One important issue in mechanism design theory is to model agents’ behaviors under uncertainty. The classical approach assumes that agents hold commonly known probability assessments towards uncertainty, which has been challenged by economists in many fields. My thesis adopts alternative methods to model agents’ behaviors. The new findings contribute to understanding how the mechanism designer can benefit from agents’ uncertainty aversion and how she should respond to the lack of information on agents’ probability assessments. Chapter 1 of this thesis allows the mechanism designer to introduce ambiguity to the mechanism. Instead of informing agents of the precise payment rule that she commits to, the mechanism designer can tell agents multiple payment rules that she may have committed to. The multiple payment rules are called ambiguous transfers. As agents do not know which rule is chosen by the designer, they are assumed to make decisions based on the worst-case scenario. Under this assumption, this chapter characterizes when the mechanism designer can obtain the first-best outcomes by introducing ambiguous transfers. Compared to the standard approach where the payment rule is unambiguous, first-best mechanism design becomes possible under a broader information structure. Hence, there are cases when the mechanism designer can benefit from introducing ambiguity. Chapter 2 assumes that the mechanism designer does not know agents’ probability assessments about others’ private information. The mechanisms designed to implement the social choice function thus should not depend on the probability assessments, which are called robust mechanisms. Different from the existing robust mechanism design literature where agents are always assumed to act non-cooperatively, this chapter allows them to communicate and form coalitions. This chapter provides necessary and almost sufficient conditions for robustly implementing a social choice function as an equilibrium that is immune to all coalitional deviations. As there are social choice functions that are only implementable with coalitional structures, this chapter provides insights on when agents should be allowed to communicate. As an extension, when the mechanism designer has no information on which coalitions can be formed, this chapter also provides conditions for robust implementation under all coalition patterns. Chapter 3 assumes that agents are not probabilistic about others’ private information. Instead, when they hold ambiguous assessments about others’ information, they make decisions based on the worst-case belief. This chapter provides necessary and almost sufficient conditions on when a social choice goal is implementable under such a behavioral assumption. As there are social choice goals that are only implementable under ambiguous assessments, this chapter provides insights on what information structure is desirable to the mechanism designer.

There are two main views of advertising – the informative view and the persuasive view. This thesis studies aspects of the informative view. One aspect of interest is whether firms can benefit from collusion on advertising even though advertising is only informative. If so, will this enhance or lower welfare? There are several reasons why firms may want to collude on adver­tising. First, the legal field is tilted in favour of nonprice collusion. Second, it is not at all obvious that collusion on price is more profitable than collusion on advertising and third, the analyses of Grossman and Shapiro (1984) and others show that profits can be increased by restricting advertising. In Paper 1, we examine firms’ incentives to collude on advertising and the implications for welfare. We find that collusion on advertising and competition on price is more profitable than competition on both price and advertising. We also find that semicollusion on advertising is detrimental to welfare. This suggests a need for monitoring, especially since it is in the interest of firms to restrict price advertising. We also compare semicollusion on price to semicollusion on advertising. We find that, in general, semicollusion on price does not lead to higher profits compared to semicollusion on advertising. Hence we lend theoretical support to the empirical literature that consistently find evidence of semicollusion on advertising rather than on price. Another important issue concerns the effect, on prices and profits, of ad-vertising only a subset of the product range. Many firms, in particular those in the retail sector, sell a wide variety of products but only advertise a few. Recent empirical evidence suggests that prices of unadvertised products are higher (Milyo and Waldfogel, 1999). Theoretically, little is known. In Paper 2, we study the effects of advertising only a subset of products. We allow for both low and high differentiation and, at the same time, we explicitly model the advertising decision. We find that the extend of differentiation between competing firms plays an important role in the analysis of loss leader pricing. When firms sell products with the same reservation price, loss leader pric­ing obtains only when differentiation is low. When products are less similar however, price competition is less intense and, as a result, firms advertise prices above marginal cost. Our loss leader pricing results enable us to shed some light on the seemingly paradoxical empirical findings in the marketing literature that loss leader pricing fails to increase store traffic, loss leader sales and hence to increase profits. We also consider a different subject – price discrimination. Although it is well understood that movements in the exchange rate have a bearing on firm profitability and hence affect firm behaviour, the role of exchange rate variability in the firm’s choice of the number of varieties to produce has (to my knowledge) not been explored. This, despite the fact that the product mix is an important aspect of firm strategy. By tinkering with the number of varieties, a firm can bolster its ability to extract consumer surplus. In Paper 3, we explore this issue. We show that variability in the exchange rate induces the firm to vertically segment markets (i.e., offer two varieties in each market). This happens because exchange rate variability affects income dispersion and hence the firm’s incentives to extract consumer surplus. To better extract surplus, the firm offers two price-quality menus, high quality variant (priced high) for top-end surplus extraction and a low quality variety (priced low) to address market coverage concerns. / <p>Diss. Stockholm : Handelshögskolan, 2005 S. 3-9: sammanfattning, s. 13-95: 3 uppsatser</p>