Is more novel research always desirable? We develop a model in which knowledge shapes society's policies and guides the search for discoveries. Researchers select a question and how intensely to study it. The novelty of a question determines both the value and difficulty of discovering its answer. We show that the benefits of discoveries are nonmonotone in novelty. Knowledge expands endogenously step-by-step over time. Through a dynamic externality, moonshots - research on questions more novel than what is myopically optimal - can improve the evolution of knowledge. Moonshots induce research cycles in which subsequent researchers connect the moonshot to previous knowledge.

I study a continuous-time moral hazard problem with learning about a two-stage project of unknown quality. The first-stage arrival time is informative but not conclusive about the project’s quality. Due to the informativeness, the optimal contract features a combination of continuation value and intermediate bonus payments as a reward. There is a negative correlation between the first success time and the share of bonus payments in the reward. Second-stage deadlines adjust to the first-stage success time: early successes are rewarded with longer deadlines in the second stage. When agent replacement between stages is possible, the principal will replace the agent if the first success arrives late.

We study Airbnb hosts' strategic pricing incentives when prices affect ratings. Two channels determine the price-rating interaction: higher prices reduce the value for money, worsening ratings, but increase the taste-based valuation of travelers, improving ratings. The price further determines the frequency of rating updates through its effect on demand. Our empirical results show a dominant value-for-money effect and that hosts lower their prices to surpass rating thresholds. We provide evidence that hosts benefit from low entry prices: offering a median entry discount of seven percent improves medium-run monthly revenues by three percent.

We study a behavioral SIR model with time-varying costs of distancing. The two main causes of the variation in the cost of distancing we explore are distancing fatigue and public policies. We show that for a second wave of an epidemic to arise, a steep increase in distancing cost is necessary. Distancing fatigue cannot increase the distancing cost sufficiently fast to create a second wave. However, public policies that discontinuously affect the distancing cost can create a second wave. With that in mind, we characterize the largest change in the distancing cost (due to, for example, lifting a public policy) that will not cause a second wave. Finally, we provide a numerical analysis of public policies under distancing fatigue and show that a strict lockdown at the beginning of an epidemic (as, for example, recently in China) can lead to unintended adverse consequences. Once the policy is lifted, the disease spreads rapidly due due to the accumulated distancing fatigue of the individuals causing high prevalence levels.

A seller serving two generations of short lived heterogeneous consumers sells a product under uncertain demand. We characterize the seller's optimal pricing, taking into account that the current period's price affects the information transmission to the next period consumers via consumer ratings. While the seller always prefers to generate more information, it is not necessarily in the consumers' interest. We characterize situations in which consumer surplus and welfare are decreasing in additional information. We provide conditions under which aggregate consumer surplus and welfare are lower with than without a rating system.

We study competitive awarding procedures of short haul railway passenger services in Germany from 1995 to 2011 by means of a newly collected data set. In particular, we use regression techniques to investigate the determinants of the number of bidders, the identity of the winning bidder and the subsidy level. We find that there are more bidders when the contract duration is high and the revenue risk low. The dominant operator is more likely to win contracts if it is the incumbent, the network is large, the contract duration is high, when used rolling stock is admitted and when there are few other bidders.