Economic Impact of the 2001 Foot and Mouth Disease Outbreak in Scotland
2. INTRODUCTION AND BACKGROUND
In 2001 the UK experienced the worst outbreak of Foot and Mouth Disease (FMD) in recorded history. Scotland did not escape the disease. The first case in Scotland was recorded on March 1 st 2001 and occurred in sheep belonging to a livestock dealer in Lockerbie. The epidemic then spread rapidly but was principally contained in the Dumfries and Galloway and Borders regions. Some 187 farms were confirmed as being infected with FMD, 1048 farms were affected by the 3km sheep and pig cull and in 28 farms animals were slaughtered on suspicion. In all, 735,000 animals were slaughtered in Scotland, with the greatest impact falling on the sheep population where 643,900 were culled. However, the disease had indirect consequences that were felt over a much wider area, affecting rural businesses and tourism across Scotland, with the result that significant differences in urban-rural effects appear likely.
There have been a number of studies on the impact of the FMD outbreak on the UK economy as a whole. The Department of Environment Food and Rural Affairs and Department of Culture, Media and Sport (2002) report examines the economic impact of FMD on the agriculture and food sectors and on other sectors profoundly affected, such as tourism. At the UK level it was reported that the overall impact of the FMD outbreak was less than 0.2 per cent of GDP but that the impact on severely infected rural areas was much more pronounced. The report also acknowledges that the impact effects have been variable, producing both winners and losers.
A paper by the Christel DeHaan Institute uses a Computable General Equilibrium Model (CGE) for the UK economy to assess the economy wide effects of FMD and the policy implications arising from these, particularly on the tourism sector. The authors find that FMD had considerable effects on farming in various regions of the UK due to the nature of intersectoral linkages in the economy but much larger effects on tourism (Blake et al, 2002).
There is also a substantial amount of Scottish-specific work on the impact of FMD. This report itself is part of a series of studies carried out by the Impact Assessment Group (IAG) into FMD in Scotland 2. The IAG draws together sectoral expertise and also includes representatives from geographic areas affected by FMD in Scotland. IAG-commissioned tracking surveys were conducted to assess the impact of non-agricultural business (George Street Research, 2001). Dumfries and Galloway were covered in a second survey that focussed on non-agricultural business as agricultural businesses were covered by other sources. Primarily these surveys found that the vast majority of business has seen no change either in turnover or in staff due to FMD, although there is obvious regional variation with more affect in the infected areas.
DTZ Pieda (2002a, 2002b) was commissioned by the Scottish Executive Environment and Rural Affairs Department (SEERAD) to undertake a study to assess the 'Economic Impact of Foot and Mouth Disease in Scotland'. Their primary aim was to synthesise all available data on the impact of FMD on communities and the economy of Scotland. Highlighting areas of insufficient data and using case studies, the report assessed the impact of FMD on 'fragile rural economies' and on particular sectors within Scotland.
The sector studies element of this two-part report evaluates the impact of FMD in Scotland on Tourism, Transport and Agri-Food. Tourism was found to be the most affected of the three, with the Agri-Food and Transport sectors suffering more from localised effects in the infected areas of Dumfries and Galloway and the Borders rather than significantly at the Scottish level.
At a more aggregate level, the FMD outbreak in Scotland has also been studied in a paper for The Royal Society of Edinburgh and the Scottish Economic Policy Network. In this paper, McDonald and Roberts (2002) use a Computable General Equilibrium (CGE) approach. This model separately distinguishes 36 commodities (including 11 farm commodities and 4 categories of tourist). It distinguishes between shocks affecting Scottish agriculture, Scottish tourism and the effects of government policies to control and compensate for the outbreak. The focus was on distributional effects. The impact of FMD was found to vary across different farm types with cattle and sheep farmers experiencing a positive overall effect, while mixed farms were much more negatively affected. A second key result is that the limited overall negative effect on households was felt most by the wealthiest 20%.
Against this background, the aim of the study is to estimate the net impact of FMD on the Scottish economy. The specific objectives of the project are:
To estimate and quantify the economic impact :
On different sectors, regions and urban/rural areas.
Of control measures.
Fulfilment of this brief requires that the research takes account inter alia of the following issues:
The direct impact on agriculture of the export ban, animal movement restrictions, livestock cull and compensation payments.
The direct impact on tourism due to the fall in visitors from overseas and the rest of the UK and the reductions in domestic tourism and rural leisure trips.
The net economic costs and impacts of carcass disposal and control measures such as compensation and welfare payments, marketing and advertising.
The indirect economic effects - including substitution and displacement - elsewhere in the economy through product market, labour market, inter-sectoral and inter-regional linkages of the direct price and output changes in agriculture and tourism.
The direct and indirect spatial impacts, particularly between urban and rural areas.
The overall net economic impact in the Scottish economy on GDP, household income, prices, employment, exports and imports and trade balance.
The social and welfare effects, including income distributional impacts.
2.1 Computable General Equilibrium modelling
The report tackles these issues through the use of Computable General Equilibrium (CGE) modelling. CGE models have a number of strengths for analysing and quantifying this type of external (or exogenous) shock to the regional economy (Greenaway, et al. 1993; Partridge and Rickman, 1998).
CGE models use sectorally-disaggregated data. This means that it is possible to look at the effects of shocks to the regional economy whose direct impacts occur in one, or a small number of, industrial sectors. Clearly in the case of FMD, agriculture and tourist industries bore the brunt of the direct effects. The total impacts can also be reported disaggregated by sector. This sectoral disaggregation means that the CGE has an advantage over regional econometric models, which tend to be very aggregative.
CGE models incorporate supply side changes. Such changes include restrictions to the total supply of particular commodities and changes in the efficiency of production in individual sectors. These are examples of effects that were produced by the FMD outbreak. This incorporation of supply-side effects means that the CGE model has an advantage over sectorally-disaggregated demand-driven models, such as Input-Output analysis.
CGE models deal with the interaction between industrial sectors and economic agents in a theory-consistent manner. That is to say, generally the quantities and prices of inputs used and outputs produced are assumed to be determined by the operation of market forces. However, imperfections to the market mechanism and government intervention can also be modelled.
Whilst sectoral detail is retained, CGE models also produce results for aggregate, macroeconomic variables, such as regional GDP, employment, unemployment and competitiveness.
CGE models are ideal for scenario analysis which identify the impact of unusual events. For these types of external shock, it is not possible to rely on previous econometric evidence.
2.2 The AMOS CGE model
The specific model that is used is AMOS, a CGE model developed in the Fraser of Allander Institute, parameterised on Scottish data. A more detailed description is given in Annex B. In the analysis of the impact of FMD, we use the period-by-period version of the model. Important characteristics of the model are:
Scottish wage rates are set through a regional bargaining function. This means that if the demand for labour rises, the real wage and total employment will rise.
Labour is mobile between industrial sectors.
In any one year, the capital stock is fixed, both in total and to particular industrial sectors. However, between years, the capital stock for individual sectors is adjusted following profitability criteria. Importantly, we assume the region is not constrained by its own savings but can borrow for capital expansion from the UK and international capital markets.
In any one year population is fixed but adjusts from year to year in line with increases in the real wage and reductions in the unemployment rate.
These model characteristics imply that there is a degree of inflexibility in the short-run that is relaxed over the long run through investment and migration.
The model is not a forecasting one and with no changes in exogenous variables simply reproduces the base-year data. This means that all the simulations here are essentially variations from the counterfactual. The base-year data are for 1999, the last year for which there is a Scottish Input-Output table that provides key information on the inter-industry output flows (Scottish Executive 2002b). In this report, the nominal variables are therefore all given at 1999 prices. Results from previous work using the AMOS model have been published in high-ranking economic and regional science journals, so that the model has been subject to extensive peer review (Harrigan et al, 1991, 1996; McGregor et al, 1996). AMOS is particularly appropriate for analysis and evaluation of government policy. It has been employed to analyse the impact on Scotland of Regional Selective Assistance, Foreign Direct Investment and the policies of Scottish Enterprise (Gillespie et al, 2001a, 2001b, 2002).
In this particular application, we have made two important adjustments to the model. First, we have changed the industrial structure of the model in order to identify the effects in the "Agriculture" and "Hotels and Catering" sectors more accurately. Second, we have incorporated a mechanism to allocate changes in activity to individual regions 3.
2.3 Adjustments to the database
2.3.1 Disaggregation of the "Agriculture" sector
In the Scottish I-O table, all farming activities are aggregated under one industry - "Agriculture". This does not provide a suitable framework to evaluate impacts of exogenous shocks such as the FMD outbreak. McDonald and Roberts (2002) note that agriculture was affected directly by FMD but the disease did not have a blanket effect on all farm types, and that some farm types suffered more than others. This study follows McDonald's and Roberts' (2002) suggestion and keeps as much detail as possible regarding the structure of agriculture. Accordingly, the I-O sector "Agriculture" was disaggregated by farm type, separately identifying the different input and output patterns of each farm type. There were seven farm types:
LFA: specialist sheep
LFA: specialist beef
LFA: cattle and sheep
The task of disaggregating "Agriculture" in the Scottish I-O table is accomplished by reconciling secondary databases from different sources. These are Farm Incomes in Scotland (Scottish Executive, 2001); the Farm Accounts Survey (FAS) database and the agricultural census (Scottish Executive, 2002a); the 1999 Scottish I-O tables (Scottish Executive, 2002b), and McDonald and Roberts (2002). Details of the data organisation and manipulation are outlined in seven separate stages and reported in Annex C.
2.3.2 Disaggregation of the "Hotels and Catering" sector
The "Hotels and Catering" sector is a major recipient of tourist expenditure. Following Jones and Roberts (2002), in the present study this sector has been disaggregated by type of accommodation. The six categories are:
The aggregate data for this disaggregation were taken from VisitScotland (2001) and the StarUK website. The pattern of inputs into the different disaggregated sectors was determined using information for Wales from Jones and Roberts (2002) and Scottish data from volume 2 of the Surrey Research Group's I-O study for Scottish tourism (Surrey Research, 1993). The information for the individual sub-sectors was made consistent with the aggregate information from the Input-Output table using the RAS technique. This is explained in greater detail in Annex C.
2.3.3 Disaggregating the elements of tourism final demand
In order to identify the impact of changes in different elements of tourism and day-trip demand, we needed to separately identify these elements of demand within the framework of the Scottish 1999 Input-Output Table. In this table, Scottish Tourism in Scotland and Day trips are included as part of Household final demand, while Rest of the UK (RUK) and Rest of the World (ROW) Tourism are represented as a combined separate entry. This required the disaggregation of RUK and ROW tourist expenditure and removal of Scottish Tourism in Scotland and Day trips from Household consumption data.
Table 2.1: Structure of Scottish tourism, 1999
1999 Size of Final Demand (m)
Source(s) of Control Total
Source(s) of Disaggregation
Rest of the World Tourism
Scottish Executive (2002b), VisitScotland (2001)
Surrey Research Group (1993), VisitScotland (2001)
Rest of the UK Tourism
Scottish Executive (2002b), VisitScotland (2001)
Surrey Research Group (1993), VisitScotland (2001)
Scottish Tourism in Scotland
Scottish Executive (2000),
Office for National Statistics (2000b),
Labour Force Survey (2001)
Scottish Executive (2000),
Office for National Statistics (2000b),
Labour Force Survey (2001)
Countryside Agency (1998)
Countryside Agency (1998), Surrey Research Group (1993)
For the purposes of this analysis we wanted to further separate these columns by location of expenditure. Again we required geographic disaggregation of expenditure into urban, uninfected rural and infected rural areas.
Table 2.1 identifies the broad structure of tourist expenditure in Scotland. It also shows the sources used to identify the aggregate expenditure totals and the breakdown of expenditure on different sectors exhibited by different forms of tourism. Note that by far the biggest expenditure comes from Daytrips, followed by RUK tourist expenditure, Scottish Tourist expenditure in Scotland and finally ROW tourist expenditure.
2.4 Regional distribution of impacts
In addition to quantifying the impact of the 2001 Foot and Mouth outbreak on the Scottish macroeconomy, a key aim of this project is to assess the extent to which the outbreak impacted on the different regions of Scotland.
As in most impact studies, the regional incidence of the directeffects of the outbreak are relatively well known. In particular, a number of impact studies, commissioned by SEERAD and others, have already detailed the extent to which the direct effects of the FMD outbreak were location-specific. In particular, the DTZ PIEDA study points out that the spatial distribution of agricultural effects were, in part, related to the three geographical policy designations adopted by SEERAD and similar to those used in this report. The same study indicated the extent to which different regions experienced a downturn in tourist numbers and day visits and also the spatial displacement of tourists within Scotland over the period. However, the spatial incidence of the overall impact of the outbreak has yet to be quantified. We begin by briefly considering alternative methods for estimating the spatial distribution of the total impact.
2.4.1 First-round impacts
Sectors both upstream and downstream from agriculture were affected by the change in farm-behaviour associated with the outbreak. In relation to these first-round impacts, a number of previous studies have considered the extent to which farm transactions are:
with other businesses located in the same area as the farm, and
with businesses located in rural areas.
One of the first of these studies was by Harrison (1993), and used the actual receipts and invoices of a number of farms participating in the Farm Business Survey to track the location of farm-related industries. Her findings indicated that farm type and size were both significant influences on the distance over which transactions occur, with the smallest farms and pigs and poultry farms having a higher degree of local transactions. Overall she found that approximately one quarter of the inputs of farms under analysis came from the most rural areas, whilst one sixth of outputs accrued to the most rural areas of the UK.
Crabtree et al (1999) adopted a similar technique to contrast the spatial pattern of agriculture and conservation related expenditures of farmers involved in the ESA scheme. The findings suggested differences in the source of inputs by input type and by location of farms. The study also estimated the location of farm household spend.
Published evidence of the spatial pattern of spending by tourism providers appears more limited. A study by Essex and Treloar (1997) investigated the local purchasing patterns of accommodation suppliers in Newquay, Cornwall. It suggested that hoteliers had very little information or knowledge about the origin of their purchases. It also highlighted the fact that very few of the suppliers servicing the tourist sectors are exclusively dependent on the tourist market or locality. Relative to the agriculture sector, however, a higher proportion of total expenditure of tourist-related enterprises will accrue to employees in the form of wages and salaries. Since employees are likely to reside locally, one might anticipate the degree of local integration of tourist sectors would be high, even if input sourcing patterns are not locally-orientated.
2.4.2 Apportioning the wider economic effects across regions: the gravity model approach
One possible way of apportioning the total impact of the outbreak to various regions of Scotland would be to use a gravity model (Richardson, 1978). This is based on the principle that the likelihood of economic transactions between two places (and therefore also the spread of the impact of economic impact of the outbreak) is a function of both their proximity and their difference in economic size. The closer together the two communities are, and the greater the size difference in economic terms, the greater the flow of goods and services.
This approach was used by Doyle et al. (1997) to assess the spatial distribution of effects associated with a reduction in agricultural support. This particular study is interesting, not only because of the use of this method, but also because the empirical application focussed on Dumfries and Galloway - the area in Scotland most adversely affected directly by the FMD outbreak. Doyle et al. used the gravity approach to apportion the total estimated impact of the reduction in agricultural support, where the total was generated through Input-Output analysis. However, there is no reason why the same method could not be used to apportion the impacts from a CGE model after having allowed for the (known) spatial distribution of direct effects and, where possible, having estimated the spatial distribution of first-round effects.
2.4.3 The industry-modelling approach to estimating regional impacts
Another, very different, method for estimating the regional impact of the outbreak is that used in the MONASH-RES model of the Australian economy (Parmenter and Welsh, 2000). The MONASH-RES model translates the output from a CGE model of the Australian economy into either regional forecasts or regional policy impacts through a set of regional equations. Within the regional equation system, industries are classified as either national (producing commodities which are readily traded between regions) or local (producing goods which are not traded between regions). Outputs from the national regions are assumed to be independent of regional demand but the output of the local industries have to adapt to satisfy the regions demands.
The MONASH-RES approach is a top down approach that requires relatively little regional data. However, it does allow some capacity for changing some of the default assumptions according to known information about the direct impacts of a shock and/or including accurate information about the regional distribution of output in national industries in particular. Moreover, historical simulations suggest the performance of the system is satisfactory.
2.4.4 Local Keynesian multiplier methods
The methods described above are essentially top-down in that they are based on the premise that the total impact estimated by the CGE model needs to be apportioned to regions. An alternative, very different approach would be to concentrate on a small area, most directly affected by the outbreak, for example, the Castle Douglas and Dalbeattie area in Dumfries and Galloway, and use a local Keynesian multiplier analysis to estimate the total income and employment effects in these regions arising from the outbreak. This modelling exercise would be quite distinct from the CGE modelling, however it may be the only appropriate method for measuring the total impact of the outbreak at this level of geographical specificity. Recent estimates of parameters required to estimate Keynesian multipliers (e.g. marginal propensity to import) could be derived from recent studies of local rural economies in Scotland (e.g. Hill et al, 2002).
2.4.5 Adopted Approach
The procedure that we have adopted is a variant of the MONASH-RES approach. Using this method we generate total employment and GDP change figures for individual regions for each element of the FMD impact. Broadly this involves a top down approach. The direct impacts are allocated to the region receiving the shock. The changes in activity in non-local industries are then distributed across the regions in proportion to the region's initial employment shares in the relevant industries. Finally the national impacts for local industries are allocated across regions in proportion to the changes in direct and non-local activity.
The criteria for giving an industry local or non-local status is the share of its output going to RUK and ROW exports. Where this share is greater than 25% we count the sector as a non-local sector. Local sectors are therefore:
Energy Distribution and Construction
Wholesale and Retail
The disaggregated Hotels and Catering sectors,
Business and Communications,
Public Administration and Education
Health and Sanitary
All other sectors are non-local. Of course, the direct sectors vary, depending on the nature of the shock.