EUROTOOLS and GENEDEC European land use modelling
These two closely related European Commission funded projects involved researchers from across
the European Union. Contracted as part of the UK research team at The University of Reading my
responsibility was to develop new modelling techniques for economic simulation of agrculture.
In the case of the EUROTOOLS project this involved the construction of a European wide model
including production and market constraints. For the GENEDEC project the aim was to assess the
impact of the Single Farm Payment in each country. To achieve this I applied a new technique
that I have developed to integrate statistical estimation within a mathematical programming
framework (see CV and testimonials page for reference to
published paper).
Mapping ethnic populations in Scotland
In association with Smiths Gore Research the aim of this project was to examine existing data sources for their suitability to predict population changes in Scotland with respect to existing and newly defined ethnic groups. Following a review of the various data sources a component cohort model was constructed to forecast future population changes under a variety of immigration scenarios. To improve the reliability of the forecasts we extended the modeling technique to calibrate the model to two separate baseline results (1991 and 2001 National Census’) rather than just one as is usually the case. In this way we were able to strengthen estimates of population flows.
Farmer behaviour towards single farm payments
Working with The University of Reading and ADAS the aim of this project was to better understand
the motivations, beliefs, intentions and behaviour towards the introduction of the Single Farm Payment
in England. A Theory of Planned
Behaviour study was conducted to elicit farmers attitudes towards a pre-defined set of possible
behaviours. Cluster analysis was performed on survey data of over 600 farmers to identify four distinct groups:
flexible strategist, dedicated producer, environmentalist, and survivor. Each of these groups exhibited
different patterns of behaviour towards the Single Farm Payment. Further to applying the Theory of
Planned Behaviour I constructed economic models to test if there were relationships between the results
of the study and business objectives such as profit, risk and labour use. The results showed that such
relationships exist although further research is required to quantify these complex interactions.
North West Waste Management Survey and Wales Waste Management Survey
Every two years surveys are conducted nationally and regionally to assess the volume and flow of waste from and to construction sites, landfills, quarries, registered exempt sites etc. The surveys for Wales in 2004/5 and the North West in 2005/6 were conducted by Smiths Gore Research. My role in these two projects was to advise on sampling and to process and analyse the data. These analyses required the use of Monte-Carlo simulation to provide confidence intervals around our estimates of waste volumes. This was necessary as the distributions of the waste involved a combination of binomial, normal and exponential normal and could not easily be solved analytically. The Monte-Carlo technique simulates the process being investigated by repeatedly generating random results based on the estimated parameters of the distributions involved. From these pseudo results reliable confidence intervals can be determined. Further to this analysis an economic model was constructed for the North West to examine future waste scenarios.
Economic viability of integrated poultry and forestry systems
This project was set up by The University of Oxford, Fair Farms and The University of Reading to assess the financial viability of integrating
poultry systems with forestry to improve the welfare of poultry. This long running project is primarily concerned
with gathering data and information on the developed system. My role was to construct a financial model to simulate the
long term financial implications of such a system under different assumptions. For further details of this work see CV and testimonial page for reference to
published paper).
Land Price Forecasts
The objective of this project was to construct models to forecast land prices and sales volumes in England, both nationally and regionally. This piece of research required a comprehensive review of all potential factors and data that could be used to predict prices. The approach taken combined Box-Jenkins Season Auto-Regressive Integrated Moving Average (SARIMA) models with principal component analysis. The SARIMA models take into account seasonality and general trends in the data as well as providing insight into how the market responds to price/sales changes. To supplement this information other factors such as house prices, retail index, farm profits, FTSE 100, earnings and base rates were incorporated. However, as these data exhibit a high degree of correlation with each other principal components were used to create a new set of factors that have the property of being uncorrelated. These new factors were then integrated into the SARIMA model to provide forecasts.
Likelihood of Possession modeling
Working with Smiths Gore Research our clients required a model to predict the likelihood of farm tenancies becoming vacant based on set of socio-economic data. To facilitate such predictions binary logistic regression models were constructed as much of the data collected was categorical. The data included such measures as education, rural interest, business structure, family history, technology etc. However, many data were missing from individual records and thus several alternative models were constructed. The best models correctly predicted farms becoming vacant or not in over 80% of cases. A bespoke program was then used to provide the best forecast for any given set of data.
Economic potential of integrating poplars into UK and European agriculture
This was another European Commission funded project involving researchers from across Europe.
My role was to develop optimisation models of forestry production to identify the best long-term
financially viable method of producing timber from poplar trees. This required optimising a
Markovian Decision Process formulated from the production constraints of a typical farm. Subsequent
to this I incorporated the findings of the optimisation process into economic models of typical
European farms to assess the likely adoption or potential for adoption of poplars.
Assessing the socio-economic impact of new breeding technologies into UK dairy and beef sectors
At the start of the 1990s there was considerable excitement about new reproductive technologies
such as sexing, embryo transfer, cloning, bovine somatotropin (BST), etc. and how they might impact upon farming and food production.
The UK government commissioned a project to examine the genetic consequences and the socio-economic
consequences to the dairy and beef sectors. The Roslin Institute in Edinburgh under the Scottish Agricultural College (the same team responsible
for "Dolly" the sheep) investigated the genetic impact of such technologies, whilst The University of
Reading were responsible for assessing the socio-economic impact. My role, with help from animal nutritionists
and economists, was to devise a suitable methodology to solve this problem. For the first stage I applied Markov chains
to simulate the phenotypic and physical changes to dairy and beef herds under the use of these technologies. As a
result of this work I became the first person to find an exact solution to the dairy replacement problem, a problem that had stood for
several decades. These Markov chain models became the bedrock of later farm models that integrated the dairy/beef herd into
the broader aspect of whole farm production. Finally I linked together all the models developed via supply and demand constraints
into a national model of agricultural production.