the upside risk of a higher
than expected futures price. In Table 2.1, the current futures price is
$155/ton. Suppose, at time zero, the option is 52 weeks from maturity. In
this case, the agent would pay $11.55/ton for the option. The price
changes as the option matures. If the futures price at maturity of the op-
tion is less than $172/ton, the option will be worthless. The agent will
throwaway the option and buy futures contracts for less than the exer-
cise price. Otherwise, the option at maturity is valued as the difference
between the futures price and the exercise price. The agent will exercise
the option to buy futures contracts at the exercise price.
Options Price
($/unit)
0 ....
~'" 8 8
Time
(weeks)
"'~ttg]N
'" :!l 0
8 a 8 ~ ol :;;
N fD ~ ....
8 8 ....... ~
v :; ~ 8 . Iii ~ ~ ~ Underlying Price
~ @ ~ .... .... ($/unit)
~
The expected return from holding a forward contract equals the ex-
pected capital gains from owning an option on forward contracts. Sub-
stituting for the expected change in the value of an option on forward
contracts, mOm, from equation (2.3) gives a partial differential equation
that is solved subject to boundary conditions.
an an (L" )
a+~V1-y +12 0/ 2 y
11 8 2m 2 2
O"Y -mow =0;
m(r,y)= {Y - y; ~ > y
0; ysy.
Options on forward contracts have an additional term, (am/ay'lii - y), in
the option pricing formula. This term compares the fixed contract price to
the changing commodity price. This differs from options on futures for
which the "contract price" is Simply the current futures price. The
equivalent term would be (ani af 'It -
f) which is always zero.
The boundary conditions are for a European put option. At the time of
maturity, if the exercise price, y, exceeds the commodity price, y, the
option is "in the money" and will be exercised. This option priCing formula
is a non-linear differential equation and no analytical solution is known.
Therefore, the solution was calculated numerically using the Crank-
Nicholson method for finite differences (Hull, 1997, p. 378; Burden and
Faires, 1997, p. 692).
40 Greg Hertzler
Option Price
($/unit)
8 '
;;i ~ 8 8 Time
~ '" ~ . 8 8
~ ... "'. 8 (weeks)
Underlying Price ~ ;! § ~ § ~ ~
($/unit) ~ ~ :e
Petzel (1984) first proposed and Bardsley and Cashin (1990) first ap-
plied the Black-Scholes formula for options on futures to the evaluation
of the benefits from government programs. The Black-Scholes formula or
its equivalent is now used to value many things from crop insurance
(Just, Calvin and Quiggin, 1999; Mahul, 1999; Stokes, Nayda and Eng-
lish, 1997) to old growth forests (Conrad, 1997). Valuing a government
The Precautionary Principle in Practice 41
30.00
25.00~111
Option Price
($/unit) 15,00
10,00
5,00
., .. 0
88 '<o~
t;i.,; ~ 8 co~?a"'" Time
., ....
- '"
...... - w,8.
..
v 00 0
~ ~ C'I (weeks)
Underlying Price - ;! ~ ~ §~ §
o
($/unit) - ~ le ~
a;
..
Exploration for resource stocks creates a call option. These stocks can
be extracted and sold if the commodity price is high enough. Originally,
the Black-Scholes formula was derived for options on financial stocks
(Black and Scholes, 1973). This can be compared with an option on re-
source stocks by combining equations (2.5) and (2.9).
42 Greg Hertzler
(s:
Y\U -8w -8s
)
-s = m{8m - 8w ) .
Y am/ay
The return above costs for finding and holding resource stocks equals
the capital gains from an option on forward contracts. Substituting for the
expected change in the value of the option on forward contracts from
equation (2.3) gives the option pricing formula.
an
-+ -
an (y V'w
( ) ) 1/ 0 2m 2 2
+ Os +5 + /2-2- Y u y -mow = 0;
Of 0' 0'
m (r, y ) = {Y - y; y > ~
0; y:s; y.
The Black-Scholes formula for options on stocks also has a term such
as (ami ay )y{8w + 8s ) for the opportunity cost of holding stocks. Resource
stocks must also be found and maintained with an additional term
{amjay)s . The prices of a European call option on resource stocks for an
exercise price of $156/ton are shown in Figure 2.4.
8 Ei ~
8 8. ..::e!!!
8 8. ...: ~ ~ "..
~ , 8 ..: ~ ~ ~ ~ - Underlying Price
~ ~
~ - ;! - ($/unit)
8 8.,.; Ie
Time 8~~
~
(weeks) 8 8.,.;:g ~ -
!i! :; ~ 8 ~ ~ § ~ ~ - Underlying Price
~ :Ii ~ - (S/unit)
:! -
In this case, less flexibility makes call options on resource stocks more
valuable than call options on financial stocks. This can be explained by
the cost of finding and maintaining the stocks. The owner of the stocks
has already invested in exploration and maintenance. If anyone else
wants to own a call option on resource stocks, they must also invest or
purchase the existing stocks at the option price.
2.3.4. Options on the Environment
Options on the environment are not a new idea. Rainfall insurance and
hail insurance are examples (Bardsley, Abbey and Davenport, 1990;
Quiggin, 1994). Weather derivatives are now available for energy com-
panies, concert promoters and farmers to insure against unfavorable
temperatures as well as rainfall. Or, consider a water right. A water right
is a call option on the flow of a stream or river. Combining equations
(2.6) and (2.11) shows that options can be written on almost any envi-
ronmental variable.
change, eOe, from equation (2.3) gives a partial differential equation for
the price of options on the environment.
oe oe ( -
-+- X -x ) + 72--X
1/ 02e 2 2
ax -eow =0;
01 oX oX2
i - X; i
e (r y ) =
> X
{
, 0; i ~X.
0.250000
0.200000
O.loUUU'J~-
Option Price
(S/unit)
0.1 nnnrln -<.-.• -
0.050000
.. 0
N'"
co(t~~"" Time
Environmental Variable
($/unit)
.
., ~ ..
.. !iI
ill !:j N (weeks)
average year is expected, the price of the water right is $0.054. If only
70% of average is expected, the price of a water right is $0.104.
For comparison, Figure 2.7 shows the prices that would be predicted if
water rights were valued as if they were put options on futures. If an av-
erage year is expected, the predicted value for the water right would be
$0.067. However, if only 70% of average is expected, the predicted value
would be $0.216, twice the value predicted in Figure 2.6.
As with all options, prices of water rights are independent of the
agent's risk aversion. In addition prices are independent of the correla-
tion between stream flow and production. However, risk aversion and
correlations affect the quantity of water rights an agent will buy. This
quantity can be found by solving the optimality conditions in equations
(2.5) through (2.11) as a system. Solving the system several times for all
users of the river will give the demands and supplies of water rights.
0.1
Option Price
($/unit)
0.1
... 0
",'"
~ r! co
"'-
ocid~~o ",<1i~- Time
o c:i ~ ;J; co o ~ ~ '" (weeks)
Environmental Variable
($/unit)
0 ci ~ ~
- ......
..,",:;j:'"
formula does not apply. A new option pricing formula is derived which
considers the provisions of the contract that underlies the option. A sim-
ple example is a forward contract with a fixed contract price over the life
of the contract. This inflexibility lowers the value of options on forward
contracts compared with options on futures. Another example is the
value of a water right. The contract level must be agreed when a water
right is purchased. Because of this inflexibility, a water right is much less
valuable than would be predicted by the Black-Scholes formula or any of
the usual methods of calculating cumulative probabilities.
However, this new option pricing formula is far from a complete model
of options on the environment. For example, exploration for resources
and creating a call option on resource stocks is a complex production
problem and not a simple investment as modeled here. Further, the price
of minerals is an endogenous variable in the system. Perhaps a major
area for future work is options on environmental variables that are en-
dogenous. Society may want to invest in options on old growth forests,
but the value of those same forests depends upon how much forest is
preserved. The benefits of an option that is exercised should be meas-
ured by consumer's willingness to pay, producer's benefits and scarcity
rent. Most of the large environmental questions, such as greenhouse gas
emissions, overpopulation, and available farmland, are in this category.
Pricing these options is not an impossible task. Notice that equation
(2.1) for the formulation of a stochastic dynamic program is, itself, an
option pricing formula, complete with boundary conditions. Expected so-
cial welfare is the value of a call option on wealth as affected by deci-
sions over the planning horizon. As one example, the precautionary prin-
ciple for greenhouse gases can be put into practice by formulating a
suitable model of society's wealth during global warming and solving
equation (2.1). The result will show how much society is willing to invest
to avoid the risks of global warming.
To derive the stochastic differential equation for wealth, begin with soci-
ety's wealth as the sum of all assets and liabilities.
W =bB+aA.
The Precautionary Principle in Practice 47
The first and second terms on the right hand side are capital gains or
losses on beginning inventories of bonds and risky assets. The third and
fourth terms are acquisitions and depreciation valued at ending prices.
Assume that the quantity of bonds can change over time by acquisitions
and the quantities of risky assets can change by acquisitions and physi-
cal degradation.
dB = Bdt;
dA = (A -A8A)dt -AaAdZA.
Acquisition of bonds is B and acquisition of risky assets is A. Risky
assets are expected to degrade by A8A , with error AaAdZA . Substituting
these changes in bonds and risky assets gives another expression for
the change in wealth.
The left hand side of this equation is the stochastic profit at the end of
each decision interval. Profit substitutes for acquisitions in the change in
wealth.
dW = db[W -aA]/b +daA -a8AAdt -aAaAdZA +JZCit +d1r.
dW = db [W - yS - mM - nN - eE ]/ b
+ dyS - yosSdt + dmM + dnN + deE + Jrdt + dJr.
Resource stocks, S, are valued at the price for all commodities in the
economy, y. Options on forward contracts, M, options on futures, N, and
options on the environment, E, have market prices m, nand e.
Society benefits from production, plus any gains from forward and fu-
tures contracts. Subtracting consumption expenditures, costs of produc-
tion and costs of finding and maintaining resource stocks gives profit.
d Jr = dyY + (y + dy) dY - dX
+ [ -qQ + yY - c (Y) - sS + (X - X) K + (h - y) H] dt
+dyY + (y + dy )dY - dX
db = bowdt;
dy = YOydt + YCJydzy ;
df = fOfdt + fCJfdz f
dY = YCJydZy ;
dX =XCJxdZx .
For society, the commodity price and production will be negatively cor-
related with covariance dydY equal to CJyCJYWyY, where WyY is the cor-
relation coefficient.
Options on forward contracts, options on futures and options on the
environment are assets that must be purchased at prices m, nand e. The
value of an option on forward contracts is a function of the commodity
price, y, and the time to maturity, r - t, where r is a maturity date in the
future. In other words, the value of an option on forward contracts is the
function m(t,y). Its differential equation is found by stochastic differentia-
tion.
1/ 8 2n 2 2]
en en + /2-2-f en
dn = [-+-fo,
a a a a, dt+-fa,dz,.
if
ie
de= [ -+/2--2 1/ ie 2 2]
8 2e X ax dt+-Xaxdzx·
a ox ox
Finally, substituting in these expectations for options on forward con-
tracts, options on futures and options on the environment gives the sto-
chastic differential equation for wealth.
References
Bardsley, P., A. Abbey and S. Davenport (1984), "The Economics of Insuring Crops
against Drought." Australian Journal of Agricultural Economics, 28:1-14.
Bardsley, P., and P. Cashin (1990), "Underwriting Assistance to the Australian Wheat
Industry: An Application of Option Pricing Theory." Australian Journal of Agricultural
Economics, 34:212-222.
Black, F., and M. Scholes (1973), "The Pricing of Options and Corporate Liabilities."
Journal of Political Economy, 81 :637-654.
Burden, R. L., and J. D. Faires (1997), Numerical Methods. Sixth Edition. Pacific
Grove, California: Brooks/Cole Publishing.
The Precautionary Principle in Practice 51
Conrad, J. M. (1997), "On the Option Value of Old-Growth Forest." Ecological Eco-
nomics, 22:97-102.
Dixit, A. K, and R. S. Pindyck (1995), Investment Under Uncertainty. Princeton:
Princeton University Press.
Grimmett, G. R., and D. R. Stirzaker (1992), Probability and Random Processes. Sec-
ond Edition. Oxford: Oxford University Press.
Hertzler, G. (1991), "Dynamic Decisions Under Risk: Application of Ito Stochastic
Control in Agriculture." American Journal of Agricultural Economics, 73:1126-1137.
Hertzler, G., J. Harman and R. K Lindner (1997), "Estimating a Stochastic Model of
Population Dynamics with an Application to Kangaroos." Natural Resource Model-
ing, 10:303-343.
Hull, J. C. (1997), Options, Futures and Other Derivatives. Third Edition. Upper Saddle
River, New Jersey: Prentice Hall.
Jarrow, R. A. (1999), "In Honor ofthe Nobel Laureates Robert C. Merton and Myron S.
Scholes: A Partial Differential Equation That Changed the World." Journal of Eco-
nomic Perspectives, 13:229-248.
Just, R. E., L. Calvin and J. Quiggen (1999), "Actuarial and Asymmetric Information
Incentives." American Journal of Agricultural Economics, 81 :834-849.
Mahul, O. (1999), "Optimal Area Yield Crop Insurance." American Journal of Agricul-
tural Economics, 81 :75-82.
Merton, R. C. (1990), Continuous-Time Finance. Cambridge Massachusetts: Black-
well.
Petzel, T. E. (1984), "Alternatives for Managing Agricultural Price Risk: Futures, Op-
tions and Government Programs." AEI Occasional Paper. Washington DC: Ameri-
can Enterprise Institute for Public Policy Research.
Quiggin, J. (1994), 'The Optimal Design of Crop Insurance." Economics of Agricultural
Crop Insurance: Theory and Evidence. D. L. Hueth and W. H. Furtan, Editors, pp
115-134. Boston: Kluwer Academic Publishers.
Stokes, J. R., W. I. Nayda, and B. C. English (1997), "The Pricing of Revenue Assur-
ance." American Journal of Agricultural Economics, 79:439-451.
3
Factors Determining Best Management
Practice Adoption Incentives and the Impact
of Green Insurance
Paul D. Mitchell and David A. Hennessy
chooses one of three production systems - the status quo system, the
BMP, or the BMP with green insurance.
The input B can be a naturally occurring substitute for x so that to > 0
and txo < 0, as when x is nitrogen fertilizer and B is nitrogen from natural
sources. Alternatively, B can measure the level of a naturally occurring
"bad" so that to < 0 and tXiJ > 0, as when x is a pesticide and B measures
the pest population. Because the natural processes and factors deter-
mining B are many and complex, the producer treats B as a stochastic
input with a distribution depending on available information.
Random production shocks independent of B are aggregated into the
random variable E: with density h(E:). Assume many local factors that are
only observed with difficulty determine E: so that it is non-contractible.
Realized yield could be used to determine E: ex post, but because yield is
subject to manipulation by the producer, accurate determination requires
prohibitively high monitoring costs and is not feasible.
3. 1. 1. Status quo
For the status quo, B follows the unconditional density v(B). The pro-
ducer's objective and first order condition are:
Maxf,.foU(Jr)v (B)h (E:) dB dE: (3.1 a)
fefou'(Jr)(fx(x,B,E:) -r)v(B)h(E:) dB dE: = 0 (3.1 b)
where, Jr = t (x,B,E:) - rx and r is the per unit price for x when normalized
by the output price. Single and double primes indicate first and second
derivatives respectively and subscripts indicate partial derivatives with
respect to the subscripted variable(s). Use x
> 0 to denote the optimal
value of x implicitly defined by the first order condition and use EUSQ to
denote the value of the objective function when evaluated at this opti-
mum. Since txx < 0 and u"< 0, the second order condition is satisfied
and EUSQ is a maximum.
3.1.2.8MP
For the BMP, the producer first collects imperfect information concerning
B, the-n chooses x. Denote this ex ante signal of the true level of B as a
and the resulting conditional density of B as q(B I a). The producer
chooses the optimal decision rule x * (a) to maximize expected utility
given a. The objective and first condition are:
Maxf,}oU(Jr)q(B I a)h(E:) dB dE: (3.2a)
56 Paul D. Mitchell and David A. Hennessy
(3.2b)
where 7T = '(x, f), c:) - rx. For any a the first order condition implicitly de-
fines an x * (a) > 0. Expected utility for this system is the expected value
of the objective at x*(a): EU SMP =
faJ..foU(7T(X * (a)))q(f) I a)h(c:)g(a) df) dc: da , where g(a) is the density of a.
Since 'xx < ° and u" < 0, the second order condition is satisfied and
EUSMP is a maximum.
Statistically the relationship between a and f) is described by the jOint
probability density b(f),a) , such that a and f) have a positive covariance.
Note that b(f),a) can be factored into q(f) I a) and g(a), the conditional
density of f) and the unconditional density of a respectively. The uncon-
ditional density v(O) observed when using the status quo production
system is obtained by integrating b(f),a) over a.
The information a provides is imperfect for reasons specific to each
BMP, but generic examples include measurement error inherent in the
information technology, a time lag occurring between the observation of
a and the realization of 0, or spatial variation in the realization of f). Sig-
nals of this sort include the late spring soil nitrogen test, which imper-
fectly estimates soil nitrogen available later that summer, and scouting
for adult insects as part of corn rootworm IPM, which imperfectly esti-
mates the corn rootworm larval population the following spring.
Green insurance covers losses due to BMP failure. Two types of BMP
failure can occur - either too little or too much of the input x is applied,
resulting in profit loss due to reduced yields or wasted expenditures. The
imperfect nature of a can mislead the producer into applying too little or
too much x, or waiting to collect and process a can result in factors such
as weather preventing or delaying application of x. Also errors can occur
because the producer does not completely or correctly understand the
BMP, or the BMP as recommended is not quite correct for the producer's
land or operation.
Producers have been primarily concerned with input deficiencies and
associated yield losses. Input surpluses are usually less costly to pro-
ducers, since yield loss is minor (if it occurs at all) and typically difficult
to detect. Agronomic research has also focused primarily on detecting
input deficiencies and measuring associated yield losses to develop rec-
ommended application rates. BMP failures that cause input deficiencies
are a concern to policy makers as well, since such failures are often easy
Best Management Practice Adoption Incentives 57
to observe and create a bad reputation for the BMP and so reduce adop-
tion. Thus in the analysis here, BMP failures that result in over applica-
tion of the input x are ignored and the focus is on failures that result in an
input deficiency.
Indemnities are based on a signal 5 from which expected yield loss due
to input deficiency as a result of BMP failure can be inferred. As a result
of measurement error, spatial variation, time lags, and other such fac-
tors, the signal 5 is an imperfect signal of input deficiency and depends
on x and 0 since they determine this deficiency. Denote its conditional
density as W(5 I x,O) . Define 5 such that high values of 5 indicate a small
input deficiency and vice versa, so that 5 is positively correlated with O.
This specification does not explicitly include a, but because x depends
on the observed a, the specification depends indirectly on a and as-
sumes that all dependence of 5 on a is mediated by x. Signals of this sort
already exist for important BMPs such as the late spring soil nitrogen test
or corn rootworm IPM. The end of season stalk test and the SPAD chlo-
rophyll meter reading measure nitrogen deficiency and the root rating
and percent of stand lodged measure corn rootworm damage (Varvel,
Schepers, and Francis, 1997; Spike and Tollefson, 1991).
Denote the indemnity received as 1(5, fl) , where fl is an index of the
level of coverage such that I~ > O. Because the policy insures against
BMP failure from input deficiency and 5 and 0 are positively correlated, te
and 15 should have opposite signs. Because 5 and 0 are positively cor-
related, high realizations of 5 are associated with high realizations of 0
and vice versa. If te > 0, then high realizations of 0 and 5 are associated
with high yields and low losses, so that indemnities should decrease in 5,
but if te < 0, then high realizations of 0 and 5 are associated with low
yields and high losses, so that indemnities should increase in 5. In either
case, te and 15 have opposite signs.
Denote the premium paid as M(fl) , which can be actuarially fair or
greater to allow the insurance provider to cover costs and earn a normal
profit. Lastly, it is assumed that the producer does not choose the level
of coverage fl, but that green insurance is a take-it-or-Ieave-it offer. Sub-
sequent research can derive and characterize the optimal fl.
The producer purchasing insurance chooses the optimal decision rule
x * (a, fl) by maximizing expected utility conditional on the observed a
and offered fl. The objective and first order conditions are:
MaxI..fofs u (n)W(5 I x,O)q(O I a)h(e) d5 dO de (3.3a)
fEfoIsu'(n)(fx(x,O,e) -r)w(5 I x,O)q(O I a)h(e) d5 dO de +
58 Paul D. Mitchell and David A. Hennessy
where, 1C = f(x,B,e) -rx +/(s,f3) -M(f3). For any a and f3, (3.3b) implicitly
defines x * (a,f3) > 0, which may differ from x * (a) due to the insurance.
Expected utility is the expected value of the objective function at
x * (a, f3): faf&fofsu(1C(X * (a,f3)))w(s I x,B)q(B I a)h(e)g(a)ds dB de da. Note
that fxx < 0 and u" < 0 are no longer sufficient to satisfy the second order
condition. Rather the second order condition is satisfied by assumption
to ensure that EU G1 is a maximum.
Table 3.1 summarizes the sequence of events and clarifies the infor-
mation set available for each production system. In each system, pro-
ducers choose optimal input use as a function of their information. For
the status quo system, input use is constant at X. For the BMP, produc-
ers choose x* as a function of the observed a. With insurance, producers
choose x* as a function of a, but adjust their choices to respond to the
insurance coverage f3. Note that for all three systems, B and a are jointly
distributed according to b(B,a). For the status quo, a is ignored so that it
seems B - v(B), where v(B) = fab(a,e)da as in (3.1), while for (3.2) and
(3.3), b(e,a) is factored into g(a)q(e I a).
Intuitively, as the signal a becomes noisier and more variable, the in-
centive to adopt the BMP should decrease. Similarly, as the information
content of the signal a improves so that it more accurately reflects true
input availability, adoption incentives should increase. In Proposition 1,
T2 = O.50';u'(,r)fxxx; and T3 = O'aou'(,r)fxoxa capture these effects. As pre-
viously discussed, T2 < 0 and T3 > O. For producers using a, these two
terms describe the necessary tradeoff between the welfare gain from the
information in a resulting from its covariance with e and the welfare loss
due to the noise and variability in the same signal.
Since the variance of a depends positively on both the variance of e
and signal noise, research that reduces either factor increases adoption
incentives. Reducing signal noise also increases the correlation with e,
further increasing adoption incentives. Improved information technology
reduces signal noise and alternative practices can reduce the variance of
e. For example, research can indicate how to manage soils to reduce
nutrient variability by using manure applications, cover crops, and differ-
ent crop rotations, while boll weevil eradication and area wide manage-
ment of corn rootworm have successfully reduced the variance of pest
pressure.
pends on Iss (the curvature of profit in 5) and risk aversion (the curva-
ture of utility in profit). Because yield is concave in 0, damage or yield
loss is convex in o. Then because 5 and 0 are positively correlated, a
reasonable assumption is that expected damage or yield loss is also
convex in 5. Given that indemnities should be positively correlated with
losses, the indemnity schedule should also be convex in 5 (Iss > 0),
though this need not be the case, since the insurer chooses the sched-
ule. If Iss > 0, producers prefer increasing the variance of 5 since it in-
creases expected profit, but due to risk aversion, producers prefer to
avoid the increased income variability. As a result, the impact of the vari-
ance of 5 is analytically ambiguous.
Intuitively, producers prefer indemnities that are tightly correlated with
actual losses, so that insurance decreases income variability by paying
the largest indemnities when most needed and vice versa. However,
since the insurance signal is imperfect, some "insurance risk" remains
because the insurance can pay indemnities that are in excess of losses
or too low and thus destabilize income. The positive correlation between
5 and 0 captures this correlation between the insurance signal and actual
losses-if 0 is low,s tends to be low so that indemnities are high, and
vice versa. In Proposition 2, T5 = O"sou'RI.t1J supports this intuition since it
is positive. As the covariance between 5 and 0 increases, the insurance
further reduces income variability, and the greater the risk aversion, the
greater the increase of adoption incentives.
Interestingly, if Iss> 0, developing new or alternative insurance signals
that are less noisy and more accurately reflect true yield losses has an
ambiguous impact on the adoption incentives of producers purchasing
the insurance. If the correlation between the signal 5 and true input
availability 0 increases, the insurance risk is reduced and adoption in-
centives increase with insurance coverage. However, if the increased
correlation is associated with a reduction in the variance of 5, then adop-
tion incentives decrease because of the convexity of the indemnity in 5. If
this convexity effect is ignored, the anticipated returns to research that
improve the insurance signal in this manner will be less than actually
realized.
The impact of the offered level of insurance coverage on adoption in-
centives depends on the how the indemnity and premium schedules are
designed. In Proposition 2, insurance coverage has a first order and a
second order effect on adoption incentives, captured by T6 and T7 re-
spectively. Intuitively the first order effect should be positive - insurance
coverage should increase adoption incentives, which requires (1/3 - M /3) >
Best Management Practice Adoption Incentives 63
3.4. Conclusion
Appendix
Proof of Proposition 1.
Status quo utility is a function of x , B, and e. Take a second order Tay-
lor series expansion at ( x(a) , B , e), where x(a) denotes the status quo
optimum x when a = a:
u(K(x,B,e)) == u(f - rX(a)) + (x - x(a))(u'(fx - r)) +
(B - B)u' to + (e - e)u' ~ + O.S(x - x(aW[u" (fx - r)2 + u' fxx] +
O.S(B - B)2[U" f,} + u' f(I(J] + O.S(e - e)2[u" ~2 + u' faJ +
(x - x(a))(B - B)[U" f,,(fx - r) + u' fx"l + (x - x(a))(e - e)[u" fe(fx - r) + u' fxel +
(B - B)(e - e)[u" f,,~ + u' feel. The utility function and its derivatives are
evaluated at f - rX(a) and the crop growth function and its partial deriva-
tives are evaluated at (x(a) , B, e). Noting that when a = a,
v(B) = q(B I a) in (3.1 a), take the expected value of both sides:
Proof of Proposition 2
References
Babcock, B.A., and D.A. Hennessy (1996), "Input Demand under Yield and Revenue
Insurance." American Journal of Agricultural Economics, 78: 416-427.
Chambers, R.G., and J. Quiggin (1996), "Non-poi nt-Source Pollution Regulation as a
Multi-task Principal-Agent Problem." Journal of Public Economics, 39: 95-116.
Feather, P., and J. Cooper (1995), "Voluntary Incentives for Reducing Agricultural
Nonpoint Source Water Pollution." U.S. Department of Agriculture, Economic Re-
search Service, Agricultural Information Bulletin No. 716. Washington, DC, May.
66 Paul D. Mitchell and David A. Hennessy
Greene, C.R., RA Kramer, G'w. Norton, EG. Rajotte, and R.M. McPherson (1985),
"An Economic Analysis of Soybean Integrated Pest Management." American Jour-
nal of Agricultural Economics, 67: 567-572.
Harper, J.K., M.E Rister, J,W. Mjelde, B.M. Drees, and M.O. Way (1990), "Factors
Influencing the Adoption of Insect Management Technology." American Journal of
Agricultural Economics 72: 997-1005.
Hennessy, D.A., and BA Babcock (1998), "Information, Flexibility, and Value Added.
Information Economics and Policy, 10: 431-449.
Hrubovcak, J., U. Vasavada, and J.E Aldy (1999), "Green Technologies for a More
Sustainable Agriculture." U.S. Department of Agriculture, Economic Research
Service, Agricultural Information Bulletin No. 752. Washington, DC, July.
Miranowski, JA, U.F,W. Ernst and F.H. Cummings (1974), "Crop Insurance and In-
formation Services to Control Use of Pesticides." U.S. Environmental Protection
Agency Research Report EPA-600/5-74-018.
Nowak, P. (1992), "Why Farmers Adopt Production Technology." Journal of Soil and
Water Conservation, 47: 14-16.
Osborn, C., D. Hellerstein, C. Rendleman, M. Ribaudo, and R. Keim (1994), "A Pre-
liminary Assessment of the Integrated Crop Management Practice." U.S. Depart-
ment of Agriculture, Economic Research Service, Staff Report AGES 9402, Febru-
ary.
Smith, V.H., and B.K. Goodwin (1996), "Crop Insurance, Moral Hazard, and Agricul-
tural Chemical Use." American Journal of Agricultural Economics, 78: 428-438.
Spike, B.P., and J.J. Tollefson (1991), "Yield Response of Corn Subjected to Western
Corn Rootworm (Coleoptera: Chrysomelidae) Infestation and Lodging." Journal of
Economic Entomology 84: 1585-1590.
U.S. Department of Agriculture, Natural Resource Conservation Service. Environ-
mental Quality Incentives Program (EQIP) Fact Sheet (1977a).
http://www.nhq.nrcs.usda.gov/OPNFB960PNeqipfact.html.
U.S. Department of Agriculture, Natural Resource Conservation Service. Environ-
mental Quality Incentives Program (EQIP) Questions and Answers (1977b).
http://www.nhq.nrcs.usda.gov/OPNFB960PNeqipQ%26A.html.
Varvel, G. E., J. S. Schepers, and D. D. Francis (1997), "Chlorophyll Meter and Stalk
Nitrate Techniques as Complementary Indices for Residual Nitrogen." Journal of
Production Agriculture, 10: 147-151.
Westra, J., and K. Olson (1997), "Farmers' Decision Processes and Adoption of Con-
servation Tillage." Department of Applied Economics Staff Paper P97-9, University
of Minnesota.
Wu, J.J., and BA Babcock (1996), "Contract Design for the Purchase of Environ-
mental Goods from Agriculture." American Journal of Agricultural Economics, 78:
935-945.
4
Uncertainty and Adoption
of Sustainable Farming Systems
David J. Pannell
Marsh et al., 2000). There is also evidence that even for innovations ori-
ented towards resource conservation, economic considerations are the
most important determinants of actual adoption decisions (Cary and Wil-
kinson, 1997; Sinden and King, 1990).
4.1.4. Impacts of uncertainty on adoption
Within the adoption process, uncertainty has several negative influences.
The key ways in which uncertainty inhibits adoption are as follows.
The fact that the final result of adopting a particular practice is highly
uncertain is an intrinsic discouragement to adoption for most people. As
noted earlier, most farmers are averse to risk and uncertainty, meaning
that they place greater weight on potential negative outcomes than on
positive outcomes. This relates to the condition above of meeting the
farmer's objectives. For some farmers, avoidance of risk and uncertainty
is an important objective.
Even if farmers are not discouraged by uncertainty per se, they may
well be discouraged by the consequences of that uncertainty, particularly
if it results in inaccurate perceptions or misinformation. This also relates
to the condition regarding farmers' objectives. If a farmer perceives incor-
rectly that an innovation is not consistent with his or her objectives, this
misperception is an impediment to adoption. The condition relating to
trialling is also relevant here. If the farmer does not conduct trials, a
chance to correct the misinformation is missed. Indeed, if the farmer is
badly misinformed, this in itself may cause the farmer to believe that a
trial is not worthwhile, trapping him or her in a state of ignorance.
Irreversibility of environmental damage is often proposed as a reason
for action to enhance conservation. This is a different motivation than
aversion to uncertainty. It relates to the concept of "option value"
whereby keeping open the options for resource use has a positive value
due to the potential for unforeseen circumstances. To the extent that an
option value is relevant to the farmer's objectives, it may influence his or
her behavior. This appears to act in favor of adoption. However, if a con-
servation practice is itself irreversible to some extent (or expensive to
reverse), there is then an option value in not adopting it. For example,
this would apply to the strategies that involve planting of trees on a pro-
portion on cropland (as recommended in parts of Australia to prevent soil
salinization in non-irrigated regions).
Uncertainty and Adoption of Sustainable Farming Systems 71
n =Gs x As + Gn x An ( 4.1)
If the farmer maximizes profit for the current period, some area of the
sustainable farming system will be grown so long as the gross margin of
sustainable farming is greater than that of traditional farming on any part
72 David J. Pannell
where,
The gross margins have time subscripts in part because they are
changing due to land degradation, and in part because the sustainable
system is likely to have up-front costs and delayed payoffs.
If the farmer chooses not to trial the sustainable system in year 1, the
profit function is:
/ = NPVt=2 .. N[Gst x Ast +Gnt x (AT - Ast)-Gsto xAsto-GntO x(A T-AstO)] (4.5)
Rearranging gives:
4.3.3. Externalities
Some land degradation problems have important "external" impacts. For
example wind erosion on one farm may impose costs on another farm,
such as "sand-blasting" of crops, or burial of fences. Externalities can con-
tribute two different types of uncertainty about the consequences of adop-
tion of land conservation practices. Firstly, a farmer may be uncertain
about who will be the beneficiary if he or she does adopt. If there is a risk
that the benefits will flow mainly to farmers other than the adopter, the in-
centives to adopt are reduced. Secondly, a farmer may be uncertain about
whether their adoption will be ineffective if other farmers do not adopt. For
example, some hydrological catchments span more than one farm, such
that several farms in the catchment contribute to rises in a saline water
table. In such cases, adoption by anyone individual may make a relatively
small contribution to preventing rises in the water table, although further
rises would be prevented if all farmers adopted.
Further, long time scales mean that uncertainty about other variables (e.g.
prices) over the relevant time scale is much greater than for a short-term
problem, further adding to the difficulty of decision making.
4.4.2. Heterogeneity of the land
In the last section, the spatial heterogeneity of land degradation problems
was recognised as an impediment to diffusion of innovations from farm to
farm. The same issue applies at the scale of a single farm. A large part of
the potential information value of a trial is derived from its relevance to
other parts of the farm. If a farmer perceives that the trial results are less
than fully transferable, the trial's benefits are reduced.
4.4.3. Minimum scale needed
For many agricultural innovations, it is possible to conduct trials on a small
scale without sacrificing much of the information content of the trial. For
example, new crop types are typically trialled on a scale that represents
just a few percent of the total area of crop on a farm. As knowledge of and
confidence in the crop increases, the scale of production increases. By
contrast, for innovations intended to prevent dryland salinity by increasing
water use, a small-scale trial may have no measurable impact. Especially
when combined with long time scales and geological heterogeneity, the
scale necessary to have an observable impact in a reasonable time may
be little smaller than full-scale adoption. Farmers would naturally be reti-
cent about leaping to such full-scale adoption given their state of high un-
certainty.
4.4.4. Observability
Related to the problem of minimum scale is the issue of observability.
Clearly, low observability of results reduces the information value of a trial.
Examples would include practices intended to reduce soil degradation in-
volving underground processes, such as soil compaction, soil acidification,
soil salinization or leaching of nutrients. Of course, some consequences of
adoption of the practice would be observable in their impacts on above-
ground plant growth. However, if the prime motivation for adoption is pre-
vention of below ground processes, above ground production may provide
a highly imperfect indicator.
4.4.5. Low covariance with traditional practices
Even if a conservation practice is easy to trial on a small scale, giving ob-
servable results quickly and providing information that is relevant to the
whole farm, the information value of the trial may be low relative to most
Uncertainty and Adoption of Sustainable Farming Systems 77
4.6. Implications
Based on this discussion, a number of clear implications can be identi-
fied. Firstly, it appears that the problem of uncertainty in adoption of land
conservation practices is much greater and more far reaching than nor-
mally recognized. The fact that farmers have been slow to take up some
innovative land conservation practices is highly understandable when
viewed within the context of the issues raised here - even without con-
sidering the range of other negative influences on adoption of these
practices (Pannell and Schilizzi, 1999).
It does appear that uncertainty is an important cause of market failure
in this case. However, it is not clear whether government intervention can
reduce the extent of this failure. On one hand, government agencies may
be in possession of information from scientific research and other
sources that is in some sense better than that held by at least some
farmers. On the other hand, even if this is true, its accuracy at particular
sites may be unknown, and assessment of its management implications
for particular farmers will certainly be outside the capacity of agencies.
Given the heterogeneity discussed here, such an assessment depends
very much on local knowledge and individual circumstances. Farmers
understand this well, and so are most unlikely to be influenced by advice
from agencies that they should adopt particular practices. Even if the
Uncertainty and Adoption of Sustainable Farming Systems 79
advice is good, it will probably not be believed, and for sound and pru-
dent reasons. Information on biophysical aspects that does not attempt
to draw management implications for individual farmers is less suscepti-
ble to this problem.
One prominent government response to land degradation problems in
Australia has been the National Landcare Program, a central feature of
which is the formation of formal farmer groups. These playa role in col-
lection and sharing of information, and in this they appear to be partially
addressing the problems of uncertainty discussed here. In particular the
following advantages of the Landcare group approach might be ex-
pected.
• It can speed the flow of information between individuals in the
group,
• It may help to facilitate jOint trials. If farmers agree to share costs,
the problem of high trial costs can be partially avoided. (In practice,
this appears to be uncommon).
• Joint trials, because they are local and farmer-run, have greater
local relevance and credibility than agency information from other re-
gions.
• Perhaps the jOint effort involved may reduce the risk of poor im-
plementation.
Although these are important advantages, it appears that there has
been excessive optimism in some quarters about the extent to which the
Landcare approach can solve the problems of information and uncer-
tainty, especially for the most intractable problem of dryland salinity. In
particular, it seems unlikely that Landcare groups could do much to ad-
dress the following problems discussed earlier.
• The contribution of externalities to high uncertainty.
• The contribution of heterogeneity to high uncertainty.
• Long time scales.
• Cases where the minimum scale needed for trials is large.
• Low observability of some trial impacts.
• Low covariance of the behaviour of the innovation with traditional
practices.
• The high cost of ceasing a trial.
It may be worthwhile for government programs intended to promote
adoption to devote resources to attempting to devise innovative methods
for addressing these aspects of uncertainty.
Another strategy that would avoid several of these remaining problems
would be to attempt to develop technologies which are profitable in their
80 David 1. Pannell
Endnotes
References
Abadi Ghadim, A.K. and D.J. Pannell (1998), "The importance of risk in adoption of a
crop innovation: Empirical evidence from Western Australia". Paper presented at the
42nd Annual Conference of the Australian Agricultural and Resource Economics
Society, University of New England, Armidale, NSW Jan 19-21 1998.
Abadi Ghadim, A.K., M. Burton, and D.J. Pannell (1999), "More empirical evidence on
the adoption of chick peas in Western Australia". Paper presented at the 43rd An-
nual Conference of the Australian Agricultural and Resource Economics Society,
Christchurch, New Zealand, Jan 20-22 1999.
Abadi Ghadim, A.K. and D.J. Pannell (1999), "A conceptual framework of adoption of
an agricultural innovation", Agricultural Economics, 21: 145-154.
Antle, J.M. (1987), Econometric estimation of producers' risk attitudes, American
Journal of Agricultural Economics, 69: 509-522.
Bardsley, P., and M. Harris (1987), "An approach to the econometric estimation of
attitudes to risk in agriculture", Australian Journal of Agricultural Economics, 31 :112-
126.
Bartle, J.R., C. Campbell, and G. White (1996), "Can trees reverse land degradation"?
Australian Forest Growers Conference, Mt Gambier, South Australia.
Cary, J.W. and R.L. Wilkinson (1997), "Perceived profitability and farmers' conserva-
tion behaviour", Journal of Agricultural Economics, 48: 13-21.
Uncertainty andAdaptian afSustainable Farming Systems 81
and B = {bt, b2, ... , brrJ. In this case set B includes activities required to
grow a traditional non-GM crop.
Option 3: Using the GM crop illegally. There are two main cases in
option 3. Firstly, a farmer may sign the contract to grow the GM crop yet
may knowingly or unwittingly not abide by all its terms and conditions.
This farmer's actions are represented by the activity set C where C = (Ct,
c2, , crrJ and the farm generates profit "C where "C = Yc-C c . The ex-
pected profit can be expressed as:
where, "C is the optimal profit generated by utilizing activity set C, the
penalty for scheme violation is Vc and the probability of detection is
P(VC)·
Secondly, a farmer may opt to not become a licensee yet the GM seed
is obtained and used illegally. In this case no contract would be signed
and the farmer's actions are activity set 0, the penalty for scheme viola-
tion is Vo and the probability of detection is p(Vo). Expected profit can
be expressed as:
(5.2)
Farm
Profit
($)
P
For the farmer to accept the gamble that their violations will be de-
tected, as shown in Figure 5.1, then p(V) and V must be suffiCiently small
Incentive Design for Introducing GM Crops 87
to provide the required incentive. If returns are less than Y'a-C'a then the
farmer would rather not adopt the GM crop.
The above decision problem highlights a few areas in which adoption
of GM crops by risk neutral farmers can be influenced and the illegal and
improper use of proprietary technology can be reduced. The options are
to increase p(V) or V or both. As shown in Figure 5.1 if V is a function of
Ya-Ca then increasing this difference will increase V. Increasing Ya-Ca
can be achieved in various ways. For example, a rigorous scrutiny of the
activity elements of set A may reveal better, fewer or cheaper ways to
grow successfully the GM crop and therefore reduce Ca. The size and
nature of the technology fees charged by the owners or licensors of the
GM technology is obviously an important component of Ca.
The proprietary technology may enable farmers to increase Ya , by
higher yields through better pest and weed control, better supply-chain
management and improved marketing. Assuming the increase in Ya is
also associated with increases in Ya-Ca and that V is positively related
to Ya-Ca as in Figure 5.1, then the greater size of V is a further disincen-
tive for illegal and improper use of the proprietary technology. Also, in
the future if price premia for GM crops arise, due to their quality im-
provements, then Ya may increase.
Increases in p(V) are possible through a range of measures such as
the licensor allocating more resources toward surveillance, rewarding
those who inform against illegal use of GM products and widely broad-
casting news about prosecutions. The purpose of such litigation would be
twofold; firstly to ensure the cost to a farmer of being detected (V) was
very high and secondly to publicize this cost and to create the impression
that the owners of the GM technology property rights were keen to detect
breaches of their proper use (Le. p(V) was not negligible). Further, in-
creases in p(V) may be possible due not to the actions of the owners of
the GM technology but rather due to the actions of either regulators or
purchasers of non-GM crops and crop products. To maintain consumer
confidence in the integrity of the non-GM status of their products some
purchasers may insist on testing the grain or product delivered to them,
thereby increasing the likelihood of detection of growers who use non-
GM marketing channels to sell their illegally grown GM crops. Failure of
growers to supply non-GM grain or product could result in fines or dock-
ages. Some government regulators may also engage in monitoring, in-
spection, policing and prosecution to safeguard community concerns
about the food and environmental safety of GM crops. Hence, with such
88 Ross Kingwell
activities p(V) and V could be sufficiently high to deter the illegal growing
of GM crops.
Increases in p(V) or V cause the lines rcc and rcD in Figure 5.1 to pivot
downwards from the points where they intersect the vertical axis (farm
profit). Eventually the farmer is restricted to choosing across a range of
Ya-Ca values to either legitimately adopt or not adopt the GM crop. Thus
it is possible to remove the problem of illegal and improper use of GM
technology by setting p(V) and/or V high enough. The effect of raising
p(V) and/or V would cause lines rcC and rcD in Figure 5.1 to eventually
pass through point P. Only at this point would the farmer be indifferent
between not adopting or legitimately adopting the GM crop.
In the preceding model the only risky decision involved the probabilistic
gambles of option 3. A risk-neutral farmer would avoid illegal and im-
proper use of GM technology if rcA > rcc and rcA > rcD. However, for a
risk-averse farmer contemplating adoption of GM crops a range of un-
certainties exist. The impact of such uncertainties will influence their
adoption decisions. For such a farmer to avoid illegal and improper use
of GM technology:
(5.3)
The problem of illegal and improper use of GM technology and the re-
sponses to it can be illustrated using the case of INGARD® cotton grown
Incentive Design for Introducing GM Crops 89
Note: a Excludes any fine associated with detection of improper or illegal use. This es-
timate assumes some cost-savings by farmers through to use of cheaper inputs and
avoidance of some management costs (eg provision of refuges).
Sources: Yield data came from Table 58, NSW lint yield 1984 to 1997, Australian Com-
modity Statistics 1998 (ABARE 1998). Price data came from Table 62, Australian raw
cotton prices 1984 to 1997, Australian Commodity Statistics 1998 (ABARE 1998). Pro-
duction costs are based on published farm surveys of Australian cotton growers (Pyke,
1998; Clark and Long, 1998). All prices and costs are in Australian dollars.
where, U(tr} is the utility function of profit and U'(tr}>O and U"(tr}<O.
Following Fraser (1991), the farmer's utility function of profit can be
represented by the constant relative risk aversion form:
where, R = -U"( 1f)1f /U'( 1f), and R is the farmer's coefficient of relative
risk aversion.
90 Ross Kingwell
Table 5.2 shows the expected profit, variance of profit, expected utility
and farm management decisions for a range of risk attitudes, given the
parameter values in Table 5.1. The results in Table 5.2 are the base
case findings. The results show the importance of risk attitude in influ-
encing farmer behavior. Risk neutral or slightly risk averse farmers would
use illegally the GM crops. Moderate and highly risk averse farmers
would accept and abide by INGARD® technology agreements. In adopt-
ing this technology this latter group of farmers would experience lower
expected profits but less profit variance. The switch in farmer behavior
from preferring illegal use to lawful adoption of the INGARD® technology,
as risk aversion increases, illustrates the potential dominating effect of
profit variance as outlined in the Appendix.
Management Choices
use is
(R=O.5) 108.6 109.1 103.3 108.9 INGARD®
slightly preferred
INGARD® use is
(R=O.9) 22.23 22.25 21.99 22.22
slightly preferred
a These are the parameter values that cause £(U(1rA» = £(U(1rD» in equation (5.3).
In words, the farmer is indifferent between accepting and abiding by the terms of the
INGARD® technology agreement or using the technology illegally.
For producers with R=O.1, only relatively small changes in the prob-
abilities of detection, the severity of fines, the cotton price and the cost of
the INGARD® technology agreements are needed to overcome the
problem of illegal use. However, it also is worth noting that for producers
with R=O.5 or R=O.9, similarly small changes in the probabilities of de-
tection, the severity of fines, the cotton price and the cost of the
INGARD® technology agreements can stimulate illegal use of the tech-
nology.
The results in Table 5.3 point to only modest changes in one or a com-
bination of the following being necessary to form an incentive-compatible
policy to address the problems of illegal and improper use: the probability
of detection, the fine severity and the cost of the INGARD® technology.
Although the cotton price will also influence producer behavior, in prac-
tice it is unlikely to be accessible to control.
To-date Monsanto has shown its preparedness to adjust its INGARD®
technology fee. Initially in Australia the fee was set at $245 per hectare.
Subsequently the fee was lowered by $35 per hectare and in 1998 the
fee was further adjusted to be effectively $155 per hectare. These
changes in the technology fee were not a response to any perceived
problems of improper or illegal use of the INGARD® technology. Rather
the fee reductions were in response to growers' mixed experience con-
cerning the profitability of the technology (Clark and Long, 1998; Pyke,
1998). However, it could be inferred that such reductions in the technol-
92 Ross Kingwell
ogy fee would have lessened the likelihood of the problems of illegal and
improper use alluded to in this chapter.
Most studies of the risk attitude of Australian farmers (Bardsley and
Harris, 1987, 1991; Abadi, 1999) reveal a range of risk attitudes, with the
majority being identified as slight to moderately risk averse. If a similar
range applies to cotton farmers then the problems of illegal and improper
use could be addressed by relatively small changes in factors such as
the technology fee and the severity of fines.
After substituting for El;r} using equation (1-1), equation (1-2) can be
re-expressed and simplified to:
(1-3)
For a farmer not engaged in any illegal use of a GM crop their ex-
pected profit is:
94 Ross Kingwell
where, p is the fixed price of grain, y is the uncertain crop yield and d is
the fixed per hectare cost of production.
Their profit variance is:
Var(nr) = p2Var(y) (1-4)
Hence the profit variance associated with the illegal use of the GM crop
exceeds that for the legal use. Thus, even if the expected profit from the
legal use of the GM crop does not exceed that from its illegal use, a risk-
averse farmer may still prefer the legal use of the crop because of the
dominance of income variance in the farmer's selection decision.
References
Abadi Ghadim, A.K. (1999), Risk, uncertainty and learning in farmer adoption of a
crop innovation. Unpublished PhD thesis, Faculty of Agriculture, University of
Western Australia.
ADHAC (2001), Gene Technology Act 2000, Australian Department of Health and
Aged Care, at http://scaletext.law.gov.au/html/pasteact/3/3428/top.htm, Oct 12,
2001.
Bardsley, P. and M. Harris (1987), "An approach to the economic estimation of atti-
tudes to risk in Australia", Australian Journal of Agricultural Economics, 31 (2): 112-
126.
Bardsley, P. and M. Harris (1991), "Rejoinder: An approach to the economic estima-
tion of attitudes to risk in Australia", Australian Journal of Agricultural Economics,
35(3): 319.
Barton, J.H. (1998), "The impact of contemporary patent law on plant biotechnology
research" in SA Eberhart et al. (Eds) Intellectual property rights III, global genetic
resources: access and property rights, pp. 85-97, Madison, WI:CSSA.
Carpenter, J. and L. Gianessi (1999), "Herbicide tolerant soybeans: Why growers are
adopting Roundup Ready varieties", AgBioForum, 2(2):65-72. Retrieved July 1999
from the World Wide Web: http://www.agbioforum.missouri.edu.
Clark, D. and T. Long (1998), "The performance of Ingard® cotton in Australia in the
1997/98 season", Cotton R&D Corporation Occasional Paper, Narribri, New South
Wales, pp. 51.
Fraser, R.w. (1991), "Price-support effects on EC producers", Journal of Agricultural
Economics, 42(1):1-10.
Fulton, M. and L. Keyowski (1999), "The producer benefits of herbicide-resistant
canola", AgBioForum, 2(2):85-93. Retrieved July 1999 from the World Wide Web:
http://www.agbioforum. missouri. edu
Incentive Design for Introducing GM Crops 95
Hanson, S.D. and G.w. Ladd (1991), "Robustness of the mean-variance model with
truncated probability distributions", American Journal of Agricultural Economics,
73(2): 436-45.
Hayenga, M. (1998), "Structural change in the biotech seed and chemical industrial
complex", AgBioForum, 1(2), 43-55. Retrieved January 1, 1999 from the World Wide
Web: http://www.agbioforum.missouri.edu.
Klotz-Ingram, C., S. Jans, J. Fernandez-Cornejo, and W. McBride (1999), "Farm-level
production effects related to the adoption of genetically modified cotton for pest
management", AgBioForum 2(2):73-84. Retrieved July 1999 from the World Wide
Web: http://www.agbioforum.missouri.edu
Kalaitzandonakes, N. and R. Maltsbarger (1998), "Biotechnology and identity-
preserved supply chains", Choices, Fourth Quarter 1998:15-18.
Latacz-Lohmann, U. and P. Webster (1999), Moral hazard in agri-environmental
schemes, Mimeo, Agricultural and Resource Economics Group, University of
Western Australia.
Lindner, R.K. (1999), Prospects for public plant breeding in a small country. Paper
presented at the ICABR conference on The shape of the coming agricultural bio-
technology transformation: strategic investment and policy approaches from an
economic perspective at the University of Rome "Tor Vergata" Rome and Ravello,
June 17-19,1999.
Pyke, B. (1998), "Ingard survey results for the second year", The Australian Cotton-
grower, 19(6): 36-39.
Riley, PA and L. Hoffman, (1999) "Value-enhanced crops: biotechnology's next
stage"; Agricultural Outlook, March 1999, pp. 18-23.
United States Department of Agriculture (USDA), Economic Research Service (ERS)
(1999), Genetically engineered crops for pest management. Retrieved June 1999
from the World Wide Web: http://www.usda.gov/whatsnew/issues/biotech
Wright, B.D. (1996), Agricultural genetic research and development policy, Conference
proceedings of the Global Agricultural Science Policy for the 21 st Century, Mel-
bourne, pp. 559-580.
6
An Economic Risk Assessment of the
Impact on Producers of Removing
Quarantine Restrictions
Robert W. Fraser
Note that in the absence of imports the relationship between price and
yield is assumed in what follows to be represented by a negative covari-
ance. This assumption is consistent with a closed economy framework.
In addition, the variance of income in the absence of imports (Varo(l))
can be approximated by, 5
Var(y) = y2Var(e)
Var(B) = variance of s.
P1<po
Var(p 1 ) < Var(po) (6.4)
COV(P1,Y1) = 0 (6.5)
Note that the presence of disease is assumed not to affect the sea-
sonal factors governing the variability of yield:
Variance(B1)=Variance(Bo), (6.7)
(6.11 )
Based on equations (6.4) and (6.6) the first term on the right-hand-side
of (6.11) is clearly negative while the second term is positive as speci-
fied. Consequently, equation (6.11) is analytically ambiguous, although it
is anticipated that in virtually all cases the magnitude of the first term will
dominate that of the second and that overall the producer's expected in-
come will be lower in the presence of imports than in its absence:
On the basis of equations (6.4), (6.6) and (6.8), the first two terms on
the right-hand-side of (6.13) are negative, while the third term is positive
as specified. The first term represents the decline in income risk with
imports due primarily to the lower expected price, while the second term
represents the decline associated primarily with less price variability.
Note also that the magnitude of both these terms is larger if disease af-
fects expected yield (Yl < Yo)' In contrast, the third term represents the
increase in income risk, which follows the removal of the negative co-
variance between domestic price and yield associated with the introduc-
tion of imports. It follows that the overall impact of the introduction of im-
ports on the variance of income is analytically ambiguous.
Moreover, these three terms do not represent all of the components of
the overall impact of introducing imports on the producer's perception of
income risk. A fourth component can be seen by approximating the pro-
ducer's expected utility of income (E(U(I))) as follows: 9
100 Robert W. Fraser
U"(E(/))
Note that this term is negative for a risk averse producer. Based on
equation (12), it is anticipated that the introduction of imports will de-
crease the producer's expected income. Consequently, the associated
change in U"(E(I)) depends on whether:
<
U'I/(E(I)) o (6.15)
>
6. 2. Numerical Analysis
In order to undertake a numerical analysis of the model developed in
Section 1 it is necessary to specify a particular form for the producer's
102 Robert W. Fraser
utility function as well as values for the parameters of the model. In what
follows use is made of the constant relative risk aversion form of utility
function referred to in the previous section:
IloR
U(I)=- (6.17)
1- R
Po = 10
a Po = 3 (Var(po) = 9)
a() = 0.3 (Var(8) = 0.09)
Yo = 100
Following the introduction of imports the Base Case values are as-
sumed to be:
PI = 6.1
a PI = 0
YI = 100
Note that PI has been specified at a level such that 90 percent of the
probability distribution of Po would exceed it. This assumption is subject
to a sensitivity analysis in what follows, as are the assumptions of zero
world price variability (a PI = 0) and zero impact of disease on expected
yield (YI = 100). Other sensitivity analysis relates to the assumed val-
ues for a Po and as·
Table 6.1 contains details of the Base Case results of producer per-
ception of the impact of introducing imports on income risk for a range of
attitudes to risk (R) and of correlation coefficients between price and
yield in the absence of imports. 11 Note that the value of this coefficient
is based on the specification: 12
(6.18)
In addition a perception index of improvement or worsening of income
risk is specified as follows:
Perception Index b
(1 ) (2) (3) (4)
a c d
E\(1X vaf\(-}{ R = 0.8 R = 1.6
Eo (1) Varo(1)
Correlation
Coefficient (p)
-0.25 0.62 0.25 1.06 1.03
The results in Columns (3) and (4) of Table 6.1 show that, in general
terms, a producer is more likely to view the impact of introducing imports
on income risk as an improvement the smaller the magnitude both of the
(negative) correlation coefficient (p) and of the producer's risk aversion
coefficient (R). The first of these findings can be explained by reference
to the diminished role of the negative covariance as a pre-import stabi-
lizer of income for lower values of p. Consequently, the reduction in in-
come risk associated with imports is more powerful for smaller p as
shown by the results in column (2) of Table 6.1. Moreover, the second
finding is explained by the indirect impact of reduced expected income
(see column (1)) on the perception of income risk (the U"(E(/)) term in
equation (6.14)). In particular, the increase in magnitude of U"(E(/))
caused by the decline in E(I) is smaller for a less risk averse producer.
This means that a given decline in E(I) and Var(l) is more likely to be
perceived as representing an improvement in income risk by a less risk
averse producer. Finally in relation to Table 6.1, it should be noted that
for all values of p between -0.57 and -0.81 for R :;: 0.8, and between -0.37
and -0.81 for R :;: 1.6, the producer's perception of income risk is that it
has worsened with the introduction of imports, despite the fact that:
104 Robert W. Fraser
Perception Index
(1 ) (2) (3)
b c d
A comparison of column (2) in Table 6.1 with column (1) in Table 6.2
shows that this change results in a weaker tendency for the introduction
of imports to reduce the variance of income. In particular, for a producer
with a value of R = 0.8 and a value of p = -0.5, the introduction of imports
in this situation is perceived as worsening income risk whereas it was
perceived as improving income risk for the situation of less elastic de-
mand in Table 6.1. Moreover, Table 6.2 shows that, compared with Table
6.1, for producers with either level of risk aversion, there is a smaller
range of values of p for which a decrease in the variance of income is
perceived as an improvement in income risk.
Next consider a situation of greater seasonal variability (00)' Table 6.3
contains details of results for a value of 08 = 0.5 (Base Case: 08 = 0.3).
Economic Risk Assessment of Quarantine Restrictions 105
Perception Index
(1 ) (2) (3)
b c d
varl (%
Varo(I) R =0.8 R =1.6
Correlation
Coefficient (p)
-0.25 0.35 1.06 0.94
-0.5 0.49 0.99 0.86
-0.75 0.81 0.91 0.76
A comparison of the results in Tables 6.2 and 6.3 shows that the im-
pact of greater seasonal variability is similar to that of a more elastic de-
mand. In both cases the change results is a weaker tendency for the in-
troduction of imports to reduce the variance of income. As a conse-
quence, in Table 6.3 there is for both levels of risk aversion a smaller
range of values of p for which a decrease in the variance of income is
perceived as an improvement in income risk. The essential difference
between the two cases is that, whereas for a more elastic demand there
is a smaller absolute gain in terms of reduced price variability, for greater
seasonal variability there is a smaller gain in reduced price variability
relative to total income variability.
Table 6.4 contains details of the results of a sensitivity analysis of the
impact of import-induced disease on expected yield. The Base Case as-
sumed that the introduction of imports had a zero impact on expected
yield (jil = 100).
The results in Table 6.4 are based on an assumption that disease in-
troduced by imports causes a 15% decrease in expected yield (YI =
85) .15 Compared with columns (1) and (2) in Table 6.1, columns (1) and
(2) of Table 6.4 show a more powerful impact of imports both on ex-
pected income and the variance of income. The first of these impacts
strengthens the negative perception of the impact of imports on income
risk (via U"{E(J))) while the second strengthens the positive perception by
directly reducing the variance of income. As a consequence, there is to a
106 Robert W. Fraser
Perception Index
b c d
Var, (1)/ R =0.8 R = 1.6
/Varo (1)
Correlation
Coefficient (p)
-0.25 0.53 0.18 1.04 1.01
Perception Index
(1 ) (2) (3)
(%
b c d
vali Varo(l)
R = 0.8 R = 1.6
Perception Index
(2) (3) (4)
b c d
Vali(l)/
/Varo(l)
R = 0.8 R = 1.6
Correlation
Coefficient (p)
-0.25 0.77 0.45 1.06 1.04
-0.5 0.78 0.67 0.99 0.93
-0.75 0.80 1.34 0.91 0.80
Whereas the Base Case featured an expected world price which 90%
of domestic prices (in the absence of imports) would exceed, the results
in Table 6.6 are based on an expected world price (PI = 7.48), which
only 80% of domestic prices would exceed (in the absence of imports). A
comparison of columns (1) and (2) in Tables 6.4 and 6.6 shows that this
change has the opposite impact on the expected level and variability of
income with imports to that of a reduction in expected yield associated
with disease. As a consequence, conflicting impacts are once again cre-
ated on the producer's perception of income risk. But, unlike the results
in Table 6.4, the results in Table 6.6 show a clear distinction from those
in Table 6.1 at both levels of producer risk aversion. In particular, the
range of values of p for which an improvement in income risk is per-
ceived decreases for both types of producers. Therefore, this finding
suggests that the lower is the expected world price relative to the ex-
pected domestic price in the absence of imports, the more likely it is that
producers will perceive an improvement in income risk with the introduc-
tion of imports.
6.3. Conclusion
The aim of this chapter has been to develop understanding of the impact
of removing quarantine restrictions on the income risk of domestic pro-
ducers. In so dOing it is intended that the existing distinction between
"economic assessment" and "risk assessment" in the analysis of quaran-
tine policy be diminished.
Section 6.1 developed a model of the impact of removing import re-
strictions on a producer's perception of income risk. This model devel-
opment identified four separate effects of the introduction of imports on
the perceived income risk of domestic producers, as well as a range of
factors, which influence the magnitude of the various effects. Section 6.2
undertook a numerical analysis of the model developed in Section 6.1
and evaluated the role of each of the factors identified as influencing the
producer's perception of the impact of imports on income risk.
On the basis of this numerical analysis the following conclusions have
been reached. In particular, a producer is more likely to perceive the in-
troduction of imports as improving income risk:
(i) the weaker is the negative correlation between the producer's price
and yield in the absence of imports
(ii) the less risk averse is the producer
(iii) the less elastic is the domestic demand for the product in question
Economic Risk Assessment o/Quarantine Restrictions 109
Endnotes
11 See Newbery and Stiglitz (1981, p. 105) for details of empirical evidence to
support these values of R.
12 See Mood, Graybill and Boes (1974, p.154).
U"(Edl)) . Vardl)
>
U "(EoCl)) . VaroCl)
<
where it should be recalled that the second term on the right-hand-side of
equation (6.14) is negative.
14 Note that for values of p nearer to -1 the role of the negative covariance as an
income stabilizer becomes dominant, and so its removal with imports results in
increased income variability (see equation (6.13)).
15 See Hafi et al. (1994).
References
Campbell, R., B. Gardiner, and H. Haszler (1980), "On the Hidden Revenue Effects of Wool
Price Stabilisation in Australia" Australian Journal of Agricultural Economics, 24(1 ):1-15.
Hafi, A., R. Reynolds, and M. Oliver (1994), Economic Impact of Newcastle Disease on the
Australian Poultry Industry ABARE Research Report 94.7, ABARE, Canberra.
James, S. and K. Anderson (1998), "On the need for more economic assessment of quar-
antine/SPS policies" Australian Journal of Agricultural and Resource Economics,
42(4):445-58.
Mood, A.M., FA Graybill, and D.C. Boes (1974), Introduction to the Theory of Statistics 3rd
Edition, McGraw-Hili, Kogakusha.
Newbery, D.M.G. and J.E. Stiglitz (1981), The Theory of Commodity Price Stabilization
Clarendon Press, Oxford.
Nunn, M. (1997), "Quarantine risk analysis" Australian Journal of Agricultural and Resource
Economics, 41 (4): 559-78.
Pope, R.D. and R.E. Just (1991), "On testing the structure of risk preferences in agricultural
supply analysis" American Journal of Agricultural Economics 73(3):743-8.
Tanner, C. (1997), "Principles of Australian quarantine" Australian Journal of Agricultural
and Resource Economics, 41 (4): 541-58.
Tanner, C. and M. Nunn (1998), "Australian quarantine post the Nairn Review" Australian
Journal of Agricultural and Resource Economics, 42(4): 445-58.
Part II
Case Studies
7
Risk Attitudes and Risk Perceptions
of Crop Producers in Western Australia
Amir K. Abadi Ghadim and David J. Pannell
It has often been suggested that risk and uncertainty have major influ-
ences on the rate of adoption of rural innovations (e.g. Lindner et al.,
1982; Lindner, 1987; Weisensel and Schoney, 1989; Tsur et al., 1990;
Leathers and Smale, 1992; Shapiro et al., 1992; Smale and Heisey,
1993; Feder and Umali, 1993). However, empirical evidence for or
against this view has been relatively scarce and weak. This is primarily
because very few of the many empirical studies of adoption of rural inno-
vations have attempted to include risk and uncertainty as explanatory
variables. Even the few existing attempts have tended to use crude proxy
variables, probably because of the practical difficulty of obtaining high
quality measures of the relevant variables, these being: farmers' percep-
tions of the impacts of the innovation on the levels of risk they face;
farmers' uncertainty about the innovation; and, farmers' attitudes to risk
and uncertainty.
A current extension program is encouraging crop farmers in Western
Australia to grow several grain legumes crops, or "pulses", that are new
to the region: chickpeas (Cicer arietinum) , faba beans (Vicia faba) and
lentils (Lens culinaris) (Hamblin, 1987; Siddique and Sykes, 1997). A
minority of farmers are trialling these crops and very few have reached
the stage of full adoption. Anecdotal evidence suggests that Australian
farmers view grain legumes as being more risky than other available en-
terprises, and that this is inhibiting their rate of adoption. Given this
situation and the fact that they are in the early stages of adoption, these
crops provided an ideal opportunity to conduct a study of the roles of risk
and uncertainty in the adoption process.
In that study the focus was on chickpeas to explore the risk attitudes
and risk perceptions of Western Australian farmers in the context of
114 Amir K. Abadi Ghadim and David J. Pannell
7. 1. 1. Attitudes to risk
Studies of risk attitudes of Australian farmers (e.g. Bond and Wonder,
1980; Bardsley and Harris, 1987) have reported a range of risk attitudes,
with a majority of farmers being classed as risk averse. Risk aversion
implies that they are prepared to sacrifice some income in order to avoid
some risk or uncertainty. The most common method for representing and
Risk Perceptions in Western Australia 115
or her utility function. According to the expected utility theory, the shape
of the utility functions of risk-averse, risk neutral and risk preferring indi-
viduals are respectively concave down, linear and convex down. The
Pratt-Arrow coefficient of absolute risk aversion is positive for risk-averse
persons and negative for risk-preferring individuals.
7.1.3. Influence of risk on adoption behavior
Some research evidence suggests that the allocation of land to innova-
tions declines with higher degrees of risk aversion (e.g. Feder, 1980;
Shapiro et al., 1992). Feder et at. (1985) reflected the view, which is
common amongst many in a large body of literature, that a risk-averse
decision-maker will take longer than a risk neutral person to decide to
adopt an innovation.
Lindner and Fischer (1981) refuted the generality of that claim. Their
analysis showed that, in some cases, risk aversion could speed up the
process of adoption. This may occur where the profitability of the innova-
tion is not highly correlated with that of the traditional or alternative en-
terprise that it potentially replaces. In this case partial adoption of the
innovation will reduce the variability of total profits simply because it be-
comes a form of portfolio diversification.
On the other hand, diversification can sometimes result in increased
risk. Specialization can result in greater efficiency and improved quality
in production and marketing of produce. This specialization can provide
some insurance against price and weather risks. Diversification may ac-
tually result in increased financial and business risk if it extends the
farmer beyond their competencies and knowledge base (Pannell et al.,
2000). In this case, the prospect of a drop in productivity from diversifi-
cation (e.g. adoption of an innovation) may discourage the farmer from
adopting as quickly as they might. Indeed, a farmer may be so highly
skilled at some technology that he or she faces substantial incentives to
persist with its use. Someone who is less skilled may find it optimal to
switch technologies regularly, and therefore enjoy long-run growth in
output (Jovanovic and Nyarko, 1996).
This argument can be extended to the situation where a divisible inno-
vation, such as a new species of a crop, is being considered by two
similar farmers who differ mainly in years of experience in farming. The
young farmer with relatively less experience in growing and managing
traditional enterprises (e.g. wheat) may find it more worthwhile to switch
than does an older farmer. This is because the older farmer, who usually
has more experience, would be very skillful and confident in growing the
Risk Perceptions in Western Australia 117
conventional enterprises. For the older farmer the opportunity cost of the
conventional enterprise is higher than for the younger farmer.
There are also other studies that have found that farmers' perceptions
of riskiness of innovations influenced the probability and intensity of their
adoption (e.g. Smale et al., 1994). Several reviews of adoption studies
have highlighted the general scarcity of empirical studies that have ade-
quately addressed the role of risk and uncertainty in adoption. This scar-
city has been attributed to difficulties in observing and measuring risk
and uncertainty. It has been argued that this omission of risk has been a
serious limitation of many of the previous adoption studies leading to
poor model specification.
7. 1.4. Measuring risk
Norris and Kramer (1990) carried out a comprehensive review of tech-
niques for measuring risk and uncertainty in agriculture. They found that
the most reliable, accurate and acceptable method is to elicit subjective
judgments of individuals about probabilities of events that are important
to them.
Elicitation of subjective judgments requires the use of specific survey
techniques to gather information from individuals who make the adoption
decisions. Surveys that enable the researchers to collect panel data offer
the most suitable tool for this purpose. The use of panel data in adoption
research offers several major advantages over conventional cross-
sectional and time-series data sets. Longitudinal surveys can be de-
signed to detail individual farmer characteristics, preferences, percep-
tions and adoption choices made by them at each point in time (Hsiao,
1986).
Given the advantages of panel data sets we chose to collect the data
for this study using a longitudinal survey. Farmers were personally inter-
viewed at their property by trained interviewers using structured ques-
tionnaires. The purpose of the questionnaire was to obtain detailed in-
formation about the characteristics of the property and the farmer. In this
chapter we discuss only a subset of results drawn from the analysis of
the data from our survey.
The sample of farmers in this study was selected from the population of
growers in the central and eastern wheat belt of Western Australia. Most
of the farms in this region have a combination of crop and livestock en-
terprises on a mix of soils ranging from sands to clay soils (Kingwell et
al., 1995). Chickpeas, the crop innovation, can be grown on neutral to
118 Amir K. Abadi Ghadim and David 1. Pannell
alkaline loams and clays. These soils form a large proportion of the re-
gion (Stoneman, 1992; Lantzke, 1992).
Lowest Highest
In this study, risk aversion was estimated with the use of subjective
probabilities in conjunction with a set of questions that prompted the
farmer to make a choice from a set of hypothetical wheat grain marketing
scenarios that exposed the farmer to different degrees of risk. The ques-
tions were specifically designed to create a realistic context for decisions
regarding the timing and the amount of the farm's wheat harvest offered
for sale under different marketing schemes. Using different approaches,
Risk Perceptions in Western Australia 121
the new one had a smaller variability associated with it. The new distribu-
tion was created by "winsorising" the original elicited distribution at a point
equal to the average of the second and third highest prices specified by the
farmer. Winsorising is similar to truncation except that the probability
weight that would be removed by truncation is reallocated to the pOint of
truncation (Fraser, 1988; Fraser and Kingwell, 1997). The low end of the
distribution was winsorised in a similar way. Table 7.1 is a numerical ex-
ample of an elicited and modified price distribution for one of the respon-
dents.
150
2 158 1 162
3 166 2 166
4 174 3 174
5 182 4 182
6 190 5 186
7 198
The farmer was then asked if he or she would consider using this
scheme for the entire production of their wheat in 1998, on the condition
that they were not committed to a pre-defined amount of grain to be de-
livered. The rationale for not asking the farmer to commit to a set ton-
nage was to remove the element of risk and uncertainty associated with
the yield of wheat so that we would be dealing exclusively with the price
risk aspect of the production of wheat. To determine whether risk prefer-
ences varied in response to the amount of money involved, the question
regarding the use of this scheme was repeated for 10, 25, and 50 per-
cent of the farmer's wheat harvest.
When this question was pre-tested, it was found that some farmers
preferred the more variable distribution to such an extent that their re-
sponse was to want to be paid to participate in this scheme. The reason
expressed by them was that, although the scheme reduced the downside
risk, it also excluded the possibility of the higher prices. Consequently,
the interviewers were asked to note the amount by which such farmers
would want to be compensated to participate in this scheme, reflecting a
degree of risk seeking.
Individuals' risk aversion coefficients could be inferred from the two
distributions and the risk premia we elicited from them. Solving for the
value of absolute risk aversion (P) satisfies equation (7.1).
CE(ll1 ; p) = CE(ll2 - R ; p) (7.1 )
where CE(Ilj ; p) is the certainty equivalent value of profit distribution i
given risk aversion coefficient p, and R is the difference in risk premia
between the two distributions, as specified by the grower.
By assuming a constant absolute risk aversion utility function of the
form,
U = a + b . exp (-pI1) (7.2)
Note: calculated from subjective yield distributions elicited in the 1995, 1996, and 1997
interviews.
Risk Perceptions in Western Australia 125
The smaller variation in the estimates of mean yields for chickpeas may
be because their perceptions were formed on the basis of information from
similar origins (e.g. the extension agents of Agriculture Western Australia).
It also probably reflects the lower mean yields of chickpeas, so that the
absolute scope for variation is less.
The sample mean yield for wheat is about one ton per hectare larger
than for chickpeas (Table 7.1). Excluding four outlier farmers, the expected
yield of chickpeas is clustered between 0.4 and 1.4 tons per hectare. The
lowest mean yield in the sample for chickpeas is 0.2 tons per hectare. One
person in the sample thought that the expected yield could be as high as
2.2 tons. Lack of adequate knowledge of the agronomy of chickpeas in this
region may be the main reason for the extremely high or low estimates of
expected yield by some individuals
Table 7.1 also shows the cluster of opinions about the expected yield of
wheat. Except for the extremes, expected wheat yields in this sample are
clustered between one ton and 2.7 tons per hectare.
The variation in the mean of the population between the three interview
years was 0.06 tons per hectare for wheat but 0.14 tons per hectare for
chickpeas. This suggests that farmers in this sample are still learning about
the performance of chickpeas. Consequently, their perceptions are likely to
be influenced Significantly by the information they collect each year. How-
ever, for a crop like wheat with a long history on these farms, an additional
year of information makes little difference to perceptions. It is worth noting
that mean yields of chickpeas for the sample population over the three
years shows a shift towards higher yields, possibly reflecting improved
knowledge about its performance, but possibly also a result of increased
skill and knowledge about suitable agronomic practices.
7.4.2. Riskiness of the old versus the new
The coefficient of variation (CV) of crop yields provides an indication of
riskiness of the crop The CV was estimated from the subjective probabil-
ity distribution of crop yields elicited from each respondent. A comparison
of the CV of chickpea yield with the CV of wheat yield can be used as an
estimate of the relative riskiness of chickpeas. All farmers in the sample
had far more experience with wheat than with any other crop, so wheat
was used in this situation as a benchmark crop.
Every year, one or two farmers gave responses that chickpeas were
less risky than wheat on the basis of yield. However, the majority consid-
ered chickpeas to be more risky than wheat. This is further summarized
by the weighted-averages shown in the last row of Table 7.3.
126 Amir K. Abadi Ghadim and David 1. Pannell
For instance, in 1995, the average CV of yield for the entire sample
was 31 percent for wheat but 49 percent for chickpeas, while in 1997 the
average CV for the sample was 29 percent for wheat but 37 percent for
chickpeas.
There is a clear indication of tight clustering of the CV values for wheat
and a wider clustering for chickpeas. This indicates the extent of uncer-
tainty and lack of knowledge about chickpeas as compared to a more
traditional crop like wheat. For instance, the majority of the farmers in the
sample (83 percent) believed that the CV of wheat yield is between 16
Risk Perceptions in Western Australia 127
response of their own wheat crops in similar climatic conditions. The re-
sults are consistent with field observations regarding the correlation be-
tween the yield of wheat and legume crops.
<0 2
0-0.025 4
0.025 - 0.05 15
0.05 - 0.075 24
0.075 - 0_1 13
0.1 - 0.25 16
0.25 - 0.5 23
0.5 - 1 4
0.08 Pooled sample mean
7.5. Conclusions
Three aspects of risk and uncertainty have been studied in the context of
a new crop currently being evaluated for adoption by a group of farmers:
farmer perceptions of the riskiness of the crop, farmer perceptions of the
covariance between the yield of the new crop (chick peas) and a tradi-
tional crop (wheat), and farmer attitudes toward risk (i.e. their risk aver-
sion or risk preference). All three elements are relevant in explaining
farmers' responses to the new crop (Abadi Ghadim and Pannell, 1999).
It was found that the riskiness of the new crop (as indicated by the co-
efficient of variation of the farmers' subjective yield distribution) was
higher for the new crop than for the old crop. This would be expected
given the farmers' lack of experience with and greater uncertainty about
130 Amir K. Abadi Ghadim and David 1. Pannell
the new crop. It was also shown that the perceived riskiness of the new
crop diminished over time, while the perceived riskiness of the traditional
crop was stable over time. The fall in perceived riskiness was presuma-
bly due to farmers gaining experience and knowledge of the new crop,
allowing them to better estimate its true performance and/or allowing
them to better identify most appropriate production methods and input
levels.
The farmers' perceptions of covariance in yield between the new crop
and a traditional crop were consistent with field evidence. The observed
moderate positive covariances are likely to be of value to farmers by al-
lowing them to interpret information about chickpea crop yields in differ-
ent seasons in the light of the responses of wheat crops in similar grow-
ing seasons.
Elicited risk preferences of farmers were consistent with previous re-
sults for Australian farmers, in that most were found to be risk averse
and that the observed range of risk attitudes was wide. The coefficients
of absolute risk aversion inferred from the elicitation process were high
relative to some results in the literature.
The study raises some interesting questions for possible future proj-
ects:
• To what extent does the diversity of perceptions that we noted in
this study reflect uncertainty versus real differences in circumstances? If
the dominant factor is uncertainty, it implies a greater potential for learn-
ing and refinement of decision-making.
• What determines how rapidly uncertainty falls as trialling pro-
ceeds?
• This chapter focused on on-farm trialling, based on a hypothesis
that it is the crucial source of information used by farmers in their adop-
tion decisions. Relatively little data was collected about off-farm sources
of information about innovations. However, an important question is,
how much more effective is on-farm trialling than off-farm information at
reducing uncertainty?
• Does the quality and relevance of off-farm information determine
the timing and intensity of on-farm trialling?
• Are there differences between various types of innovations with
respect to the importance of risk and uncertainty in their adoption? The
innovation. studied here is one with a rapid financial payoff from suc-
cessful adoption. But some innovations, such as land conservation
practices, have a much slower payoff, and often their trials reveal infor-
Risk Perceptions in Western Australia 131
mation more slowly (e.g. see Pannell, Chapter 4 in this volume). To what
extent does this affect the rate and level of adoption?
Acknowledgments: This project was partly funded by the Rural Industries Research
and Development Corporation (RIRDC). We also wish to thank the Cooperative Re-
search Centre for Legumes in Mediterranean Agriculture and Agriculture Western
Australia for their support. Finally, we are grateful for the kind contribution of the
farmers who partiCipated in this study.
References
Abadi Ghadim, A. K., M. Burton, and D.J. Pannell (1999), "More empirical evidence on
the adoption of chick peas in Western Australia", paper presented at the 43rd An-
nual Conference of the Australian Agricultural and Resource Economics Society,
Christchurch, New Zealand, Jan 20-221999.
Abadi Ghadim, A. K., and D.J. Pannell (1999), "A conceptual framework of adop-
tion of an agricultural innovation", Agricultural Economics, 21: 145-154.
Abadi Ghadim, A. K., and D.J. Pannell (1998). "The importance of risk in adoption
of a crop innovation: empirical evidence from Western Australia", 42nd Annual
Conference of the Australian Agricultural and Resource Economics Society,
University of New England, Arimdale, New South Wales, pp. 15.
Alreck, P. L., and R.B. Settle (1995). The Survey Research Handbook. Sydney:
Irwin.
Anderson, J. R., J.L. Dillon, and B. Hardaker (1977). Agricultural Decision Analy-
sis. Armidale: University of New England.
Arrow, K. J. (1970), Essays in the Theory of Risk-Bearing. London: North-
Holland.
Bardsley, P., and M. Harris (1987), "An approach to the econometric estimation
of attitudes to risk in agriculture", Australian Journal of Agricultural Economics,
31: 112-126.
Bar-Shira, Z. (1992), "Nonparametric test of expected utility hypothesis", Ameri-
can Journal of Agricultural Economics, 74: 523-533.
Bond, G., and Wonder, B. (1980). "Risk attitudes amongst Australian farmers",
Australian Journal of Agricultural Economics, 24: 16-34.
de Vaus, D. A. (1991), Surveys in Social Research. North Sydney: Allen and Un-
win.
Feder, G. (1980), "Farm size, risk aversion and the adoption of new technology
under uncertainty", Oxford Economic Papers, 32: 263-2893.
Feder, G., R. E Just, and D. Zilberman (1985), "Adoption of agricultural innova-
tions in developing countries: a survey", Economic Development and Cultural
Change, 33: 255-297.
Feder, G., and D.L. Umali (1993), "The adoption of agricultural innovations: a
review", Technological Forecasting and Social Change, 43: 215-239.
Fraser, R. W. and R. Kingwell (1997), "Can expected tax revenue be increased
by an investment-preserving switch from ad valorem royalties to a resource
rent tax"? Resources Policy, 23: 103-108.
132 Amir K. Abadi Ghadim and David 1. Pannell
Fraser, R. W. (1988), "A method for evaluating supply response to price under-
writing", Australian Journal of Agricultural Economics, 32: 22-36.
Hamblin, J. (1987), "Grain legumes in Australia", 4th Australian Society of Agron-
omy Conference, , La Trobe University, Melbourne, Victoria, pp. 65-81.
Hardaker, J. B., R.B. Huirne, and J.R. Anderson (1997), Coping with Risk in Agri-
culture. Wallingford: CAB International.
Hertzler, G. (1997), "A new theory for explaining the paradoxes in decision making
under risk and for measuring time and risk preferences", in R. B. M. Huirne, J. B.
Hardaker, and A A. Dijkhuizen (eds.) Risk Management Strategies in Agriculture.
State of the art and future perspectives. Wageningen, Wageningen Agricultural Uni-
versity, pp. 65-86.
Hey, J. D. (1979), Uncertainty in Microeconomics. Oxford: Martin Robertson.
Hsiao, C. (1986), Analysis of Panel Data. Vol. 1. 1 vols. Cambridge: Cambridge
University Press.
Jovanovic, B., and Y. Nyarko (1996),"Learning by doing and the choice of tech-
nology", Econometrica, 64: 1299-1310.
Kingwell, R. S., A.K. Abadi Ghadim, S.D. Robinson, and J.M. Young (1995),
"Introducing Awassi sheep to Australia: an application of farming systems
models", Agricultural Systems, 47: 451-471.
Lantzke, N. (1992). "Soils of the Northam Advisory District - the Zone of Ancient
Drainage", Bulletin. Department of Agriculture Western Australian.
Leathers, H. D., and M. Smale (1992). "A Bayesian approach to explaining se-
quential adoption of components of a technological package", American Jour-
nal of Agricultural Economics, 68: 519-527.
Lindner, R., and M. Gibbs (1990), "A test of Bayesian learning from farmer trials
of new wheat varieties", Australian Journal of Agricultural Economics, 34: 21-
38.
Lindner, R. K. (1987), "Adoption and diffusion of technology: an overview", in B. R.
Champ, E. Highley and J.V. Remenyi (eds.), Technological Change in Postharvest
Handling and Transportation of Grains in the Humid Tropics, ACIAR Proceedings
No. 19, ACIAR, Canberra, pp. 144-151.
Lindner, R. K., and A.J. Fischer (1981), "Risk aversion and the innovation time
lag", University of Adelaide, Australia.
Lindner, R. K., AJ. Fischer, AJ. ArnOld, and M. Gibbs (1982), "Measuring beliefs
about future events and innovation profitability", Department of Economics,
University of Adelaide, Australia.
Machina, A. (1981), "Rational decision making versus "rational" decision model-
ing", Journal of Mathematical Psychology, 24: 163-75.
Norris, P. E., and R.A Kramer (1990), "The elicitation of subjective probabilities
with applications in agricultural economics" Review of Marketing and Agricul-
tural Economics, 58: 127-147.
Pannell, D.J., R.L. Malcolm, and R.S. Kingwell (2000), "Are we risking too
much? Perspectives on risk in farm modeling", Agricultural Economics, 23: 69-
78.
Pratt, J. W. (1964), "Risk aversion in the small and in the large", Econometrica,
32: 122-136.
Schoemaker, P.J. (1982), "The expected utility model", Journal of Economic Lit-
erature, 20: 529-63.
Shapiro, B.I., B.W. Brorsen, and D.H. Doster (1992), "Adoption of double-
cropping soyabean and wheat", Southern Journal of Agricultural Economics,
24(3): 33-40.
Risk Perceptions in Western Australia 133
Siddique, K.H.M., and J. Sykes (1997)" "Pulse production in Australia past, pres-
ent and future", Australian Journal of Experimental Agriculture, 37: 103-111.
Smale, M., and P.W. Heisey (1993), "Simultaneous estimation of seed-fertilizer
adoption decisions", Technological Forecasting and Social Change, 43: 353-
368.
Smale, M., R.E. Just, and H.D. Leathers (1994), "Land allocation in HYV adoption
models: an investigation of alternative explanations", American Journal of Agri-
cultural Economics, 76: 535-546.
Smidts, A. (1990), Decision Making Under Risk. A Study of Models and Meas-
urements Procedures with Special Reference to the Farmers Marketing Be-
haviour. Wageningen Economic Studies 18. Wageningen: Pudoc.
Stoneman, T. C. (1992), "An introduction to the soils of the Merredin Advisory
District- description, illustration and notes on nine common soils", Bulletin.
Western Australian Department of Agriculture.
Tsur, Y., M. Sternberg, and E. Hochman (1990), "Dynamic modelling of innova-
tion process adoption with risk aversion and learning", Oxford Economic Pa~
pers, 42: 336-355.
Weisensel, W. P., and R.A. Schoney (1989), "An analysis of the yield-price risk
association with specialty crops", Western Journal of Agricultural EconomiCS,
14: 293-299.
Wright, G. and P. Ayton (1994), "Subjective probability: what should we be-
lieve?", in G. Wright, and P. Ayton (eds) , Subjective Probability, Chichester:
John Wiley and Sons, pp. 163-184.
8
Valuing Pest Control:
How Much is Due to Risk Aversion?
Terrance M. Hurley and Bruce A. Babcock
How yields are affected by a Single pest, European corn borer, can best
be modeled by assuming that harvested yield depends on two factors:
the level of yields that would occur without corn borer damage (the pest-
free yield); and, the level of corn borer damage. The pest-free yield is a
random variable, depending on many factors, including weather, applied
136 Terrance M. Hurley and Bruce A. Babcock
inputs, and the amount of damage from other pests (including other in-
sects, weeds, and disease). The level of corn-borer damage depends on
corn-borer population and control efforts. Corn borer populations in most
of the U.S. Corn Belt vary widely from year to year, thus damage is also
a random variable. In addition, most corn farmers do not treat for corn
borers, thus damage control efforts without Bt corn are zero.
Abstracting from the level of applied inputs (assumed to be set at ex-
pected-profit maximizing levels), and assuming that the profit-maximizing
level of damage control is zero, harvested yield depends on the level of a
Single pest infestation as follows:
a-l(b )p-l
1 Yf -Yf
(8.2)
fey f) = B(a, [J) ba+p+1
1 1
f(Yh I p) = B(a,[J) S(p)
Valuing Pest Control 137
S(p)b
EU( 7th I p) = EU( 7th) = f U((y - c»Af(Yh )dy . (8.5)
o
Ph S(p)b
= f f U((y - c)A)f(Yh I p)dydp (8.6)
PI 0
138 Terrance M. Hurley and Bruce A. Babcock
f !(Yh I p)h(p)dp
Ph
!(Yh) = (8.7)
PI
The integration needed to find the non-Bt corn density and all subse-
quent integrations used to find expected utility levels were carried out using
Monte Carlo integration techniques.
A comparison of the density functions in Figures 8.1 and 8.2, shows that
adoption of Bt corn shifts the density to the right. This shift to the right clearly
shows the benefits of Bt corn. There is a higher expected yield and a lower
140 Terrance M. Hurley and Bruce A. Babcock
probability of very low yields. But the rightward shift is not a simple mean
shift because maximum and minimum yields are not affected. The maxi-
mum yield is not affected because of the assumption that the yield capacity
of Bt corn and non-Bt corn are identical. And there is no reason to expect
that the minimum yield of Bt corn is any different than non-Bt corn.
1000
900
800
700
..,.
u 600
c::
:J 500
~ 400
300
200
100
0
0 50 100 150 200
Vie Id (bu/ac)
1000
900
800
700
Conventional corn
...
- - - - - - - - - - --------------- -----~~ --- .,.~--:--------------
>- 600
..
u
c
:::I
tI'
500
300
200
100
0
0 50 100 150 200
Yield (bu/ae)
Besides a shift in the density, there is also a spread in the density. This
spread is more apparent in Hall County than in Boone County. This in-
crease is indicated by lower peaks in the Bt corn densities. The cause of
this increase in spread is that Bt corn has a larger positive affect on yield
when corn borer-free yield is high than when it is low. That is, the assump-
tion of proportionate damage (a common assumption in nearly all pest
control studies) means that elimination of a damaging agent has a smaller
effect when pest-free yields are low than when they are high.
Table 8.1 presents the results of this analysis. The risk aversion coeffi-
cients are reported in the first column. The next two columns express this
level of risk aversion as a ratio of the corresponding risk premium to the
standard deviation of returns under Bt corn. This ratio shows the reason-
ableness of the range of risk aversion coefficients used in this study. The
last two columns report the change in CER after adopting (costlessly) Bt
corn. This change in CER is the per-acre maximum a producer would be
willing to pay for Bt corn (in terms of bushels of corn).
At a risk aversion level of 0.0 (risk neutrality) the maximum willingness to
pay is simply the change in expected profit. This amounts to 5.03 bu/ac in
Iowa and 11.2 bu/ac in Nebraska. This amount reflects the change in ex-
pected yield that would occur from adoption of Bt corn. The change in ex-
pected yield varies directly with expected corn borer pressure. If current
corn borer pressure is higher than that used in this analysis then the
change in expected yield would be greater than that reported in Table 8.1.
8.4. Implications
At first glance the results indicate that a risk-averse producer values the
pest control benefits of Bt corn less than the risk-neutral producer. This
implies that given a fixed treatment cost, a risk-averse producer is less
likely to adopt Bt corn than his risk neutral counterpart. That is, increases
in risk aversion decreases treatment. Does this mean Bt corn is a risk-
increasing technology?
Valuing Pest Control 143
Pope and Kramer (1979) define an input to be risk increasing if the risk-
averse firm uses less of the input than a risk neutral producer. A discrete
analog of this definition would define a risk-increasing technology as one
that would be more readily adopted by a risk-neutral firm than a risk-averse
firm. Or, equivalently, a technology could be said to be risk increasing if
increases in risk aversion decrease the willingness to pay to adopt the
technology. The results in Table 8.1 indicate that this definition leads one
to conclude that Bt corn is indeed a risk increasing technology.
But an examination of Figures 8.1 and 8.2 shows that a characterization
of Bt corn as a risk-increasing technology is extremely misleading. Bt corn
decreases the probability of achieving low yields and increases the prob-
ability of achieving high yields. How could such a technology be charac-
terized as risk-increasing?
A concave utility function simply means that a producer prefers yield sta-
bility over yield variability because the utility gain from positive yield shocks
is less than the utility loss from negative yield shocks. As shown in the fig-
ures, Bt corn shifts the distribution of corn yields to the right and increases
the variability of yield. Both risk averse and risk neutral producers value
equally the shift to the right in the distribution. But risk-averse producers
lose utility, relative to the status quo, because the standard deviation of
and the leftward skewness of yields increases with Bt corn. This disutility
occurs even if the increase in standard deviation is caused by an increase
in upside risk. It is in this sense that expected utility theory characterizes
Bt corn as risk-increasing.
References
Babcock, B. A, E. K. Choi, and E. Feinerman. {1993}, "Risk and Probability Premiums
for CARA Utility Functions." Journal of Agricultural and Resource Economics, 18:
17-24,
Babcock, B. A, and D. Hennessy (1996), "Input Demand Under Yield and Revenue
Insurance." American Journal of Agricultural Economics, 78: 416-27.
Calvin, D. D. (1996), "Economic Benefits of Transgenic Corn Hybrids for European
Corn Borer Management in the United States." Public Interest Document Supporting
the Registration and Exemption from the Requirement of a Tolerance for the Plant
Pesticide Bacillus thuringiensis subsp. kurstaki Insect Control Protein as Expressed
in Corn (Zea mays L.),.
Fox, G. and A Weersink (1995), "Damage Control and Increasing Returns." American
Journal of Agricultural Economics, 77(1}: 33-39.
Hennessy, D. A (1997), "Damage Control and Increasing Returns: Further Results."
American Journal of Agricultural Economics, 79(3}: 786-91.
Just, R. E., and R. D. Pope (1979), "Production Function Estimation and Related Risk
Considerations." American Journal of Agricultural Economics, 61: 276-84.
Kalaitzandonakes, N. (1999), "A Farm-Level Perspective on Agrobiotechnology: How
Much Value and for Who?" AgBioForum 2(2}: 61-64.
Lichtenberg, E. and D. Zilberman (1986), "The Econometrics of Damage Control: Why
Specification Matters." American Journal of Agricultural Economics, 68: 261-173.
Nelson, C. J. and P. V. Preckel (1989), 'The Conditional Beta Distribution as a Sto-
chastic Production Function." American Journal of Agricultural Economics 71 (2):
370-378.
Pope, R. D. and R. A Kramer {1979}, "Production Uncertainty and Factor Demands
for the Competitive Firm." Southern Economic Journal; 46(2): 489-501.
Saha, A, C. R. Shumway, and A Havenner (1997), "The Economics and Economet-
rics of Damage Control." American Journal of Agricultural Economics; 79(3}: 773-85.
9
The Influence of Price Risk on
Set-aside Choice in the EU
Hild Rygnestad and Robert W. Fraser
9.1. Methods
In this section a theoretical model for price risk is developed based on a
mean-variance framework. This price risk is then built into the simulation
model from Fraser and Rygnestad (1999) and Rygnestad and Fraser
(1996) through a utility function based on the expected profit level, the
profit variance and the individual farmer's attitude to price risk. Note that
as only output price risk is considered in this analysis, all costs are
known with certainty.
Price risk can be described through a probability density function. And
as presented in Fraser (1994) the support system in the EU can lead to a
price distribution as in Figure 9.1.
Frequency
Price
p E(pw)
Figure 9.1 assumes that the underlying world price for wheat has a
normal distribution with an expected value, P, and a known variance,
var(p), and that the intervention price, j3, underwrites the expected world
price. Because the farmer is guaranteed a price of at least j3, the price
support scheme leads to a higher expected producer price, E(pw) , and a
smaller variance, var (Pw).
The expected wheat price and the price variance can be described as
in equations (9.1) and (9.2), respectively (Fraser, 1994):
. . . - ali Z ( P)
E ( Pw ) = F ( P ) P + ( 1 - F ( P ) ) [ p + (1 _ F (p)) J
(9.1 )
var (p ) = [ 1- F ( P)] u? 1 _[ Z ( P~ ]2 + ( P- p) Z ( P~
w P (1- F ( p)) Up (1- F ( p))
(9.2)
- apZ(p)
E
= p+ 1-F(p)
(9.3)
The Influence of Price Risk on Set-aside Choice in the EU 149
Thus, the total variance of profit is the sum of profit variance from each
land quality with adjustments for the covariance. The covariance between
profit from different land qualities can be described as (Heimberger and
Chavas, 1996):
Substituting equation (9.5) into equation (9.4) leads to the formula for
calculating the variance of profit:
(9.7)
150 Hild Rygnestad and Robert W. Fraser
R = _U"( I( ) I( .
U'( I( )
The yearly expected utility of profit is discounted over the six year pol-
icy period to arrive at a Net Present Value (NPV) of expected utility to
compare different set-aside schemes:
NPV=~l' rE(V(I(,))
L..,=l (1+r /,-1) (9.10)
(9.12)
(9.13)
On a farm with homogeneous land quality j (Le. Itj = 1 and ati = lti = 0),
equations (9.12) and (9.13) simplify to:
(9.14)
ovar (1() 2 2 0Y
_--,---...::.t~= 2 L (1- at j) var ( Pw ) - (9.15)
ONj ON
By substituting equations (9.14) and (9.15) into (9.11) the optimal nitrogen
rate for land quality j can be found:
Note that, compared to the nitrogen rate without risk, the new optimal
rate is not only dependent on the input price and yield response function
parameters. It is also dependent on farm size, the set-aside rate, compen-
satory payments, regional yield, variable, semi-fixed and fixed costs as well
as the expected level and variance of the output price and the farmer's at-
titude to risk. If, on the other hand, the farm has heterogeneous land qual-
ity, the optimal fertilizer rates must be determined by solving a set of si-
multaneous equations due to the covariance (see equation (9.6)).
The leaching function is maintained from the original model:
N',
X t = :1:'7= d ( I tj - atj ) x j e°.7 ( n: -1) + atj Yj 1 (9.17)
Table 9.1 shows parameter values for prices for season 1998/99. The
world and intervention prices are obtained from publications, and the above
formulas lead to an expected wheat price of 126.13 EU R/t.
The economic model used in this paper is based on Rygnestad and Fraser
(1996). The essential modifications are the use of yield response functions es-
timated from the results of Danish scientific field trials (Landsudvalget for Plan-
teavl, 1985-98), and the use of Danish cost and income structures for the
1998/99 growing season, including the appropriate subsidy income (see Table
9.3). Note that whereas the parameter values for the average land quality yield
response function (see Table 9.2) are estimated directly from the results of the
field trials, the parameter values for the poor and good land quality functions
are established by subjectively modifying these direct estimates 4 . In addition
note that the definition of utility in equation (9.9) is not applicable if the expected
profit level is negative. However, with the above base scenario, the estimated
profit from wheat and set-aside are negative on many land quality combinations
- as is also the case in Danish farm statistics on the average net profit of wheat
production (SJFI, 2000 b). A less partial analysis would include other crops as
well as livestock production, thereby leading to a positive farm profit. According
to farm statistics, the average profit on full time farms in Denmark was close to
The Influence of Price Risk on Set-aside Choice in the EU 153
Land qualities
Unit Poor Average Good
mj 5.03 8.03 10.03
0 0.49 0.51 0.52
bj 10.24 10.26 10.27
Note: Own estimations based on field trial data from Landskontoret for Planteavl (1985-1998).
Land qualities
Unit Poor Average Good
Fertilizer costs, Pr (If EUR/ha 101 134 150
Production costs, Pp m EUR/ha 167 267 333
Fixed costs EUR/ha 838 838 838
Total costs EUR/ha 1,106 1,239 1,321
The fertilizer price is the average price for commercial fertilizer in 1998 weighted by nitrogen content.
Agricultural conversion rate: 1 EUR =7.43 DKR.
Sources: SJFI (2000 b); Commission of the European Communities (2000) ; Statistics Denmark
(2000).
mentation of the Agenda 2000 reform and assuming no changes to the world
price for wheat.
Table 9.4: Standard Leaching Rates for Three Land Qualities, Denmark 1
Land qualities
Unit Poor Average Good
Standard leaching winter wheat, x kg/ha 39
Standard leaching Rot set-aside,r kg/ha 31
Standard leaching NR set-aside, r kg/ha 19
Table 9.5: Prices and Risk after Implementation of the Agenda 2000 Reform
expected utility for the two options to shift towards unity. This is illus-
trated by Figure 9.2 which is a two-dimensional representation of land
quality variation between poor and average, and which shows for R=0.6
the ratios of expected utility for all land quality combinations before and
after the introduction of price risk. But more significantly, for some land
quality combinations this shift results in the ratio of expected utility
switching from less than unity (i.e. a preference for the Non-rotational
option) to greater than unity (i.e. a preference for the Rotational option).
Moreover, the significance of this "switching" feature of the numerical re-
sults is only emphasized by adjusting the specification of wheat price risk
from that prevailing in 1998-99 (see Table 9.1) to that following from the
Agenda 2000 agreement (see Table 9.5). As indicated by a comparison of
these tables, the implementation of the Agenda 2000 reform would see a
substantial increase in the level of price risk faced by EU producers (i.e.
the CV of producer prices increases from 0.10 to 0.158). And as shown in
Figure 9.3, this increase in price risk results in a strengthening of the ten-
dency for producers with heterogeneous land qualities to switch from pre-
ferring the Non-rotational set-aside option to preferring the Rotational op-
tion. Consequently, this analysis suggests that implementation of the
Agenda 2000 reform would see farmers switching away from a preference
for Non-rotational set-aside towards a preference for Rotational set-aside.
1.015
1.010
it
z
t
e. 1.005
~
~ 1.000
~II
'I--R=O
- 6 - R=0.6
~ 0.995 !
'0
>
IL 0.990
'0
0
i:
II: 0.985
0.980 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _---'
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
% Poor land (reat average)
1. 004 1
« 1.002 '
z
l 1.000 +--,----~__=;::J::::::;::::::*~::::;-==-____:::T:::::,--i
~ . -+- NPV ratio, R=O
:g 0.998
__ Basis, R=0.6
:; ___ Agenda 2000, R=0.6
¥ 0.996
.=-
'0 0.994
0
;},
0.992
0.990
30% 33% 35% 38% 40% 43% 45% 48% 50%
% Poor land (rest average,
Figure 9.S: Ratio of Expected Utility for Different Levels of Poor Land
and Risk Aversion, with the Agenda 2000 Reform.
The Influence of Price Risk on Set-aside Choice in the EU 157
Table 9.6: Relative Expected Utility (Net Present Value) with Price Risk,
Heterogeneous Land Quality and Different Levels of Risk aversion
Land quality
40% poor, 45% average and Expected util- Output ratio Leaching
15% good soil ity ratio ratio
Risk aversion = 0 0.9967 0.9932 1.0360
Risk aversion = 0.6 0.9998 0.9931 1.0356
Agenda 2000 & 1.0001 0.9928 1.0355
Risk aversion = 0.6
9.4. Conclusion
The primary aim of this chapter has been to extend the theoretical model
of set-aside choice outlined in Rygnestad and Fraser (1996) to include
the role of price risk and risk aversion on this choice in the context of the
EU set-aside policy.
It has been shown that the overall effect of the recognition of price risk
and risk aversion is to dilute the significance of the role of heterogeneous
land quality in determining a farmer's preference between the two set-
aside options. In particular, the inclusion of price risk and risk aversion
typically shifts the ratio of the expected utility of the two options towards
unity for all land quality combinations.
More significantly, this shift results in a preference switch from the
Non-rotational set-aside option to the Rotational option for a range of
farms with heterogeneous land quality. In addition, this "switching" fea-
ture is exacerbated by the inclusion of price risk at a level consistent with
the implementation of the Agenda 2000 reform. Specifically, it was shown
for the case of cereal farming in Denmark that the average land quality
combination across all farms is very close to those combinations with the
"switching" feature. Moreover, those farms, which do switch to set-aside
158 Hild Rygnestad and Robert W. Fraser
options will exhibit improved output control, but significantly worse nitrate
leaching.
Consequently, it was concluded that the implementation of the Agenda
2000 cereal reforms could see a substantial switch of farmer preference
towards the Rotational set-aside option and, associated with this switch,
improved output control but also increased nitrate leaching. These find-
ings have clear implications for nitrate abatement policies currently oper-
ating within the CAP.
Endnotes
4 The adjusted R2 for this estimation was 0.54. with all parameter values signifi-
cant at the 5% level. The basis of subjectively modifying these direct estimates
was by reference to the results from Danish field trials for low, average and high
yielding land (Landsudvalget for Planteavl, 1985-1998).
References
Agra Europe (1993a). Changes to EC set-aside will increase 'slippage'. No. 1550, July
9, P/1-P/2.
Agra Europe (1993b). Commission paper exposes weaknesses of set-aside. No. 1543,
May 21, P/1-P/2.
Ansell, D.J. and SA Vincent (1994). An evaluation of set-aside management in the
European Union with special reference to Denmark, France, Germany and the UK.
Centre for Agricultural Strategy Paper 30.
The Influence of Price Risk on Set-aside Choice in the EU 159
Babcock, B.A., W.E. Foster, and D.L. Hoag (1993). Land quality and diversion deci-
sions under U.S. commodity Programs. Review of Agricultural Economics, 15(3}:
463-471.
Brown, C. (1993). CAP reforms in historical and international perspective. Department
of Agriculture, University of Queensland, Agricultural Economics Discussion Paper
193.
Commission of the European Communities (2000). The agricultural situation in the
Community 1999 report. Office for Official Publications of the European Communi-
ties, Luxembourg.
Council of the European Communities (1999). Regulation amending regulation (EC)
No. 1766/92 on the common organisation of the market of cereals and repealing
regulation (EEC) No. 2731/75 fixing standard qualities for common wheat, rye, bar-
ley, maize and durum wheat. Official Journal of the European Communities, No.
1253/1999, L 160/18-20, 26 June.
Danmarks Statistik (2000). Priser pa vigtige handelsgf2ldninger (Prices of important
commercial fertilisers). Danish Statistics, Statistiske efteretninger 2000:3, Copenha-
gen.
European Commission (1999). Berlin European Council: Agenda 2000, conclusions of
the Presidency. Newsletter.
Firbank, L.G., H.R. Arnold, B.C. Eversham, O.J. Mountford, G.L. Radford, M.G. Telfer,
J.R. Treweek, N.R. Webb, and T.C.E. Wells (1993). Managing set-aside land for
wildlife. Institute of Terrestrial Ecology, Research publication no. 7.
Fraser, R. W. (1993). Set-aside premiums and the May 1992 CAP reforms. Journal of
Agricultural Economics, 44(3}: 41 0-417.
Fraser, R. W. (1994). The impact of price support on set-aside responses to an in-
crease in price uncertainty. European Review of Agricultural Economics, 21: 131-
136.
Fraser, R. W. and H. Rygnestad (1999). An assessment of the impact of implementing
the European Commission's Agenda 2000 cereal proposals for specialist wheat-
growers in Denmark. Journal of Agricultural Economics, 50(2}: 328-335.
Heimberger, P.G. and J.P. Chavas (eds.) (1996). The economics of agricultural prices.
New Jersey: Prentice Hall.
Hoag, D. L., B.A. Babcock, and W.E. Foster (1993). Field-level measurement of land
productivity and program slippage. American Journal of Agricultural Economics,
February, pp. 181-189.
Krause, M. A., J.H. Lee, and W.W. Koo (1995). Program and nonprogram wheat acre-
age responses to prices and risk. Journal of Agricultural and Resource Economics,
20(1}: 96-107.
Landbrugets Radgivningscenter (1999). Handbog i plantedyrknin~ 1999 (Handbook for
arable farming 1999). the Danish Agricultural Advisory Centre, Arhus.
Landsudvalget for Planteavl (1986-1998). Oversigt over landsforsf2lgene. Forsf2lg og
undersf2lgelser i de landf2lkonomiske foreninger (Summary of agricultural experi-
ments). the National Committee on Crop Production, Arhus.
Magid, J., N. Christensen, and E. Skop (1994). Vegetation effects on soil solution
composition and evapotranspiration - potential impacts of set-aside policies. Agri-
culture, Ecosystems and Environment, 44: 267-278.
Newbery, D.M.G. and J.E. Stiglitz (eds.) (1981). The theory of commodity price stabili-
zation. A study in the economics of risk. Oxford: Clarendon Press.
160 Hild Rygnestad and Robert W. Fraser
ance programs in the Corn Belt (Iowa, Illinois, Indiana, Ohio, and Mis-
souri). Specific objectives include estimating the effects of production
risks on farmers planting decisions, and then measuring how cropping
patterns would be affected by two revenue insurance programs. The ef-
fects of production risks on planting decisions are first measured by es-
timating a system of acreage response equations that use risk measures
as independent variables. The effects of revenue insurance are then es-
timated by simulating the effect of changing production risks and ex-
pected returns on cropping patterns.
The economic and fiscal performance of insurance as a means of farm
income protection is investigated in several studies (Turvey 1992a
1992b; Gray, Richardson, and McClasky, 1994; Hennessy, Babcock, and
Hayes, 1997). Research also focuses on the effect of crop insurance on
chemical use at the intensive margin. For example, Horowitz and
Lichtenberg (1993), in an analysis of Midwestern corn farmers, find that
crop insurance increased fertilizer use by 19 percent and pesticide ex-
penditures by 21 percent. In contrast, Smith and Goodwin (1996) and
Babcock and Hennessy (1996) conclude that crop insurance decreases
fertilizer and chemical use. These studies, however, do not include the
effect of crop insurance on cropping patterns, which may be of greater
importance in affecting total chemical use and environmental quality than
per-acre input use. Wu (1999) examines the effect of federal crop insur-
ance on cropping patterns and chemical use and find that the "extensive
margin" effect of crop insurance can dominate the effect of crop insur-
ance on chemical application rates, leading to an increase in total chemi-
cal use and nonpoint source pollution. Wu (1999), however, focuses on
crop rather than revenue insurance programs on cropping patterns.
In the next section, we present a model to examine the effect of pro-
duction risk on farmers' acreage decisions in the presence of a revenue
insurance program, followed by a discussion of empirical specification,
data, and estimation procedures. The empirical results and policy impli-
cations are discussed in the last two sections.
flecting the random state of nature and stochastic market factors that
affect crop yields and prices. w is normalized such that 0 :$ W:$ • De-
note the density function of w as g(w). Let TC i (s, W) be the per-acre
profit from growing crop i on type-s land. The assumption that growing
crop 1 is more profitable and riskier on soil type s implies that
ETC 1(s, w) > ETC 2 (s, w) and var(TC I (s, > var(TC 2 (s, w» Let liCs) w».
be the proportion of type-s land allocated to crop 1. Then the total acre-
age of crop 1 is
1
Al =f r1 (s)f(s)ds (10.1 )
o
(10.2)
where J(a) is the premium rate, and c{ (R,a,s) is the per-acre production
cost on type-s land under revenue insurance. Note that revenue insur-
ance may change the profit distribution for crop 1 in two ways: a) by
truncating the revenues, and b) by altering the farmer's input use which
in turn affects production costs and profits. 2 Thus, the production cost
also depends on R and a. Even if the farmer does not receive any in-
demnity payment, profit may still be affected by the insurance payment
because of the effects on input use.
164 lunJie 1. Wu and Richard M. Adams
E1C{ (s) - E1C2 (s) - RACov(II, 1C{ -1C2) + fJ(s) - yes) = 0, (10.5)
where, 1C{ (s) is the per-acre profit from crop 1 under revenue insurance
(defined in the Appendix), RA '" -U"(EII)jU'(EII) is the Arrow-Pratt abso-
lute measure of risk aversion, and fJ(s) and yes) are the Lagrange multipliers
for the constraints of 0 s '1(s) s 1and satisfy the Kuhn-Tucker conditions:
P(sh (s) = 0, yes )(1 - f) (s» = O. The third term in (10.5) measures the
farmer's risk premium to grow crop 1. If the additional profit from growing
crop 1, E[1C{ (s) -1C2(S)] , is greater than the risk premium, yes) is positive.
From the Kuhn-Tucker condition, 1j(S)=1, i.e., the type-s land will be allo-
cated to crop 1. On the other hand, if the additional income from growing
crop 1 is less than the risk premium, the land will be allocated to crop 2.
Equation (10.5) indicates that revenue insurance affects crop mix in two
ways. First, it affects crop mix through its impact on farmers' expected
profit. This result is especially relevant because federal crop insurance has
historically been subsidized in the U.S. For example, in March 2000, the
U.S. Congress passed a bill to subsidize insurance premiums by up to 60
percent. The subsidy would increase farmers' expected profits for insured
crops, which would result in more land allocated to these crops. Second,
revenue insurance affects crop mix through its impact on the risk premium.
As shown in the appendix, the risk premium under revenue insurance is
smaller than without insurance. Thus, the farmer will be more likely to plant
crop 1 even if the insurance program is actuarially fair. The more risk
averse the farmer, the larger the effect of the revenue insurance on land
allocation.
Production Risks, Acreage Decisions, and Revenue Insurance Programs 165
(10.6)
(10.7)
(10.8)
(10.9)
with i=1 ,2 and where subscripts j and t are added to indicate county and
year because county-level, time-series data are used to estimate the
equations (see the discussion of data below). .Li= Aijt = TA j is the
total acreage of cropland and potential cropland In c6unty j, which is
assumed to be constant over time. TA j is defined as the maximum of
the 1982, 1987, and 1992 acreages of cropland, pastureland, and
rangeland in county j. The years of 1982, 1987 and 1992 were chosen
because agricultural censuses were conducted in these years and data
were available. One advantage of the logistic specification is that it en-
sures that predicted land use proportions remain between zero and one
and sum to one. In addition, the logit model is a flexible functional form,
and has been shown to outperform other flexible functional forms, in-
cluding the translog (e.g., Lutton and LeBlanc, 1984).
From the logistic regression model, the acreage elasticity of crop i with
respect to an independent variable x in county j in year t can be derived
to be (see Greene, 1993, p. 697):
(9.10)
where, '7kx is the coefficient of x in the equation for crop k, and Sijt is the
share of crop i in county j in year t. Since the acreage elasticity depends
on several estimated coefficients, it is important to provide measures of
statistical significance for the estimated elasticities.
10.2.2. Estimating the effects of revenue insurance on crop mix
Federal revenue insurance programs reduce farmers' production risk by
guaranteeing a revenue floor. The resulting censored distributions of
crop revenue affects both the expected value and variance of revenue.
Since the effects of censoring are best understood in the context of a
normal distribution (Chavas and Holt, 1990), we examine the effect of
Production Risks, Acreage Decisions, and Revenue Insurance Programs 167
RI = {ali if R<aR
(10.11)
R if R"i! aR.
The expected value and variance of RI are (Chavas and Holt, 1990)
where, h = (aR - E(R»/V(R)i/2 , and ¢O and <1>0 are the density and dis-
tribution functions of the standard normal, respectively.
Given the effect of revenue insurance on the expected value and
variance of revenue, the change in the share of crop i in county j under
revenue insurance is estimated by 3
e Wijt
(10.14)
1+ k 2
e Wkjl
where,
2 I
Wijl = 7Joi + }: [1JikiE(~jl)+1J2kr(~jt)]+'l3iCol(,Rljt,R2jl)+17 4i Zi,
k=i
I
Wijl = 7Joi +
}:2 I nI
[1Jiki E(!?kjl) + 1J2kr(~jt)] + 'l3iCol(.l?ljl,R2jl) + 17 4i Zi'
I
k-i .
The correlation between corn and soybean revenues is determined by
weather conditions and by corn and soybean markets. Because corn and
soybeans require similar weather and soil conditions and corn and soy-
bean prices are highly correlated, revenues from growing corn and soy-
beans are also highly positively correlated. This positive correlation re-
duces the effect of revenue variances on the corn-soybean mix because
when revenue from growing corn is low, revenue from growing soybeans
168 lunJie 1. Wu and Richard M. Adams
also tends to be low. The correlation variable was included in the acre-
age response equations to reflect this relationship between corn and
soybean. Since revenue insurance neither affects weather nor directly
affects corn and soybean markets, we assume that the correlation vari-
able is not affected by revenue insurance in the simulations.
10.2.3. Data
Historical data on county crop acreage collected by the National Agri-
cultural Statistic Service (NASS) were used for this analysis (downloaded
from the NASS website). The pooled time-series and cross-sectional
data cover counties in the Corn Belt from 1975 to 1994.
Following Gardner (1976), the expected prices for corn and soybeans
were specified as the average futures prices in the planting season,
which were estimated as the average of the first and second Thursday
closing prices in March on the Chicago Board of Trade (CBT) for Decem-
ber corn and November soybeans. During the study period, the expected
price for corn was influenced by government price supports, which
should be reflected in the futures price for corn. In addition, the Acreage
Reduction Programs (ARP) directly affect acreage decisions. Thus, the
ARP rate for corn was included in the model as an independent variable.
The ARP rate was taken from Green (1990). A dummy variable for 1983
is included to account for the effect of the payment-in-kind (PIK) program
offered in that year. Time-series data on input prices were also included
as independent variables, which include farmer wage rates and the
prices (indices) paid by farmers for seed, fuel, and chemicals. All prices
are normalized by the index of prices paid by farmers for all inputs in-
cluding interest, taxes, and wages (U.S. Department of Agriculture, 1971-
1997).
The perceived variances of corn and soybean prices are estimated
following Chavas and Holt (1990). Specifically,
3
where, the weights wJ are .5, .33, and .17, and E t - J - 1 is the expectation,
at planting time in period t-j, of the price for crop i at harvesting in period
t-j. Sensitivity analysis indicates that the results of this analysis are in-
sensitive to the choice of the weights.
Given the expected prices and yields and their variance measures, the
expected revenue and the variance of revenue for corn and soybean are
estimated by the following expressions (Bain and Engelhardt, 1992, p.177):
Production Risks, Acreage Decisions, and Revenue Insurance Programs 169
The mean and variance of corn and soybean yields during the study pe-
riod were estimated following Chavas and Holt (1990). Specifically, corn
and soybean yields were first regressed on a trend variable for each
county. The resulting predictions were taken as expected yields, and the
estimated residuals were used to generate the variance of yields and the
correlation between price and yield. The non-truncated correlation be-
tween price and yield was estimated to be -0.293 for corn and -0.149 for
soybeans. Historical data on corn and soybean yields for each county in
our study region were taken from the National Agricultural Statistics
Service.
The land quality variables were derived by linking the National Re-
source Inventories (NRI) with the Natural Resource Conservation Serv-
ice's (NRCS) Soil5 database. The NRI is conducted every five years by
the NRCS to determine the status, condition, and trend of the nation's
soil, water, and other related resources at more than 800,000 sites
across the continental U.S. Each NRI site is assigned a weight (called
the expansion factor or xfactor) to reflect the acreage the site represents.
Each NRI sample site is linked to the NRCS's SOIL 5 database, provid-
ing detailed soil profile information from soil surveys. From the data, av-
erage measures of soil properties for the top soil layers were estimated.
These include average organic matter percentage, clay percentage, soil
pH, and permeability at each NRI site. The data also include information
about soil texture and land capability class. By using the expansion
factor at each NRI site, the average land quality for each county was es-
timated.
Weather data from 1975 to 1992 were obtained from the Midwestern
Climate Center. The mean and variance of temperature and precipitation
during corn and soybean growing seasons were estimated from these
weather data and included in the crop choice model.
The dependent variables, lagged by one year, were also included in
the model to reflect the impact of crop rotation on crop choice. If soy-
beans were planted in the previous year, the current year's crop is more
likely to be corn because a corn-soybean rotation is the most commonly
used cropping system in the Corn Belt. A trend variable is included to
170 JunJie J. Wu and Richard M. Adams
cally, the effects on crop mix of programs guaranteeing 50, 75 and 100
percent of the average revenue during the study period for each crop in
isolation, and then in combination are evaluated. This is achieved by using
Corn Soybeans
Variable Coefficient t-Statistic Coeffi- t-Statistic
cient
the results from the estimated relationships between crop acreage and
the explanatory variables to simulate changes in county-level crop
acreages under each insurance option. The resulting changes in crop
172 JunJie J. Wu and Richard M. Adams
Corn Soybeans
With Respect To Elasticity t-Statistic Elasticity t-Statistic
The second, and most important, aspect of the empirical analysis con-
cerns how these variables act collectively to determine the aggregate
crop acreage and crop mix within the 421 counties examined here under
conditions of a simulated crop revenue insurance program. The basis for
this simulation of crop acreages, and hence changes in crop mix, is
contained in equation 10.11. The results, for each crop separately and
then for a revenue insurance program encompassing both crops, is re-
ported in Table 10.3.
The effects of a revenue insurance program for either crop, in isolation,
are as expected. Specifically, a revenue insurance program for corn re-
sults in an increase in corn acreage under the 50, 75 and 100 percent
programs. A "corn only" program results in soybean acreage reduction
for the 50, 75, and 100 percent options as soybeans acreage is diverted
to corn. When only soybeans are covered, soybean acreage increases.
However, corn acreage also increases. This is due to the rotational con-
174 lunJie 1. Wu and Richard M. Adams
Note: The simulations are based on the 1993 price and acreage levels. The insur-
able reve nue levels are assumed to be the average of corn and soybean
revenue from 1985 to 1994. Percentage changes are in parentheses.
In the case of a revenue insurance program that covers both crops, the
results differ across the program options. For example, under the 75 per-
cent revenue option, corn acreage increases by 12 percent, while soy-
Production Risks, Acreage Decisions, and Revenue Insurance Programs 175
bean and "other" acreage declines. Under a 100 percent revenue cover-
age option, corn acreage increases by nearly 23 percent, and by neces-
sity soybean and other crop acreage decreases to accommodate the ex-
pansion in corn acreage. It should be noted that these changes in acre-
age are understated due to the failure to consider moral hazard possibili-
ties. For example, revenue insurance may reduce farmers' use of risk-
reducing inputs. Qualitatively, the "intensive margin" effects would in-
crease the crop mix effect of insurance. That is, adjusting input uses in
the presence of the revenue insurance will increase the profit from the
insured crop, making it more attractive relative to the uninsured alterna-
tive. 8 To adequately address the role of input adjustments requires farm-
level data and specifications, which is not possible using the county-level
specifications and data employed here. However, other research indi-
cates a substantial moral hazard issue in designing federal crop insur-
ance programs (e.g., Chambers, 1989; Goodwin, 1993). Wu (1999) esti-
mated the crop-mix effect of federal crop insurance programs using farm-
level data of crop insurance participation in the central Nebraska Basin
and found a much larger adjustment in crop mix.
One implication of the theoretical model developed in this chapter (and
discussed elsewhere in the literature) is that revenue insurance pro-
grams may affect the environmental performance of agriculture by
changing crop mix. For example, corn and soybeans account for over 70
percent of total cropland acreage in the Corn Belt and use more chemi-
cals than other crops in the study region. Thus, switching from other
crops to corn or soybeans implies negative environmental conse-
quences, such as greater soil erosion or increased use (and runoff) of
agricultural chemicals. More importantly, counties with the largest per-
centage increase in corn acreage also have the greatest potential envi-
ronmental problems, as reflected in the land characteristics data. Spe-
cifically, the average slope for farmland in counties where the percentage
increase in corn acreage is above the median is much steeper than the
average slope for farmland in other counties (4.2 percent slope vs. 2.7
percent slope for all counties). In addition, the percentage of high quality
land (land of NRCS soil class I and II) is lower in counties where the per-
centage increase in corn acreage is above the median; 55 percent of the
land is classified as high quality soils in these counties vs. 70 percent in
counties where the percentage increase in corn acreage is below the
median. The presence of a corn revenue insurance program thus is likely
to lead to an increase in production from environmentally sensitive land,
with associated increases in soil erosion and other types of externalities.
176 JunJie J. Wu and Richard M. Adams
10.3. Conclusions
Endnotes
The choice of the coverage level a and land allocation {rl (s)} influences
the range of w in which the farmer receives an indemnity payment. Sup-
pose an increase in w increases the revenue potential. Then the critical
value of w below which the farmer receive indemnity payment, w C , is
defined by
I
fRI (s, wC)rl (s)f(s)ds = AaRI . (A1)
o
Thus, the Lagrangean function for the maximization problem in (10.4)
can be written as
we I
where II(rl (s),a, w) is defined by (10.3), Jl and v are the Lagrange multi-
pliers for the constraint of Os a sa, and fil(s) andrl(s) are the Lagrange
multipliers for the constraints of Os rl (s) s 1.
The first-order conditions for the optimal land allocation can be derived
by differentiating (A2) with respect to 'i(s) for a given s:
w"
-- =
OL U (IT(rj (s),a, we )-av"- + fU'(IT)--g(w)dw
m
orj (S) itj(s) orj (s)
o
c I
- U(II(Ij(s),a, 11"c ~ + JU'(II)~g(W)dW + fil (s) - rl(s) = 0. (A3)
il)(s) il)(s)
w'
The first term and the third term in (A3) offset each other. By substi-
tuting (10.3) into (A3) , we get
~
orj (s)
=
W
f 'U'(I1)[aR; - Cj(R,a,s) -lea) -JZ"2(S, w)k(w)dw
o
°
I
where, fi(s) = fil (s)/ EU' and res) = rl (s)/ EU'. By substituting the first-order
Taylor expansion of the marginal utility function U'(II) "" U'(ElI) +
U"(ElI)(II - ElI) into the third term in (A6), it can be simplified to
Now, we show that the risk premium (RP) under the revenue insurance
is smaller than without the insurance program. By definition,
where the equality holds only when the farmer is risk neutral (i.e.,
RA = 0). The difference is always non-positive because both co vari-
ances in the last brackets are negative. Intuitively, as weather improves,
the indemnity payment Jr{ - Jrl goes down, whereas the total profit goes
up. So the first covariance is negative. Also, as weather improves, 1rl
goes up, and the difference between profits with and without revenue
insurance goes down. So the second covariance is also negative.
References
Babcock, A.B., and D. Hennessy (1996), "Input Demand Under Yield and Revenue
Insurance." American Journal of Agricultural Economics, 78: 416-27.
Caswell, M., and D. Zilberman (1985), "The Choice of Irrigation Technologies in
California." American Journal of Agricultural Economies, 67: 224-234.
Chambers, R.G. (1989), "Insurability and Moral Hazard in Agricultural Insurance Mar-
kets." American Journal of Agricultural Economics, 71: 604-16.
Chavas, J.-P., and M.T. Holt (1990), "Acreage Decisions Under Risk: The Case of
Corn and Soybeans." American Journal of Agricultural Economics, 72: 529-38.
Gardner, B. L. (1976), "Futures Prices in Supply Analysis." American Journal of Agri-
cultural Economics, 58:81-84.
Gray, A. W., J. W. Richardson, and J. McClasky (1994), "Farm Level Impacts of
Revenue Assurance." Unpublished manuscript, Department of Agricultural Eco-
nomics, Texas A&M University.
Green, R.C. (1990), "Program Provisions for Program Crops: A Database for
1961-90." Agriculture and Trade Analysis Division, Economic Research Service,
U.S. Department of Agriculture. Staff Report No. AGES 9010. March.
Greene, W. H. (1993), Econometric Analysis. 2nd edition. New York: Macmillan.
Goodwin, B.K. (1993), "An Empirical Analysis of the Demand for Multiple Peril Crop
Insurance." American Journal of Agricultural Economics, 75:425-34.
Hardie, I.W. and P.J. Parks (1997), "Land Use with Heterogeneous Land Quality: An
Application of an Area Base ModeL" American Journal of Agricultural Economics,
79: 299-310.
Hennessy, DA, BA Babcock, and D.J. Hayes (1997), "Budgetary and Producer
Welfare Effects of Revenue Insurance." American Journal of Agricultural Econom-
ics, 79: 1024-34.
180 JunJie J. Wu and Richard M. Adams
Holt, M.T. and S.R. Johnson (1989), "Bounded Price Variation and Rational Expecta-
tion in an Endogenous Switching Model of the U.S. Corn Market." Review of Eco-
nomics and Statistics, 71: 605-13.
Horowitz, J.K. and E. Lichtenberg (1993), "Insurance, Moral Hazard, and Chemical
Use in Agriculture." American Journal of Agricultural Economics, 75:926-35.
Houck, J. P., and M. E. Ryan (1972). "Supply Analysis for Corn in the United States:
the Impact of Changing Government Programs." American Journal of Agricultural
Economics, 54: 184-191.
Johnston, J. (1984), Econometric Methods. Third Edition. New York: McGraw-HilI.
Lichtenberg, E. (1989), "Land Quality, Irrigation Development, and Cropping Patterns
in the Northern High Plains." American Journal of Agricultural Economics,
71 :187-194.
Lutton, T.J., and M.R. LeBlanc (1984), "A Comparison of Multivariate Logit and
Translog Models for Energy and Nonenergy Input Cost Share Analysis." Energy
Journal, 5:35-44.
Maddala, G.S. (1983), Limited-Dependent and Qualitative Variables in Econometrics.
Cambridge: Cambridge University Press.
Shonkwiler, J.S., and G.S. Maddala (1985), "Modeling Expectations of Bounded
Prices: An Application to the Market for Corn." Review of Economics and Statistics
67: 697-701.
Smith, V.H., and B.K. Goodwin (1996), "Crop Insurance, Moral Hazard, and Agricul-
tural Chemical Use." American Journal of Agricultural Economics, 78: 428-38.
Turvey, C.G. (1992a), "Contingent Claim Pricing Models Implied by Agricultural Stabi-
lization and Insurance Policies." Canadian Journal of Agricultural Economics,
40:183-98.
Turvey, C.G. (1992b) "An Economic Analysis of Alternative Farm Revenue Insurance
Policies." Canadian Journal of Agricultural Economics, 40: 403-26.
U.S. Department of Agriculture (1971-97), Agricultural Statistics, Washington, D.C.
Wu, J. (1999), "Crop Insurance, Acreage Decisions, and Non-Point Pollution." Ameri-
can Journal of Agricultural Economics, 81: 305-20.
Wu, J. and B. W. Brorsen (1995), "The Impact of Government Programs and Land
Characteristics on Cropping Patterns." Canadian Journal of Agricultural Economics,
43: 87-104.
Wu, J. and K. Segerson (1995). "The Impact of Policies and Land Characteristics on
Potential Groundwater Pollution in Wisconsin." American Journal of Agricultural
Economics, 77:1033-47.
11
The Effects of Crop Insurance and
Disaster Relief Programs on Soil Erosion:
The Case of Soybeans and Corn
Barry K. Goodwin and Vincent H. Smith
and crop residues); and the P-factor which describes erosion control prac-
tices in place (such as terracing, contour farming, etc.). The soil-erosion
index is influenced by cropping patterns and production practices, and,
thus, we utilize simultaneous-equations estimators to evaluate determi-
nants of soil erosion as well as the effects of erosion on production prac-
tices, including participation in the federal crop insurance program.
Out of necessity, therefore, the data used in the study are drawn from
several different sources. Unpublished countywide data on premium
rates and participation in federal crop insurance programs for corn and
soybeans were obtained from the Federal Crop Insurance Corporation
Office of Risk Management. Data on county level crop plantings, the K-
factor, and the soil erosion index were obtained from the NRI. Informa-
tion on disaster payments at the county level were obtained from the En-
vironmental Working Group Agricultural Disaster Data Base. Data on
other government payments, fertilizer and other chemical expenditures,
and crop revenues were obtained from the U.S. Department of Com-
merce Regional Economic Information System county tables. Finally,
data on county yields and acreage were obtained from the National Agri-
cultural Statistical Service (NASS).
A complete data set that included land use variables and the endoge-
nous soil erosion index was constructed for three Agricultural Census
years - 1982, 1987 and 1992. A continuous data set that included the
universal K-factor, insurance program participation data, and relevant
economic variables was constructed for the period 1982 through 1992.
We utilize the USDA Economic Research Service's farm resource region
that represents the main corn and soybean producing region designated
as the Heartland. This includes all of Illinois, Indiana, and Iowa as well as
large portions of Missouri and Ohio, and smaller parts of Ohio, Kentucky,
Minnesota, Nebraska, and South Dakota. Farm resource regions are de-
fined to include relatively homogeneous counties. Detailed descriptions
of the variables used in the analysis are presented in Table 11.1 together
with sample means and standard deviations.
Rate Per Acre Real premiums paid per insured acre 0.0519 0.0177
(Soybeans)
Rate Per Acre Real premiums paid per insured acre 0.0715 0.0230
(Corn)
Erosion Index NRI index of wind and water soil erosion 4.5418 2.9088
11.2.1. Do farmers with land that is inherently more erodible have higher
participation rates in federal crop insurance programs?
To obtain some preliminary inSights about this question we estimated
standard single equation ordinary least squares and state level fixed ef-
188 Barry K. Goodwin and Vincent H. Smith
Participation Liability
Crop Revenue -0.0119 0.0741- 0.0543 0.1550- -0.0036 0.0583- 0.0365* 0.0846-
-
~
"'tl
Proportion (0.0337) (0.0311) (0.0373) (0.0348) (0.0210) (0.0200) (0.0227) (0.0271)
Universal K-Factor 0.0002 -0.0013 0.0104- 0.0042- -0.000004 -0.0010 0.0062** 0.0026'
(0.0018) (0.0018) (0.0022) (0.0021) (0.0016) (0.0011) (0.0014) (0.0013)
~
Price Elasticity at -0.6483 -0.5758 -0.9909 -0.9196 -0.6274 -0.6255 -1.1736 -1.0544 00
\0
Means
190 Barry K. Goodwin and Vincent H. Smith
11.2.2. Does the federal crop insurance program result in higher levels of
soil erosion?
The findings reported in Table 11.2 only provided initial insights about
the issue of whether farmers with more highly erodible land participate in
the federal crop insurance program because they are estimated using
single equations techniques and examine only the inherent erodibility of
land and not actual soil erosion. The issue of whether soil erosion in-
creases as a result of the federal crop insurance program is more appro-
priately evaluated in a simultaneous equations model, in which farmers
are assumed make decisions about program participation for crops and
input use (both with respect to land and agricultural chemicals) at the
same time. As previously noted, actual erosion is likely to be endoge-
nous with respect to production and conservation practices as well as
such policy parameters as crop insurance, disaster relief, and the con-
servation reserve program.
Thus, a four-equation model was estimated in which the endogenous
variables are the county-level soybean and corn MPCI participation
rates, the level of expenditures on fertilizers and other chemical inputs,
and the soil erosion index. The latter takes indirect account of effects
resulting from land use in soybean, corn, and other crop plantings, pas-
ture, and other agricultural uses, because changes in land use drive
changes in the soil erosion index at the county level. The four-equation
simultaneous systems model was estimated using three-stage least
squares for different sets of explanatory variables, including state spe-
cific effects. Results, which were invariant to the inclusion of state-
specific fixed effects, are reported in Table 11.3.
In the simultaneous equations model, the participation equations for
corn and soybeans include the same set of variables as do the single
equation models in Table 11.2, except that the universal K-factor is re-
placed by the (endogenous) soil erosion index in the simultaneous equa-
tions model for which results are reported in Table 11.3. In this model,
there is no evidence of any causal relationship from estimated soil ero-
Crop Insurance and Disaster Relief Programs 191
aNumbers in parentheses are standard errors. Single and double asterisks indicate statistcal
significance at the a = .10 and .05 levels, respectively. Number of observations is 625.
Crop Insurance and Disaster Relief Programs 193
This analysis evaluates the relationship between soil erosion and partici-
pation in the federal crop insurance program for corn and soybeans in six
important states. The results can be summarized as follows. First, par-
ticipation in crop insurance programs tends to be lower in counties where
land is inherently more erodible. Second, over time participation in crop
insurance programs has increased more rapidly in areas that are inher-
ently more erodible, suggesting that participation could actually be higher
in such areas at some pOint in the future. Finally, when endogenous soil
erosion is considered simultaneously with crop insurance participation for
corn and soybeans as well as with fertilizer usage patterns, the results
indicate that insurance purchases by corn producers are associated with
increased soil erosion while insurance purchases by soybean producers
are associated with less erosion.
These results are pertinent to the issue of "extensive margins;" that is,
acreage that is marginally productive and typically considered to be more
highly erodible. However, it is important to note that these margins exist
at two different levels. The first involves production in areas of the coun-
try not usually thought of as important in production of a particular crop.
An example might include dryland wheat production in West Texas. A
second extensive margin involves the marginal land located in major
producing regions. Our analysis is directed toward the latter margin as
we believe the effects of commodity-specific programs such as the MPCI
insurance program are likely to be more significant in such areas. 6
Although our research provides important implications regarding soil
erosion and agricultural pOliCies, much work remains to be done. The
NRI database contains precise measures of the productive capability of
land.? Future research is needed to examine the cultivation capability of
land and its relation to insurance program participation and crop produc-
tion. Further, other crops, including wheat and grain sorghum, may also
have important soil erosion effects that should be investigated in this
194 Barry K Goodwin and Vincent H. Smith
context. Finally, the margin in areas with less significant levels of pro-
duction also merits further analysis.
Endnotes
References
Babcock, B. and D. Hennessy (1996), "Input Demand Under Yield and Revenue In-
surance." American Journal of Agricultural Economics, 78: 416-427.
Goodwin, B. K. (1993), "An Empirical Analysis of the Demand for Multiple Peril Crop
Insurance." American Journal of Agricultural Economics, 75: 425-34.
Goodwin, B. K. and T. Kastens (1993), "Adverse Selection, Disaster Relief, and the
Demand for Multiple Peril Crop Insurance." Unpublished research report, Kansas
State University.
Crop Insurance and Disaster Relief Programs 195
As stated in the Introduction, the motivation for this book has been
prompted by two main developments in relation to agriculture over the
last decade: (i) the increasing integration of agricultural and environ-
mental policy as society's perception has grown of the relationship be-
tween agricultural activity and the quality of the environment; (ii) the
growing role for risk management skills and instruments in agriculture as
farmers have been increasingly exposed over this period to market as
well as other sources of risk. And, although at first glance it may be
thought that this book comprises a diverse range of contributions to a
very broad topic, a closer scrutiny shows that two key themes run
strongly through both the theoretical and case studies parts of the book.
The first of these, addressed by the theoretical contributions of Paul
Mitchell and David Hennessy (Chapter 3) and David Pannell (Chapter
4), and in the case studies of Amir Abadi and David Pannell (Chapter 7)
and Bruce Babcock (Chapter 8), relates to the importance of a clear un-
derstanding of the factors governing the adoption of environmentally
sensitive farming practices in uncertain conditions. The chapter by David
Pannell is particularly inSightful in this context in its highlighting of three
key "states of farmer awareness" that must be achieved for adoption to
take place:
Moreover, David argues that although farmers typically act out of "self
interest", this self-interest is potentially considerably broader than simply
the profit motive. In addition he argues that uncertainty has a number of
negative influences on these "states of farmer awareness". Conse-
quently, he suggests that the opportunity to trial an innovation takes on
particular value when these features of and impediments to the adoption
process are recognized. Specifically, the opportunity to trial allows the
Conlusion 197
References
European Commission (2001) "Risk Management Tools for EU Agriculture" Working
Document, Agriculture Directorate-General, January.
MAFF (2001) "Risk Management in Agriculture" Discussion Document, Economics
and Statistics Group, Janauary.
USDA (1999) "Managing Risk in Farming: Concepts, Research and Analysis"
Agricultural Economics Report No. 774, Market and Trade Economics Division and
Resource Economics Division.
Author Index
Newbold, C. 146
Ladd, G.W. 89 Norris, P. E.117, 118
Lantzke, N. 118 Norton, G.W. 66
Latacz-Lohmann, U. Nowak, P. 53
Leathers, H.D. 67, 113 Nunn, M. 96, 109
LeBlanc, M.R. 166 Nyarko, Y. 116
Lee, J.H. 159
Lichtenberg, E. 3, 134, 142, Oliver, M. 96
161,162,181 Olson, K. 53
Lindner, R.K. 31, 67, 68, 69, Osborn, C. 63
81,83,113,116,127
Lockie, S. 67 Pannell D.J. 67, 68, 78, 80, 81,
Long, J. 182 114,129
Long, T. 89, 92 Pardey, P.G. 81
Luenberger, D. G. 14 Parks, P.J. 161
Lutton, T.J. 166 Perry, J. 1
Petzel, T.E. 40
Machina, A. 115 Pindyck R. S. 31
Maddala, G.S. 165, 167 Pluske, J. 69
Magid, J. 146 Pope, R.D. 109, 135, 143
Mahul, 0.40 Pratt, J.W. 115
Malcolm, R.L. Preckel, P. V. 135
Malmquist, S.14 Pyke, B. 89
Maltsbarger, R. 83
Marsh, S.P. 70 Quiggin, J. 3,11,12,15,18,19,
McBride, W. 95 20, 21, 22, 24, 25, 26, 27, 40,
McClasky, J. 162 43, 54
McFadden, D. 16
McPherson, R.M. 66 Radford, G.L.159
Merton, R.C. 36, 38 Rae,J.67
Miranowski, J.A. 54 Rajotte, E.G. 66
Mirrlees, J. 24, 25 Rendleman, C. 66
Mjelde, J.W. 66 Reynolds, R. 96
Mood, A.M. 109 Ribaudo, M. 66
Mountford, O.J. 159 Richardson, J. W. 162
Myers, R.J. 69 Riley, P.A. 82
Nayda, W.1. 40 Rister, M.E. 66
Roberts, D. 146
Nelson, C.J. 135 Robinson, S.D. 132
Newbery, D.M.G. 109, 148 Rogers, E.M. 74
202 Author Index
Tanner, C. 109
Subject Index
Programming, 31
Quarantine restrictions, 96
Uncertainty
about conservation innova-
tion, 74
and risk, 118
Utility
expected, 29, 115
marginal, 33
maximization
Wealth,46