Visitor Expenditure: The Case of Cycle Recreation and Tourism

Abstract

The paper seeks to contribute to our understanding of the economic impact of sports tourism using the case study of a cycle network in the North East of England, UK, used for tourism, recreation and utility purposes. It explores the foundations of economic impacts of such a network and focuses on underlying behavioural responses of cyclists and their spending. The paper develops a model of expenditure on the basis of 383 travel diaries. The findings confirm that incomes, group sizes and durations of activity are integrally linked determinants of expenditure. The expenditures and durations of cycle trips are linked to preferences for longer journeys. This has implications for planners of routes to attract all types of cyclists from the most casual leisure trip to racing cyclists. Furthermore, the research findings infer that as extra-network and intra-network tourism groups cycling on the network do not behave differently they therefore should both be targeted by sports and tourism agencies.

Keywords:

• Cycle Tourism
• Incomes
• Group Sizes
• Duration
• Expenditure

Introduction

There is now a growing literature on sports tourism where sport is increasingly discussed either as a motivation for travel or as a discrete tourism activity. Perhaps more importantly sports tourism is recognized as a distinct and unique phenomenon that requires more detailed investigation than hitherto (Standeven & de Knop, 1999; Weed & Bull, 2004; Higham, 2005; Gibson, 2006). This approach contrasts sharply with the traditional emphasis in the sports and tourism literature, which has focused on the economic impact of sport tourism and, in particular, sports events (Preuss, 2004; Gratton et al., 2000, 2005). The latter, of course, fits very largely with a broader literature on the economic impacts of tourism (Dwyer et al., 2004; Mules, 2005). One of the limitations of such studies is that they have often referred to discrete events or cases and thus have tended to be disconnected with the wider literature on sports tourism. While some authors have been critical of the economic impact literature in relation to sports tourism, this paper seeks to contribute to our understanding of this in relation to cycling in a region of the UK.

This paper offers an examination of the determination of user expenditures associated with a cycle network developed for sport, recreation, health and utility. There are, therefore, two main aims of the paper. The first is to explore the foundations of the direct economic impacts of such a network. In so doing the paper addresses several conceptual problems in estimating the expenditures of route users, extending and developing a model currently used in the literature. The second aim of the paper is to examine the behaviour underpinning expenditures in connection with both recreational cyclists defined as those originating within the network, but whom are nonetheless engaged in tourism activity, and also tourism cyclists defined as those originating from outside the network. The latter definition of tourists is the one adopted by regional development agencies in England when assessing economic impact, as discussed further below. However, route marketing and enhancement of economic impact may require variable strategies if the behaviours of types of users are different. Section 2 provides a brief review of the literature as a contextual discussion. Section 3 then describes the context of the research. Section 4 outlines the theoretical framework of the paper. Section 5 outlines the research and research design. Section 6 describes the data, the methods of analysis and presents the main results. Conclusions then follow.

Literature Review: Some Conceptual Issues

An important feature of the more recent sports tourism literature is that

Sports tourism is a social, economic and cultural phenomenon arising from the unique interaction of activity, people and place. (Weed & Bull, 2004, p. 37)

This definition indicates how a specific activity, i.e. cycling, can have a distinct character illustrated by the

… comparison of cycling to work through the pollution of a major city with cycling through beautiful countryside in one of the National Parks throughout the world … (Weed & Bull, 2004, p. 37)1

These quotations raise two important points. The first is that cycling (and thus logically walking) as a means of travel directly connects the activity and travel features of tourism; they become inseparably linked (Bull, 2006). In this regard, subtle distinctions can apply. For example cycling in the countryside could be undertaken in order to get to work. This element of the day could be nevertheless recreational activity as well as a functional trip. Likewise, any recreational cycling by local people from their own immediate environment could be viewed as tourism activity if it offers a similar experience to that enjoyed while on holiday elsewhere. This raises the interesting question of whether or not the motivations and behaviours of these types or segments of cyclists differ. A second related point is that the quotations suggest that motivations and behaviours may differ according to the interaction of people connected with the activity. For example, it might be expected that group behaviour is distinct from single-person behaviour. Both of these issues have implications for the marketing and planning of cycle routes in seeking to enhance economic advantage. Is there a need for braided promotional strategies for each type of user in terms of their origin and group composition?

As far as the economic impact of sports-tourism is concerned, the literature has a number of characteristics. The first is that it essentially defines tourists as visitors to a pre-defined area (normally administrative boundaries) of impact over a specific timescale. This is to avoid including ‘deadweight’ expenditures in the analysis, which merely redistribute economic activity rather than bringing additional expenditure into a local economy. It should be noted that considerable complexities are discussed in this regard concerning, for example ‘time switching’ and ‘crowding out’ (UK Sport, undated; Preuss, 2004). The second is that there has been considerable discussion about how the measurement of impacts should proceed in both tourism and sport (Dwyer et al., 2004, 2006; Tyrrell & Johnston, 2006). There have been two principal lines of discussion. The largest proportion of the literature concerns the relative merits of the appropriate method of multiplier calculation in order to identify the indirect and induced effects following an injection of (sports) tourism expenditure. There is some consensus that a computable general equilibrium approach is apposite with larger scale analysis although Input–Output models, despite critical reviews to date, are also in wide use for local or sub-regional studies (Dwyer et al., 2004; Blake, 2005; Sun, 2007).

The second strand of literature is connected with how best to estimate the initial (sports) tourism injection of expenditure into the destination or event (Frechtling, 2006). It is argued that large gaps remain in our understanding of the determinants and form of visitor expenditure (Wilton & Nickerson, 2006). For example, a source of potential bias is that both the group size and duration of visitation are sometimes ignored in simple gross or per capita estimates of direct expenditure.2 This is in the sense that the individual provides the basis of expenditure data collection and these individual measurements are then aggregated. The measurement error of recording individual versus group or party expenditure has been discussed by several authors in the recent literature (Downward & Lumsdon, 2003; Stynes & White, 2006; Loomis, 2007; and Fredman, 2008). The point is that it can be hypothesized, as noted above, that groups behave differently to the sum of their parts and consequently it is group expenditures that should be directly measured. A further lacuna in the literature is that the duration of the leisure activity has typically been treated as an exogenous factor in the analysis, from which expenditure calculations per flow of time might be estimated (Downward, 2004). Intuitively, however, as activities that allocate time, sports, tourism and sports-tourism clearly involve potentially mutual decisions about both expenditures and time allocation.

It is these latter issues that are primarily investigated in this research. The paper seeks to explore the potential symbiotic link between the duration of sports tourism activity and expenditure associated with the activity by drawing upon a parsimonious model of expenditures that recognizes the need to focus on groups and which also accounts for the trip and route preferences of visitors. The paper thus extends and develops the work of Downward & Lumsdon (2004), Loomis (2007) and Fredman (2008) but also assesses if local recreational behaviour, in terms of expenditure, is different to that originating from outside the area. Drawing upon economic theory, this paper argues that, as a form of leisure, sports-tourism activity involves the choice to consume time as well as monetary resources. As a consequence, the ‘duration of the activity’ may be understood as one element of the expenditure of total resources. The relationship between these two fundamental elements, time and monetary expenditure may, consequently, provide additional input into policy aimed at eliciting maximum economic impact from the tangible monetary expenditure.

These issues are addressed by examining the duration of activity and expenditure associated with a series of cycle trails promoted for tourism purposes in the North East (NE) of England in the United Kingdom. It has been prompted by the growing literature regarding potential community and commercial benefits gained from cycle tourism (Cope, et al., 2000, 2003; Bowker et al., 2007).

The Study Area

The North East England Regional Tourism Strategy 2005–2010 (One NorthEast, 2005) has ten main objectives, of which the key four are:

1. To attract more domestic and overseas tourists to the region;

2. To increase visitors' average spend;

3. To increase visits throughout the year, not solely in the main holiday season;

4. To grow the distribution of tourism across the region.

The tourism sector in the North East of England generates an estimated 8.5 million visitor overnight stays and visitor expenditure of £3.4 billion per annum within the regional economy (One NorthEast, 2008). Much of this is related to city-based tourism at locations such as the regional capital of Newcastle-upon-Tyne or historic cities such as Durham. However, the strategy also sets out the case for the development of markets for long distance walking and cycling routes building on an earlier cycle-tourism strategy Gearing Up for Growth (ACK, 2002). The development of rural tourism within the North East has focused on attracting visitors to a rich variety of landscapes near to these urban areas, as well as on the wide range of heritage assets such as Hadrian's Wall World Heritage site (One NorthEast, 2006).

Four key routes surveyed as part of this study sit within a larger cycle network developing within the region. They were selected principally because they were designed as thematic tourism routes that reflect the topography and cultural heritage of the North East region. The routes were:

1. C2C (Sea to Sea) Cycle Route (including routes in Sunderland)

2. Coast and Castles Cycle Route

3. Hadrian's Cycleway

4. Pennine Cycleway (northern section)

The characteristics and length of the routes are listed in Table 1. In the UK, routes tend to be developed using traffic-free or off-highways sections wherever feasible. They also utilize highways and other facilities where the cyclist has to mix with other traffic.

Table 1  Route Length by Route Type Classification (km)

Route	Urban		Rural		Total
	Traffic-free	On road	Traffic-free	On road
C2C (east)	34.9	6.1	53.1	32.7	126.8
C2C (west)	13.1	2.3	29.5	115.1	160.0
C2C (all)	48.0	8.4	82.6	147.8	286.8
Coast and Castles	25.1	13.4	35.4	77.1	151.0
Hadrian's Cycleway (east)	24.8	7.7	11.7	59.4	103.6
Hadrian's Cycleway (west)	14.6	8.6	22.4	85.2	130.8
Hadrian's Cycleway (all)	39.4	16.3	34.1	144.6	234.4
Pennine Cycleway	1.0	2.6	12.3	167.8	183.7

Theoretical Framework

Following Downward & Lumsdon (2000, 2004), to understand the expenditure flowing from cycle-tourism in the subject area, a starting point for the research is the economic theory of consumer demand. Maximizing utility from a budget constraint involving the income level of consumers and the prices of alternative commodities produces a standard demand relationship. If q refers to the quantity demanded of a good or service, p to the relative price of the commodity, M to the consumer's income, T to the consumer's tastes, and t to a given period of time, equation (1) describes the theory.

Quantity demanded of a good or service over a particular time period, and for a given set of consumer preferences, will be dependent upon relative prices between a good or service and its substitutes (or complements) as well as incomes. Because the relative prices of goods and services will be constant at any specific point in time, it is often more practical to measure demand in more aggregate terms for groups of goods and services. Consequently, demand can be represented by expenditures as identified in equation (2).

This representation of demand is known as an ‘Engel curve’. It focuses attention on the relationship between expenditure and income for given tastes (Deaton & Muellbauer, 1986). This theory of demand suggests that for given sets of visitor preferences and for a given duration of time, incomes and spending will co-vary as flows of economic activity. Essentially, it is this reasoning that underpins the monetary expenditure components of economic impact studies.

However, as recognized by Becker (1965, 1976), the basic economic model of consumer behaviour is problematic in that it does not account for the fact that the consumer is also a producer. The individual, uses the resources of ‘time’ and ‘market goods’ (essentially monetary expenditures) in ‘household production’ to generate the basic commodities that yield utility from consumption. As all economic activities involve time and other goods, purchased via markets, economic agents essentially make choices involving the relative intensity of these inputs in both producing and consuming commodities. In this regard, the commodities that are consumed are functions of goods bought in markets and time. There is an obvious application here for sports tourism in that leisure time is invested in production and consumption, but so are goods that are purchased on the market. In this regard, sports tourism can be viewed as a composite commodity, with a composite demand.3

Moreover, whilst the approach developed by Becker (1965, 1976) is motivated primarily with households in mind, conceptually speaking the point is that groups of individuals might invest in personal capital, skills and capabilities, or social capital and reputation to facilitate the consumption that provides the greatest utility for them. This helps to explain why demand, ostensibly by individuals, is often structured according to broader socio-economic characteristics, in which individuals share consumption profiles (see Downward & Riordan, 2007, for an analysis of sport and recreational consumption).

In the current context, therefore, it is to be expected that groups of different sizes will consume tourism in different ways according to their shared group preferences and that both the time element and market-good expenditure element of consumption might be expected to differ for different types of groups. This is consistent with the arguments of Weed & Bull (2004) and consequently developed by Bull (2006) in relation to racing cyclists as sports tourists. Thus, equation (3) provides an exposition of the relationship in which a composite commodity Z, that is a latent variable, comprising expenditure and time, is consumed by different groups j depending upon the incomes and preferences of the groups.4

Recognizing that the duration of a leisure activity is, in itself, potentially a key decision variable implies that the interaction between expenditures and duration needs to be explored when examining expenditure. Economic impact studies ignore this problem and model expenditures in isolation.

Figure 1 presents the logic of the current investigation, with arrows suggesting the direction of relationships. Two dependent variables, expenditure, and the duration of activity, are potentially related, as described above, and mutually determined by group size, income, route and trip characteristics as independent variables. In each case, therefore, an equation can be specified that models the decision contingent upon the trip and route characteristics to measure motivations; that is, the preferences and tastes of cyclists, and the group sizes and income. If the expenditure equation also includes the duration of trips, whilst the trip-duration equation also includes expenditure, feedback between the decisions can be investigated. Distinguishing between groups as tourists to the region or local residents making recreational trips can then be used to identify the net economic injection to the area according to the protocol of economic impact studies.

Figure 1 Conceptual Model of Expenditure and Trip Duration.

Research Context and Design

Data were collected from 2001–2006 as part of a larger scale monitoring strategy for cycling in the North East of England with a design originally piloted for monitoring the North Sea Cycle Route (NSCR) and which has been discussed fully in Cope et al. (2004). Given that the research context was a network of long-distance cycle routes, two main issues affected the research design. These were the selection and combination of appropriate research methods and, subsequently, sampling. In the former case, two research tools were employed. As part of the broader monitoring activity, the multi-user routes were investigated by an intercept survey to capture information on numbers in user groups, and the purpose of the trip. A closed response format was adopted. A travel diary was also offered to those intercepted, to be returned by pre-paid post, when the trip was complete. This had two related functions. The first was to record more sensitive information such as group incomes and expenditures, which needed to be collated for the group. The second was to capture values of actual behaviours, such as expenditure and trip duration, which had not yet occurred as the diary was issued. The diary method is well tested in the context of media research and to a lesser extent in tourism studies (Thornton et al., 1997). The use of diaries raises issues of sample selection in building a model and is discussed further below. Figure 2, however, describes the research design.

Figure 2 Research Design.

The second major issue concerned the sampling approach from which to collect data and its periodicity. Drawing on the concept of ‘gravity’ modelling, i.e. that trip generation will reflect the density of populations in sources and destinations, it was decided to capture data on each route of the network at points of access to and egress from a typical chain of different centres of gravity for each route in the network. Data were then collected at selected weekends and weekdays over spring to autumn so as to include the traditional summer vacation period. Data collection was, to a large extent, constrained by the administrative process of collecting monitoring data but the aim was to capture the stylised characteristic of each route in the network at consistent periods of time, rather than simply to randomly sample across the network. This also allowed routes within the network to be identified and differences in behaviour modelled. Table 2 describes the variables that were measured in the study.

Table 2  Variable Description

Variable	Theoretical contribution	Survey instrument	Measurement scale
Group expenditure	Dimension of consumption (production) for economic impact	Diary	Scale
Group size	Context of consumption (production)	Intercept survey	Scale
Trip duration	Dimension of consumption (production)	Diary	Scale
Group income	Income constraint	Diary	Scale
Tourism group	If majority of group is strictly > half were resident outside the NE	Intercept survey	Nominal
Particular purpose	Preferences, i.e. of trip	Intercept survey	Nominal
Short circular trip	Preferences, i.e. of trip	Intercept survey	Nominal
Short out and back	Preferences, i.e. of trip	Intercept survey	Nominal
Day cycle	Preferences, i.e. of trip	Intercept survey	Nominal
Short break	Preferences, i.e. of trip	Intercept survey	Nominal
Longer linear cycle	Preferences, i.e. of trip	Intercept survey	Nominal
Longer circular cycle	Preferences, i.e. of trip	Intercept survey	Nominal
Coasts and Castles	Preferences, i.e. of network route	Intercept survey	Nominal
Sunderland	Preferences, i.e. of network route	Intercept survey	Nominal
C2C	Preferences, i.e. of network route	Intercept survey	Nominal
Hadrian's Wall	Preferences, i.e. of network route	Intercept survey	Nominal
Pennine Trail	Preferences, i.e. of network route	Intercept survey	Nominal

Data, Methods of Analysis and Results

Over the period of the research a total of 3104 diaries were issued. However, the usable number of diaries returned, i.e. those that recorded relevant incomes, durations and expenditures, was only 373, representing an effective response rate of 12%.5 This is relatively low, but such a wastage rate in the absence of interviewer completion is always likely to be high. Nonetheless, the main emphasis of the research was to generate as close to a statistically reliable sample for analysis of the network as a whole and the response produced a sample close to the recommended size of n = 377 with a margin of error of 5% and 95% confidence interval.

Two important caveats should be attached to this claim however. The first is that because the population of cyclists from which the sample is drawn is essentially unknown, there is no opportunity to allow for response bias in the data collection. The second issue concerns the sub-sample characteristics of the aggregate sample. The second and third columns of Tables 3 and 4 report the relative proportions of the issue of diaries, that is groups accepting a diary, and the effective response rates for individual routes in the network and the years in which sampling took place respectively. Column four in each table calculates an index in which the ratio of diaries issued to diaries returned is expressed with a base of 100.6 Column 5 then provides the standard normal test statistic for a test of the proportion of the returned diaries for each route or year (out of the total returned) in comparison to the proportion of diaries issued for each route or year (out of the total number issued).7 The null hypothesis is that the proportions were equal. In Table 3, the index values show that the Coasts and Castles route and the Pennine trail had more diaries returned as a proportion than were issued as a proportion, with the opposite case for the other trails. However, a statistical difference was only identified for Coasts and Castles and Hadrian's Wall with a significance level of 5%. In Table 4 the index values suggest that in the years between 2001 and 2004 inclusive there was a greater proportion of returned diaries than issued, with 2005 and 2006 being the opposite. However, statistically significant differences at the level of 5% only apply to years 2003, 2005 and 2006. These results suggest a degree of fragility of the sub-samples as representatives of temporal and route demand. Consequently, this suggests that the total sample has more reliability, but nonetheless the results would be better seen as generating indicative findings through a case-study rather than a basis upon which to build forecasts, particularly for segments of the network and particular years.8

Table 3  Relative Route Response Rates

Route	Diaries Issued % (n = 3104)	Diaries returned n = 373	Index Base = 100	Z
Coasts and Castles	32.31	42.09	76.77	3.83*
Sunderland	10.60	8.04	131.78	−1.82
C2C	20.20	19.30	104.65	−0.44
Hadrian's Wall	28.70	21.98	130.57	−3.13*
Pennine Trail	8.18	8.58	95.38	0.27
Total	100	100
*Denotes a 5% statistical significance

Table 4  Relative Period Response Rates

Year	Diaries Issued % (n = 3104)	Diaries returned % (n = 373)	Index Base = 100	Z
2001	6.86	9.65	71.10	1.82
2002	20.20	22.52	89.70	1.07
2003	10.02	17.96	55.78	4.00*
2004	22.45	23.86	94.11	0.64
2005	17.17	10.72	160.12	−4.02*
2006	23.29	15.28	152.42	−4.30*
Total	100	100	100
*Denotes a 5% statistical significance

Table 5 provides descriptive information on the other components of the analysis, the dependent variables of spending and duration, the group size numbers and trip characteristics. The data reveal some cyclical average expenditures suggesting that demand has fluctuated. However, it also appears that this has coincided with greater group sizes and incomes, the presence of more tourism groups and greater duration of trips. Significantly in the first row, the figures in brackets suggest that the number of groups reporting zero spending has also fallen. The data also reveal that over time the trip characteristics have changed, with some growth in longer cycling activities but a reduction in day cycling. Whilst the first result is commensurate with the growth in tourism groups' durations and incomes, the latter result clearly could be connected with the small size of subsamples. Regression analysis was consequently used to tease out the interaction between these variables and to generate a model of expenditures and duration of cycling.

Table 5  Characteristics of the Sample

Variable	2001	2002	2003	2004	2005	2006	Total
Mean group expenditure (£)	80.5	200.32	120.8	164.9	467.0	314.0	212.0
(% cases £0)	(30.6)	(29.8)	(31.3)	(29.2)	(2.5)	(1.8)	(22.8)
Mean group size	2.0	2.5	1.3	2.1	2.4	2.0	2.0
Mean duration (hours)	21.0	12.5	27.5	16.0	33.3	39.3	23.2
Mean group income (£000)	47.4	60.0	40.0	44.7	81.8	58.0	54.0
Tourism group (%)	0.0	21.4	4.5	9.0	57.5	61.4	23.3
Particular purpose (%)	2.0	0.0	1.0	3.0	0.0	0.0	1.6
Short circular tip (%)	8.0	6.0	17.0	19.0	1.0	0.0	13.0
Short out and back (%)	5.0	12.0	15.0	17.0	1.0	0.0	13.0
Day cycle (%)	8.3	31.0	22.4	21.3	7.5	0.0	17.7
Short break (%)	19.4	11.9	6.0	13.5	37.5	31.6	17.7
Longer linear cycle (%)	5.6	19.0	10.4	13.5	40.0	50.9	22.0
Longer circular cycle (%)	22.2	8.3	11.9	6.7	10.0	15.8	11.3

Following the logic summarised in Figure 1, the variables measuring group expenditure and also the duration of the group trip were regressed on trip characteristics as well as route identifiers (in order to capture group preferences), group sizes and incomes. This would enable a representation of the economic impetus to spend either time or money on cycling in group settings. However, in the expenditure equation, duration was also included as a potential determinant, and vice versa in the duration equation. This was to recognise that the choice to consume may be conditional on the former. To distinguish tourists groups, consistent with economic impact studies, a dummy variable was used to identify if the majority of the group originated from outside the network area. Finally, a variable measuring the year of the survey was also included to recognize the pooled nature of the data, and to capture any specific effects for any given year.

For the expenditure equation, two estimators were employed. The first was Ordinary Least Squares (OLS), which was also the case for the duration variable. As indicated in equation (4), OLS estimates the impact of a vector of independent variables x upon the dependent variable y by estimating a vector of coefficients, β, subject to a random error, υ, which is assumed to be (at least asymptotically) independently and normally distributed, all over the groups i and for time periods t. In this analysis, the standard errors for statistical inference were estimated as Huber-White robust standard errors to control for any heteroscedasticity in the broadly cross-sectional sample.

In the expenditure case, Tobit analysis was also undertaken to allow for the possibility that any recorded zero expenditures could represent constrained opportunities to spend money as opposed to voluntary actions not to spend.9 This could arise as a result of supply-side opportunities being unavailable. As indicated in equation (5), Tobit estimates the effects of the vector of independent variables, x, upon the ‘latent’ variable y* where this variable is only observed for the cases in which y is less than or greater than some censoring value. In other words, the value of the latent variable ‘collapses’ onto a specific value for some cases, which in this context is zero pounds sterling.

To obtain the final estimated equations, the general specification for each model was refined. This was done using a general-to-specific approach by testing restrictions that coefficients upon individual variables, such as year, or spending, or groups of variables, such as the routes, or trip characteristics, were insignificantly different from zero. The theoretical analysis above determined which sets of variables entered the restriction tests. Table 6 presents the tests of restrictions, with statistics marked with an asterisk being those in which the null hypothesis of no (joint) significance of the variable(s) could be rejected with a 5% significance level. The table reveals that the variables measuring group incomes and size, as well as the trip preferences of cyclists appear as the significant determinants of expenditure, but not duration. Further, the results suggest that it is only the trip and route characteristics that are significant determinants of the duration of the cycling activity.10

Table 6  Variable Deletion Tests⁶

Variable	Expenditure	Duration
Year	F(1, 357) = 0.95	F(1, 357) = 1.93
Group Expenditure	n/a	F(1, 357) = 0.09
Group Size	F(1, 357) = 6.51*	F(1, 357) = 0.64
Trip Duration	F(1, 357) = 0.09	n/a
Group Income	F(1, 357) = 5.02*	F(1, 357) = 0.05
Tourism Group	F(1, 357) = 0.06	F(1, 357) = 0.01
Trip Preferences	F(6, 357) = 11.97*	F(6, 357) = 11.65*
Route Preferences	F(4, 357) = 2.24	F(4, 357) = 3.55*

Tables 7 and 8 present more detailed analysis of the results and also the regression coefficient estimates. Table 7 reveals that the Tobit analysis confirms that trip preferences, group sizes and incomes contribute to direct economic impacts, as suggested by the OLS analysis. There is a difference in the quantitative magnitude, however, and this difference in the coefficients reflects the potential bias in the OLS analysis as a result of having ‘too many’ zero expenditure values because of constraints on possible expenditure options. The Tobit coefficients thus represent what the ‘desired’ effects on expenditure could be without the constraints and can be thought of as an upper bound on expenditure. Naturally, the coefficients suggest larger (in absolute terms) effects of the variables. In contrast, Table 8 presents two sets of marginal effects. The first ‘censored’ effects identify the change in actual observed expenditure following from a change in the independent variables conditional on expenditure being greater than zero. The second ‘probability’ effects measure the impact that the independent variables have on the probability of being a non-zero expenditure group. This is equivalent to examining where expenditures are more likely to come from should any constraints on expenditure relax.

Table 7  Expenditure

Independent variable	OLS estimated coefficient	t-ratio	Tobit estimated coefficient	t-ratio
Group income	1.85	2.12*	2.04	7.84*
Group size	29.71	2.37*	32.82	5.18*
Short circular trip	−72.50	−1.83**	−116.40	−2.34*
Short out and back	−52.54	−1.39	−75.96	−1.54
Day cycle	−35.84	−0.88	−7.53	−0.16
Short break	34.59	0.77	90.46	1.95
Longer linear cycle	134.69	3.04	193.01	4.23
Longer circular ride	94.48	2.06	154.89	3.20
Constant	−42.84	−0.55	−97.47	−2.22
R ^2	0.4720	Pseudo R ^2	0.0662
F(8, 364)	29.78	LR χ ^2 (8)	269.28
*P ≤ 0.05; **P ≤ 0.1.

Table 8  Expenditure Marginal Effects

Independent variable	Censored estimated coefficient	Z	Probability estimated coefficient	Z
Group income	1.00	2.29*	0.040	2.37*
Group size	16.08	2.53*	0.008	2.58*
Short circular trip	−48.95	−2.24*	−0.290	−1.90**
Short out and back	−33.63	−1.45	−0.190	−1.27
Day cycle	−3.66	−0.13	−0.020	−0.13
Short break	49.27	1.37	0.180	1.72
Longer linear cycle	114.51	2.78*	0.340	4.43*
Longer circular ride	93.50	2.13*	0.270	3.73*
*P ≤ 0.05; **P ≤ 0.1.

The first set of marginal effects support the qualitative predictions of both the OLS and raw Tobit results but suggest lower (absolute) impacts. The results can thus be thought of as a lower bound of expenditure. Significantly, all of the results suggest that longer trip types contribute most to expenditure relative to trips for a particular purpose, and that shorter trips reduce expenditure likewise. In the case of the lower bound estimates, an additional group member then generates approximately £16 more expenditure, whilst an increase in group real income by £1000 leads to an increase in spending of £1. In the upper bound estimates, the values would be approximately £33 and £2 respectively for increases in group size and real income. The results broadly suggest that longer trip types, and larger groups, are the key to generating higher expenditures. Significantly, the probability of being in a positive expenditure group is shown to be much greater for longer trip types than group size. The implications of these results are that first network supply-side opportunities are (naturally) very important in bringing about a positive economic impact from such cycle tourism activities. Second, that longer trips are the key to generating expenditure as well as group size and income. The latter results help to validate the model of expenditure, but also that there is clearly a need to target group sizes and higher income level cycle groups to shift them into spending categories and to elicit further spending.

Other important results worth noting are that neither the year or route characteristics are significant, nor is the tourism group variable. This suggests a relatively stable pattern of expenditure behaviour for recreational cycling regardless of the point of origin of the cyclist, and over time. This suggests that the underlying behaviours of the route users are similar. Thus, it is plausible that promotional activity to stimulate economic impact could be common across both intra and extra-network users.

Interestingly, the ‘actual hours of duration’ variable is insignificant as a determinant of expenditure. However, the results in Table 9 reveal that duration of the trip is shown to be significantly related to the trip characteristics, which are significant determinants of expenditure. This does suggest that both duration and monetary expenditures are directly linked to preferences for longer trips and, as such, adds support for the theoretical perspective noted above. The route characteristics are also shown to be jointly significant. As these characteristics were each measured as a dummy variable, the resulting equation can be thought of as a hedonic model which decomposes trip duration into its constituent parts and which can be compared in a relative sense. This is because the constant in the regression equation captures the value of duration independently of any of the characteristics and represents a hypothetical baseline trip of one and a half hours. Compared to this, therefore, trips on the Coasts and Castles route would be, on average over 12 hours longer. Likewise, longer circular rides are approximately 60 hours greater.

Table 9  Duration

Independent variable	Estimated coefficient	t-ratio
Coasts	12.66	3.04*
C2C	1.59	0.57
Hadrian	7.13	2.10*
Pennine	8.87	1.88
Short circular	−5.24	−1.36
Short our and back	−5.09	−1.41
Day cycle	−4.17	−1.18
Short break	8.40	2.14*
Longer linear cycle	34.83	5.20*
Longer circular cycle	59.53	5.16*
Constant	1.51	0.35
R ^2	0.3237
F(10, 362)  19.88
*P ≤ 0.05.

Figure 3 summarizes all of the results schematically. What this reveals is that both income and group size are key ubiquitous determinants of sports tourism expenditure as evidenced in relation to cycling. This is consistent with other literature. Moreover, the duration of trips and the actual expenditures that emerge are related, but through the preferences, that is motivations, of groups for different durations of trips. This is regardless of user segmentation and purpose of trip as suggested by others (Bull, 2006). This means that whilst the cyclist might commit both time and expenditure to their cycling activity, the former can influence the latter, but not vice versa and that cycling can be thought of as necessarily comprising the consumption of time, but not necessarily so expenditure; they are recursively related.

Figure 3 Expenditure and Duration.

This has implications for policy and planning; route usage can be planned more effectively with economic impact as a motivating factor. Eliciting economic impact requires focusing upon cycling packages that increase the duration of rides but also look to encourage larger groups and higher income segments. The variation of possible expenditure profiles, yet commonly validated model, also suggests that attention be paid to redeveloping these opportunities such that cyclists can spend more in community facilities such as cafes, restaurants and shops en route. For example 60-hour circular routes, or 30-hour linear packages could be developed rather than indeterminate lengths prescribed through opportunistic factors. Finally, the research indicates that these strategies could be common to both intra-network cycle tourists as well as extra-network cycle tourists, the latter coveted by regional development agencies to generate economic impact.

Conclusions

This paper has examined the relationship between the duration and monetary expenditures of cyclists using a network of routes in the North East of England. Drawing upon a statistically reliable sample it has been confirmed that both the incomes, group size and durations of activity are integrally linked determinants of expenditure, and thus have the potential to contribute to direct economic impact as suggested in both the early literature (Snepenger & Milner, 1990) and more recently (Dwyer et al., 2006). A theoretical innovation of the research, however, is that it has been shown, drawing upon economic theory, that there is an emphasis on the composite nature of goods, as involving both the inputs of time and monetary expenditure by consumers producing the goods that they consume, and that both the expenditures and durations of cycle trips are linked to preferences for longer trips. It is through the latter, in addition to group incomes and sizes that expenditure emerges.

In contrast, duration is not directly affected by expenditures or the particular route characteristics. The study concludes, therefore, that planning such tourism facilities such as multi-user trails for economic impact requires targeting those with preferences for longer trips and routes that facilitate this, taking into account the incomes of user segments and group size. Significantly, for networks such as in this study, the research findings infer that extra-network and intra-network tourism groups do not behave differently and consequently could be targeted by extensions of the same strategies, thereby increasing route usage and the related benefits of cycling and economic impact.

On a final note, the paper offers an understanding of the economic behaviour of participants involved in the soft definition of sports tourism (Gammon & Robinson, 2003). This is a departure from the previous sports tourism literature, which has focused principally on the economic impacts accruing from a sports event or events. As Bull (2006) notes, there is a need to expand our research approaches to evaluate the economic dimensions of the sports tourist experience rather than simply concentrating on behaviour. This will be the likely direction of future research.

Notes

Ontologically one is recognising a distinction between the transcendental features of the act of riding a (the same) bike from that act being undertaken for different purposes and in different contexts that both condition and define experiences.

The economic impact literature consistently shows that overnight stays generate disproportionately more expenditure than day visitors. The implication is that expenditure is connected with the duration of tourism activity, with accommodation counted as part of that activity. This study addresses this issue in a slightly different sense by focusing on the time actually engaged in a specific tourism activity.

The intrinsic relationship between the duration of tourism visitation and expenditure is implied in travel cost analyses of the economic valuation of visitor attractions. Pioneered by Knetsch (1963) and Clawson & Knetsch (1966) distance of travel, which has a cost associated with it, is used to proxy the price paid, for example, in visiting a recreational location. The point that is made here is that this is a special case of demand in which only the time element is considered. The framework provided by Becker is more general.

The variable is a latent variable because the activity is created by the act of consumption out of the employment of bought goods and equipment and application of time and endeavour.

It should be noted that this included genuine zero expenditure responses by groups, a matter that is discussed further in the next section.

We are grateful to an anonymous referee for suggesting this analysis.

The ‘one-sample’ test was undertaken with the proportions of diaries associated with the total issued acting as the population. The two-sample test could not be undertaken even though the issued diaries were a sample, because the two samples would be dependent. In each case, with Sunderland being the lower boundary, with n = 30, the sample sizes were ≥30. This is necessary as the test is a binomial approximation to the normal distribution.

This is tacitly acknowledged in the aim of the paper in trying to test a particular theoretical model of user expenditures and the sense of regression analysis which, in the absence of experimental conditions, can help to weight insights about the total sample according to what is actually observed in specific sub-samples connected with, for example, the year of sampling, the route or type of user etc.

It is important to recognize that these were not missing values but genuine zero expenditures. In a Tobit analysis, the coefficients are estimated subject to a probability that non-zero values of the dependent variable are observed. They do not, therefore, just identify an effect of a change in an independent variable on the dependent variable for cases that are not censored, i.e. above zero pounds sterling in this case.

In part, these results help to justify the estimators chosen. If simultaneity had been identified between duration and expenditure then this would have introduced bias that would have required correcting for.