Embedded Files
This paper studies welfare effects of dynamic pricing in an environment with competitive sellers of differentiated and perishable goods. With dynamic pricing, sellers can extract more surplus from buyers by frequently updating the prices of unsold goods in response to changing market conditions. However, this type of pricing may also intensify competition among sellers, since sellers have incentives to undercut each other. Due to these two opposing forces, welfare outcomes for buyers and sellers are ultimately ambiguous. I develop a dynamic structural model to estimate demand and supply of listings on Airbnb. The model predicts that with dynamic prices, on average, transaction prices fall by 6.2%, while social welfare increases by 0.1%. I find that in the majority of markets, the competitive effect dominates, which leads to a 4.3% drop in profits for sellers, 3.3% drop in platform revenue, and 3% increase in consumers' welfare.
In this paper, I address two major issues associated with the estimation of demand for differentiated products with a possibility of stockouts. First, in the case for Airbnb, each guest potentially faces a unique choice set since each option has a finite capacity. I use data from small-sized and isolated urban areas which allows me to reconstruct the set of options that was available to each buyer. I find that improperly accounting for availability of choices for guests produces estimates for price elasticity which are on average 16% more inelastic. Second, unobserved characteristics of listings on Airbnb play a crucial role in decisions of guests. Therefore, to address the price endogeneity in demand estimation I use and contrast two different approaches. The first approach relies on the panel structure of the data. Hence, including the time invariant listing fixed effects captures the unobserved heterogeneity of listings. The second approach uses BLP type instruments to correct for price endogeneity. The instrumental approach is implemented using the control function method. I find that both the fixed effect estimator and the instrumental approach yield significantly more elastic demand estimates for lodging in comparison to the estimates derived without addressing the price endogeneity. The price elasticity for an entire listing is -1.5, while for a shared listing the price elasticity is -1.3.
Google Sites
Report abuse