29
Whole farm modelling Farmer decision-behaviour 14 th November 2008, U. Reading Daniel Sandars Research Officer Natural Resources Management Centre

Whole farm modelling farmer decision behaviour

Embed Size (px)

DESCRIPTION

 

Citation preview

Page 1: Whole farm modelling farmer decision behaviour

Whole farm modelling Farmer decision-behaviour14th November 2008, U. ReadingDaniel SandarsResearch OfficerNatural Resources Management Centre

Page 2: Whole farm modelling farmer decision behaviour

Introduction

• 1) Background to the Silsoe Whole Farm Model and the policy challenge

• 2) Extension from linear profit maximisation to non linear utility maximisation

• 3) Progress towards implementing the RELU-Birds preference models.

• 4) Reflections on the scientific challenges ahead

Page 3: Whole farm modelling farmer decision behaviour

Farm LPs

• Whole farm planning LPs have two subtly different roles; Prescriptive uses guide an individual farmer to better decisions whereas predictive uses help understand how farmers response to choice or change. For the policy maker we are still doing prescriptive OR!!

• Profit maximisation has been effective for predicting the aggregate response of farmers to change.

• …even though there might be evidence that this does not describe how individuals behave!

Page 4: Whole farm modelling farmer decision behaviour

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

0.1 1 10 100 1000 10000 100000 1000000 10000000

Arable area, ha

Per

cen

tag

e ab

s re

lati

ve e

rro

r

Page 5: Whole farm modelling farmer decision behaviour

Soils and Weather

Workable hours

Profitability (or loss)

Crop and livestock outputs

Environmental Impacts

Possible crops, yields, maturity

dates, sowing dates

Silsoe Whole Farm ModelLinear programme, important features timeliness

penalties, rotational penalties, workability per task, uncertainty

Machines and

people

Constraints and

penalties

Page 6: Whole farm modelling farmer decision behaviour

Heavy clay, 800 mm annual rainfall

0

50

100

150

200

250

Ho

urs

Sandy loam, 500 mm annual rainfall

-

50

100

150

200

250

Ho

urs

Workable hours v. tractor hours

Period, fortnights Period, fortnights

Page 7: Whole farm modelling farmer decision behaviour
Page 8: Whole farm modelling farmer decision behaviour
Page 9: Whole farm modelling farmer decision behaviour

Introduction

• 1) Background to the Silsoe Whole Farm Model and the policy challenge

• 2) Extension from linear profit maximisation to non linear utility maximisation

• 3) Progress towards implementing the RELU-Birds preference models.

• 4) Reflections on the scientific challenges ahead

Page 10: Whole farm modelling farmer decision behaviour

The standard LP model

• xij are what could be produced, such as different crops,

with profit cj and resource consumption aij per unit

• bi are resource constraints, such as land area

njx

mibxa

ts

xcZ

j

n

jijij

n

jjj

,...,2,1,0

,...,2,1,

.

max

1

1

Page 11: Whole farm modelling farmer decision behaviour

Voluntary conservation behaviour

• How would free conservation education influence farmer behaviour?

• What types of policy intervention do farmers find unacceptable?

• Biodiversity arises from hotspots rather than the average?

Page 12: Whole farm modelling farmer decision behaviour

Multi-criteria methods

Discrete choice problems Continuous choice problems

Methods Multi-criteria Decision Making, Analytic Hierarchy Process, Outranking methods, etc

Goal programming, Compromise programming, Multiple Objective programming

Features Elicits a rich picture of attributes. Formal problem structuring methods. Interactive with a few motivated decision makers

Simple view of attributes. Few examples of formal problem structuring methods. Examples of non-interactive uses

Role Mostly prescriptive solutions, but have seen AHP claim to predict the outcome of the US presidential election

Most examples prescriptive

Page 13: Whole farm modelling farmer decision behaviour

Utility Theory

• Jeremy Bentham (15 February 1748–6 June 1832)

• Auto-Icon University College London

Page 14: Whole farm modelling farmer decision behaviour

What objectives/ Goals?

• Ask farmers? Few examples of robust repeatable methodology!

• From the farm planning literature? Many examples of using attributes that other people used!

• From the psychological literature?

• We used a mixture of both

Page 15: Whole farm modelling farmer decision behaviour

Multiple-object LP

• zk are component objectives, such as profit, risk, biodiversity

• wk are a set of weights used to form a single composite objective

qkw

njx

mibxa

ts

xcz

zwZ

k

j

n

jijij

n

jjjqq

q

kkk

,...,2,1,0

,...,2,1,0

,...,2,1,

.

max

1

1

1

Page 16: Whole farm modelling farmer decision behaviour

rl

qkw

njx

mibxa

ts

qkxcz

qkzwZ

kl

j

n

jijij

n

jjjklkl

q

kkllkl

...2,1

,...,2,1,0

,...,2,1,0

,...,2,1,

.

...2,1,

...2,1,max

1

1

1

Suppose we have a set of r decision makers, one of which is our normative ideal, each with view on how an action will change an objective (biodiversity) cjkl and the extent to which they are prepared to trade that objective off against profit wkl

Decision makers!

Page 17: Whole farm modelling farmer decision behaviour

Non linear preferences

Value function

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 2 4 6 8 10 12

No birds/ha

Val

ue

of

bir

ds

to a

dec

iso

n m

aker

Page 18: Whole farm modelling farmer decision behaviour

Decision makers 2

rl

qkw

njx

mibxa

ts

qkxcz

qkzvwZ

k

j

n

jijij

n

jjjklkl

q

kklkllkl

...2,1

,...,2,1,0

,...,2,1,0

,...,2,1,

.

...2,1,

...2,1,)(max

1

1

1

Where vkl is a piecewise linear value function coefficient

Page 19: Whole farm modelling farmer decision behaviour

Separable programming

0

20

40

60

80

100

120

0 2 4 6 8 10 12z ki

vki

A(1,1)

B(4,16)

C(7, 49)

D (10,100)

1,,,0

333

513315

4321

4321

4321

ki

ki

Z

V

If any δi is >0 then all preceding =1and all following =0

Page 20: Whole farm modelling farmer decision behaviour

Introduction

• 1) Background to the Silsoe Whole Farm Model and the policy challenge

• 2) Extension from linear profit maximisation to non linear utility maximisation

• 3) Progress towards implementing the RELU-Birds preference models.

• 4) Reflections on the scientific challenges ahead

Page 21: Whole farm modelling farmer decision behaviour

Entering weights

• Separable programming - lambda form (piece wise linear approximation) - Additive utility

Page 22: Whole farm modelling farmer decision behaviour

Program output screen

Page 23: Whole farm modelling farmer decision behaviour

Weight distributionattributes (metrics)

0

0.05

0.1

0.15

0.2

0.25

0.3

0 5 10 15 20

Attribute

No

rma

lise

d w

eig

ht

centroidobserved

Page 24: Whole farm modelling farmer decision behaviour

Introduction

• 1) Background to the Silsoe Whole Farm Model and the policy challenge

• 2) Extension from linear profit maximisation to non linear utility maximisation

• 3) Progress towards implementing the RELU-Birds preference models.

• 4) Reflections on the scientific challenges ahead

Page 25: Whole farm modelling farmer decision behaviour

Ruth GassonFarmers Goals

• Instrumental• Growth, Income, working conditions, security

• Expressive• Pride, self respect, creativity, achievement,

aptitude• Social

• Prestige, belonging, tradition, family, community• Intrinsic

• Physical effort, sense of purpose, independence, control, the outdoors

Page 26: Whole farm modelling farmer decision behaviour

Issues

• Most measures are appalling ambiguous proxies for the concept contained in the goal that they are representing.

• Redundancy amongst attributes.• The swing weight method does not force sacrifice

and thus over states the importance of non-primary goals.

• -indirect estimation methods do we have the data?• -orthogonal elicitation methods – do we have the

resources and the patience of farmers?

Page 27: Whole farm modelling farmer decision behaviour

Survey resultstrade offs

• Extreme

• -£25,279 to see another bird species

• -£2 mean profit to reduce profit deviation by £1

• £55,000 to give up a day off

• £661,826 to give up a days rough shooting

• £771,000 to fill out another set of forms?

Page 28: Whole farm modelling farmer decision behaviour

Conclusions

• We can optimise a richer utility based predictive model of farmer behaviour, but can we specify, model, parameterise, and validate it.

• Hard…there are many open questions

• It is worth doing scientifically and simply being able to offer better or different insights than the alternatives available to policy makers is reward enough.

Page 29: Whole farm modelling farmer decision behaviour

Other events

• The OR Society Special Interest Group on Agriculture and Natural Resources (chair Prof. Tahir Rehman (U. Reading) and secretary Daniel Sandars (Cranfield))

• Relaunch 2nd April 2009 @ Reading University

• The EURO working group on OR in Agriculture and Forestry Management (Co-ordinator Dr Lluis Plà)

• 5th Meeting EURO XXIII July 5th-8th (Bonn)

• EURO Summer School July 25th-August 8th (Lleida)