Using Analytical Tools for
Agricultural Development
John McKenzie, founder of Innovastat, has extensive experience in teaching and
consulting in the areas of Monte Carlo simulation, optimization under uncertainty,
forecasting, and inferential statistics. He is teaching a graduate class at the University
of Colorado at Boulder this fall semester in the Information and Communication
Technology for Development Program at the ATLAS Institute. His class, held in the
computer lab, is using the Palisade’s DecisionTools Suite to model development
strategies. To contact him: john.mckenzie@colorado.edu (303) 530-2809
©Innovastat All Rights Reserved
1
©Innovastat All Rights Reserved
2
HAVE RISK TOOLS
WILL TRAVEL
Email: john.mckenzie@colorado.edu
Boulder, Colorado
©Innovastat All Rights Reserved
3
Colorado
4
Boulder
5
McKenzie Farms
Since1893
©Innovastat All Rights Reserved
6
7
8
9
Local Farmer
McClean County, Illinois
Richest Soil in the World
10
One significant technology that has not been adequately
implemented in development work is the use of decision making
tools incorporating risk and uncertainty by farmers. Farmers with
limited resources need to maximize their chances of success, not
just for their own survival but for the common good. The results of
poor choices or an inability to plan for unusual events in the farm
sector can have far-reaching impacts on local economies. Yet at the
present moment there are few approaches incorporating risk for
decision making by small farmers.
A
B
S
T
R
A
C
T
I developed tools and techniques that create farm plans that include
decisions that minimize risk, taking into account weather
conditions, commodity price fluctuations, input price changes, and
even cultural characteristics.
©Innovastat All Rights Reserved
11
My initial steps have been to educate consultants, funding agencies, and
government officials in the field of risk and decision analysis through
formal training in higher education classrooms and computer labs, as well
as informally at conferences and meetings I have been asked to attend.
Funding agencies need to be shown the usefulness of risk analysis for the
small farmer so that changes in the established ways of decision-making A
B
can be elevated to a higher plateau.
S
T
R
A
C
T
My method is not imposing a top-down solution but investigating and
analyzing what currently exists at the farm or community level in the
context of risk. Implementation of decision analysis tools first involves
learning and collecting information from small farmers in the field and
developing farm plans.
Using the Monte Carlo simulation and other appropriate optimization
software, I then work directly with the farmers to implement the strategies
that most effectively minimize their risk. At every stage the farmer is
consulted and involved in the process of making his/her farm successful.
©Innovastat All Rights Reserved
12
Surprisingly, methods such as Monte Carlo simulation and
optimization under uncertainty – employed routinely throughout the
corporate world – are not being applied to solve small farmer
problems.
A
B
S
T
R
A
C
T
My work has been to utilize analytics and advanced decision
strategies with farmers. I believe that using this technology will give
the subsistence farmer the greatest chance of success (e.g.
maximizing the certainty of achieving a particular goal), and will
provide incomes that are greater and more stable from season to
season. Without the benefit of these tools to assess and manage risk,
small farmers face conditions that add significantly to their risk and
reduce their likelihood of success, sustainability or profitability.
©Innovastat All Rights Reserved
13
Beyond assisting individual small farmers, this method could greatly
enhance decision-making at the public policy level. The results of
the farmer input and the individual farm plans will reveal the
components of various systems and will demonstrate that there may
be unintended consequences of public policy that thwart a
successful outcome.
A
B
S
T
R
A
C
T
When such constraints are identified, then relevant policy solutions
that are effective, both in cost and impact, need to be explored and
implemented. These policies can be in terms of research, education,
improved targeting and delivery of farm subsidies, changes in laws,
improvements in infrastructure or credit, etc. These
recommendations are preferable to opting for a costly and inefficient
back of the envelope approach when these problems seemingly
become too complex to solve.
©Innovastat All Rights Reserved
14
Financial Times 8/3/11
©Innovastat All Rights Reserved
15
What is the Problem and Why
Should You Care?
• Malthusian Problem
Disaster in Somalia
– Exponential growth of population
– Linear growth or no growth of resources or productivity
• Food Availability and Cost
– Food Crisis of 2007-2008 – riots all over the world,
high wheat and rice prices
– Mozambique, Sep 2010 – 30% increase in bread prices
riots leaving seven dead. Rising prices partly a result of
Russian ban on exports of their wheat. Can’t participate
in the market if your income is too low.
– Two edge sword – create opportunities for exports
©Innovastat All Rights Reserved
16
World Food Problems
• Chronic Hunger
– 925 million people will face chronic hunger this year
United Nations Food and Agriculture Organization (FAO)
• Hidden Hunger
– Two billion will suffer from hidden hunger, the lack
of vitamins and minerals that affects mental and
physical growth.
©Innovastat All Rights Reserved
17
Globalization and Food Security
• Growing incomes are putting pressure on commodity
prices and farm inputs.
– Growing meat consumption
– PotashCorp in Canada
• Global agricultural markets are volatile
– Core CPI does not include food and energy prices
– Rationale for subsidies in the U.S.
• Food Production will need to increase by 70% by
2050. (Sustainability and Security of the Global Food Supply Chain,
Rabobank Group 2010)
Millennium Development Goals
©Innovastat All Rights Reserved
18
Climate Change
• Farmers have noticed it.
• Weather changes, extreme events (recent floods in
Thailand, rising temps.
• Colorado runoff has changed
• More impact in the developing world
– Sub-Sahara
• Need research in to increase productivity
including drought resistant varieties.
• Adaptive Strategies - $7 billion is needed in
improvements to offset climate changes calorie
reductions. (IFPRI)
©Innovastat All Rights Reserved
19
The Problem:
We Are Not Allocating Our Resources Efficiently
The West has invested more then $2 trillion during the last 50
years on foreign aid to help the world’s poor.
What have been the results -what is there to show for it?
Many say not much.
According to the World Bank, the majority of Information,
Technology, and Communication Development (ITCD)
projects fail. Failure Fairs are well attended.
A market failure? Anti-poverty, global hunger, food and income
issues are public goods, as they are not being provided in the
market place. Are governments and NGO’s addressing this market
failure?
20
Why Are We Not Allocating
Our Resources Efficiently?
Big head and big belly syndrome
Close may be considered good enough.
Lack of accountability
William Easterly, author, speaks to
• Experts that plan versus experts that search
Hubris of those in charge – resistance to change, over
confidence in one’s abilities, gut feelings
Hasn’t been demonstrated – needs to be driven by results.
No training in analytical techniques
21
Why Are We Not Allocating
Our Resources Efficiently?
Big head and big belly syndrome
If you don’t understand something, then you may be likely to
dismiss the technique and its application.
Resources are expended evaluating projects after the fact but less
at forecasting the efficacy of the project pre-implementation.
People want the “number.” – (determinisitc v. probablisitc)
No tangible product to sell but a process or idea.
Need to hide the technology.
22
What Are Your Options Or Consequences
In Cases of Bad Decisions?
• Stakes are higher for small farmers than for large
corporations for bad decisions.
• Farm resources can be limited – no deep pocket.
• “Layoffs” of resources is limited. Only choice
may be to shut down for a farmer.
• Asset Specificity - inflexibility to switch
resources.
• Decision making is more critical for farms and
limited resource companies than for large
corporations.
23
Problems of Policy Makers
Policy makers, as well, have limited
resources. Unfortunately, most government
and NGO decision makers do not possess
expertise in sophisticated analytical
techniques and view farmer problems and
their solutions as simple and linear when in
fact the components of these farming
systems are more simultaneous,
interdependent, and involve varying levels
of risk. These decision makers often throw
up their hands and opt for a costly and
inefficient back of the envelope approach
when these problems seemingly become too
complex.
©Innovastat All Rights Reserved
Bounded
Rationality In the
decision making process,
people speak of bounded
rationality according to
the work of Herbert
Simon. It states that
people have the inability
to process all the
alternatives of a problem.
That they can only look
at a half dozen solutions.
This is a cop out to
justify not using a
anlalytical process to
solve problems.
24
What if we could allocate our
resources better?
More bang for the buck
More people are helped
Greater poverty reduction
Increased incomes
Risk reduction
We have the expertise, data, and analytical
tools to achieve these goals today.
©Innovastat All Rights Reserved
25
Technology Adoption
Risk Tools are Technologies
The agricultural sectors in Latin America & Africa
have adopted technologies such as the introduction
of efficient irrigation systems and hybrid seeds.
However, one significant advancement that has
not been adequately implemented is the use of
decision making tools incorporating risk and
uncertainty.
Farming systems are composed of highly complex biological
systems as well as the complex social interactions. These complex
systems require complex methods to analyze and solve problems.
©Innovastat All Rights Reserved
26
Decision Making Technologies
“Though at heart most business
problems are information problems,
almost no none is using information
well. But here on the edge of the
twenty-first century, the tools and
connectivity of the digital age now
give us a way to easily obtain, share,
and act on information in new and
remarkable ways.”
“…I work on planning under
uncertainty. That’s the big field
as far as I’m concerned; that’s the
future. Maybe I’m the only one
who says that. … all the
problems that are solved under
deterministic means have that
fundamental weakness- they don’t
properly take uncertainty into
account”
Business @ The Speed of Thought,
Bill Gates
George Dantzig – “Father of Linear
Programming”
OR/MS TODAY October 1999
27
What Are These Decision
Making Technologies?
• Established Methods
– Accounting
– Enterprise Analysis
– Deterministic Optimization
Descriptive
Deterministic
• Ones That We Should Adopt
–
–
–
–
–
Monte Carlo Simulation
Simulation Optimization
Stochastic Optimization
Forecasting Under Uncertainty
Real Options
Analytical
Stochastic
28
What are some of
the risks in Agriculture?
Endogenous
Exogenous
•
•
•
•
•
Production Risks
Market Risks
Financial Risks
Legal Risks
Human Resource Risks
Specialists advise farmers on
items such as what crops to
grow and cultural methods to
use. Nevertheless, these
suggestions are not based or
evaluated on local risk
conditions. As an illustration,
experts may recommend
purchasing insurance but at
what price?
29
What Do You Mean When You Say….
Qualitative Assessment
• “Planting winter wheat around here is more or less a sure
thing with very little risk?”
• The price of corn should be around $4.50 per bu. in
September.
• “The water supply looks like it will be above average.”
• “If it snows and rains some more, then maybe river water will
be available. It would seem highly unlikely that we would get
no additional river water but it may be scant.”
May 2002 - Boulder and White Rock Newsletter sent to shareholder.
30
Innovaag
InnovaAg develops farm plans that include decisions that
minimize risk taking into account weather conditions,
commodity price fluctuations, input price changes, cultural
characteristics*, etc. These plans will give the farmer the greatest
chance of success (maximizing the certainty of achieving a
particular goal), and provide incomes that are greater and more
stable from season to season. Minimizing the fluctuations or
volatility in farm income will help the farmers avoid catastrophic
failure and allow them to remain on the land and continue
farming.
* Sociological inputs, e.g. being able to cooperate, chances of a
neighboring farmer being able to help in a timely manner. We use a
sociological indicator, the consumer confidence index in our
macroeconomic decision process.
©Innovastat All Rights Reserved
31
Community Benefits
Whole communities are the beneficiaries of the
output of the InnovaAg process. Greater incomes
will allow farmers to have more participation in
the marketplace and thus provide better for their
families in terms of housing, nutrition, health care,
and education for their children. The farmers
themselves, will decide how to best spend the
extra income that is generated, be it for family
expenditures or for farm improvements.
©Innovastat All Rights Reserved
32
Policy Makers
Beyond assisting individual small farmers, InnovaAg
will greatly enhance decision-making at the national
level. The results of the farmer input and the individual
farm plans will reveal the components of their systems
and will demonstrate that there are constraints that
thwart a successful outcome. When such constraints
are present, then relevant policy solutions by the
government need to be explored and implemented.
These policies can be in terms of research, education,
improved targeting and delivery of farm subsidies,
changes in laws, improvements in infrastructure or
credit, etc.
33
©Innovastat All Rights Reserved
Impediments That May Need Correction
Some endogenous, others exogenous
• Bad roads, lack infrastructure, corruption,
lack subsidies for factor inputs (fertilizer
and ag chemicals, lack of price floors, no
quotas, no tariffs, poor access to markets,
no cell phones, few tractors, poor
equipment, no irrigation systems, lack of
access to credit, poor varieties of seed,
intermittent electrical service…
©Innovastat All Rights Reserved
34
An Illustration of Potential Solutions
Resulting from the Model Output
• Cell Phones
– Info on prices (information transfer makes free
markets work better)
– Getting buyers and sellers together.
– Technology transfer and adoption
• Kenya Farmers Helpline
– Financial services
• Cash transfers
©Innovastat All Rights Reserved
35
More Illustrations
•
•
•
•
•
Transnational cooperation
Railways, roads, storage, refrigeration
Waste reduction through the food chain.
Access to markets
Diversification
You don’t know what
appropriate solution to
implement unless you have
figured out what the problem is
Intuition may help you create
the model and the model my
validate your intuition.
©Innovastat All Rights Reserved
36
Back to Modeling -in the news
• Malawi – Disastrous harvest of 2005 resulted in
1/3 of population needing food aid– Since then
seed and fertilizer have been subsidized (despite
free market pleas from the World Bank, their
advice was to develop export cash crops to
develop an income stream) Result: the country
become a net food exporter in two years.
• Ghana – subsidies have helped increase food
production 40%
©Innovastat All Rights Reserved
37
FARM PLANNING UNDER UNCERTAINTY
YUMA COUNTY
• Sensitivity Analysis
• Monte Carlo Simulation
• Optimization Under
Uncertainty
Case Study
38
The Future of Agriculture
• It is possible that the output of the model
will show that production agriculture is
not a sustainable activity in the long run.
• Farm in Yuma County, Colorado
– Without subsidies
– With subsidies
39
Maximize Profits
Crops To Grow
with Subsidies
Number of
Acres
Dryland Wheat
Irr Sugar Beets
Irr Alfalfa
Irr Dry Beans
1920
130
260
650
Without subsidies, the model tells us not to farm.
40
Process
InnovaAg is not imposing a top-down solution but
investigating and analyzing what currently exists at
the farm or community level in the context of risk.
Implementation of decision analysis tools first
involves learning and collecting information from
small farmers in the field and developing farm plans.
Training in applied risk analysis appropriate for small
farmers with little education will be offered. Tailormade plans will enable each farmer to decide the
optimal course of action based on his/her individual
goals and risk preferences.
©Innovastat All Rights Reserved
41
PROYECTOS AGRÍCOLAS
BAJO INCERTIDUMBRE
Aldea de Tres Sábanas
42
Collect the data
43
©Innovastat All Rights Reserved
44
45
ENSAYO
RENDIMIENTO
PRECIO
GANANCIAS
1
$25.00
$2.50
$63
2
$24.00
$3.00
$72
3
$38.00
$3.50
$133
4
$38.00
$2.50
$95
5
$22.00
$3.00
$66
6
$33.00
$3.50
$116
7
$38.00
$3.50
$133
8
$38.00
$3.00
$114
9
$24.00
$2.50
$60
10
$25.00
$3.00
$75
46
47
Two Different Schools of Thought
• 40,000 feet above in the airplane
– Few assumptions that are generalized
• Outcome is more sensitive to assumptions
• 80/20 rule
• On the land with the farmers
–
–
–
–
Decentralization of the decision process.
Dissagregation of the inputs
The system is made up of more parts
Less prone to overall misdirection.
48
Case Study One
Continental Divide
Pacific
Atlantic
Inflows = f(Snow, rain, humidity, temp)
©Innovastat Corporation All Rights Reserved
49
Upper Colorado
River Basin
Rio Grande
River Basin
Missouri River Basin
Arkansas River Basin
©Innovastat Corporation All Rights Reserved
50
Upper Green
North Platte
White-Yampa Rivers
South Platte
Republican
Colorado Headwaters
Smokey Hill
Middle Arkansas
Gunnison
Upper Arkansas
Upper Colorado-Dolores
Rio Grande Headwaters
Lower San Juan
Upper San Juan
Upper Cimarron
Upper Rio Grande
©Innovastat Corporation All Rights Reserved
51
Case Study:
Surface
Creek
U
%
Surface Ck
U
%
U
%
U
%
U
%
U
%
U
%
U
%
U
%
See Appendix
©Innovastat Corporation All Rights Reserved
52
U
%
U
%
Locating, Mapping and Conducting Rehabilitation Assessment
2/10/05 to 8/24/07 (30 months)
Published and Distributed Reports to 14 Drainage Districts
Case Study Two
Las Animas
Lamar
Rocky Ford
La Junta
Lump all districts together or analyze
them separately?
Granada
Process
The initial steps required would be to educate
funding agencies and government officials in the
field of risk and decision analysis through formal
training in the classroom and computer lab. They
need to be shown the usefulness so that there is a
chance that changes in the established ways of
decision-making can be elevated to a higher plateau.
Once they grasp the concept, they will find it
indispensible in their work. People may not believe in
an idea or process if they do
not understand it.
©Innovastat All Rights Reserved
54
Benefit/Cost
• The optimal outcome would be for these risk and
decision technologies to be accepted by the
development community and policy makers.
Although they are willing to promote technologies
such as better seeds and irrigation systems they
have, to date, been reluctant to adopt the many
tools of decision making. With adoption, the
money earmarked for development projects would
have a greater impact per dollar spent and a
benefit/cost analysis can be undertaken to
determine that effectiveness.
©Innovastat All Rights Reserved
55
Evaluation
The optimal outcome for farmers is a substantial
increase in their productivity, financial success,
and incomes and a significant decrease in the
volatility of these incomes from year to year. The
farmer results can be measured by receiving
feedback from the farmers and by evaluating preInnovaAg and post -InnovaAg incomes,
agricultural production, and farm improvements,
etc. Also, two similar groups, one with the
InnovaAg implementation and the other one
without can be statistically compared.
©Innovastat All Rights Reserved
56
Normality
Important points to
remember
• The distribution is an equation and this equation is trying to explain the
behavior of crop yields. Perhaps no equation can do that.
• The distribution relies on parameters for its input. A normal distribution
requires a mean and a standard deviation. The mean may not be an accurate
representation of the central point of the data. The mean is usually derived
from a sample and perhaps, more observations are required in order to
adequately describe the data and to limit the margin of error.
• When we summarize data into parameters like a mean and then further try to
force the data into an equation, accuracy is lost and the distribution may not
explain the real data adequately and truthfully.
57
Histogram of Yields of Dryland
Wheat 1980-2010
Important points to
remember
©Innovastat All Rights Reserved
58
Important points to
remember
©Innovastat All Rights Reserved
59
Important points to
remember
The “average”
value of 45,010
acre feet doesn’t
occur very often.
The data appears non-normal (unlike a bell shaped curve) and bimodal (there are two humps to the curve)
60
Water Supply Forecasting Using Nonparametric
Assumptions For the Holbrook Canal
61
Correlations
•
•
A diversified economy is needed to
reduce risk – just like a farm – need
inputs that are not positively correlated.
The consequences of not
inputting the appropriate
correlation coefficients in
your model can lead you
astray.
Modern Portfolio Theory
– Models with less than
perfect correlation reduce
risk.
– Models with much
negative correlation
reduces your exposure to
risk
Important points to
remember
Var(A+B) = Var(A) + Var(B) +2Cov(A,B)
62
Optimization Under Uncertainty
epiphany
3 a (1) : a usually sudden manifestation
or perception of the essential nature or
meaning of something (2) : an intuitive
grasp of reality through something (as
an event) usually simple and striking (3)
: an illuminating discovery
Important points to
remember
Principles of Parsimony and Simplicity
Occam’s Razor
“Many branches of pure and applied mathematics are in great need of computing
instruments to break the present stalemate created by the failure of the purely analytical
approach to nonlinear problems” (John von Neumann)
63
Benefits of Stochastic Optimization or
Optimization Under Uncertainty
The ability to solve problems w/o calculus.
Important points to
remember
The ability to solve problem that traditional methods cannot accomplish. We
assume the normality of functions when in fact a better choice of words would be
“we pretend.”
Calculus was invented by
Leibnitz and Newton in
the late 1600s
Calculus requires a
continuous and
differentiable functions
64
Monte Carlo Simulation
For Dryland Wheat
•
•
•
•
•
•
The model is the dryland wheat enterprise budget
We change two of the variables – yield and price
Add the correlations.
Please refer to Simulation
Add Decision Trees
Workbook
Add Optimization
Add Forecasting
65
66
67
Portfolios, Portfolios, Portfolios
Portfolios are commonly viewed as a group of individual
financial investments. Nevertheless, the definition of portfolio
can be extended to include a collection of various other
investments or courses of action that you may undertake.
Fortunately, with Monte Carl simulation and optimizer
software, “simple” models can generate sophisticated and
intuitive results.
You can use concepts of modern portfolio theory to solve your
specific models outside the realm of traditional models.
68
Portfolios
•
•
You can gain an appreciation of optimization under
uncertainty in the context of portfolios.
We can explore the features of @RISK Optimizer for
various portfolio optimizations. .
• Portfolio of stocks
• Portfolio of projects
• Portfolio of an oil field
• Portfolio of pharmaceuticals
• Portfolio of farms
69
Synergy
• The interaction of two
or more agents or
forces so that their
combined effect is
greater than the sum of
their individual effects.
"Synergy means behavior of whole systems unpredicted
by the behavior of their parts."
- R. [Richard] Buckminster Fuller (1895 - 1983)
70
Optimization Under Uncertainty
• Maximize food security.
• Maximize the certainty of a certain return. (Some
farmers are risk averse and some may be risk
loving) (Can be non-parametric)
• Minimize the risk of a certain return.
You constraint
• Maximize the return for a given risk.
would contain a
• Other goals:
budget constraint.
– Minimize carbon footprint
– Minimize energy use
©Innovastat All Rights Reserved
71
Using Analytical
Tools for
Agricultural
Development
©Innovastat All Rights Reserved
Thanks for
attending!
72