« Previous | Contents | Next »
Listen
Economic Impact of the 2001 Foot and Mouth Disease Outbreak in Scotland
6. LIMITATIONS OF THE MODELLING APPROACH
In this report we have attempted to identify and to quantify the Scottish-wide impacts of the FMD outbreak using a specially adapted variant of the Scottish Computable General Equilibrium (CGE) model, AMOS. Whilst there are clear strengths in this approach, there are limitations too, and these need to be identified and taken into account when interpreting the results.
6.1 All economic models simplify reality
The most general point is that economic models, of necessity, simplify reality. They focus on what are taken to be key causal relationships driving particular economic interaction whilst ignoring those variables which are thought to have a negligible effect. For example, it is often necessary to assume all agents within a particular category (firms in individual sectors, households in individual regions) are identical. Whilst we know that this is not strictly true, the additional information is either not available or would generate model results that are too complex to handle effectively. In short, there is no benefit from a 1-to-1 map. Of course, the simplifying assumptions should be made with care. But in practice there is always a trade-off between greater detail (which is often interpreted as representing greater "reality") and model tractability and data availability.
6.2 Model parameterisation
The CGE model requires many individual relationships to be specified in a precise manner. This often means both identifying exactly how individual markets are assumed to operate and fixing the value of key parameter values, such as the various demand elasticities. The model also has a fully identified supply side so that individual production relationships have to be chosen.
Where possible we have used Scottish-specific data to parameterise this model. We have also taken great care in attempting to separately identify sectors of the economy that were thought to be particularly affected by the FMD outbreak. However, a full econometric parameterisation of the model is simply not possible. Moreover, data restrictions mean that it is often necessary to take parameter values that have been determined using UK regional data, or from similar studies in other countries. In some cases, values are simply best-guess estimates. In this particular study, key assumptions that underpin the model affect the product market, labour market and the investment decisions.
6.2.1 The product market
In the variant of the model used in these simulations, all commodity markets are taken to be perfectly competitive. This simply means that no individual firm (or consumer) can influence the market price and that all firms are taken to be attempting to maximise profits in each time period. This assumption is restrictive but not as restrictive as it might at first appear. For example, we do not ignore the presence of government subsidies or taxes on production or other effects of government intervention into markets. In fact many of the simulations we perform are attempting to replicate these interventions. Also we do not assume that there is a world market for any of the commodities identified in the model. Products from different countries are not taken to be perfect substitutes for one another.
6.2.2: The labour market
This is perhaps the most difficult market to successfully model. In the variant of AMOS used in this report we impose:
A unified Scottish labour market. That is to say, the wage rate in all sectors modes in line with one another, and there is no skill disaggregation.
The Scottish real wage is determined by bargaining, where the wage level is negatively related to the Scottish unemployment rate.
On the supply side, changes in aggregate employment are made up:
changes in the unemployment rate.
changes in the participation rate.
migration into and out of Scotland in response to tightness of the labour market.
There is perfect mobility of labour between industrial sectors in response to variation in labour demand.
These assumptions mean that where labour demand rises in one industrial sector, there is some crowding out of employment in other sectors. Essentially the real wage rises, leading to employment falling in some sectors not strongly directly or indirectly affected by the demand increase. This means that, in the short-run anyway, the net increase in aggregate employment is typically much lower than the employment increase in the sectors directly affected. This is a result that contrasts with the typical Input-Output or Keynesian multiplier approach (McGregor
et al, 1996).
However, the reverse occurs where the initial shock to the economy is a reduction in labour demand (which is typically the situation that we encounter in these simulations). The fall in employment leads to a weakening of the labour market, a fall in the real wage and a marginal improvement in the competitiveness of sectors not directly or indirectly strongly affected by the initial reduction in demand. In this case the net employment fall will, at least initially, have a much smaller absolute size than the direct reduction.
The shift of activity into other sectors as a result of labour market "crowding out" or "crowding in" can have implications for the effect of GDP. If labour is taken to move from lower to higher added value sectors as a result of the initial reduction in demand, GDP can potentially rise, even though employment falls.
6.2.3 Investment
The demand for a change in capacity in an individual sector is determined in the AMOS model as an adjustment to a perceived mismatch between the desired and actual capital stock. Changes in output and labour costs are the prime determinants of capital stock adjustment.
6.3 Short-run adjustment
The AMOS model assumes that the economy is initially in equilibrium and any exogenous change, say the export ban, sets in motion adjustments to the endogenous variables, such as the output and employment levels in individual industries. One might want to challenge the general assumptions concerning the adjustments in the capital and labour market. It is econometrically very difficult, given the limited available data, to be sure of the detailed determinants of regional employment and investment. However, in this case the employment and capital stock adjustments might be limited for another reason.
The FMD outbreak might have been thought, from the outset, to have been a short run phenomena. If this is so, because of hiring and firing costs, workers might have been retained, rather than made redundant, by firms directly hit by the FMD outbreak. Hiring and firing costs would include factors such as skill retention, worker goodwill, search costs etc. The existence of such costs would limit the initial reduction in employment and the subsequent redeployment of released workers in other sectors. Similarly, the potentially complex set of capital stock adjustments set in train as a result of FMD in Scotland might not have occurred if firms perceived the problem as being purely temporary.
There appears to be evidence that employment adjustment in the agricultural sector is limited and the change in the labour supplied in this sector is made through changes in the hours worked (or intensity of work). It is also a low wage sector, so that wage rate reductions within the sector itself might be limited by minimum wage legislation. (The average wage in the disaggregated agricultural sectors is given in Table D.3 in Annex D
11). If this is so, then some of the wage effects identified in the model are likely to be overstated.
6.4 The Scottish specific focus
In this report, we identify the impact of FMD in one region of the UK, Scotland. However, the disease also affected agriculture and tourism simultaneously in the Rest of the UK (RUK). There are clearly strong trading links between the two economies. Ideally the impacts on the Scottish and RUK economies should be modelled simultaneously. Although we have attempted to minimise the distortions in the results that arise from analysing Scotland in isolation, some inaccuracy is unavoidable.
6.5 Regional disaggregation
The AMOS model gives results for the Scottish economy, disaggregated to individual sectors. A fully consistent regional disaggregation within Scotland, would require information on commodity, migration and commuting flows between regions within Scotland and a modelling of the labour market at a much lower level of spatial aggregation than that of the Scottish economy as a whole. These data and analysis do not exist at present.
The procedure that has been adopted to identify the regional impact of the FMD outbreak essentially distributes the Scottish impacts across regions on the basis of the initial spatial impact of the direct effects and the local or national/international orientation of individual sectors. As such, this top-down method is very much a second-best procedure and a bottom up modelling approach would be very much preferred.
The procedure we have adopted distinguishes between local and national/internationally traded sectors. We assume that local industries are affected solely by local activity, either through consumption or intermediate demand. The activity in national/international sectors, on the other hand, is determined by national forces. An alternative would have been to adopt a gravity approach. In that case, the activity of all sectors in one region would be affected by the activity in all other regions. Those sectors that we have identified as local sectors would be less influenced by purely local affects. However, those sectors which we have identified as national/international sectors would be more influenced by purely local effects.
6.6 The modelling of the direct shocks
All practical applications of a CGE approach to modelling policy intervention involves some ingenuity in attempting to use the exogenous variables available to match the action of policy instruments. In this case, for example, modelling the impact of the cull of trading animals and the subsequent compensation payments is tricky. Whilst the cull keeps the animals off the market, it does not stop their production. We have therefore modelled the impact of the cull as a supply restriction, but the compensation for trading animals as a demand shock. The compensation therefore covers the cost of production of the output that did not reach the market. Clearly, whilst this does not perfectly capture the effects of this policy combination, it attempts to get at the key forces at play.
Also where the system-wide impacts of changes in exogenous variables are being modelled, the overall results are only as good as the estimates for the exogenous changes. In this report, we have had to make a number of assumptions about the impact of FMD on exogenous demand, such as the change in tourist activity and the direct impact of the export ban that occurred as a result of the FMD outbreak and subsequent restrictions. The sensitivity of the results to changes in these assumptions has been given in Section 4 of the report. However, similar sensitivity of outcomes would follow changes in assumptions concerning other direct impacts associated with FMD.
Finally, for certain shocks, the desired regional and sectoral distributions of the initial impacts are simply not available. In these cases, we have to make the best estimates possible. For example, we take the regional distribution of RUK Tourism, Scottish Tourism in Scotland, and Daytrips to have the same initial geographical dispersion. In the absence of better information, this is the best that can be achieved but clearly some imprecision attaches to the results.
« Previous | Contents | Next »