Deniz Selman Ph.D. Candidate Department of Economics University of Pennsylvania 3718 Locust Walk ~ Philadelphia, PA 19104 tel: 703.627.1493 denizs@econ.upenn.edu curriculum vitae (pdf)
research
Optimal Sequencing of Presidential Primaries (job market paper)
full text: (pdf). extended abstract: (pdf).
ABSTRACT: The American presidential nomination process consists of a series of elections (primaries) in which states vote at different times. This paper focuses on the problem faced by a political party which wants to choose the optimal temporal structure for its primaries. I consider an environment in which a sequential election may generate voter herding, and address both when and how the party can benefit from social learning to maximize the probability of selecting the highest quality candidate. By choosing whether to have a sequential election and---if so---the actual sequence in which states vote, the party can control whether momentum effects will be present and guarantee that any voter herding will be ex ante beneficial to the goal of selecting the highest quality candidate. When candidates are expected to have equal partisan support, simultaneous voting outperforms all sequential elections. When one candidate has more partisan support, a sequential election can outperform simultaneous voting; under this condition, voter herding compensates for the partisan imbalance. This result is a novel example of a beneficial information cascade, in contrast with the socially inefficient cascades in the standard herding literature. In a sequential election, it is optimal for the states voting early in the primary season to be those which: (i) are smaller, (ii) have fewer partisans, (iii) have more informed voters, and (iv) display more voter diversity.
JEL classification: D71, D72, D82, D83.
Political Knowledge and the 2004 Presidential Election (joint with E. Harvill)
full text: (pdf).
ABSTRACT: Political knowledge is an important factor in understanding voting behavior, particularly when voters' policy positions conflict with those of their favored candidate. In this paper, we use the National Annenberg Election Survey (NAES) for the 2004 U.S. election to directly estimate voters' ideological positions, candidates' ideological positions, and voters' knowledge of candidates' ideological positions. We then analyze the relationship between these estimates and the individual's vote in the 2004 presidential election. The NAES asks voters for their own stances and for the candidates' stances on specific policy issues. Because candidates' stances on these policies are known, the data allow us to (1) estimate voters' knowledge levels using item response theory, and (2) simultaneously estimate the positions of voters and candidates on the same spectrum. Typically, in the empirical literature, individuals' ideological positions are taken from self-reported placement on an ideological spectrum. Here, we use Poole and Rosenthal's (1985, 2000) spatial model to estimate ideological positions, treating individuals' opinions as Poole and Rosenthal treat roll-call data for legislators. This method recovers the underlying ideological space while avoiding the perception issues associated with self-placement. Assuming that voters prefer to vote for the candidate whose ideological position is closer to their own, we find that more knowledgeable individuals are far less likely to vote for the candidate farther from them. Moreover, significantly more voters vote "mistakenly" for Bush than for Kerry. Though a majority of voters have ideological positions closer to Kerry, Bush receives more votes due to this pattern of mistakes.
JEL classification: C13, D72, D82, D83.
Coordination Frictions and Heterogeneity in Markets with Bidding
full text: (pdf).
ABSTRACT: Two heterogeneous buyers with commonly known preferences must choose which one of two different goods--a high value good and a low value good--to bid on when the goods are sold through simultaneously held first-price auctions. We find that efficiency of equilibrium allocation depends on allowing sellers to announce and commit to reservation prices before the auctions are held. This is achieved by first characterizing a unique buyer equilibrium for a generic subset of the parameter space when reservation prices are exogenously set to zero. This equilibrium exhibits coordination frictions and results in a potentially inefficient allocation of the goods. The high value good is sold for sure, though it could be to either buyer and could even be at the reservation price. The low value good is either not sold or sold exactly at the reservation price to the buyer with a lower marginal valuation for the high value good. We then characterize buyer equilibria for all reservation prices. Endogenizing seller behavior by allowing sellers to announce and commit to reservation prices results in a unique market equilibrium of the game in which buyers perfectly coordinate, the efficient allocation is achieved, and both allocation and payoffs are equivalent to that of the price-posting case. The contrasting results between exogenous and endogenous reservation prices stems from the following: when the same side that searches (here, the buyer side) is responsible for price formation, coordination frictions are still present because buyers have incentive to compete over the higher value good and hence could both bid on it. However, when the passive side of the market (here, the seller side) is in effect responsible (here, by setting binding reservation prices), the two-sided heterogeneity results in perfect coordination by the buyers in choosing their goods.
JEL classification: C78, D44.