<<

. 4
( 11)



>>

As a consequence they are unsuited to determining the absolute number of people
who will visit a site. Therefore, our visitor arrivals model has to be composed of
variables that have relevance across the population and can be readily transferred
between sites.
To date there has been relatively little research regarding the level and determi-
nants of visits to woodland in the UK. Furthermore, of those few studies which
have examined this issue, most have looked at national recreational demand (Willis
and Benson, 1989; Whiteman, 1991) rather than that at any particular forest site.
This chapter draws in part upon material presented in Bateman et al. (1999c) and we are grateful to the Regional
Studies Association for permission to use this material.
1 The GIS procedures employed here are presented in a non-technical descriptive manner. Further details of the
commands used are presented in Bateman (1996).


91
92 Applied Environmental Economics

One notable exception is provided by the work of Colenutt and Sidaway (1973)
who model the demand for day-trip visits to the Forest of Dean. Here a combined
on-site and household (postal) survey was used to collect information regarding trip
origins and the factors determining visits. Analysis of these data revealed that by
far the most important factor determining arrivals was travel time, to the effective
exclusion of other explanatory variables.
The Colenutt and Sidaway result is important because it suggests that an arrivals
function may be estimated relating travel time to the probability of a visit taking
place. The analytical power provided by a GIS makes it possible to apply such a
function to detailed population data, such as those provided in the UK Census, in
order to predict arrivals at any existing or hypothetical site.2 Obviously, in practice,
the validity of taking an arrivals function estimated at one site and applying it to
another needs to be carefully assessed in terms of the accuracy of the predictions
made. Such a test is carried out and presented subsequently.


Recreation demand: the Thetford Forest study
The objective of this study was to estimate an arrivals function for a given forest
which could then be applied across our Welsh study area. The base data for our
initial investigation were obtained as part of the Thetford 2 study described in
Chapter 3. Here individual journey distance and duration measures, adjusted for
the availability and quality of the road network, were calculated for use within the
ITC valuation study discussed previously. However, such individual-level variables
were inappropriate for use within our arrivals function where travel times were
required for all points across the study area rather than just those which were
the outset origins of surveyed visitors. We therefore needed to convert our travel
time road network data into complete coverage travel time zones which would have
relevance to visitors and non-visitors alike. To obtain this continuity of coverage
the vector (line) data derived for each individual segment of the road network had
to be rasterised.
Rasterisation is a process of converting vector features (here roads) to cells on a
regular grid, in this case of 500 m — 500 m squares,3 covering the extended East
Anglian area from which visitors originated. In this study the travel time values
assigned to points along roads were reassigned to the grid cells which contained
those points. A ˜majority ¬lter™ was run recursively across the entire study area
to smoothly ¬ll in the gaps between roads, providing values for all grid cells and
2 Kliskey (2000) also uses a GIS-based approach to generate models of recreation potential, reporting an empirical
analysis of recreational snowmobiling in British Columbia.
3 This produced a total of 161,195 cells for our entire study area, of which 58,364 were directly ¬lled through
the rasterisation process (i.e. they contained roads), the remainder being assigned values through the process
described in the text.
Recreation: predicting visits 93



Travel time (minutes)
<= 9.9
10“29.9
30“59.9
60“89.9
>= 90


Main roads


Study site



0 10 20 30 40 50 km


Figure 4.1. Travel time zones for the Thetford Forest study.

producing a continuous travel time surface centred upon the site and fanning out
to ¬ll the entire study area. The majority ¬lter worked by means of a ˜moving
window™ (usually eight by eight cells in extent),4 where the centremost empty cell
was assigned the value held by the majority of already assigned cells in the speci¬ed
window. This approach worked well for the vast majority of cells. However, a few
gaps remained in areas very remote from any roads where the ¬lter window did
not contain any cells ¬lled directly by the rasterisation process. These grid cells
were given the values of their nearest neighbours.5 For further discussion of these
procedures see Bateman et al. (1995b, 1999c) and Brainard et al. (1997).
Once all grid cells had been assigned a value they were grouped into convenient
categories. Inspection of the calculated travel times showed that the extended road
network encompassed values up to 120 minutes. Within this range, thirteen time
zones were de¬ned. Given the concentration of visit origins around the site, the
innermost zones (between 0 and 30 minutes) were tightly de¬ned at ¬ve-minute
intervals, after which ten- and eventually ¬fteen-minute bands were used (between
30 and 60 minutes and 60 and 120 minutes, respectively). Figure 4.1 illustrates

4 At the edge of the study area the window could feasibly reduce to as few as four cells (only ¬lled cells are
incorporated into the ¬lter). The possibility of an edge distortion does exist but, given the very large number of
cells used in the entire Thetford dataset, any such distortion would be extremely minor.
5 Bateman et al. (1995b) undertook an analysis of vector and raster road speeds for this study. This shows that,
within the Welsh study area considered subsequently, vector travel times were somewhat shorter than raster
equivalents. Following this analysis, raster values were multiplied by a value of roughly 1.2 to ensure parity
with the more accurate vector values.
94 Applied Environmental Economics

Table 4.1. Observed and predicted visitor rates

Time Actual Zonal Observed visit Predicted visit Predicted
zone1 (1) visits2 (2) pop™n3 (3) rate4 (4) rate5 (5) visits6 (6)
5 13 954 0.0136268 0.0103972 9.9
10 31 21,596 0.0014355 0.0027285 58.9
15 8 13,326 0.0006003 0.0012476 16.6
20 10 14,377 0.0006956 0.0007160 10.3
25 26 26,811 0.0009698 0.0004655 12.5
30 38 58,416 0.0006505 0.0003274 19.1
40 46 191,009 0.0002408 0.0001879 35.9
50 65 405,831 0.0001602 0.0001222 49.6
60 17 375,134 0.0000453 0.0000859 32.2
75 48 776,817 0.0000618 0.0000559 43.4
90 15 562,508 0.0000267 0.0000393 22.1
105 7 253,762 0.0000276 0.0000292 7.4
120 ” ” ” 0.0000225 ”
150 ” ” ” 0.0000147 ”
180 ” ” ” 0.0000103 ”
210 ” ” ” 0.0000077 ”
240 ” ” ” 0.0000059 ”
300 ” ” ” 0.0000038 ”
360 ” ” ” 0.0000027 ”
500 ” ” ” 0.0000014 ”

Notes: 1 Upper limit of travel time zone measured in minutes of vehicle travel.
2
Number of party visits recorded during survey (no repeat visits in sample).
3
Number of households within each travel time zone as recorded in the 1991 Census.
4
Column (2) divided by column (3).
5
Visit rate predicted from the best-¬tting arrival function (discussed subsequently).
6
Predicted visit rate multiplied by zonal population (number of visiting parties).


resultant travel time zones, although for clarity of reproduction these have been
amalgamated to ¬ve categories.
Once travel time zones were de¬ned the relevant zone for each survey respondent
was identi¬ed by matching the outset origin of each of the surveyed visitors to the
travel time surface. Results from this exercise are presented in the ¬rst two columns
of Table 4.1. Here column (1) shows the upper limit of each travel time zone (in
minutes of vehicle travel to the site) and column (2) records the number of party
visits to the site from each zone during the period of the survey6 (other columns
are discussed subsequently). Of the total sample of 351 parties, 324 (92.8 per cent)

6 The possibility of repeat visits during the survey period was recognised. This was tested for and proved not to
be a feature of the survey sample.
Recreation: predicting visits 95

originated from time zones encompassed by our GIS road network. This provided a
suf¬cient sample both to estimate an arrivals function and to extrapolate it beyond
the limits of our road network.
The desired arrivals function would predict visits as a function of travel time.
However, to achieve this it was necessary to account for varying population densi-
ties across our time zones (i.e. we needed to calculate a visit rate in terms of party
visits per capita). Accordingly a population grid surface was interpolated which co-
incided in geographic extent with the travel time surface. Totals for persons usually
resident in Enumeration Districts (EDs; the ¬nest level of detail available) were
extracted from 1991 Census7 data using the SASPAC software (London Research
Centre, 1992) and grid references for ED centroids were obtained from ¬les held
at Manchester Computing Centre.8 Further discussion of the population surface
concept is provided in Bracken and Martin (1995) and Martin (1996b).
Allocation of residential populations to the 500 m — 500 m grid cells composing
the travel time zones was achieved through a volume-preserving algorithm, using
a form of the SBUILD program described by Martin (1990). A mask image was
used to prevent allocations outside the study area and initial input to the software
consisted of 6,675 centroids with a population of 2,723,971. The surface produced
by SBUILD (after cell totals were rounded to the nearest integer) contained a total
population of 2,724,133 suggesting that the program produces accurate population
estimates, at least at the aggregate level. Detailed inspection indicates that the char-
acteristics of urban areas are well represented in the population surface and the
only criticism which might be made is that some areas classed as ˜unpopulated™
undoubtedly contain isolated properties. This type of de¬ciency is, however, virtu-
ally inevitable given reliance upon data for areal aggregates such as Enumeration
Districts and in the context of this research is not thought to represent a signi¬cant
problem.
Population totals for our de¬ned travel time zones were straightforward to calcu-
late within the Grid module of Arc/Info. By allocating each of the surveyed parties
to a travel time zone, summing to derive a total, and dividing by the resident pop-
ulation, a zonal visit rate was calculated. Results from this exercise are shown in
Table 4.1. Here column (3) records the zonal population derived as discussed above.
Column (4) divides visits from each zone, in column (2), by zonal population to
give our observed visit rate. This represents the dependent variable in our arrivals
function. The contents of columns (5) and (6) are described subsequently.

7 Crown Copyright, ESRC/JISC purchase.
8 A check on the accuracy of grid references was then conducted by calculating mean centres and standard
distances for the EDs within each ward. This process revealed a few gross errors in grid references, which were
corrected.
96 Applied Environmental Economics

Table 4.1 indicates the expected, strongly negative, relationship between travel
time and visit rate.9 An examination of this relationship revealed that a double-log
model provided a good ¬t to the data.10 Equation (4.1) summarises the resulting
arrivals function.

’ 1.46 ’ 1.93 ln TZ
ln VR (4.1)
(’2.41) (’11.39)

where:

VR = observed visit rate (number of party visits from zone i divided by
zonal population)
TZ = travel time zone (minutes)
R2 (adj.) = 92.1%. Figures in brackets are t-statistics.

Investigations into potential omitted variables and correlation of residuals failed
to reveal any signi¬cant problems with Equation (4.1). Given the strength of this
relationship we felt con¬dent in extrapolating our arrivals function to more distant
travel time zones. Columns (4) and (5) of Table 4.1 list observed and predicted
visitor rates, while columns (2) and (6) report actual and predicted visitor numbers.
The arrivals function predicted 317.8 party visits from the ¬rst twelve travel time
zones during the sampling period. This compares with an actual ¬gure of 324, an
error of less than 2 per cent.
Our arrivals function refers to those visitors interviewed during the sampling
period. One of the main reasons for conducting our survey at Thetford rather than
at a Welsh site was that it is one of the very few forests for which accurate daily
and weekly visitor records are available (weekly data being held for several years).
This information enabled us to allow for those visitors to Thetford who were not
interviewed during our sampling period and also to establish that a very stable
relationship exists between annual and survey period visits (Bateman, 1996, gives
full details of this analysis).11 This allowed us to extrapolate our sample-period
arrivals function to an annual basis. Comparison of predicted with actual annual
visits showed a discrepancy of just over 1 per cent.


9 Note that observations from the furthest time zone (120 minutes) were omitted from our analysis (full details
of which are given in Bateman, 1996) as this zone was not completely encompassed by the road network (see
Figure 4.1).
10 The small number of observations means that we should exercise some caution here. The double-log form
narrowly outperformed a semi-log (dependent) model, while other forms ¬tted the data poorly. This is similar
to the ¬ndings of Colenutt and Sidaway (1973) who report results for both forms although it is not made clear
which is superior.
11 This analysis reveals a consistent pattern of visits over the year, in which arrivals were well predicted by
seasonal factors, extreme weather events and national holidays.
Recreation: predicting visits 97


Applying the arrivals function: predicting arrivals in Wales
Before considering the detail of this analysis it is worthwhile to remind ourselves
of its place within the overall study. The objective of the research presented in this
section is to yield a map of predicted arrivals to actual or potential woodland sites
for a regular grid across our Welsh study area which can then be monetised using
the values derived in Chapter 3. This money value map will provide the ¬rst element
in the analysis of woodland bene¬ts. Maps of the timber and carbon storage value
of woodland, derived in Chapters 6 and 7, can then be readily added to this to yield
our estimate of the total bene¬ts of woodland. These results can then be compared
to the map of agricultural values derived in Chapter 8 to allow us to conduct a
spatial CBA of the net bene¬ts of conversion of land from agriculture to woodland
in Chapter 9.
Our ¬rst concern in the present analysis was to test the validity of our arrivals
function against the actual number of visits made at a sample of Welsh woodland
sites. A study area boundary was de¬ned and coincident road network and popu-
lation surfaces constructed in a manner similar to the Thetford analysis. In order
to allow for distant travellers to potential woodland sites along the Welsh border,
the study area was de¬ned so as to reach deep into England.12 Appropriate county
boundaries were obtained from the Bartholomew database. Road data were ex-
tracted, clipped and corrected as described in Chapter 3. B-roads and minor roads
outside Wales were deleted, except where their omission created signi¬cant gaps
in road topology. Roads that were just outside the de¬ned study area were also
included (notably the M6 motorway outside Coventry) if their absence seemed
likely to have a signi¬cant impact on calculations of population accessibility. The
resulting road network is illustrated in Figure 4.2.
Roads were again rasterised onto a 500 m — 500 m regular grid. The value as-
signed to each cell was the class of the road segment (as recorded in the Bartholomew
database) with the greatest cumulative length running through the grid square. As
a consequence, a long section of road that just clipped the edge of a cell took
precedence over a short segment of road that actually had the greatest length within
the grid square. This was a feature of the rasterising algorithm and could not be
readily circumvented. Urban boundaries were rasterised and overlaid onto the road
network to allow separation of urban from rural roads.
Population data and centroids for Enumeration Districts were again obtained
from Manchester Computing Centre. The study area encompassed 30,311 Enumer-
ation Districts with a total resident population of 13,821,562. Once centroid grid
12 The study area comprised the following counties and areas: Avon, Cheshire, Clwyd, Dyfed, Gloucester, Greater
Manchester, Gwent, Gwynedd, Hereford & Worcester, Merseyside, Mid Glamorgan, Powys, Shropshire, South
Glamorgan, Staffordshire, West Glamorgan, West Midlands and Anglesey & Holyhead. Minor islands off the
coast of Britain were removed.
98 Applied Environmental Economics




Motorway
A“road
B“road


0 25 50 75 100 125 km




Figure 4.2. Digital road network for Wales and the English Midlands. For cartographic
reasons English B-roads and all minor roads are omitted from the map.

references had been checked, the SBUILD program was again used to generate a
population surface at 500 m — 500 m grid cell resolution. The program again per-
formed well, yielding a total population estimate of 13,821,361 people. Figure 4.3
illustrates the resulting surface.
With the Welsh travel time zone algorithm and the relevant population surface
de¬ned, an actual versus predicted test of our arrivals function was possible. At
the time of this analysis the Forestry Commission only held visitor data for ¬ve
sites in Wales. Furthermore, in conversation with of¬cials it became apparent that
two of these were closed for unusually long periods during the year while a third
contained several special attractions not normally found at forest sites which raised
Recreation: predicting visits 99




Population per 500 m grid cell
< 49
50“99
100“249
250“499
500“999
>= 1,000
0 25 50 75 100 125 km
Welsh Border

Figure 4.3. Population density surface for Wales and the English Midlands. (Source of
population data: 1991 Census, Crown Copyright, ESRC/JISC purchase. The population
density values were calculated using the SBUILD software with 1991 Census Enumeration
District centroids.)

visitor numbers above those normally expected for such a location.13 This sample
size and associated complications meant that the desired standard of testing was
not feasible (a problem which was not adequately addressed until additional sites
subsequently became available and an extended test across more than thirty sites
was carried out as described subsequently in this chapter). However, it was decided
to undertake a simple comparison of predicted and actual arrivals at each of the ¬ve
Welsh sites available. For each of these sites, arrivals were predicted by (i) using the
13 These include a museum, catering facilities and a variety of organised recreational activities.
100 Applied Environmental Economics

rastering algorithm to de¬ne zones and travel times; (ii) interrogating the SBUILD
population surface to obtain an estimate of the population in each zone and; (iii)
applying this information through the arrivals function given in Equation (4.1)
in conjunction with the sample period/annual visitor conversion factor calculated
during the Thetford survey. Equation (4.2) simply relates actual to predicted visits
per annum.14
ACTUAL 0.903 PREDICTED (4.2)
(4.420)
where:
ACTUAL = actual arrivals at site (party visits per annum)
PREDICTED = predicted arrivals at site (party visits per annum)
R2 (adj.) = 83.0%. Figures in brackets are t-statistics.

Equation (4.2) indicates that, despite the limitations of this analysis, the arrivals
function performs as expected with the slope coef¬cient for PREDICTED not being
signi¬cantly different from 1. Given this result and the lack of data for further testing
we concluded that the arrivals function did provide at least a defensible predictor
of annual arrivals at a typical woodland site (i.e. one with similar basic facilities to
that found at Thetford).
We were therefore able to make a case for applying the arrivals function to a reg-
ular grid of points across the study area and so predict expected annual recreational
visits to actual and hypothetical woodland sites across Wales.15 An important prac-
tical issue, however, is the appropriate grid size for such an analysis. Even with the
use of a raster structure and other efforts to shorten processing, determination of
travel time zones for a representative grid covering the whole of Wales represented
a signi¬cant computational exercise. Using available computing facilities each site
took between ¬fteen and thirty minutes to process (depending on workload). As-
suming the former time, calculation of a 1 km grid surface for the entire area of
Wales (some 20,500 cells) would take over 200 days of continuous processing;
clearly a coarser sampling scheme was required.
The issue of grid size was investigated by de¬ning two transects across Wales.
The ¬rst of these ran due east from the coast near Aberystwyth to the English
border and was composed of thirteen sites, each separated by 2.5 km, and another
¬ve sites at 5 km spacing. The second transect ran from a similar origin due south to
a point just outside Swansea and was composed of sites all at 5 km intervals. Travel
14 Analysis con¬rmed that any constant was not signi¬cantly different from zero.
15 Such estimates do not take into account the substitution effects which would arise in any speci¬c area if a
number of woodlands were planted in that locality. The object of the current exercise is to identify those areas
where the establishment of a wood would be bene¬cial. The impact of supply-side changes is considered
subsequently.
Recreation: predicting visits 101




= 5 km grid square centroids

0 25 50 75 100 125 km


Figure 4.4. 5 km grid points used to generate the predicted woodland visitors surface.

time zones and zonal populations were de¬ned for all of these sites and predicted
visits estimated using the arrivals function. Inspection of these predictions showed
that both the 2.5 km and 5 km resolution sites were sensitive to changes in local
population density and the quality of surrounding road infrastructure (details in
Bateman et al., 1995b). The detail afforded by the 5 km grid system indicated that
such a resolution was adequate in re¬‚ecting the major contrasts in predicted visitor
numbers engendered by population density and road availability/quality. Clearly a
2.5 km grid would give greater information regarding rates of change. However,
given the very considerable processing demands of such a grid, and the acceptability
of results from the 5 km resolution sites, such an approach seemed unnecessary.
Accordingly travel time zones were calculated for a 5 km grid for the whole of
Wales. The base map of grid points used to generate subsequent visitor potential
surfaces is illustrated in Figure 4.4.
Regardless of the chosen resolution, certain sampling problems are dif¬cult to
alleviate. Inconsistencies arise from the interaction of the road network with the
sampling pattern. Cell values depend upon how far a sampling point falls from
any kind of road. Two areas equally far from population and with comparable road
102 Applied Environmental Economics

infrastructure might have different estimated travel times (and therefore predicted
visit numbers) if in one of the areas the sampling point falls right on a road and
in the other the sampling point is far from any road. There is no straightforward
way around this arbitrariness. However, the ¬ndings for the two transects (and
subsequently the entire area of Wales) were reassuringly sensible and predictable,
suggesting that these inconsistencies had not had any signi¬cant impact.
Travel times were calculated for each of the 5 km grid sites as follows. A window
was de¬ned around each site and the site rasterised. An allocation process, using
a cost impedance grid based on road characteristics (see Brainard et al., 1997),
was run to ¬nd the shortest path linking the site and each other cell in the raster
surface. The impedance necessary to reach each of these locations was assigned
to corresponding cells in an output grid. This provided, in minutes of travel, a
time-surface output which was then classi¬ed into time zones. Information on total
residents for each of these areas was subsequently extracted from the rasterised
population surface and recorded in a separate ¬le. This process was then iterated
across all sample sites in the 5 km grid.
Once time zones and zonal populations had been calculated for all grid points,
woodland recreation demand (in terms of total party visits per annum) was predicted
using the arrivals function. Figure 4.5 illustrates the resulting predicted woodland
visitors surface.
Figure 4.5 strongly re¬‚ects the in¬‚uence of population distribution upon the pre-
diction of recreational woodland visits. In southern Wales the in¬‚uence of cities
such as Swansea and Cardiff and the densely populated ˜valleys™ area results in rel-
atively high visitor predictions. Similarly, in the north-east, the in¬‚uence of nearby
English cities such as Manchester and Liverpool is very clear. Conversely, in mid
Wales and western coastal areas, the sparse population results in very low visitor
arrival estimates. Population impacts tend to be compounded by the distribution of
higher quality transport infrastructure. This in¬‚ates the already high arrival numbers
generated by the proximity of large centres of population. However, infrastructure
effects are perhaps best demonstrated in areas of relatively low population density
such as coastal, mid and north Wales. Figure 4.6 shows this area in detail, super-
imposing the relevant major road network. Here we can see that the presence of a
major road creates a heightened potential visitor corridor as it facilitates visits by
individuals from relatively distant travel time zones.


Mapping predicted recreation values
In Chapter 3 we derived various estimates for the unit value of a party visit to a recre-
ational woodland. In particular we emphasised a lower-bound value of £1.82 per
Recreation: predicting visits 103




< 35,000
35,000 to 59,999
60,000 to 99,999
100,000 to 149,999
>= 150,000



Motorway
Dual carriageway
Single carriageway



0 25 50 75 100 125 km



Figure 4.5. Woodland recreation demand in Wales: predicted annual total party visits
per site. (Source: Bateman et al., 1999c.)




C
< 35,000
35,000 to 59,999
BF
60,000 to 99,999
B

Dual carriageway
Single carriageway
D

C = Caernarfon
M
BF = Blaenau Ffestiniog
B = Bala
D = Dolgellau
M = Machynlleth
A
A = Aberystwyth

0 10 20 30 40 50 km


Figure 4.6. Woodland recreation demand in north-western Wales: predicted annual total
party visits per site. (Source: Bateman et al., 1999c.)
104 Applied Environmental Economics

a. CV meta-analysis b. ITC




< £60,000
£60,000 to 99,999
0 25 50 75 100 125 km
£100,000 to 199,999
£200,000 to 299,999
>= £300,000

Figure 4.7. Predicted value of total annual woodland recreation demand per site using two
valuation estimates: (a) lower-bound values based on cross-study analysis of CV values;
(b) upper-bound values based on ITC study.



party per visit derived from our cross-study analysis of CV results and an upper-
bound estimate of £3.59 per party per visit obtained from our ITC analysis. GIS
capabilities were used to apply these values to our estimates of the number of
annual party visits to a given (real or hypothetical) woodland to yield predictions
of the total annual recreational value of sites. Figure 4.7 illustrates the maps of
recreational value produced by this exercise.
The distribution of values within each of the maps shown in Figure 4.7 mirrors
that of the base demand map (Figure 4.5). However, the fact that our upper-bound
valuation is nearly twice that of our lower-bound estimate is well illustrated here.
The degree to which this variability constitutes a cause for concern is uncertain.
If we are con¬dent of these bounds then, in a cost-bene¬t context, if the lower-
bound value is suf¬cient to justify a switch from other land uses into woodland,
further precision may be unnecessary. Similarly, if even upper-bound values are not
Recreation: predicting visits 105

large enough to justify such conversion, then again these estimates are suf¬cient
for decision analysis. Only if the cost-bene¬t balance lies within these bounds is
further precision required. Given this, then, at least as an exercise in methodological
development, use of these estimates seems justi¬ed.


Extensions
The work described above details the extent of our research to date on the case
study area of Wales and is used as the basis of the cost-bene¬t analysis presented
in Chapter 9 of this volume. It also represents our only attempt to date to generate
maps of arrivals and recreation values for large areas, embracing both existing and
potential recreation sites. However, our recent and ongoing research concerning
other areas of Britain extends our methodology for modelling visits and values. In
this section we brie¬‚y review this work to provide the reader with a ¬‚avour of the
directions in which this research is developing.
In work described in Lovett et al. (1997), Bateman et al. (1998) and Brainard
et al. (1999) we examine how both the number of arrivals and the value of those
visits to woodland sites alters according to a range of attribute characteristics. These
include:

(i) travel costs, described by the accessibility of the site to the potential visiting population
(i.e. taking into account the spatial distribution of the whole of the British population
in relation to the study site)
(ii) the socio-economic and demographic characteristics of that potential visiting popu-
lation (allowing for the possibility that, say, richer households or those with more
children may visit such sites more often)
(iii) the availability of substitute woodlands described by an inverse weighted distance to
all other British woodlands from all possible visitor outset origins
(iv) site quality characteristics (for example, presence and size of a car park, length of
woodland walks, etc.).

These models represent a substantial extension to those described previously, both
because of the additional explanatory variable considered and because they permit
the estimation of site-speci¬c coef¬cients, yielding estimates of consumer surplus
for each individual site. This allows for the possibility that the value of the recre-
ational experience varies between sites.
The ¬rst stage of this analysis involved calculation of a variety of variables de-
scribing items (i) to (iv) above for Thetford Forest. These variables were obtained
from a variety of sources. Travel distances and times were calculated and pop-
ulation distribution obtained using the GIS as described previously. Data on the
Table 4.2. Of¬cial recreational visit numbers, predictions of arrivals and consumer surplus estimates for
twenty-seven English woodlands
Of¬cial estimate of visits Predicted visits Per party consumer Site consumer surplus
Site name (per annum) (per annum) surplus (£ per visit) (£ per annum)
Dunwich 18,980 15,957** 1.56 24,828
Two Mile Bottom 22,636 22,678** 2.72 61,676
Kielder Castle 24,243 56,747* 3.57 202,767
Forest Drive 31,641 26,200** 3.57 93,616
Warksburn 3,794 5,351* 7.42 39,706
Bogle Crag 14,924 47,475 5.38 255,408
Grizedale 85,181 81,015** 3.48 281,824
Noble Knott 7,543 35,407 3.51 124,149
Whinlatter 55,797 60,838** 3.36 204,571
Blackwater 39,338 37,518** 5.19 147,813
Bolderwood 22,963 28,503** 4.86 182,318
Moors Valley 165,552 157,561** 4.14 652,149
Bucknell 21,360 45,526 1.63 74,117
Salcey 77,650 75,644** 2.23 168,735
Wakerley 51,490 42,354** 2.06 87,456
Dalby 130,151 77,804* 3.31 257,260
Chopwell 42,298 54,251* 6.36 344,846
Hamsterley 76,796 71,770** 3.50 251,462
Simonside 12,430 32,526 2.94 95,462
Blidworth Bottom 54,547 41,844** 3.15 131,776
Blidworth Lane 52,754 45,103** 3.16 142,394
Blidworth Tower 37,596 45,288** 2.91 131,660
Chambers Farm 23,605 22,808** 1.92 43,836
Goyt, The Street 84,279 73,400** 2.63 193,058
Normans Hill 30,936 35,975** 2.66 95,748
Thieves Wood 72,276 45,617* 2.66 121,474
Sherwood Centre 38,919 42,325** 1.78 75,430

Notes: * = predictions within 50% of of¬cial estimates;
** = predictions within 25% of of¬cial estimates.
Recreation: predicting visits 107

socio-economic and demographic characteristics of the population were obtained
from the UK Census and this information was spatially assigned using the GIS. Dis-
tances from each possible outset origin on a regular grid across Britain to each poten-
tial woodland recreation substitute site16 were calculated and an inverse weighting
scheme applied (with weights being empirically derived by analysis of the outset
origin of visitors to Thetford in relation to substitute availability from those ori-
gins) to give prominence to those nearer to each potential outset origin. Finally, site
quality characteristics were obtained from the Forestry Commission.
These data were then used to estimate a model to predict visits (and values) for
Thetford Forest. This was then transferred to predict arrivals and recreation values
at twenty-seven English woodlands17 for which of¬cial estimates of visits were
available (although the Forestry Commission freely admitted that these estimates
were somewhat approximate). Results from the transfer exercise are detailed in
Table 4.2, contrasting of¬cial estimates with predictions derived from our transfer
function from which estimates of per party and per annum consumer surplus are
obtained and detailed.
Considering Table 4.2, our extended transferable model provides estimates which
are highly correlated with those of the Forestry Commission (p < 0.001).18 Given
the lack of a gold standard for determining the accuracy of either set of estimates,
this seems an adequate basis for future research and arguably provides an acceptable
planning tool. Certainly this was the opinion of the Forestry Commission which
recently asked the authors, together with their colleague Andy Jones (also at the
University of East Anglia), to apply this methodology to a larger dataset of nearly
11,000 interviews conducted at forty sites across Britain. When completed, this
analysis will be combined with a second, recently ¬nished study (again with Andy
Jones) commissioned by British Waterways, examining over 5,000 interviews con-
ducted at ¬fty-three inland waterway sites across Britain. These studies further ex-
tend the methodology set out above by incorporating wider sets of socio-economic,
site quality and substitute availability variables (for example, non-woodland sub-
stitutes such as waterways, beaches, built heritage and urban attractions are con-
sidered). At the time of writing, results from these studies were being prepared for
publication. However, in both cases similar messages were clearly given by the
data, which are of particular relevance to the work described in this volume. While

16 Potential substitute sites were taken from a variety of sources including satellite imagery, the Institute of
Terrestrial Ecology land use map and a joint Countryside Commission and Forestry Commission large-sample
household survey of outdoor recreation.
17 Brainard et al. (1999) consider a further six subsites, that is sites within a larger forest with multiple sites.
However, our transfer model was unreliable for such applications, i.e. it only predicts for visitors to a distinct
forest rather than for areas within a given forest.
18 A regression test relating of¬cial estimates to our transfer predictions showed a coef¬cient which was not
signi¬cantly different from 1 with a constant which was not signi¬cantly different from zero.
108 Applied Environmental Economics

issues such as the socio-economic characteristics of potential visiting populations,
the availability of substitutes and site characteristics are all signi¬cant predictors
of visits and values, all of these variables are dwarfed by the signi¬cance of travel
costs. It seems that the business world mantra of ˜location, location, location™ being
the vital determinant of demand applies equally well to the demand for open-access
recreational public goods. Indeed omission of all other explanatory variables yields
relatively small estimation errors. Given the strength of this result we feel that the
analysis presented in preceding sections remains a valid input to the CBA conducted
in Chapter 9 of this volume.


Limitations of the predicted recreation values
We now return to our analysis of the case study area of Wales. While we feel
that the recreation value maps illustrate the methodological potential of applying
GIS techniques in this context, it is important to conclude this chapter with a brief
discussion of a number of potential limitations and further issues which would have
to be addressed before the full decision-making potential of this approach could be
realised.


The supply side
Our analysis only considers the demand side of the woodland recreation ˜market™.
The recreation value maps indicate the recreation demand for a typical woodland
established at any of the 5 km grid intersections of the base map (Figure 4.4).
They do not tell us about the supply side of this market. There are two major ways
in which the supply side interacts with demand to determine actual visits. First,
the existing distribution of woodland will already have soaked up some of our
predicted demand. Second, as new forests are planted and (with some time lag)
recreational services become available, so demand becomes satis¬ed. If supply
exceeds demand in any one area such that non-congested recreation sites already
exist, then the demand for new sites will be lower than that predicted in Figure 4.7
which ignores the distribution of existing sites.
To a substantial degree these concerns are incorporated within the extension
work described earlier through the addition of substitute availability variables.
However, this work also shows that it is travel time and cost which remain by far
the strongest determinant of visits and values. Therefore, while there is clearly scope
for using this research to re¬ne our visit prediction maps, the same research suggests
that the results summarised in Figures 4.5 to 4.7 remain valid approximations of
underlying relationships and are acceptable as an element within our subsequent
CBA assessment.
Recreation: predicting visits 109


Applicability of the Thetford Forest period to annual conversion factor
As part of our arrivals function calculations we had to convert from the survey
period to an annual basis. One concern here is whether the conversion factor used
is valid for other sites or unique to Thetford Forest. In order to test this fully we
would ideally need data regarding the annual distribution of visits both at Thetford
and at any site to which we wish to extrapolate. Unfortunately, as described in
relation to Table 4.2, of¬cial estimates are still only rough approximations and
robust values are currently unavailable for our Welsh study area. Gillam (pers.
comm.)19 suggests that seasonality patterns are likely to be roughly similar across
England and Wales and only differ in very remote areas such as the north of Scotland
where seasonal peaks are likely to be relatively more pronounced. On the basis of
this information, and in the absence of any contrary evidence, we feel that we have
adopted a defensible approach to this issue.


Comparability of recreation in Thetford Forest with that in Wales
The major demographic and infrastructure differences which separate Wales from
our East Anglian survey site are explicitly accounted for in our arrivals function
which allows for both population density and road network quality. Two remaining
issues are pertinent here. First, does our survey site provide similar recreational
services to those of our visitor potential map? By de¬nition, the answer here is
yes, because we are looking at the creation of similar service sites where the major
recreational attraction is open-access walking and its associated activities. However,
in the absence of data concerning site quality and facilities this approach will not
be appropriate for predicting arrivals to non-standard real or hypothetical sites.
Second, does the psychological perception of woodland recreation differ between
East Anglia and Wales? In considering this we must separate it from the supply-
side problem commented upon above. Once such a distinction is made we see
no reason to suspect any inconsistency here (although it cannot be ruled out), an
assertion reinforced by the earlier work of Colenutt and Sidaway (1973) in the
Forest of Dean (on the Welsh border) which reports similar visitation patterns to
those observed in our own analyses.


Conclusions
The analysis presented in this chapter has used a variety of GIS techniques to
model visits to a speci¬c woodland and then apply the resultant arrivals function to
19 Simon Gillam (Chief Statistician, Forestry Commission) noted that the Thetford Forest estimates were believed
to be among the most reliable available and therefore provided a reasonable basis for this analysis.
110 Applied Environmental Economics

produce estimates of visitation to similar woodlands across our Welsh study area.
The estimates have then been converted into money values using the valuation
studies presented previously. Results conform well to prior expectations showing
predicted demand to be linked to population distribution and site accessibility.
A number of problems have been identi¬ed in the course of this analysis. Both
the per visit values and visit number estimates were not sensitive to certain site
characteristics. However, the extensions described above provide a methodology
for addressing these problems and the results of this recent work suggest that
the errors created by such omissions are acceptably small, travel costs being the
overriding determinant of visits and values.
Given this we can defend our analysis both on methodological and empirical
grounds. Furthermore, the adoption of a sensitivity analysis approach, using upper-
and lower-bound valuation assumptions to create an envelope of recreational values,
represents a substantial improvement over the common omission of such values
from land use planning. In subsequent chapters we augment these with further
forest values before making a comparison of aggregate values with those from
conventional agriculture in the Welsh study area.
5
Timber valuation




Introduction
In this chapter we assess both the social and private value of timber production. This
is the major market-priced output of woodland. Furthermore, while recent policy
statements from both the National Assembly for Wales (1999, 2001a; Forestry
Commission (FC), 2001a,b) and UK government departments (Department for
the Environment, Transport and the Regions (DETR), 2000) emphasise the need
to adopt a holistic approach to managing woodlands, explicitly recognising their
multipurpose nature, timber production remains, nevertheless, a key element of
such a strategy, playing an important role in rural economies (FC, 1998).1
The economic and policy imperative to include timber production within any
cost-bene¬t analysis of land use change involving forestry is therefore clear. How-
ever, the estimated value of this production depends crucially upon the real price
of timber. Because plantation returns are long delayed, any (even small) change
in real prices will have a major impact upon net present value (NPV) sums.2 In
order to assess this, the chapter opens with a brief history of commercial forestry in
the UK designed to acquaint the reader with the recent, major and trend-breaking
increase in domestic timber supply. In the subsequent section both the supply and
demand sides of the UK market are modelled so that a balanced view on future
prices can be derived. These conclusions are reinforced by time-series analyses of
price movements.

1 Note that, while this document explicitly refers to English woodlands, the recent Cabinet Of¬ce report to the
Prime Minister (Cabinet Of¬ce, 2000) makes it clear that this is the ¬rst of three strategy documents.
2 NPV is the sum of discounted net bene¬ts (bene¬ts minus costs) over the lifetime of a project (here a plantation).
For further discussion, see Price (1987b); also, see Reed and Haight (1996) who introduce stochastic elements.
In practice, felling and management decisions may be highly complex. This was recognised even in the classic
optimal rotation model proposed by Faustmann (1849) (see Chang, 1998; Deegen, 2000). However, this decision
becomes even more complex when forest-owners are motivated by objectives other than pro¬t maximisation
(see, for example, the discussion of owner™s amenity bene¬ts by Tahvonen, 1999; or of recreational hunting by
Akabua et al., 2000).


111
112 Applied Environmental Economics

Whilst timber value is clearly important, private planting decisions are often
determined by the availability of shorter-term grants rather than long-delayed felling
bene¬ts, so we devote a section to reviewing the various subsidy schemes available.
The next section brings together the preceding discussions regarding prices and
grants and information on plantation costs and tree growth to produce the base
rotation3 models upon which our timber valuations are calculated.
The long time horizons inherent in woodland investments bring us to the vexed
question of discounting. We discuss the principle of discounting and provide a
brief review of the literature regarding the ˜correct™ discount rate with respect to
both social cost-bene¬t analysis and private investment appraisal. We conclude that
as no single, clearly correct discount rate can be identi¬ed, a sensitivity analysis
approach is required.
The subsequent section provides investment appraisal results from the viewpoint
of a private individual (the farmer) and this is then extended to provide a limited
social cost-bene¬t analysis of the timber product of a plantation (i.e. ignoring
those externalities dealt with elsewhere in this research). In both cases, NPV and
annuity equivalent (de¬ned subsequently) results are reported, the former being the
usual fare of the forest economist while the latter are comparable with competing
agricultural outputs.
Assessment of all possible woodland tree species was not feasible given both
time constraints and a lack of data concerning less popular species. Furthermore,
preliminary analysis indicated that costs and bene¬ts of different conifers would
be reasonably similar,4 the same being (broadly) true of broadleaves. Therefore,
two ˜indicator™ species were selected for analysis: Sitka spruce (conifer), and beech
(broadleaf).

Historical background
Pre-1945
In terms of land use, British forestry has always been the poor cousin of agriculture.
Although the prehistoric ˜natural™ condition of the land was primarily as forest, the
in¬‚uence of man has consistently been to clear-fell and convert land to agricultural
use. Even by the time of the Domesday Book only 15 per cent of England remained
under trees.5 This deforestation trend continued for most of the last millennium
with particularly heavy losses occurring in the sixteenth and seventeenth centuries
when adoption of advanced husbandry techniques and subsequent enclosure of
common land allowed agriculture to con¬ne forestry to marginal areas and private
3 A rotation is the full lifespan of a plantation from planting to felling.
4 This is of course a relative statement. Differences do exist and are important at the micro level. However, the
magnitudes of costs and bene¬ts are similar enough for this to be a defensible assumption in this study.
5 Pers. comm. Colin Price, Department of Agricultural and Forest Sciences, University of Wales, Bangor.
Timber valuation 113

parklands, the latter often being operated on a non-commercial basis for private
amenity values (Rackham, 1976). By 1900 only 4 per cent of England, 5 per cent
of Wales6 and 2 per cent of Scotland and Ireland was under forestry, these being
by far the lowest levels in Europe (Rackham, 1976; National Assembly for Wales,
2001a).
At the start of the twentieth century the UK was almost completely dependent
upon imports for its timber supply. This strategic weakness was exposed by the
German naval blockade of Britain during the First World War. With timber a major
input to the UK™s vital coal industry it was felt that the creation of a strategic
domestic timber supply was essential to the future security of the country and,
in 1919, the Forestry Commission was established. Although strategic security
of supply was the FC™s initial objective this was quickly supplemented by further
aims such as the commercial production of timber, the stimulation of employment in
areas of rural depopulation and the provision of public bene¬ts such as open-access
recreation and wildlife habitats.7
Public sector forestry in the UK has from the outset followed an erratic course. A
strong initial political will to establish a secure national timber supply ensured that
the 1920s were a period of major afforestation, reversing the trend (if not the effects)
of the previous millennia. However, as memories of wartime shortages receded and
world timber prices slumped, the 1930s saw planting ¬gures fall well behind the
30,000 ha annual target envisaged at the creation of the FC. This slump was offset
to some extent by the Commission™s own promotion of forestry as a response to
rural depopulation trends and a government initiative ˜to create a settled force of
woodsmen and their families whose livelihood would be enhanced from their own
tenanted smallholdings™ (Philip, 1976). Nevertheless, the 1930s still saw an overall
contraction of new planting.


Post-1945
Figure 5.1 illustrates total, FC and private sector annual planting from 1945 to 2000,
providing a starting point for our discussions of the development of both public and
private woodland during this period.

Public sector forestry
The end of the Second World War marked the start of the most sustained period
of UK forestry expansion in recorded history (see Figure 5.1). Initially, national

6 An overview of the history of Welsh forestry from prehistoric times to the present is given in National Assembly
for Wales (2001a).
7 In recent years the FC has also defended its existence as a source of import savings and reduction in agricultural
subsidy. Pearce (1991) shows the import substitution argument to be invalid.
Figure 5.1. Forestry Commission, private sector and total annual forestry planting, Great Britain 1946“2000. (Source: Forestry Commission,
(1979, 1985b, 1988a, 1989, 1990, 1993, 1994a, 1997, 2001c.)
Timber valuation 115

security concerns and high prices again dominated policy objectives. The post-war
adoption of a planned approach to the economy, ¬rm prices and the expansion of
the world timber trade ensured that FC planting accelerated to a peak of over 28,000
ha per annum in the decade following the war. The period from the mid 1950s to
the early 1970s was characterised by fairly stable public sector planting of about
24,000 ha annually. This was helped by a government decision to allow the FC to
operate at a favourably low rate of return compared to other state investments. A
discount rate of only 3 per cent8 was required of the Commission compared to rates
of between 5 per cent and 10 per cent for other state-owned enterprises.9
The year 1971 marked a signi¬cant peak for the FC with plantings exceeding
28,000 ha for the ¬rst time since the early 1950s. However, that year also marked
a turning point beginning a downward trend in FC planting which has continued
for three decades to the present day. The 1970s were a dif¬cult period for the UK
economy with the oil crisis and domestic economic problems (in particular relatively
high in¬‚ation and poor trade balances) leading to heavily depressed growth rates.
This put pressure on all areas of public ¬nance, to which the FC was not immune.
Contractions in FC employment (Thompson, 1990) accompanied reductions in
planting and by 1979 annual planting had dropped to 11,800 ha, i.e. about 40 per cent
of the 1971 level.
The election in 1979 of a Conservative government, pledged to the reduction
of the public sector in favour of private enterprise, meant that the decline in new
planting seen in the 1970s has been extended throughout the 1980s and up to the
present day, a trend which was unaffected by the election of a Labour government
in 1997. New planting in Wales ceased in 1993, followed by England in 1996 and
Scotland in 2000, this latter year being the ¬rst since the Second World War in
which the FC had not undertaken any new planting.
Arguably, a more serious threat to the absolute scale of FC operations is the
disposal of its estate into private ownership, a trend which, as Table 5.1 shows, can
be traced back to the early 1980s when an extensive programme of land sales was
implemented.10 Prior to this, the overall size of the estate had grown every year
since its creation to stand at 1,264,000 ha in 1981 but this has declined in every
year since to stand at 1,052,900 ha in 2000, a reduction of 16.7 per cent.

8 Even lower rates of return were required from plantings carried out in Northern Ireland. From 1989 the FC
was set a target rate of return of 6 per cent but, as this is virtually unattainable without explicit valuation of
non-market bene¬ts, the Treasury initially allowed new investment decisions to be taken at a 3 per cent rate with
the resultant shortfall being written off as Forestry Subsidy (H.M. Treasury, 1991: Annex G). Felling decisions
remain at a 5 per cent discount rate to retain compatibility with existing FC appraisal systems.
9 From 1989 this has been set at 8 per cent for commercial public sector enterprises, with a discretionary rate
of 6 per cent applied to projects with returns accruing to the public sector. This latter rate applies to forestry
management decisions, with the exception of Forestry Enterprise which is permitted to use a 3 per cent discount
rate as an explicit subsidy (Pearce and Ulph, 1998).
10 See statement by the then Secretary of State reproduced in Appendix V of FC (1985a).
116 Applied Environmental Economics

Table 5.1. Forestry Commission holdings: Great Britain 1978“2000 (™000 ha)

FC Awaiting Total Total FC Annual change in
1
estate2 total FC estate3
Year-end plantation planting etc. forest
1978 862.5 83.4 945.9 1,253.2 ”
1979 868.2 84.0 952.2 1,256.3 3.1
1980 884.0 78.4 962.4 1,263.4 7.1
1981 895.7 70.2 965.9 1,264.0 0.6
1982 905.5 59.4 954.9 1,258.7 (5.3)
1983 908.7 54.0 962.7 1,250.9 (7.8)
1984 901.7 47.6 949.3 1,209.2 (41.7)
1985 900.5 34.3 934.8 1,181.0 (28.2)
1986 897.5 30.2 927.7 1,165.5 (15.5)
1987 899.7 23.4 926.4 1,156.4 (9.1)
1988 898.5 20.6 919.1 1,149.4 (7.0)
1989 898.2 17.2 915.4 1,144.2 (5.2)
Productive Other
woodland4
land
1990 863.5 34.3 11.2 909.0 1,139.5 (4.7)
1991 858.5 34.5 9.8 902.8 1,133.1 (6.4)
1992 855.3 34.8 5.6 895.8 1,127.5 (5.6)
1993 845.4 37.1 5.1 887.6 1,115.4 (12.1)
1994 826.6 44.0 3.2 873.8 1,099.5 (15.9)
1995 815.6 45.4 2.0 862.9 1,089.8 (9.8)
1996 804.6 46.7 0.8 852.0 1,080.0 (9.8)
1997 795.4 48.3 0.4 844.1 1,073.0 (7.0)
1998 784.7 49.7 0.7 835.0 1,061.8 (11.2)
1999 779.5 50.7 0.6 830.8 1,057.3 (4.5)
2000 774.2 51.1 0.6 825.9 1,052.9 (4.4)

Notes: 1 Between 1979 and 1984 this ¬gure was disaggregated into ˜land awaiting planting™
and ˜scrubland, etc.™ with the latter being about 10 per cent of the former during that
period.
Total forest + other non-woodland (nursery land + agricultural land + unplantable +
2

forestry workers™ holdings).
3
Numbers in brackets indicate reductions in the FC estate.
4
Recreational land, etc.
Source: Forestry Commission (1979, 1985b, 1989, 1990, 1993, 1994a, 1996, 1997, 2001c).


Despite numerous ministerial pronouncements on safeguarding public access to
land sold by the FC, in almost all cases privatisation has led to the exclusion of
the public (Goodwin, 1995).11 This is particularly serious given that it has been in
areas of high population where the proportion of FC woodlands privatised has been

11 In the period from October 1991 to November 1995, of 35,233 ha privatised only 506 ha (1.4 per cent) had
freedom of access guaranteed (Goodwin, 1995).
Timber valuation 117

highest (Lean, 1996).12 Despite recent claims that the present government has ˜put a
stop to large-scale sales of forest land by the Forestry Commission™ (Cabinet Of¬ce,
2000: p. 90), disposals have continued under the present Labour administration
although arguably at a lower rate than under the previous Conservative government.
Given the poor record of ensuring access to private woodlands, this does appear to
run contrary to of¬cial policy initiatives to promote rural recreation and access as set
out in the recent Rural White Paper (DETR, 2000) and suggests a potential failure
of ˜joined-up government™ in this area. However, this trend of loss of open-access
land in terms of a reduced FC estate is slightly ameliorated by the increased area of
˜other woodland™ within the estate, the majority of which is used for recreational
purposes. Nevertheless, these gains are more than offset (at least in quantitative
terms) by the scale of disposals, suggesting that this remains a cause for concern.
Given this background, recent policy documents promoting the growth of mul-
tipurpose woodlands (FC, 1998) can perhaps best be interpreted as indicating con-
tinued support for private sector expansion (discussed subsequently) and a reorien-
tation of public sector forestry away from conifer monocultures and towards mixed
and broadleaved woodland. This seems most clearly to be the case in our case study
area of Wales. In its most substantial woodland policy document since devolution
in 1999, the National Assembly for Wales (2001a: p. 45) recently stated that:
Since substantial areas of coniferous forest will be harvested during the next 30 years, there
will be an opportunity to reshape these woodlands to deliver wider bene¬ts to society. The
National Assembly™s estate can play a leading role in this process since it is made up of
over 80% coniferous species, compared to only half in private woodlands.

The same document continues by providing an insight into the National Assembly™s
de¬nition of the multipurpose woodland they seek to promote over their chosen
policy horizon of ¬fty years: ˜. . . multi-purpose woodlands managed for recreation,
landscape and wildlife as well as for timber production . . . absorbing carbon dioxide
and so helping ameliorate climate change™ (p. 45).
Such a de¬nition ¬ts well with the cost-bene¬t analysis conducted in this vol-
ume, and this suggests that the results reported in subsequent chapters may have
particular resonance within the current policy environment both across Britain and,
particularly, within our study area of Wales.
Finally, while expansion of the FC has de facto halted, other publicly funded
forests are in process of being established, most notably the National Forest cur-
rently being developed under the auspices of the Countryside Commission in central
England. First proposed in 1987 and de¬ned as a series of woodlands comprising
12 For example, between 1981 and 1996, 91 per cent of FC woodlands in West Yorkshire were privatised;
72 per cent in Durham; 67 per cent in Kent; 53 per cent in Humberside and 43 per cent in Essex (Lean, 1996).
However, one countervailing trend has been the growth of charity-funded woodlands (although these are not
always open-access) such as those operated by the Woodland Trust (Smith, 1996).
118 Applied Environmental Economics

some 30 million trees planted over an area of about 200 square miles (516 km2 ),
the National Forest is intended to bring economic and quality-of-life bene¬ts to
a relatively depressed area (Countryside Commission, 1987, 1993; Cloke et al.,
1996). However, examination of the National Assembly for Wales™ (2001a) Strat-
egy for Trees and Woodlands shows that there are no explicit plans to develop
similar projects either within or near to our study area of Wales.

Private sector forestry
From the outset, direct government intervention through the agency of a state
forestry service has been complemented by the stimulation of a private forestry
sector through the provision of tax relief and other incentives to private individuals
who invest in timber production.13
Despite these incentives, inexperience meant that initial private sector involve-
ment was very restrained. However, from the late 1950s a proliferation of ¬rms spe-
cialising in facilitating private forestry investments considerably eased the practical
problems of such investment. These companies located land, arranged purchases,
planting and felling, and took care of the tax liability and refunding formalities,
thus allowing those for whom tax relief was an attractive proposition to become
forest-owners without ever having to visit a plantation or see a tree.
In this way, post-war planting of private woodlands expanded at a steady rate from
1945 to the early 1970s (see Figure 5.1). However, as with the FC, the 1970s were
a period of relative decline for the private forestry sector. As the OPEC oil-shock
sent the world economy into recession, so the UK™s forest-owning elite no longer
had the excess taxable income to divert into forest tax-havens. However, these were
just the people who bene¬ted from the private sector boom of the 1980s and by
1989 the planting of private woodlands was at its highest ever level. In the search
for cheap afforestable land14 many sites of great ecological value were destroyed
(Royal Society for the Protection of Birds, 1987). This factor, and a national outcry
against such tax avoidance,15 caused the government to act and withdraw such tax
relief with effect from late 1989 (UK Parliament, 1988).
The removal of tax relief had an immediate impact upon private sector planting,
which almost halved from 1989 to 1990. The reason it did not fall further was
primarily the existence of a system of planting and maintenance subsidies (discussed
subsequently) designed to appeal to land-owners and, to a lesser extent, farmers,
rather than to those in search of tax havens. These appear to have generated a
13 Details of these tax relief schemes are given in Bateman (1992).
14 Unlike most other planting costs, land purchase was not tax-deductible. This led investors to plant on cheap,
but often highly unsuitable, wetland areas, destroying valuable natural habitats to produce very poor but highly
tax-deductible plantations (Royal Society for the Protection of Birds, 1987).
15 Culminating in a disparaging Observer front-page magazine feature on the hundred largest forest-owners in
Britain (Lean and Rosie, 1988). See also The Times (1988) and Bloom (1988).
Timber valuation 119

reasonably steady expansion in British private woodlands of just over 15,000 ha per
annum throughout the period 1990“2000.16 However, unlike Forestry Commission
operations, much private woodland development falls outside the scope of policy
in¬‚uence, making objectives such as the promotion of multipurpose woodland more
dif¬cult to achieve (Selman, 1997).


Historical background: summary
In forestry terms the UK has only recently expanded its domestic supply. Although
this grew rapidly in the post-war period, new planting by the FC is now at a total
standstill, superseded by private planting at a relatively constant (if, in national
terms, low) rate. However, current government policy argues that a holistic assess-
ment of the multipurpose nature of woodland suggests a strong case for further
expansion (FC, 1998; DETR, 2000). Certainly, compared to its continental neigh-
bours, the UK lags behind in terms of its forest resource. After eighty years of
expansion, less than 11 per cent of the land area of Great Britain is under woodland
while about 77 per cent is under agriculture.17 This compares with EU averages of
25 per cent and 60 per cent respectively.18 However, this disparity of itself does not
constitute a valid case for continued expansion of UK domestic timber supplies. In
order to assess this we need ¬rst to consider long-term market conditions, and it is
to this issue that we now turn.


The UK timber market and long-term prices
Softwoods
The UK™s consumption of wood products far outstrips its domestic production, the
resultant shortfall being met through timber imports. Indeed, wood products are
consistently within the top ¬ve import items by value. Much of the empirical work
presented in this volume concerns the early 1990s, a period when the UK consumed
roughly 45 million m3 of wood products annually,19 at a cost of £6.3 billion, of which
approximately 83 per cent was softwood products (Forestry Industry Committee
of Great Britain (FICGB), 1992; FC, 2001c). The past decade has seen an overall
modest increase in demand to about 47 million m3 in 1999 at an import cost of
about £6.7 billion (United Nations Development Programme (UNDP) et al., 2000;
FC, 2001c). With both global and domestic demand for lumber forecast to increase
by 20“40 per cent by 2010 (Brown, 1999; Matthews and Hammond, 1999) and

16 Of this about two-thirds was concentrated in Scotland.
17 18
Authors™ calculations based upon FICGB (1992), UNDP et al. (2000) and FC (2001c). Ibid.
19 Measured in wood raw material equivalent (WRME).
120 Applied Environmental Economics

potentially double from present levels by the middle of the twenty-¬rst century
(FICGB, 1992; Watson et al., 1998), some commentators have forecast increases
in future real prices for timber.20
With respect to softwood prices, we see two major ¬‚aws in this argument. First,
the present level of UK production represents only the early stages of an ongoing
substantial expansion of domestic supply engendered by the sustained high levels of
planting in the inter-war years and the period from the late 1940s to the 1970s. This is
set to continue, with production reaching an estimated peak of nearly 20 million m3
by the early 2020s and then tailing off (as a result of the curtailing of Forestry
Commission expansion since the 1970s) to a plateau of about 12 million m3 by the
2050s. Second, and more importantly, this expansion of domestic supply has been
echoed by an increase in the availability of softwood import supplies (UNDP et al.,
2000).21 World coniferous roundwood production rose from 1,096 million m3 in
1971 to a peak of 1,307 million m3 in 1986, slipping back only slightly to a level
of 1,295 million m3 in 1991 (Whiteman, 1995)22 since when the area of softwood
felled in most developed countries has generally been exceeded by the area replanted
(UNDP et al., 2000). When combined with arguments regarding ongoing technical
change,23 these factors seem to suggest that real prices for softwood are unlikely
to increase in the foreseeable future. Indeed, heavy felling by some Baltic nations
during the late 1990s resulted in a fall in UK real timber prices for softwoods over
the course of the decade (Forest Enterprise, 2001).
A number of commentators have examined the issue of whether real timber
prices have changed signi¬cantly over time, the majority concluding in favour of
constant real prices (Doran, 1979; Price and Dale, 1982; Pearce and Markandya,
undated; Bateman and Mellor, 1990; Bateman, 1996; UNDP et al., 2000). In an
in-depth analysis, Whiteman (1995) undertook a time-series analysis of real soft-
wood prices from 1870 to 1989. His best-¬tting time-series model for this period
indicated stable real prices (excluding shocks) prior to the Second World War, a
shift to a higher level during the war and a continuation at a higher, but again
constant (excluding shocks), level after the war. Whiteman™s best estimate was
therefore for a constant real softwood price for the foreseeable future. Our own
analysis of a shorter time-series from the Second World War up to the early 1990s
also supports an assumption of constant real prices, with a single shock to the sys-
tem during the commodity price boom of the 1970s (Bateman and Mellor, 1990).
20 This argument is reinforced by concerns regarding acid-rain damage to forests (Ewers et al., 1986; Bergen
et al., 1992; Pearce, 1993; FC, 1994b). However, estimates indicate that this is unlikely to have any signi¬cant
impact upon timber supply and consequent prices (Bateman, 1996).
21 This trend is exempli¬ed by the case of Sweden where, since the 1930s, timber growth has consistently
outstripped cutting (Wibe, 1992).
22 These measurements are in underbark volumes.
23 Two forms of technical change can be identi¬ed: (i) improved plantation husbandry; (ii) increased availability
of timber substitutes (particularly in the construction industry; see Leigh and Randell, 1981).
Timber valuation 121

Table 5.2. High forest by general species: Forestry Commission and private
woodland in Great Britain 1947“2000 (™000 ha)

Forest type 1947 1965 1980 1994 2000
Mainly coniferous high forest 397 922 1,317 1,516 1,584
Mainly broadleaved high forest 380 352 564 615 837
Total 777 1,274 1,881 2,131 2,421

Source: Figures for 1947, 1965 and 1980 are from the occasional Census of Woodlands
(FC, 1987; reproduced in Pearce, 1993). Figures for 1994 are from FC (1994a) and include
some extrapolation from the 1980 Census. Figures for 2000 are from FC (2001c).


Extending this analysis to the present day (by incorporating data up to March 2001
from Forestry Enterprise, 2001) suggests that the slump in real prices observed in
the late 1990s is not statistically signi¬cant over this longer period (although of
course it would eventually become so if it were sustained for a suf¬ciently long
time).
In conclusion, the consensus view fails to support the hypothesis of future in-
creasing real prices for UK softwoods and we therefore adopt an assumption of
constant real prices in the subsequent analysis (although we do note the possibility
of unforeseen shocks challenging such an assumption).

Hardwoods
While global reserves of coniferous forest have been reasonably stable or have
even grown over the past two decades, the post-war era has seen some decline in
temperate hardwoods and a dramatic fall in tropical hardwoods. Considering ¬rst
the British case, the twentieth century saw a continuation of a centuries-old decline
in the area of ancient broadleaf woodlands. In England and Wales this stood at just
142,000 ha in 1933 yet had more than halved by the mid 1980s (Nature Conservancy
Council (NCC), 1984). The bulk of this loss arose from conversions to mainly
conifer plantations, with the remaining losses generally attributable to agricultural
encroachment (NCC, 1984; Council for the Protection of Rural England (CPRE),
1992). However, the planting of new broadleaved woodlands has meant that, since
the 1960s, the overall area of broadleaved high forest has consistently risen, as
shown in Table 5.2.
While newly planted broadleaved woodland does not have the ecological value
of ancient woodland, it does represent an encouraging trend. However, as in the
case of softwoods, the UK is far from self-suf¬cient in hardwoods. The present
level of UK domestic hardwood (round and sawn) consumption is about 2 million
m3 per annum. This exceeds domestic production, which fell over the past decade
122 Applied Environmental Economics

from 1.2 million m3 in 1991 to 0.8 million m3 in 1999 as a result of low planting and
high felling early in the century (FICGB, 1992; FC, 2001c). While this represents
a much higher self-suf¬ciency rate than for softwoods, and production is forecast
to rise to 1 million m3 per annum in 2001 and remain at that level for at least
twenty years (FC, 2001c), nevertheless the UK is highly import-dependent and
consequently subject to ¬‚uctuations in the world market.
Global stocks of hardwoods have fallen dramatically in the post-war period,
primarily as a result of deforestation in the developing, tropical countries of the
world in which such trees are predominant. The causes of this deforestation are
complex and interlinked and include increasing consumption pressures from both
the developed and the developing world (Whiteman, 1995; Global Environment
Facility, 1998; World Bank, 1999),24 population and poverty pressures in the
developing world (World Resources Institute, 1994; United Nations Population
Division (UNPD), 1998), sustained growth in demand for fuel-wood, to the point
where it is currently estimated that half of all global wood consumption is as fuel
(UNDP et al., 2000), and forest burning for agricultural expansion and other reasons
(Myers, 1990; Elvidge et al., 1999; World Commission on Forests and Sustainable
Development, 1999).
The total loss of global forests to date is uncertain but may be as high as
50 per cent (Bryant et al., 1997). What is more certain is that annual net hard-
wood extraction rates rose from about 0.8 per cent at the end of the 1970s (Doran,
1979) to 1.8 per cent a decade later (Myers, 1990) but have fallen back slightly
over the course of the 1990s (Food and Agriculture Organization (FAO), 1997;
UNDP et al., 2000). Current losses are the subject of considerable controversy but
probably exceed 130,000 km2 annually (FAO, 1997; Matthews et al., 2000; Tucker
and Townsend, 2000; UNDP et al., 2000), a rate which means that by 2010 only
Brazil and the Democratic Republic of Congo will have any signi¬cant remaining
areas of rainforest and both of these will be under unsustainable long-term pressure.
Given that the rainforests represent the richest global environment for biodiversity
(Davis et al., 1994; Olson and Dinerstein, 1998), the potential exists for species
extinction on a scale unprecedented in human history (MacNeill, 1990; Pearce and
Warford, 1993; World Resources Institute, 1994; Old¬eld et al., 1998; UNDP et al.,
2000).
Setting aside the terrible ecological consequences of this destruction, the unsus-
tainable nature of current global hardwood extraction has been seen by some as
likely to lead to increases in associated real prices. Indeed, published estimates of
such increases can be found, ranging from a credible 0.5 per cent to what we regard
24 While the majority of population growth is in the developing world, a citizen of the developed world consumes
up to ten times the ecosystem goods and services of a citizen in the developing world (Global Environment
Facility, 1998).
Timber valuation 123

as an unfeasibly high 4 per cent annually.25 However, this is balanced by opposing
views such as that of Whiteman (1995), who argues that while consumption may
increase, ˜it should be possible to improve forest management to meet these de-
mands, which would then keep timber prices relatively stable™. Certainly the rate
of growth in British hardwood planting detailed above should mean that, providing
there is not an unforeseen sharp rise in domestic demand, the current rate of UK
self-suf¬ciency may improve, giving something of a domestic buffer against future
reductions in global supplies.
This is an area of uncertainty, disagreement and relatively little in-depth research.
While we feel that there is a considerably stronger case for real price increases in
hardwoods than softwoods, any such rise is likely to be some way off. Furthermore,
as our wider study examines the potential for conversions out of existing agricultural
land use and into forestry we prefer to adopt conservative assumptions with regard
to changes in the future value of both land uses. Adopting any positive rate of real
price increase for hardwoods would translate into substantial increases in projected
timber values. Instead we prefer to adopt the zero real price rise assumption of
Whiteman (1995) and accept that any transpiring price increases will improve the
potential for land use conversion to forestry. Such an assumption also allows us
readily to revise our calculations in the light of subsequent improved information.
We now turn our attention away from prices towards the other major source of
timber-based revenues: grants and subsidies.


Grants
Given the long-delayed nature of forestry returns, government incentives have al-
ways played a major role in UK private sector planting decisions. The earliest
incentives coincided with the establishment of the Forestry Commission when,
in 1919, a scrub clearance and ground preparation grant was introduced. A sec-
ond planting grant scheme, introduced in 1927, established an enduring trend for
broadleaves to be given preferential subsidy rates over conifers, re¬‚ecting an early
recognition of non-strategic/production objectives within forestry policy.
Following the Second World War a variety of Forestry Commission administered
schemes were introduced. Examination of these reveals a gradual movement in
forestry policy objectives from simply maximising timber production to initiatives
giving equal emphasis to timber, environmental and recreational goals (Johnson and
Nicholls, 1991; Bateman, 1996; Winter, 1996; MacFarlane, 2000; DETR, 2000;
FC, 1998, 2001b). However, until the late 1980s the overriding force behind the
expansion of private sector forestry was tax concessions. The scrapping of most

25 Estimates are from Johnston et al. (1967), Doran (1979), Burnham (1985) and Hart (1987).
124 Applied Environmental Economics

of these concessions in the 1988 Budget (Lynch, 1989) thrust the role of grants
centre-stage as the main means of state support for forestry in the UK, a role which
has persisted to the present day.
The majority of woodland grants are administered by the Forestry Commission
although other funding bodies are also important and both are considered below.
The review presented in this section focuses primarily upon the rates of grant
in operation during the period of the early 1990s for which our empirical model
operates. However, in comparison with the major changes introduced at the end
of the 1980s, the basic structure of this grant aid has varied relatively little over
the past decade although grant amounts have generally increased (Winter, 1996;
MacFarlane, 2000). A number of recent policy initiatives have stressed plans for
expanding forestry both in Britain as a whole (FC, 1998; DETR, 2000) and in
our case study area of Wales (National Assembly for Wales, 1999, 2001a; FC,
2001a,b). However, these documents have not put forward policies for an overhaul
of the grant-aiding system but rather suggest that levels of aid may be increased in
the future (although no speci¬c announcements have been made to date). Therefore,
the structure of the model presented here for the early 1990s remains valid today
and for the foreseeable future although rates of grant are already dated as they are
continually under review.


Forestry Commission administered grants
Throughout the 1980s the FC emphasised its reorientation away from the simple
pursuit of timber output and towards wider objectives (FC, 1985c). Such policy was
embodied in the introduction, in 1988, of the Woodland Grant Scheme (FC, 1988b)
which, alongside the stimulation of timber production and rural employment, ex-
plicitly set out to enhance landscape, create wildlife habitat, provide longer-term
recreation and sporting facilities and encourage the conservation and regeneration
of existing woodlands. Rates of support under the Woodland Grant Scheme (WGS)
were revised in 1990 as listed in Table 5.3.
Payments under the WGS were made in three instalments: 70 per cent at planting,
20 per cent after ¬ve years and 10 per cent after a further ¬ve years (subject to
satisfactory establishment). In addition to this a Better Land Supplement (BLS) was
payable for planting on arable/improved grassland cultivated (including ploughing)
within the previous ten years. BLS was £400/ha for conifers or £600/ha for broad-
leaves, all payable at planting.
Further enhancement of this package was provided in 1992 by the introduc-
tion of the Woodland Management Grant (WMG). This provided an annual ad-
dition to the WGS, payable after the ¬rst ten years of establishment in return
for the setting down and execution of ¬ve-yearly management plans designed to
Timber valuation 125

Table 5.3. Woodland Grant Scheme payments (£/ha)

Area planted Conifers Broadleaves
0.25“0.9 ha 1,005 1,575
1.0“2.9 ha 880 1,375
3.0“9.9 ha 795 1,175
10 ha+ 615 975

Source: Johnson and Nicholls (1991).

Table 5.4. Woodland Management Grants

Period of eligibility Rate of grant
Type of WMG (age of wood in years) (£/ha per annum)

Standard: conifer1 11“20 10
Standard: broadleaf1 11“40 25
Special1,2 11 onwards 35
Supplement for small woods3
Standard: conifer 11“20 5
Standard: broadleaf 11“40 10
Special4 11 onwards 10

Notes: 1 All these grants are also payable as additions where the owner is a farmer under
the Farm Woodland Scheme, as compensation for agricultural output forgone (but not for
establishment costs).
2
Higher rates are available for woodlands of special environmental value (nature
conservation, landscape or public recreation). The owner will be expected to maintain the
wood™s character. These grants are available for any forest older than ten years. However,
they may be extended to younger or even proposed forest if the Forestry Commission is
satis¬ed that there is demand for such a provision.
3
Available as additions for all woodlands of less than 10 ha (of correct age).
4
Available for any woodland (over ten years) of less than 10 ha where the woodland is of
special environmental value.
Source: Johnson and Nicholls (1991).

increase the environmental value of the woodlands concerned. Table 5.4 details
WMG payments.
The year 1991 also saw the FC introduce the Community Woodland Supplement
(CWS), a further addition to the WGS (and WMG) designed to promote recreational
woodlands ˜within 5 miles of the edge of a town or city and in an area where the
opportunities for woodland recreation are limited™ (FC, 1991). In implementation
this has been interpreted very broadly so that relatively small communities of just
a few thousand people are considered suf¬cient to justify payment of CWS. At
its introduction the scheme consisted of a single payment of £950/ha payable at
126 Applied Environmental Economics

planting. All woodlands qualifying for CWS were allowed WGS and WMG, the
latter being paid at the enhanced ˜special™ rate.
In addition to the above, from 1992 the FC offered a single £100 ¬‚at rate payment
for each new woodland (irrespective of size) conditional on the drawing up of a
management plan (FC, 1991).26


Other grant schemes
In 1988 the then Ministry of Agriculture, Fisheries and Food (MAFF) introduced the
Farm Woodland Scheme (FWS) to provide annual income support to farmers who
establish woodlands on what was previously agricultural land (MAFF, 1987a).27

<<

. 4
( 11)



>>