Both the Health Spa industry and its non-profit competitors (YMCA, country clubs, etc.) sell their services, almost without exception, in the form of a non-cancellable, non-refundable membership agreement, by which the purchaser acquires the right to unlimited use of the ordinary services of the facility for some substantial period of time, usually a year or more. This practice has recently been criticized by the Federal Trade Commission; the FTC has proposed to require the industry (but not its non-profit competitors) to offer customers who wish to cancel a pro-rata refund (minus a small cancellation fee) for the unused portion of their membership. The purpose of this paper is to try to explain the current practice; one conclusion that follows from the explanations I will offer is that the proposed regulation is, from an efficiency standpoint, undesirable.
The puzzle is a simple one. If customers who cancel are given a refund equal to the marginal cost of the visits to the spa which they would otherwise make, some will choose to cancel who otherwise, faced with a zero marginal cost for continuing to use the spa, would have continued to consume a service worth less to them than it costs to produce. The resulting efficiency gains could be divided between customers and firm by increasing the membership fee by an amount less than the value of the cancellation option to the customers but more than its cost to the firm. It should, therefore, pay the firm to offer only contracts with cancellation options. The observed pattern is almost the exact opposite .
The explanation offered by the commission is that the industry uses extensive advertising and high-pressure sales tactics to persuade the customers that the services offered by a spa are much more attractive than they in fact are; non-refundable contracts are then necessary to prevent defrauded customers from getting their money back. The commission has also suggested that in some cases the spa deliverately provides low quality service in order to discourage use and so permit a larger membership for a given capital plant. Both sets of allegations have, of course, been denied and extensively rebutted by spokesmen for the industry. For my present purposes it is sufficient to note two difficulties with the commission's argument. First, although some spas might act in the way the comnission suggests, unless the customers of health spas are extraordinarily irrational it would then pay other spas to offer higher quality services at a higher price, guaranteeing their bona-fides by offering a partial or complete refund to dissatisfied customers. Second, the nonprofit providers of similar services, who in general do not employ extensive advertising or salesmen, sell their services in precisely the same way as does the health spa industry - that is to say, in the form of non-refundable memberships. So the commission's allegations, whether true or false, do not provide an explanation for the observed fact that services of this sort are, almost without exception, sold in this form.
The explanation offered an behalf of the industry is that most of the costs associated with a new member occur within the first month of his membership. While some of the calculations offered are suspect (they generally ignore the capital cost of the facilities, which is surely one of the main costs that ought to be spread out over the period of membership) the conclusion is supported by the observed pattern of rates charged; memberships for long periods cost substantially less per month than do memberships for short periods, and the per month cost of a renewal membership is generally about a quarter of the per month cost of a one year initial membership. While this explains why the industry objects to offering a pro-rata refund on cancelled contracts it does not explain the absence of cancellation provisions in which the amount refunded corresponds to the cost to the firm of providing services for the remainder of the contract period.
There are two alternative explanations which do explain the existence of non-cancellable contracts. The first and simpler is that the customers prefer them (more precisely, that the expected value of a cancellation provision to the customer, although it may be positive, is less than its cost to the firm) because the customers believe that the product of the firm is something that they ought to buy, but if given a continuing choice will not. As Gordon Tullock has shown, it is possible, provided that one's internal discount function is not of the simple e-rt form usually assumed by economists, to rationally wish to commit oneself at one time in such a way as to limit one's choices at some future time. A familiar example is the Christmas Club offered by many banks; customers commit themselves (not in this case irrevocably) to be forced by the bank to save a certain amount each month, and they in effect pay the bank for the service of reducing their alternatives. Similarly, many of the health spa's customers may believe that they "ought" to exercise, get healthy, etc., but that, given a choice, they will not stick to an exercise program. By buying a membership they commit themselves. They have paid their money and have to get something for it - or as an economist, recognizing that sunk costs are sunk costs, would put it, by signing a non-refundable contract they have irrevocably reduced the cost to them of using the spa.
A second and more interesting explanation is also possible. Assume that it is impractical (presumably because of monitoring costs) for the health spa to sell its services on a "per visit" basis. This is plausible, first because in fact neither health spas nor their non-profit competi- tors are observed to sell their services in that form, and second because the marginal cost per visit appears, judging by rates charged on renewal memberships, to be very low (on the order of two dollars a visit), so that even relatively small monitoring costs might well make monitoring inefficient.
Another explanation, which I have been given by at least one person in the industry, is that customers are over-optimistic about how often they will use the spa; the firm knows that, on average, they will use it less often than they expect. This would explain, without invoking monitoring costs, why the spa cannot sell its services on a per-visit basis; at what the firm regards as a fair price, given its probabilities, an unlimited membership will always appear more attractive to the customer.
This is industry's version of the Federal Trade Commission's explanation, the difference being that the blame is put on the over-optimistic customer instead of the salesman. It depends on an assumption - consistent error - of which economists must always be suspicious. One could perhaps argue that even if customers do not consistently overestimate how much they will use the spa's services, a spa which offers only contracts for unlimited usage will only get those customers who do. It will be misled into believing that they represent the entire market, and so continuing its policy. While this argument might apply to some firms, it is hard to believe that it would apply to all, especially since those firms which did behave that way would be leaving a large market of nonoptimist customers unsatisfied and so providing opportunities for their wiser competitors.
In any case, assume that, for one reason or another, per visit sales are impractical. Assume also that there is considerable variation in the amount of use of the health spa by different members. If the potential members themselves know how much they will use the spa, the industry is faced with a problem precisely analogous to adverse selection in health insurance. Since the firm cannot distinguish in advance those customers who will make extensive use of its facilities from those who will not, and since the assumed monitoring costs prevent it from charging differential prices to customers according to how much they actually do use the facilities, it must sell the right to use its services at the same price to different customers - a price corresponding to the cost of the average level of usage of the facilities. To those customers who expect to be heavy users this price is attractive, to those who expect to be light users it is not. The light users do not become customers. The average level of usage is then higher than the average level which would have applied for the potential customers as a whole (since only the heavier users become actual customers) which makes the price even higher, which discourages more of the light users. In an extreme case the process can be carried to the point of almost complete market failure, with only a single customer (the one expecting to be the heaviest user) choosing to buy the service. More generally some equilibrium point is reached at which some of the lighter users are excluded from the service despite the fact that the value of the service to them is larger than the cost of providing it to the firm. The outcome is therefore inefficient.
Adverse selection for health insurance can be prevented by selling group policies to groups which have been selected on some basis other than their belief about their future health. An analogous solution exists in the health spa case but it depends on one further assumption - that the customers only discover whether they are going to be heavy or light users of the facilities after some period of use. The adverse selection problem is a product of differential knowledge - specifically the buyer's informational advantage over the seller (the symmetric problem is discussed in Ackerloff (1970)). By making a binding decision at a time when he does not have differential knowledge the customer can elimi- nate adverse selection and receive as his reward some part of the resulting efficiency gains
The argument may be made clearer by a numerical example. Suppose there is a customer who believes, correctly, that he has a fifty percent chance of deciding to use the spa once a week and a fifty percent chance of deciding to use it three times a week. Further, suppose that each time he uses it he receives a benefit which he values at $11 and that each use costs the firm $9. I am here abstracting away from the actual and much more complicated pattern of costs incurred by the firm; I am also ignoring the possible efficiency losses from overuse of the spa, given that the marginal cost to the customer is zero.
The spa offers the customer a non-cancellable contract providing unlimited use of the spa at $20/week for a year. The customer accepts; his expected consumer surplus is $2/week and the expected producer's surplus is the same. Ex post he either loses $9/week or gains $13/week; the firm either gains $Ii/week or loses $7/week.
Suppose now that the contract is made cancellable. After a short while the customer who uses the spa only once a week cancels. Knowing that this will happen, the firm adjusts its prices upwards to $30/week. Those customers who turn out to be heavy users receive a surplus of $3/week; those who are light users cancel and receive no surplus. The ex ante expectation of a new customer under these circumstances is a consumer surplus of $1.50/week, as is the ex ante expectation for the producer surplus of the firm. There is thus a net loss as a result of the substitution of a cancellable for a non-cancellable contract.
A different way of putting the same argument is to observe that the firm is offering the customer two contracts plus a bet. The first contract is to provide use of the spa once a week for $10; the second is to provide use three times a week for $30; the bet is a ten dollar bet that the customer will use the spa only once a week. Both of the contracts yield net benefits to both buyer and seller; the bet is a fair one and hence, assuming risk neutrality, costs nothing. Since in either eventuality the customer owes the firm $20/week the firm does not have to distinguish between heavy users and light users -- which by assumption it cannot do. Given that it cannot distinguish, such a combined offer is the only way in which it can sell the first of the two contracts; if it is compelled to offer only refundable memberships the combined offer becomes impossible since the customer is, in effect, in a position to reneg on the bet if he loses) and the first contract cannot be sold.
I have offered two explanations for why non-refundable memberships are offered. I have not explained a closely related feature of industry practice, the failure of most firms to offer higher priced refundable memberships as an alternative choice. While a few firms offer a short membership with an option to renew, corresponding in effect to a longer cancellable membership, most do not. Within the context of the arguments I have offered, the obvious explanation is that the advantages of the non-cancellable membership are sufficiently large to make cancellable memberships, at their equilibrium price, unattractive to most customers; it therefore does not pay most firms to offer them. While my explanation is consistent with the observed scarcity of short term or cancellable memberships it does not predict it; I have as yet been unable to come up with any alternative explanation that does.
I have provided explanations ofwhy non-cancelable contracts for providing continuing service might exist and be efficient; I have not shown that those are in fact the reason such contracts do exist in the health spa industry. Further research might involve comparing the prices of short and long term contracts where both are offered and examining the pattern of use by customers in the health spa industries and the practices of firms in similar industries (tennis clubs, for example) where a considerable amount of monitoring is necessary for other reasons (reserving courts) and where additional monitoring may therefore be very inexpensive.
Akerlof, G. A. "The Market For Lemons: Qualitative Uncertainty and the Market Mechanism", Quarterly Journal of Economics, Vol. 83, No. 3,
(1970), pp. 488-500.
Association of Physical Fitness Centers, "Fact Sheet," (1975)
Benston, George J., "Submission To The Federal Trade Commission On The Impact Of The Trade Regulation Rule Relating To The Health Spa Industry, Proposed and Published By The Federal Trade Conmission On August 18, 1976, In the Federal Register" University of Rochester, Rochester, NY November 7, 1977.
Bond, Frank, "In the Matter of Proposed Trade Regulation Rule Concerning Health Spas," Transcript of Testimony given on December 14, 1977 at the FTC Public Hearing No. 215-50 at 26 Federal Plaza, New York.
Federal Trade Commission Staff (?), "Health Spas: An Economic Analysis of Refund Policy Alternatives." Draft: November 7, 1977.
Sachs, Reynold M., "An Economic and Financial Analysis of the Impact of the Federal Trade Commissions Proposed Trade Regulation Rule Relating To The Health Spa Industry On Consumers And The Industry", Washington, D. C., November 7, 1977.
Tullock, Gordon, The Logic of the Law, New York: Basic Books Incorporated, 1970.
 I should state, for the benefit of those readers who believe in the importance of author bias, that I encountered this puzzle while serving as a member of the board of directors of a firm that owns and operates health spas.
 I will use the term "health spa" in the sense suggested by the Association of Physical Fitness Centers: "The modern health spa is usually housed in a permanent facility and offers programs of physical fitness for both sexes and trained personnel for the purpose of instruction. Addi- tionally, it offers swimming pools, whirlpools, saunas, steam rooms, and massage services. It is not uncomnon to find facilities for tennis, miniature golf, food and beverage facilities, and health bars, and regu- larly scheduled classes in exercises martial arts, and other skills related to physical activities. In some larger spas, a wide range of other social activities for members including dances, parties, and classes in various skills, may also be provided. Some spas offer retail services including health foods, gym equipment, and other exercise-related material. APFC (1975, pp. 1-2).
 In some cases there is an additional charge for the use of special services -- massage, for example. Cancellation of a contract for some specific reason, such as medical reasons or a move to an area where no facility is available, is generally permitted.
 From the tables in Sachs (1977) it appears that of the twenty firms he surveyed only four offered memberships shorter than 12 months and all of those also offered 12 month contracts, 24 month contracts, or both. According to the APFC "Fact Sheet" less than 5% of industry revenues come from contracts of less than 12 months term.
 The Initial Notice proposing the regulation was published on August 18, 1976 in the Federal Register (40 FR 34615): the Final Notice was published on May 24, 1977. The proposed cancellation fee is 5%.
 The only explicit economic analysis of the efficiency gains or losses of moving to cancellable contracts from the conmission's standpoint which I have been able to find is a document entitled "Health Spas: An Economic Analysis of Refund Policy Alternatives" which was obtained from the com- mission under the freedom of information act. It is described as a draft and dated 11-7-77; no author is given. The author assumes throughout that although refundable contracts would be more attractive to customers than the current non-refundable contracts, they would be sold at the same price and at that price demand for them would be the same! Hence, although he recognizes the obvious efficiency argument in favor of refundable contracts, it does not occur to him that there is any inconsistency between this argument and the failure of firms to offer such contracts.
 Benston (1977), Sachs (1977), Bond (1977), APFC "FACT SHEET."
 Sachs (1977), Benston (1977)
 Sachs (1977) tables 29 and 30. A least squares fit to the average prices reported for 3 month, 12 month, 24 month and 36 month memberships yields the result: Total Price ~ $132 + (number of months) x $12. This is a very approximate estimate of the cost function, since there may be systematic relations between differences in the cost functions of different spas and different lengths of contracts offered. The average rate for re- newal is about $5/month (almost always for a 12 month period).
 Additional arguments apply to the special case where memberships are sold for a Spa that has not yet been completed. The firm, in de- ciding whether to complete the Spa, requires an estimate of future demand; sales of non-cancellable memberships will presumably give better information for this purpose than sales of cancellable memberships, since the latter may represent customers who have a high probability of cancelling. In addition, non-cancellable memberships under such circumstances repre- sent a device for spreading risk.
 Tullock (1971, pp. 22-28).
 This point is discussed briefly in Sachs (1977, pp. 77-78).
 This argument does not explain the absence of cancellable contracts; cancellation with refund of marginal cost costs the firm nothing and is worth something to the customer, however optimistic he may be.
 Akerlof (1970).