Stirling March 24 ’09 A combinatorial optimisation approach to non-market environmental benefit...

StirlingMarch 24 ’09

A combinatorial optimisation approach to non-market

environmental benefit aggregation via simulated populations

Stephen Hynes, Nick Hanley, Cathal O’Donoghue


Background

This research considers the use of spatial microsimulation in the aggregation of regional environmental benefit values.

Use matched survey and Census information to produce regional aggregate WTP figures

An application to corncrake conservation on Irish farmland


The Corncrake


The Datasets The Irish National Farm Survey

Collected as part of the Farm Accountancy Data Network of the European Union (FADN) Farmers asked WTP info in conjunction with the usual farm activity info.

The Census of Agricultureidentifies every operational farm in the country and collect data on agricultural activities undertaken on them

The 2006 NFS contains 1,177 nationally representative farms and the Census contains info on 145,000 farms.


WTP Question in NFS A payment card showing the bid amounts of €10,

€20, €30, €40, €50 and €60 and farmers were asked: “of these bid amounts which would be the maximum you would be willing to pay (€) each year into a conservation fund to aid in the restoration of this bird and bring the singing male population back up to a sustainable population of 900 birds”

Given the nature of the CV elicitation format we use an interval regression to model WTP


The combinatorial optimisation problem Want to find an optimum configuration (i) of farms satisfying:

where denotes the minimum error between the actual census tables of size, system and soil type and the simulated tables constructed using the configuration of NFS farms

But there is a maximum possible number of farm configurations in the matching process and also a computation time constraint

In order to solve this combinatorial optimisation problem we employ what is defined as an approximation algorithm which yields an approximate solution in an acceptable amount of computation time.

),(min)( iEiEERi

oopt

optE


Process for a Single ED Choose a configuration (i) of NFS farms to represent the Census

SAP tables for a single ED. Another configuration j can be obtained by randomly selecting a

number of records in configuration i and replaced them with ones chosen at random from the universe of NFS records.

The number of records to be replaced is defined as T. Letting the probability that configuration j will

be the next configuration of farms in a predefined sequence of configurations (the java program sets the number of iterations) is given by 1, if <0 and by if >0.

The acceptance of a new configuration is decided by drawing random numbers from a uniform distribution on [0, 1] and comparing these with . This process continues, with T being lowered at each step until the maximum number of iterations has been hit or the error falls within the desired setting.

),()( iEjEEij

ijE TEij /exp ijE

TEij /exp


Simulating Annealing ProcessOn completion of the matching process we

have a list of farm ids from NFS for each ED in the country

We then simply merge all the information associated with the farms in the NFS (including WTP for corncrake conservation) to characterise our ED population of farms

Now have farm activity information and WTP for each farm in each ED


WTP Aggregation

We calculate the aggregate environmental value of the corncrake conservation program in 3 alternative ways.

1. 2.

3.

n

iSIMiTPW

1

ˆNFSPnWT ˆ

n

iSIMigionalTPW

1_Re

ˆ


Results-CV Interval Regression

Variable NFS Model

Size of Farm (acres) -0.043 (-0.02)** Family Farm Income (€/1000) 0.075 (0.03)*** Age of Farm Operator 0.027 (0.05) Organic Nitrogen Production (kg/hectare) -0.036 (-0.02)** REPS farma 2.072 (1.26)* Total crops and pasture as a fraction of Total Farm Size -4.97 (-4.85) Constant 14.73(5.31)*** Log likelihood -274263

Likelihood Ratio 2 (6) test 18.55


WTP estimates for the 4 alternative estimation methods

Average WTP Total environmental value of a

Method of Analysis Per Farm (€) corncrake conservation program (€) Payment Card Interval Regression for NFS sample 10.65 (10.49, 10.82) 1,544,857 NFS Payment Card Interval Regression applied to simulated farm population 10.40 (10.39, 10.42) 1,508,592 Payment Card Interval Regression estimated using simulated farm population 9.58 (9.56, 9.59) 1,388,195


Total WTP estimates per County for the 4 alternative aggregation methods

County NFSPnWT ˆ

n

iSIMiTPW

1

ˆ

n

iSIMiREGIONALTPW

1_

ˆ

Extensive Counties Galway 141,293 145,010 151,238 Donegal 89,939 91,996 98814

Intensive Counties Tipperary S 40,289 37,170 33,616 Cork 162,572 149,369 135,506


Conclusions

Using the combinatorial optimisation one can take into account the spatial heterogeneity of the target population in the aggregation process.

The synthetic spatial micro-data can be combined with other GIS datasets for follow on analysis.

Usage in other stated and revealed preference valuation techniques


Thank You!

Hynes, S., Hanley, N. and O’Donoghue, C. (forthcoming). A combinatorial optimisation approach to non-market environmental benefit aggregation via simulated populations, Land Economics

Hynes, S. and Hanley, N. (2009). The ‘‘Crex crex’’ lament: Estimating landowners willingness to Pay for Corncrake Conservation on Irish Farmland, Biological Conservation 142: 180-188.

Stirling March 24 ’09 A combinatorial optimisation approach to non-market environmental benefit...

Documents

Transcript of Stirling March 24 ’09 A combinatorial optimisation approach to non-market environmental benefit...