Research

Working papers

A developer seeks to persuade a welfare-maximizing bureaucracy, with rotating officials, to award a larger fraction of a contract to her. Officials' decisions are subject to a bureaucratic norm, whereby a decision can be only based on evidence that is either recorded by her predecessor or is directly presented to her. Thus, Bayesian inference is \emph{restricted} when a predecessor fails to record evidence, and bureaucrats can exploit this to induce the developer to conduct more informative experiments. In a class of information design problems where the static values of persuasion are zero to the bureaucracy and strictly positive for the developer, I show that there are two possibilities in the dynamic game. Either the developer conducts a more informative experiment and the official decides immediately, giving the bureaucracy a positive value; or there is delay and the gradual release of information. In either case, the developer is worse off due to the bureaucratic norm. 

The seller bargains with two buyers to make a deal with each of them, using an alternating offer protocol. The bargaining begins with one buyer, with the second entering at a future date. If the seller has a concave utility function defined over the payment from buyers, the two bargains affect each other. When the seller's utility exhibits decreasing absolute risk aversion, a higher price in the first bargain increases the price in the subsequent bargain. Even if two players are identical except for the arrival date, they will make different payments to the seller. The shape of the utility and the arrival date determine whether there is a first or second-mover advantage. The existence of one buyer can be beneficial to the other. Furthermore, though agreements in our model are reached on different dates, the usual limit results do not approach that of the sequential Nash bargaining solution. Finally, we extend the model to a vertical market, in which an upstream seller supplies the buyers with a critical input. The buyers compete with each other in a downstream market. We find that, even if the buyers are symmetric Cournot competitors, the equilibrium of the model is asymmetric, with one buyer paying more than the other. Furthermore, prior to entry by the second firm, the price set by the incumbent varies with the expected entry date. Standard models would not predict this.

I study a symmetric two-player game of strategic experimentation where both players have private information. I find that two-sided private information improves welfare, both at the ex ante and interim stages, by mitigating the free-rider problem. Furthermore, in some states of the world, there may be over-experimentation, i.e., players may experiment more than the social planner would under complete information.

Based on the cheap talk model with naive receivers who take the message at face value in Ottaviani and Squintani (2006), I endogenize the probability of the receiver blindly believing in the sender by allowing the sender to increase this naivety probability at a cost. When the probability chosen is observed by receivers, receivers can benefit from this ability of the sender, and the fully revealing equilibrium is possible. But this ability of the sender damages information transmission and removes the fully revealing equilibrium if the probability is not observable. These results can explain how information is conveyed in advertising when the advertiser can design the content of advertising as well as use extra expenditure to affect the consumers' gullibility.