Can cost increases increase competition? Asymmetric information and equilibrium prices.

(5) Assuming to the contrary that [[delta].sub.1] < [[delta].sub.2] provides lender 1 not only with an information advantage but also with a cost advantage over lender 2. Because this is a model of pure price competition, this large advantage would in most instances shut lender 2 out of the market entirely.

(6) This assumption is analogous to that used by Villas-Boas (1999) in his analysis of competition with customer recognition. See also Greenwald (1986) for an application to labor market competition. We analyze the case where both lenders can make counteroffers to their old customers in Section 5.

(7) This assumption guarantees that lender 1's private information actually matters. For [[delta].sub.2] < [[delta].sub.1]-1/2, lender 2's cost advantage is so large that lender 1 gets entirely shut out of the market.

(8) This result has been established in models of banking competition by Broecker (1990) and Rajan (1992), among others. More generally, this result is fairly standard from the literature on common value auctions where only one bidder possesses private information (e.g., Engelbrecht-Wiggans, Milgrom, and Weber, 1983).

(9) This is a sufficient condition, but dearly not a necessary one, as [F.sub.2] can be increasing in [[delta].sub.1] even absent the restriction on the lower bound [r.bar].

(10) The assumption that [theta] is uniformly distributed greatly simplifies the analysis, but is not necessary for our results to hold. What is required is that an increase in lender 1 's cost of funds leads to an increase in the average quality of borrowers applying to lender 2. This requirement is satisfied by any continuous and strictly increasing distribution function.

(11) It should be noted that, for [lambda] [member of] ([[lambda].bar], [bar.[lambda]]), the equilibrium with the competitive fringe is identical to that found in Section 3, so that all results related to marginal changes in [[delta].sub.1] go through identically irrespective of the exact structure of the markets.

(12) In the context of the model presented earlier, this is like assuming that [lambda] = 0 so that there are no unknown borrowers, and that the informed lender cannot make counteroffers to its known clients. The introduction of a switching cost "smooths" out the payoff functions and allows for the existence of an equilibrium in pure strategies.

(13) Unlike the analysis in the previous sections, here there is a distinction between the case where lender 1 competes against a single lender versus against more than one lender. Absent the zero profit condition imposed by competition within the fringe, the uninformed lender's interest rate need not go down when [[delta].sub.1] increases.

(14) As a couple of recent articles from the Wall Street Journal highlight, firms not plagued by these difficulties are "bound to poach employees and clients of troubled companies, intensifying a battle already under way among the existing firms" (Dugan, 2005). More specifically, incoming CEO Michael Cherkasky adopted a strategy of eliminating marginal customers, so as to "jettison clients that aren't profitable and exit the people who support them" (McDonald, 2005).

(15) It should be noted, however, that it is unclear whether these costs associated with litigation necessarily imply an increase in the marginal costs for the beleaguered brokers. Nevertheless, some observers have argued that smaller, newer brokers may benefit from a relatively lower cost of retaining qualified staff, which implies a lower marginal cost of doing business (Insurance Journal, March 18, 2005).

(16) This assumes that the lender's cost increase represents a pure transfer between the lender and the suppliers of the input, such as depositors, so that there is no deadweight loss as a result of the cost increase.

Giovanni Dell'Ariccia, International Monetary Fund; gdellariccia@imf.org. and Robert Marquez, Arizona State University; rsmarquez@asu.edu.