Economic impacts of wind farms on Scottish tourism: report

2 Outline methodology

2.1 The estimation of expenditure change

2.1.1 Introduction

In chap 3 estimates of the total level of tourist expenditure in our chosen regions are given. Economic Impact occurs when the level of economic activity, normally in the form of a change of expenditure, changes. This section is concerned with the critical identification of the percentage of the expenditure that will be lost or gained as a result of tourists being negatively (or positively) affected by wind farm activity.

An Economic Impact Analysis framework involves an estimate of the economy before and after a specific event. Normally the "after" is immediately following the innovation but, particularly where activity is expected to grow, the "after" period could be any specific time in the future. The framework produces two time related problems. First, in the case of wind farm development, there is no single point but a continuing series of innovations. In addition there is no certainty about which developments will obtain consent and when they will commence. For the purposes of this exercise we have assumed:

• That all project with current applications will proceed;
• no other projects will occur;
• they will all be complete at an analysis point that has no specific time attached.

The second problem arises because whilst tourists can stipulate a likelihood of return that is fairly accurate, they do not know when that will occur and indeed are likely to underestimate the time. If the likelihood of return drops by say 20% as a result of wind farm development and that likelihood covers a five year period, then it will take five years before the total drop has occurred. Again to minimise problems of re-order distributions and biased time estimates the economic impact analysis is conducted at an unspecified point in time when all developments and all outcomes have worked through the system.

2.1.2 The Theoretical Framework

In this research we assume two models of behaviour relating to two distinct situations. Firstly we model the tourist travelling in Scotland enjoying the attractions and scenery. An unknown percentage of these will observe one or more wind farms and as a result, for these, there may well be a change in the likelihood of returning to the area. In effect there will be a shift of the demand curve. It is worth noting that there is some evidence in the literature of positive impacts of attractions at a very localised level, probably as a result of their rarity (e.g. mountain biking, visitor centres, walking). The most obvious developments are information centres that offer an inexpensive wet weather destination to the holiday tourist. In addition large wind farms offer an extensive car free road network in the hills often with extensive views over the area. The Land Reform Act suggests such areas should be available to walkers and cyclists and could well be a tourist asset if properly promoted.

This report has not explicitly attempted to identify the potentially positive impacts of wind farms as a tourist attraction at the size of local area levels used in the case study areas; in part because the substitution effects are so substantial - if the tourist did not go to the wind farm they would go somewhere else instead. However this analysis if applied to any tourist attraction be it a bird sanctuary, a castle or a theme park, would suggest minimal economic impact. But the number, range and quality of attractions available in an area do have an impact and in complementing that package a wind farm centre might have an effect significantly greater than implied by a conventional impact analysis. Such an analysis would be of considerable interest.

However, we feel that our methodology goes some way to capturing any residual positive impacts that may exist after these displacement effects, as any tourist that feels that a wind far m m ight act as a tourist attraction could indicate an increased likelihood of return to the area under our questionnaire design.

The second model relates to accommodation directly exposed to wind farm developments. There are two extreme positions we can identify. In the first we assume that the supply of beds is fixed and the price falls due to a decrease in demand. This is likely to be the short term position. As discussed in section 2.1.6, this leads to situation where the drop in price is equal to the drop in the mean willingness to pay.

The second model relates to accommodation directly exposed to wind farm developments. There are two extreme positions we can identify. In the first we assume that the supply of beds is fixed and the price falls due to a decrease in demand. This is likely to be the short term position. As discussed in section 2.1.6, this leads to situation where the drop in price is equal to the drop in the mean willingness to pay.

In the alternative scenario we assume that the hotelier charges at a level that covers costs and normal profits and that these do not change with the view. Consequently if the value of the room falls we would expect in time the number of rooms available in the affected area to fall with price maintained. The expenditure change will be the result of change in sales and the accommodation model relates this change in sales to the estimated change in willingness to pay.

Particularly over the longer term, the concept of two discrete models, one for the travelling tourist and one for accommodation is far too simple. Any change in demand is likely to have an effect on prices charged and the average expenditure of tourists will inevitably include some of the affected accommodation expenditure. Our estimates therefore have to be seen as indicative with a range which has a minimum given by travelling tourists only and a maximum defined by the sum of the accommodation and travelling effects.

It is acknowledged that the impact on some of those most affected such as long distance walkers, are not included in this analysis. Because the numbers and average expenditure of these groups are low we are confident that any negative economic impact will be extremely small. However, we do believe that this area is worthy of further study.

2.1.3 Forecasting the Numbers Exposed to Wind Farms

Wind farm developments only affect a proportion of tourists and an even smaller proportion of the accommodation. It would seem obvious that a key question relates to the proportion of tourists exposed and yet we were unable to find a single study that attempted to make such an estimate. In part we suspect this relates to the absence of appropriate skill sets in typical tourism and economic consultancies and the limitations of available data.

In appendix A we discuss in detail the use of the industry standard Arc- GIS software to identify the Zone of Visual Impact ( ZVI) collectively for the wind farms in each of the study areas, the length of road in each of the ZVIs and the number of bed spaces within these areas. Appendix B discusses the data sources available for estimating the number of tourists on the specified roads and the classification of the whole of the tourist body into three classes; Unexposed, Medium Exposure and High Exposure. These procedures require a number of quite contentious assumptions and consequently we conduct, as with the expenditure effects, sensitivity analyses and a range of estimates.

The "order of magnitude" estimates that emerge from this process are, in our view, robust and extremely enlightening. As a result we believe that similar analyses should become a part of the planning process to provide objective measures of the local and tourist population affected and the impact on the tourist infrastructure.

2.1.4 Forecasting the Behaviour of Tourists Exposed to Wind Farms

Methods for forecasting behaviour are normally classified as either quantitative or qualitative. Although quantitative approaches are preferred (Scott Armstrong, 2003) they are dependent upon the existence of adequate relevant data for analysis. In this case any model would need to take into account factors such as exchange rate fluctuations, economic growth, demographic changes and even airport security congestion in order to identify any wind farm effect. In addition the detail of the data would need to match the detail of the impact. As an example we would need time series data for at least ten years on the specific areas of the Highlands affected by wind farms rather than for the Highland and Island Tourist Board area as a whole. The only quantitative study attempted was the Cornwall Tourist Board (2000) study and predictably no significant impact could be found. Any effect, if it existed, was effectively swamped by the other factors of demand.

The two appropriate qualitative methods are broadly Intention Surveys and Expert Opinion. Both have been used, sometimes together (e.g. System3, 2003). Scott Armstrong (2001) continually emphasises that qualitative approaches are subject to bias and that structure is fundamental to success. In his seminal 2 1985 work, he identifies Expert Opinion as possibly the most inaccurate (Scott Armstrong 1985). This relates, in part, to the surprising finding of research by Griggs(1958), Levy and Ulman(1967) et al that experts forecast no better than trainees and were more susceptible to bias and anchoring ⁶. It is clear that surveys of the opinions of those involved in tourism are not likely to be as accurate as surveys of the intentions of tourists themselves. If the approach is to be considered then the construction of a Delphi group, covering all relevant disciplines, is likely to generate far more accurate forecasts. The Steering Group associated with this project would be a good example of such a group.

Morwitz (2001,2006) and Scott Armstrong et al (2000) examined the forecast performance of intentions surveys and the requisite conditions needed for accuracy. These were summarised in Scott Armstrong (1985) thus:

• Event Important
• Respondent has Plan
• Respondent Reports Correctly
• Respondent can fulfil plan
• New information unlikely to change plan

The most important type of trip from both the tourist view and in terms of expenditure is the summer vacation. This is important, is planned and is in control of the respondent. The information set is inevitably dependent upon the forecast horizon. As the horizon recedes into the distance unknown but significant events, such as births, deaths and marriages that affect plans are more and more likely.

The way the respondents report their intentions is important. Morwitz(2001) found that likelihood was more accurate that yes/no type responses. She also found that there was a consistent under-estimation of the time before the repeat event e.g. if the respondent was asked the likelihood of purchasing the good or service in the next five years then this corresponded most closely to the likelihood of purchase in the next seven. As discussed earlier this problem has been side stepped by locating the time point for the economic analysis at some unspecified time in the future when effects have worked through the system.

Given that the conditions for accurate assessment are largely met this still leaves the question of how accurate. Assessment of accuracy is difficult because of problems such as time delays and dealing with likelihoods. Armstrong et al (2000) conducted a meta study comparing published intentions type forecasts with trend extrapolation and with a combination of both. Unlike Lee et al (1997) they found that intentions data significantly improved trend forecasts and if there was a choice intentions data might be preferable. For the telephone service they found the mean absolute error to be around 3%. This seems very acceptable. However we are primarily interested in change which may well be of the same order of magnitude ⁷. Again we provide potential ranges of responses.

2.1.5 The Relative Effect

For each tourist subgroup j the intercept survey provides an estimate of the before and after likelihoods of return (r and s) under different levels of exposure k, r _jk and s _jk. We assume that tourists who have not previously been to Scotland, continue at the same steady rate. The percentage of the tourists in an area with high, medium and no exposure p _k are also known from the survey. Chap 3 gives the expenditure by each sub group x _j. Consequently we calculate the change in expenditure by SS(r _jk - s _jk)*p _k *x _j . Table 2-1 and Table 2-2 illustrate the process

Table 2-1 Likelihood of Return Example

		% Likelihood of Return			Group Spend £m
		High	Medium	None
Holiday	Before	80	80	80	£650
	After	60	70	80
Long Day	Before	90	90	90	£350
	After	80	90	90
% in Category		5	25	70

In the table above the likelihood of return for the two types of tourists, holiday makers and those out for a long day trip are identified when they had high exposure, medium exposure and no exposure. As we would expect the no exposure likelihoods are always the same. The total spend for each group in the area is also given.

To obtain the second table we multiply the difference between the likelihoods in each category by the percentage of the group in that category and the expenditure of the group. For example holiday makers who had high exposure had a 20% fall in likelihood and high exposure occurred for 5% of the group. Thus we would anticpate a 20% * 5%*£650m =£6.5m fall in the tourist expenditure for holiday makers staying overnight who had high exposure to wind farms.

Table 2-2 Assessment of Expenditure Example

	High	Medium	None	Total £m
Holiday	6.5	16.25	0	£22.75
Long Day	1.75	0	0	£1.75
Total	8.25	16.25	0	£24.50

This example leads to a total 2.45% reduction in expenditure. A critical factor in this example is the large number of tourists that are simply not affected by wind farms.

2.1.6 The Change of Expenditure in the Accommodation Sector

It is clear that individuals value the scenery and the introduction of "industrial" infrastructure, be it wind turbines or other large metal structures such as electricity pylons or masts, reduces that value. There has been a long tradition of assessing the change of value by examining the change in willingness to pay. Figure 2-1

Figure 2-1

Assuming the demand for a room is linear and in part dependent upon the scenery and is given by the Demand Curve D1. At a given price P1 the consumer surplus is given by the triangle D1,B, P1. = ½ ßQ ² where ß is the slope of the linear demand curve. From a sample of consumers the mean WTP extra would be M1-P1= ½ ßQ ₁ ie. ß = 2*(M ₁-P ₁)/Q ₁.

The short term is represented by supply inelasticity (Q1) and a fall in price from P ₁ to P ₂ as hoteliers publicise special offers in a bid to fill the bed spaces. Given the constant supply the consumer surplus (represented by D2,C,P2) will be constant (=½ ßQ ₁²) and the mean willingness to pay extra (M ₂ - P ₂) also constant. Thus

The proportionate change in expenditure = (M ₁-M ₂) /P ₁ .

In a similar way, in the longer term, supply contracts towards Q ₂ and price moves back to P ₁. In effect we would expect marginal suppliers, whose have dropped prices in an attempt to fill beds, to drop out of the market as requirements for investment in refurbishment become apparent. Q ₂ = 2*(M ₂-P ₁)/ ß and thus we obtain Q ₁-Q ₂ = 2*(M ₁-M ₂)/ ß. and the proportionate change in expenditure is given by (P ₁*(Q ₁-Q ₂))/(P ₁*Q ₁). Given ß = 2* (M ₁-P ₁)/Q ₁ we obtain

The proportionate change in expenditure = (M ₁-M ₂)/(M ₁-P ₁)

The before and after mean WTP is given by the Internet Survey and consequently we can assess the before and after (and percentage change) in accommodation expenditure in the affected rooms. Taking this percentage change, the percentage of rooms affected and the accommodation expenditure in the area we obtain an estimate of the expenditure change ⁸.

2.2 Economic impact analysis

The full effect on regional income and employment of each (gross or net) pound of the change in tourist expenditure depends, among other things, on what the tourist purchases and the strength of the direct effect, the indirect effects and the induced effects. These effects are briefly explained below.

The Direct Effect is simply the increase in local income and employment arising from the initial tourist expenditure. Through a combination of taxation and the purchase of supplies from outside, a proportion of this initial expenditure will be immediately lost to the area, and effectively can be ignored. However, a proportion of expenditure will remain within the area. It is this proportion which creates the direct effect. For example, the direct employment effect of tourist expenditure on, say, accommodation is simply the proportion of employment in hotels that is dependent on that expenditure. The direct income effect of accommodation expenditure is the wages and profits paid by hotels to local households.

It should be noted that some categories of expenditure have a minimal direct impact. For example, only about 5% of spending on petrol has a direct effect locally; 95% 'bounces off' through tax, duty and the purchasing of inputs from outside. If the only expenditure incurred from a day trip to a hill or forest area is the petrol at the local garage then the direct effect will be minimal. In contrast, accommodation expenditure has a strong direct effect. The composition of tourist expenditure is thus important in determining the magnitude of the direct effect on local incomes and employment.

There are Indirect Effects arising from the Direct Effect. For example a hotel may purchase butcher supplies locally. This supports the wages of the local butcher's staff, the butcher's own income from self employment and perhaps the rent charged by the shop owner. It also contributes to employment in the butcher's shop. These effects are known as the first round indirect effects. There are further indirect rounds to be considered. The butcher may purchase some of his supplies from a local abattoir, thereby supporting the wages of abattoir staff and the abattoir's profits. It also contributes to employment in the abattoir. There will be further rounds of, albeit successively smaller, indirect effects. For example the abattoir may purchase livestock from local farmers, who in turn may purchase building services from local companies. The combined impact of the direct and all the rounds of indirect effects are modelled by what is termed "Type I" multiplier analysis. Among other things, this analysis would calculate the total Type I household income in the area (measured by Gross Value Added (G.V.A.)) and employment (measured by Full Time Equivalents ( FTEs)) dependent on tourism..

As described, both the direct effect and every round of indirect effects increases household incomes in the area in the form of wages, profits, rents and income from self employment. Thus, the income of a diverse range of households will be increased as a result of tourist spending (e.g. hotel workers, hotel owners, butcher's staff, the butcher, butcher's landlord, the abattoir staff, owners of the abattoir, farm workers, the farmer, building workers etc….). In each spending round a proportion of these incomes are spent on locally produced goods and services, creating further local income and employment. This is the Induced Effect. "Type II" multiplier analysis incorporates these induced effects into the analysis, enabling the estimation of the corresponding Type II total income Effect (Type II GVA) and Type II total employment (Type II FTEs). In this report we only record the outcome of the Type II analysis.

The strength of the direct, indirect and induced effects depend on such things as inter-firm linkages within the regional economy, taxation policy, and the proportion of local income normally spent within the region. These parameters themselves will be dependent on the size of the region. Specifically, the smaller the area the less likely local business and retailers will purchase locally produced supplies (weak indirect effects). Also, the smaller the area, the less likely local households will purchase locally produced goods (weak induced effects).

In modelling the regional economy, this study is using the Detailed Regional Economic Accounting Model ( DREAM®) developed by CogentSI. This model is described in chapter 7.