Partial refunds can be an attractive proposition for service providers. They induce early customers to give up a service they no longer need in exchange for a refund, and the firm can then sell the service again. Yet partial refunds are not uniformly adopted by service industries or even by providers within the same industry.

United Airlines, for instance, does not offer partial refunds even though most of its competitors do. This observation prompted Liang Guo of HKUST to investigate when partial refunds are effective, and how they affect the interaction between competing firms.

"Anecdotal evidence suggests partial refunds vary across providers generally characterised by limited capacity and buyer uncertainty. For example travel-related services usually provide them but other services such as athletic facilities and concerts clubs are normally resistant. I wanted to look at why some providers might offer refunds and others wouldn't," he said.

He developed a model to work out the optimal policy choices for two competing service providers. The providers had two types of customers ?early-arriving, low-valuation customers, and late-arriving, high-valuation ones. They could both offer zero refunds, both offer partial refunds, or one could offer zero and the other partial refunds.

The best choice depends on the capacity of the firms and the action of their rival. With zero refunds, the firms are in effect local monopolies due to their limited capacity: the reservations sold in the advance period cannot be re-sold and the firms cannot sell to all potential demand from late-arriving customers.

A partial refund policy, on the other hand, means the firms can use the cancelled service units to meet demand from late-arrivers and therefore have a higher payoff.

Interestingly, this may introduce competition because the two firms have to compete for advance shoppers.

"One major result here is that a partial refund policy may induce a firm to advance-sell more service units [in anticipation of cancellations], which may in turn intensify the price competition in the advance period," Guo said.

This is especially likely when there are a larger number of advance shoppers, as a firm may undercut its rival in anticipation of greater gains. However, that means intensified competition, which could ultimately dampen the enthusiasm for partial refunds.

"The results could explain why the insights of profitability in a monopoly setting that have been reported elsewhere, may not necessarily transfer to a competitive environment. Competition may in fact decrease firms?incentive to offer partial refunds," he said.

"The choice of refund policy hinges on the importance of an efficiency-enhancing versus a competition-enhancing effect. When capacity is sufficiently small, efficiency-enhancing dominates the choice and both firms in equilibrium follow identical partial refund policies. When it's sufficiently large and competition-enhancing dominates, they follow zero refund policies," Guo said.

"In an asymmetric equilibrium, neither effect is dominant. Here, it's profitable to exploit the efficiency-improving effect, but only when the rival doesn't follow suit, otherwise the competition would be too intense."

Guo suggested that the different refund policies between travel-related service and other services, such as concerts or health centres, might be due to the fact that buyers of the other services have options ?they can stay home or do a different activity. Meanwhile, the asymmetry among airline refund policies may be because there is some, but not much room to exploit the efficiency-increasing effect of partial refunds.

"The central insight for service managers is that offering a partial refund is a 'double-edged sword'. On the positive side, it encourages consumers to cancel unused reservations and allows more effective use of the limited service capacity. However, it may also intensify competition among service providers and hence nullify the efficiency gain. Therefore, in designing refund policies, managers should invest more scrutiny to investigate the consequent impact on the strategic interaction among rival firms," he said.