Abstract: Consumers search on a platform to find a product they like. The platform observes which products consumers inspect and buy. Based on these observations it ranks products to maximally learn, in the long term, which product consumers like. We find that a monopoly platform first experiments with rankings and later only ranks products that early consumers bought. This guarantees that later consumers are pickier, helping the platform to learn what consumers really like. The more dissimilar consumer tastes, the more consumers search themselves and the platform learns about products. Competition restricts what platforms learn.
Abstract: Consumer-tracking technology offers new tools for price discrimination in digital markets. We examine the impact of sellers using this technology to adjust prices according to a buyer’s prior search length in a competitive search market where buyers differ in patience. We find “Coasian equilibria” wherein sellers reduce prices for buyers with longer search lengths which in turn requires them to reduce prices for buyers with shorter search lengths. In commonly studied environments, Coasian equilibria not only yield higher welfare for every buyer than all uniform-pricing equilibria, but are also the only equilibria when some mass of buyers are arbitrarily patient.
Abstract: I analyse a model of partially directed search where consumers decide which firm to visit based on correct, but incomplete, information about firms' prices, as are advertised prices in the presence of stock-outs. Equilibria exist where firms set two prices, a low promotional price and a higher regular price, with positive probability. The promotional price can be below the firms' marginal cost, rationalising unprofitable promotions prevalent in reality. My model predicts that unprofitable promotional prices are lower in markets characterised by stiffer competition and higher search costs, in line with differences between the US and Europe.
Abstract: In this paper, we study a competitive search model with demand uncertainty: when the good is fashionable, there are more buyers around. Homogeneous sellers receive the same signal about demand and post prices with commitment. Homogeneous buyers know if the good is fashionable, observe all prices, and direct their search freely. Price dispersion may arise in equilibrium due to sellers’ imperfect information. Thus, our model offers a potential explanation for the puzzling failure of price comparison websites to eliminate price dispersion.
In this paper, we study a competitive search equilibrium in a market for an indivisible good. Buyers privately observe a common shock to their valuation of the good which can be either high or low. This creates an adverse selection problem about an aggregate state for the sellers. Meetings between buyers and sellers are bilateral. We show that in this environment when the difference between the two valuations is not too large, the equilibrium features pooling. That is, the sellers find it optimal to post a single price, foregoing the possibility of screening the valuation of buyers. The separating equilibrium, when it exists, gives buyers a choice between paying a high price for the good or buying a ticket for a lottery with the good as its prize.
Public engagement: Christie Conference 2021, Teachers' Day 2022
Abstract: This paper theoretically studies price discrimination based on search costs. "Shoppers" have a zero and "nonshoppers" a positive search cost. A consumer faces a nondiscriminatory "common" price with some probability, or a discriminatory price. In equilibrium, firms mix over the common and the shoppers' discriminatory prices, but set a singleton nonshoppers' discriminatory price. Consumer welfare increases if price discrimination is restricted enough. An individual firm's profit can increase in the number of firms. These results have important implications for regulations that limit the tracking of consumers (e.g., EU's GDPR, California's CCPA) and for evaluating competition online.
I study a sequential search model where buyers face an unknown distribution offers and learn about the distribution from other buyers' actions. Each buyer observes whether a randomly chosen buyer traded in the previous period. I show that a cyclical equilibrium exists where the informational content of observing a trade fluctuates: a trade is good news about the distribution in every other period and bad news in the remaining periods. This leads to fluctuations in the volume and probability of trading. They fluctuate more if the unknown distribution is bad rather than good. A steady-state equilibrium where buyers are more likely to continue searching than in the cyclical equilibrium is less efficient than the cyclical equilibrium. A market that starts at date one converges to the cyclical equilibrium for some parameter values.
I study stationary cutoff-strategy equilibria of a dynamic market model where buyers sample sellers sequentially from an unknown distribution. Buyers learn about the distribution from the sampled sellers and a "trade signal''. The trade signal reveals whether a randomly chosen seller traded yesterday. Observing a trade (as opposed to no trade) is good news about the distribution. Buyers who observe a trade use a higher cutoff than buyers who observe no trade, despite buyers' learning from sampled sellers that puts a countervailing pressure on the cutoffs. The trade signal may reduce market efficiency, while an appropriate exogenous signal increases efficiency.
This paper extends the standard sequential search model by allowing the agent who compiles the choice set via search (the "searcher") to differ from the agent who chooses from the set (the "chooser"). I show for a general joint distribution of the agents' preferences that the searcher’s optimal policy is a threshold rule. In contrast to the standard model, the threshold is weakly decreasing in time (i.e., exhibits the "discouragement effect"), although the search horizon is infinite and the search environment stationary. I characterise the threshold and discuss the testable implications of the discouragement effect. The characteristics of my model differ from two single-agent search models that feature a time-varying threshold (convex search costs or deadline). In particular, my model features a threshold that decreases endogenously over time and never generates return to an item rejected earlier, in contrast to the other models.