Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

Notes

Some Problems in Estimating the Demand for


Outdoor Recreation
Kenneth E. McConnell

Increased pressures to provide realistic, empirical to the recreation site is as important a cost variable
measures of value for outdoor recreation have gen- as the travel cost itself in determining the quantities
erated considerable research dealing with the de- demanded of outdoor recreation. As Brown and
mand for outdoor recreation. The thrust of this Nawas state, "Given the importance of increased
research is to provide nonmarket measures of the distance on the negative factors of travel time and
value of recreational facilities which are compara- alternative recreational opportunities, the inclusion
ble to market values for competing activities. Mea- of a separate variable, such as travel time per dis-
surement of the value of recreational facilities has tance zone, would appear to be needed" (p. 24).1
been beset by data shortcomings and theoretical The purpose of this section is to demonstrate that,

Downloaded from http://ajae.oxfordjournals.org/ at Yale University on June 29, 2015


problems, and, as indicated in an article by J. A. for estimating demand functions consistent with
Sinden recently published in this Journal, some of utility maximization, the researcher should con-
the perennial research problems remain. sider not simply the time in transit but the total time
The purpose of this paper is to derive recreation required for completing the recreation activity. The
demand functions consistent with utility maximiza- model developed here is not meant to describe real-
tion to address specific issues in empirical research. ity completely. It abstracts from many important
Outdoor recreation demand functions are used to determinants of the demand for recreation in order
make inferences about the consumers' surplus and, to analyze the impact of the constraint on time. For
implicitly, about the social welfare derived from this approach to be useful, recreationists must act
particular sites. Hence, it is important that the em- as if they are constrained by the time available for
pirical research be as consistent as is feasible with them to recreate.
the theoretical structure from which the concept of We begin with a simple statement of the general
consumers' surplus is derived. In discussing the problem. Suppose the consumer chooses among n
relationship between recreation demand functions recreation activities, rj, j = 1, ... , n, and a compos-
and consumers' surplus, Cesario and Knetschstate ite bundle of goods, x, so as to maximize the utility
that' 'whereas the assumptions of constant margi- from the consumption of goods and recreation per
nal utility of income over the relevant range and unit of time. The consumer's choices are con-
utility maximization are necessary, ... they gener- strained by his income and the total time available
ally do not pose particular practical difficulties" (p. to him. Let Utx , r) be the utility indicator with r
700). This paper will demonstrate explicitly the denoting an n-dimensional vector.
practical difficulties implied by the assumptions of We will examine the case where the consumer
utility maximization. can choose the amount of time he works and also
This paper considers three specific issues in out- the case where the time at work is fixed. Let w
door recreation demand: (a) the appropriate mea- represent the amount of time worked. To include
sure of the value of time as a cost of participating in both work and recreation in the same time con-
outdoor recreation, (b) the logical units of mea- straint, they must be converted to the same units.
surement for a recreation activity when estimating The quantity variable for the recreation activity, r.,
demand functions via the travel cost method, and is measured in the number of trips of a specified
(c) the appropriate functional form of the outdoor length. Let aj represent the number of units of time
recreation demand function, particularly the rela- required to participate in one unit of rj. Hence, a.r,
tionship between the price and income variables. is the total time required to consume rj trips of
outdoor recreation activity.
When the consumer can choose the amount of
The Value of Time in the Demand for time to work, the time constraint is
Outdoor Recreation n

Since the work of Cesario and Knetsch, it has been


(I) L a.r, + w = T,
j=1

widely accepted that the time required in traveling where T is the total time available. If the consumer
Kenneth E. McConnell is an assistant professor in the Depart-
ment of Resource Economics at the University of Rhode Island. 1 The work of Gary Becker suggests that the total time spent in

Contribution Number 1590 of the Rhode Island Agricultural an activity is the appropriate measure of the opportunity cost of
Experiment Station. R. Cummings, V. Duff, two anonymous time. Empirical work in recreation, to my knowledge, has been
reviewers, and the editor have contributed helpful comments. limited to measuring the opportunity cost of travel time only.
330
McConnell Estimating Recreation Demand 331

cannot choose the amount of time he works, then that m» /-LIX. Here m- has the dimensions
the time constraint is ~u~ility / ~~tility or (f1incomeldtime). It fol-
n time Income
(2) Ij=l
a.r, = T*,
lows that mT can be interpreted as the opportunity
cost of scarce recreation time, measured in dollars
of income. It is the equilibrium value of an addi-
where T* is the fixed amount of recreation time tional unit of recreation time in any recreational
available. In each constraint, a, is the amount of activity. Equation (7) can be written
time required for one unit of rj. For example, sup-
pose that T is measured in days. If rj is the number (8) aUlarj = X(cj + ajmr).
of day visits to the beach, a, = 1. If rk is the number It is useful to compare equation (8) with equation
of three-day camping trips, ak = 3. (5). According to equation (8), the cost of a unit of
The consumer must spend within the constraints thejth recreation activity is travel costs (Cj) plus the
implied by the equation, scarcity value of time (ajfflT) in income. It is impor-
(3) F(w) = px + !
j=l
Cjrj,
tant to note that even in the case where the con-
sumer foregoes zero earnings, he still considers the
scarcity value of time as part of the cost of the

Downloaded from http://ajae.oxfordjournals.org/ at Yale University on June 29, 2015


where p is the price of the composite good, C j is the recreation activity because the time could have
net variable costs, including travel and transfer been used in other recreational activities.
costs, of the jth activity, and F(w) is the income The implications of this section can be stated in a
generated by working w units of time. Solving equa- general way. Let m» be the marginal monetary
tion (1) for time worked and substituting in equation value of an extra unit of time, be it earnings, F'(w),
(3), we can write the constrained maximand, or simply the dollar value of an extra unit of recrea-
n tion time. Then the demand for outdoor recreation
(4) U(x, r) - X(px + I Cjrj
can be written
j=l
n
(9) rj = Jj(Cj + ajmT, y).
- F(T - I
j=l
ajrj» , Here y is money income and c, + ajmT is the unit
cost of the experience, where the unit is a trip.
,where the consumer chooses the quantities of Because ajmr is included in the measure of unit
recreation activities and the composite good. costs, it is clear that understanding the opportunity
Maximizing equation (4) with respect to rj yields cost of total time is important for accurate mea-
sures of the economic value of outdoor recreation.
(5) aUlarj = X(Cj + ajF'(w», Equation (9) has important implications for em-
where c, is the travel and transfer cost of a trip, pirical research in outdoor recreation. With regard
F'(w) is the marginal earnings ofa unit of work, and to survey questionnaires, it suggests the following
a, is the number of units of work that the consumer questions.
must forego in order to enjoy one trip of recreation (a) Could the recreationists have worked for pay
activity, rj. Hence, a.F'tw) is the earnings foregone during the period of the recreation visit?
by enjoying a trip of recreation activity, r.. It is the i) If no, proceed to question (b); F'(w) = O.
earnings over the whole recreation period, as indi- ii) If yes, how much could have been
cated by a., the number of periods of work required earned?
by the recreation trip and not just the time in tran- (b) What might the recreationist have done if not
sit. The important conclusion is that the cost of on this trip?
consuming a unit (trip) of an outdoor recreation i) If nothing, then /-L = 0 and mr = O.
activity is the sum of travel costs and the value of ii) If something, attempt to determine wil-
earnings foregone over the entire trip. lingness to pay for second most favorable
Suppose that the recreationist maximizes utility activity.
subject to a fixed amount of recreation time, as Significant responses to this series of questions
given by equation (2), and a fixed income, denoted would indicate that recreationists act as if their time
by y. Instead of equation (4), the maximand will be were scarce. However, it is possible to obtain zero
value for the opportunity cost of time for some
(6) U(x, r) - X(x . p + !
j=l
Cjrj - y)
people (i.e., m- = 0), because it is probable that a
zero value for mr indicates some persons have no
alternative source of earnings and will not use the
time for other recreation activities.
Equation (9) also has important implications for
the specification of the demand function. The tradi-
The first order condition with respect to rj is
tional approach has been to estimate the parame-
(7) aUlarj = XCj + JUlj. ters of the function:
The multiplier, /-L, is the marginal utility of (10) number of trips
recreation time. Let us define a variable mr such = f(travel costs, income, distance).
332 May 1975 Amer. J. Agr. Econ.

The distance variable in equation (10) has been from the site. For models of other functional forms
used as a proxy for time in transit. Equation (10) (i.e., nonlinear), it is not possible to determine, on
has been difficult to estimate satisfactorily, how- an a priori basis, the effect of the omission of the
ever, because travel and transfer costs are often opportunity cost of time on the estimated slope of
almost a linear transformation of distance. The re- the demand function and the value of consumers'
sults of this section suggest that the researcher surplus.
should make estimates of the value of total time The importance of this section is that it suggests
consumed on the recreation trip, add this value to that the recreationist may consider the cost of the
the travel and transfer costs incurred, and estimate total time spent on the recreation activity as part of
the parameters of the relation. A distance variable the cost. This conclusion is at variance with the
can be included in equation (9) to account for the accepted role of time cost in a recreation activity,
different relative travel costs of substitutes faced by which has been limited to the cost of time in travel-
recreationists from different distance zones and to ing to the site. The inclusion of the additional value
allow for the filtering affect of distance on informa- of time changes the structure of the estimating
tion about a site. We can now write: equation as well as the nature of the information to
be gathered by the researcher.
(II) number of trips
= f(travel costs + time costs, income, distance).

Downloaded from http://ajae.oxfordjournals.org/ at Yale University on June 29, 2015


It should be observed that the distance variable in
equation (II) has a substantially different interpre- Units of Measurement for the Travel Cost Method
tation from distance in equation (10).
We can conclude from this analysis that using the Researchers in the field of recreation economics
travel cost method on a given set of observations, have used several quantity measures in estimating
while ignoring the opportunity cost of time, under- travel cost demand functions. The most common
estimates the marginal value of the recreation. units are trips, user-days, and user-hours. For
Something can also be said about the estimates of example, in Sinden's work, the quantity of outdoor
the consumers' surplus when equation (10) rather recreation demanded is "the number of hours
than equation (II) is estimated. Suppose the model of a specific recreation activity . . . per family
estimated is r = Clbl + Db 2 + vb«. when the true member during the 12 months prior to the inter-
relationship is r = (Cl + C2){31 + D{32 + y{33 + u , view" (p. 62). For Burt and Brewer, "quantities of
where C 1 = travel costs, C2 = time costs, D = outdoor recreation services are measured in units
distance, r = recreation trips, y = income, u = of visitor days" (p. 826). Knetsch uses number of
random variable, b, = least-squares estimators of visits.
{3r{3j = true coefficients. Under fairly reasonable Basically, the two measures of recreation de-
assumptions, it can be shown that Eb 1 < {31. 2 manded are trips and user-days, user-days being
Thus, on the average, the slope of the quantity of merely a linear transformation of user-hours." De-
recreation consumed with respect to the pricevari- riving the demand functions from a utility maximi-
able is underestimated when the opportunity cost of zation framework demonstrates that user-days are
total time is ignored. not an appropriate measure for the quantity of
For the linear per capita demand function, un- recreation demanded when using the travel cost
derestimating the slope with respect to the price method.
variable results in an underestimate of consumers' Utility maximization implies that
surplus." Hence, we can infer that a demand curve (12) .:lutility/.:lquantity
derived from the travel cost method which ignores (.:lcosts)
the opportunity cost of time will result in a down- = marginal utility of income
(.:lquantity) .
ward biased estimate of the consumers' surplus
Researchers implicitly use a first-order condition
2 The bias from omitting C2 is a form of specification bias. (See like equation (12) to solve for quantity as a function
Ramsey.) It can be shown that PI - Eb, = -[j3I/oo2] (covtc., C2) of marginal costs. It is clear, however, that in many
var b i + cov(D, C2) cov(bl> b2 ) + cov(y, C2) cov(b l , ba», where 002
is the variance of the error term. A priori income and time costs
cases the marginal cost of a user-day is independent
are likely to be highly correlated, while there is no reason to of the level of travel costs, so that specifying the
believe that COV(CI, C2) and COV(C2, D) are non-negative. If these demand for user-days as a function of travel costs
relations hold, the bias will be positive. If everyone's time is has no basis in theory.
valued the same, C2 is a constant and the coefficients are not
biased. If the correct specification is nonlinear as implied in a A simple example will illustrate the difficulties of
subsequent section, the importance of the bias from a linear model analyzing user-days with the travel cost technique.
is reduced. The costs of recreating can be divided into fixed
a If Eb, < PI> then lIPI < lIEb l • Since lIPI is the slope of the costs per trip (travel costs) and net variable costs
demand curve with the price on the vertical axis, quantity on the
horizontal axis, and lIEb l its estimator, we can say that the true 4 The variable user-days here is meant to include not only days
curve is more inelastic than the estimated curve. Consumers' on the site but also days consumed getting to the site. Researchers
surplus is derived by raising the price from zero until quantity have not been consistent in this respect. Burt and Brewer include
becomes zero. Both curves begin at the same point on the quantity time consumed in reaching the site in their analysis of user-days
axis because the quantity taken at a zero price is known. Hence, whereas Sinden's measure of user-hours includes only time at the
the more inelastic curve yields the greater consumers' surplus. site.
McConnell Estimating Recreation Demand 333

per user-day times the quantity of user-days. Sup- impact of income changes on the demand for out-
pose that, for a particular recreationist, fixed costs door recreation. An issue recently addressed by J.
per trip are $10 and net variable costs per day are A. Sinden concerns the independence of the price
$5. It seems clear that the recreationist will adjust and income slopes in the estimation of the demand
his user-days in such a way that the marginal value model. Sinden compares the estimation of the fol-
of a user-day is $5, as long as total conditions are lowing equations:
satisfied. The marginal cost of a user-day is inde-
pendent of the travel costs, once the recreationist (13)
has undertaken the trip. Hence, to be consistent and
with utility maximization, the demand for user-days
should be estimated as a function of net variable (14) r, = f30 + f31Cj for each income group.
cost per day including time costs. Here rj is the number of visits, y is income, and c,
The inconsistency of user-days with the travel is a measure of costs, including time costs, of a
cost method would not be a matter of concern if visit. Referring to formulations exactly analogous
researchers did not use the resulting functions to to equations (13) and (14), Sinden states, "Despite
derive measures of consumers' surplus. The com- a review of pertinent literature, no a priori reason-
putation of consumers' surplus requires knowledge ing could be developed to suggest that one equation

Downloaded from http://ajae.oxfordjournals.org/ at Yale University on June 29, 2015


of the per capita demand functions. The demand would be more relevant than the other" (p. 65). If
function is essentially a relationship between per one assumes that outdoor recreation demand func-
unit cost and quantity demanded. The relationship tions have been derived from utility maximization,
between travel costs and user-days cannot truly be a considerable amount can be said about the a priori
called a demand curve, even though it is possible to specification of the demand functions.
estimate such a relationship statistically. 5 The specific issue addressed by Sinden concerns
It is difficult to determine on an a priori basis the influence of income on the slope of the demand
what impact the inconsistency between user-day for recreation with respect to the price. This
and the travel cost method would have on estimates influence depends upon the functional form as-
of consumers' surplus because it is questionable sumed. Since estimates of consumers' surplus vary
whether the quantity estimated is really consumers' greatly with the functional form of the demand
surplus. It seems likely that the more meaningful curve, the issue is of more than passing interest.
relationship between trips and travel costs is more The difference between Sinden's specifications,
inelastic than the relationship between user-days given by equations (13) and (14) above, is that for
and travel costs because recreationists from the equation (13), the price slope is independent of the
further distance zones are likely to reduce their trip level of income (i.e., a2r)ac jay = 0). For equation
numbers and expand their user-days per trip. (14), the price slope depends upon the level of in-
User-days are clearly important for many pur- come (a2r jlacjay =J 0). Equation (13) is a special case
poses. The capacity of a recreation site is best of equation (14).
defined in terms of user-days per day. If one had If one derives the income slope in a utility
good observations on net variable costs per day, maximizing framework, it can be shown that
including time costs which vary for individuals, it n
might be possible to estimate the demand for user-
days. (15) arjlay =.I
k=l
Ujk-lCklc'U-lC,
The results of this section imply that the trip or
visit is the most logical measure of quantity con- where U jk- 1 are the elements of U- 1, the inverse of
sumed in the travel cost method of estimating the the matrix of the second-order derivatives of the
demand for recreation. Analysis with user-hours or utility functions, and C is the vector of all costs and
user-days may be appropriate in many cases. How- prices faced by the recreationist. 6 It is clear from
ever, it does not appear logically correct to make equation (15) that the income slope depends upon
inferences about consumers' surplus under the as- the costs (c) for all specifications of the utility func-
sumption that travel costs are the marginal costs of tion. Hence, a2rjlacjay =J 0 and by the symmetry of
a user-day or a user-hour.
6 Assume that CJ includes the travel and transfer costs and time
costs per unit of thejth activity and that those costs are constant.
A Priori Specification of Outdoor Recreation For simplicity of notation, we can drop the composite good, x ,
from the analysis. Then differentiating the first-order conditions,
Demand Functions equation (5), with respect to income gives

Since David Seckler originally raised the issue, f


L... U Jk -
ark
-
ax.
cJ - = 0 j = 1, n
there has been considerable discussion about the k-l ay iJy

and
5 The relationship can be estimated statistically because raising
travel costs reduces the number of trips. With fewer trips, the
number of user-days declines also because recreation time for
most people occurs only with days off from work, weekends, and
vacations. or in matrix form, where UJk is the matrix of second-order partials
334 May 1975 Amer. J. Agr. Econ.

cross partials, a2rj/ayacj I- O. The price slope de- ical issues of research in outdoor recreation
pends upon the level of income. We may conclude economics. The three basic conclusions of this
that the more general specification, equation (14), is paper are (a) that the appropriate time variable in
more in keeping with utility maximization. Demand the demand for outdoor recreation is the value of
analysis which has as its goal inferences about con- the total time consumed by the recreation activity
sumers' surplus should deal with demand functions rather than simply the time in transit, (b) that the
which permit the price slope to vary with the in- unit of measurement consistent with the travel cost
come level. method is the trip or visit and not user-days, and (c)
The attempt to make the functional form of em- that there is a priori reason to specify the demand
pirical demand functions consistent with utility function such that the price slope can change as
maximization has important implications for policy. income changes.
First, for any given distribution of income, a differ-
ent functional form for the demand curve implies a [Received September 1974; revtston accepted
different value for the consumers' surplus. Second, January 1975.]
as Stoevner and Brown pointed out for a particular
case (p. 152), some nonlinear specifications imply
that estimates of consumers' surplus change as the

Downloaded from http://ajae.oxfordjournals.org/ at Yale University on June 29, 2015


distribution of income changes. Keeping demand References
functions consistent with utility maximization will
give analysts a common point of departure. Becker, Gary. "A Theory of the Allocation of Time."
In sum, there is a priori reason to specify the
Econ. J. 75 (1965):493-517.
econometric structure of a demand model in such a
Brown, William G., and Farid Nawas. "Impact of
way that the price coefficient may vary as income
Aggregation on the Estimation of Outdoor Recrea-
varies. The a priori specification implied by utility tion Demand Functions." Amer. J. Agr. Econ. 55
maximization precludes additive functional forms. (1973):246-49.
The result is important because it permits analysts Burt, Oscar R., and Durward Brewer. "Estimation of
to determine which demand functions are not con-
Net Social Benefits from Outdoor Recreation."
sistent with utility maximization, instead of assum-
Econometrica 39 (1971):813-27.
ing consistency in all cases. Cesario, Frank J., and Jack L. Knetsch. "Time Bias in
This paper has developed an analysis of the de- Recreation Benefit Estimates." Water Resour. Res.
mand for outdoor recreation in a utility maximiza- 6 (1970):700-704.
tion framework for the purpose of analyzing empir- Knetsch, Jack L. "Economics of Including Recreation as
a Purpose of Eastern Water Projects," J. Farm
of the utility function and arjay is the vector of income slopes, Econ. 46 (1964):1148-57.
Ramsey, James B. "Classical Model Selection through
Specification Error Tests." Frontiers in Econome-
trics, ed. PaulZarernbka, pp. 13-48. N.Y.: Academic
Press, 1974.
By the rules for inverting a portioned matrix, it can then be shown Seckler, David W. "On the Uses and Abuses of
that Economic Science in Evaluating Public Outdoor
Recreation." Land Econ. 42 (1966):485-94.
Sinden, J. A. "A Utility Approach to the Valuation of
where A is a matrix occupying the first n by n places of the inverse Recreational and Aesthetic Experiences." Amer. J.
and -U-1C(C'U-1C)-1 is a column vector occupying the n + 1 Agr. Econ. 56 (1974):61-72.
column of the inverse. Solving for any particular ar/iJy yields the
expression given in the text. This result implies that a demand
Stoevner, H. H., and W. G. Brown. "Analytical Issues
function of the form r = g(c) + h(y), when gO and h(') are in Demand Analysis for Outdoor Recreation: Re-
arbitrary functions, is not consistent with utility maximization. ply." Amer. J. Agr. Econ. 50 (1968):151-53.

You might also like