Imperfect competition and quality signalling.

Second, one could posit that quality is a firm-wide attribute and that firms produce other products that benefit from high quality. This is the approach taken by Daughety and Reinganum (2005), who suggest that, for example, a technology could be used to produce both branded therapeutic drugs and generic multivitamins which are sold in bulk to private-label distributors. A "better" technology would improve safety when used to produce therapeutic drugs, and would yield higher output when used to produce generic multivitamins. In this case, although high quality is disadvantageous in the market for the product with the safety attribute, it can still be advantageous for the firm overall.

Third, the model we are exploring might concern a new product whose quality is initially uncertain but will eventually become common knowledge. Thus, one can view this model as pertaining to an "introductory period," and augment the model with a "long-run" period in which quality is common knowledge. This introductory period need not be very brief, if consumers buy relatively infrequently, arrive sequentially, and do not observe past prices; this is the approach used by Bagwell and Riordan (1991) to explain the phenomenon of high and declining prices over a product's life cycle. Although high quality is disadvantageous in the introductory period, it is advantageous in the long run.

Fourth, one could view this model as part of a stationary equilibrium, in which firms draw their technologies (or labor forces) anew each period. There is a common basic observable quality level (which is reflected in the demand curve through, for example, the maximum willingness to pay, [alpha]). In addition, there is a random (uncontrollable) shock each period, which affects both quality and costs. Thus, in each period, each firm has a chance [lambda] of being an H-type firm and a chance 1 - [lambda] of being an L-type firm. By Proposition 4, in high-value markets with a sufficiently high proportion of H-type firms, both H-type firms and L-type firms make higher profits with this stochastic technology with private information about quality as compared to the profits that would arise if all firms were known to have low quality. Thus, the firms enjoy a collective benefit from this stochastic technology (though a firm that draws an H-type technology in a given period suffers a temporary disadvantage). Even though there might be an individual temptation to invest in R&D (or relocate) to lower one's own frequency of being the H-type, if this kind of investment (or its consequence) is detectable by rivals, then it might be deterred by the anticipation that rival firms would follow suit and all firms would end up (worse off) in the equilibrium with known low quality.

As mentioned above, our analysis takes the number of firms as given. If the firm is a multiproduct firm, or if entry is limited by rent-generating attributes such as "know-how" or patented technologies, then the number of firms operating in this particular industry is not determined by a free-entry condition, and may be taken as exogenous. Alternatively, suppose that firms can enter upon payment of an entry cost, denoted F, before learning the quality of their product. If all firms in the industry obtain positive operating profits regardless of their realized types and their perceived types (as is ensured by Assumptions 1 and 2), then the ex post distribution of firms will still be anticipated to involve a fraction [lambda] of H-type firms and a fraction 1 - [lambda] of L-type firms. Ignoring entry costs, ex ante expected profits are given by [lambda][[PI].sub.H]([alpha], [beta], [gamma], [delta], [lambda], k, n) + (1 - [lambda])[[PI].sub.L]([alpha], [beta], [gamma], [delta], [lambda], 0, n), where all of the parameters on which the equilibrium profits depend are indicated. One question is whether the number of firms can be determined in equilibrium, and still be consistent with Assumptions 1 and 2. To see that it can, we suggest a three-step procedure (this admittedly awkward argument is necessitated by the complex dependence of the equilibrium profits on the various parameters). The steps are: (i) choose a target industry size [??]. Recall that for high-demand markets (that is, for sufficiently high values of [alpha]), both [[PI].sub.H]([alpha], [beta], [gamma], [delta], [lambda], k, n) and [[PI].sub.L]([alpha], [beta], [gamma], [delta], [lambda], 0, n) are decreasing in n; (ii) choose a value of [lambda] in (0, 1), values of([alpha], [beta], [gamma], [delta]) that are consistent with Assumptions 1 and 2 (for n in a neighborhood of [??]) and involve [alpha] sufficiently large to obtain interim profits decreasing in n for both firm types (for n in a neighborhood of [??]), (19) and a value of k < [delta]; and (iii) let the fixed cost of entry be F [equivalent to] [lambda][[PI].sub.H]([alpha], [beta], [gamma], [delta], [lambda], k, [??]) + (1 - [lambda])[[PI].sub.L]([alpha], [beta], [gamma], [delta], [lambda], 0, [??]). Then [??] is an equilibrium with unrestricted entry. Each firm that enters obtains zero expected profits (net of the entry cost), which is the same as it would earn by not entering; moreover, no additional firm would want to enter because ex ante expected profits are decreasing in n (by construction).

4. Applications to tort reform and professional licensing

In this section, we use the model to address two issues. First, we show that the result concerning an increase in the loss [delta] may mean that recent proposals for tort reform may increase (not decrease) the number of consumers harmed and thus the number of lawsuits. Second, we argue that professional licensing regulations that serve to lower the loss due to low quality, [delta], may provide an additional benefit to consumers by increasing competition, thereby reducing both prices.

Tort reform. The foregoing model suggests that tort reforms currently being suggested, such as "capping" (limiting) damages awards or setting higher evidentiary standards for plaintiffs to win cases, may perversely help low-quality firms, possibly at the expense of high-quality producers. To see this, consider the following direct extension of our model. In what follows, we assume that damages are verifiable, but that (as has developed in the law), tort law preempts warranty when dealing with harm (see Daughety and Reinganum, 1995). We therefore model the consumer loss, [delta], as composed of two parts, compensated damages ([[delta].sub.c]) and uncompensated damages ([[delta].sub.U]). If the consumer's losses consist of both "economic losses" (such as hospitalization costs) and "noneconomic losses" (such as emotional distress arising from the harm having occurred), then a statutorily imposed cap on recovery of noneconomic damages would mean that [[delta].sub.U] > 0. Such caps have been in operation at the state level, and have recently been under consideration in the U.S. Congress. (20) Damages that are compensated must be paid for by the firm, so this means that the low-quality firm's unit cost of production is now positive: [C.sub.L] = [[delta].sub.c]. To maintain the usual incentives for revelation, assume that [[delta].sub.C] < [C.sub.H] = k < [delta] [equivalent to] [[delta].sub.C] + [[delta].sub.U]. Finally, for simplicity, we ignore other losses that might arise due to the filing of lawsuits, settlement bargaining, or trial. (21)

As in the previous section, one can show that (for the modified model, holding [delta] constant) for high-value markets, the L-type's price, quantity, and profit are increasing in [[delta].sub.U] for sufficiently high [lambda]. Thus, the effect of a cap on noneconomic damages due to torts arising from products provided in high-value markets, where there is a sufficiently high proportion of high-quality producers, is to enhance the prospects of the low-quality firms! It is probably not unreasonable to think of medical services as a high-value market, so such caps may lead to an increased number of malpractice lawsuits, as the price distortion associated with signalling shifts more consumers toward lower-quality medical providers (and more harms). Although this discussion formally focuses on caps on damages awards, a simple variation would provide the same result for increases in evidentiary standards (that is, raising the evidentiary requirements that a plaintiff must meet in order to prove a defendant's liability for a tort), as this would increase the expected uncompensated loss.