 The question we are interested in is how people make timing decisions in situations where they have little experience or little data on which they can base their decision. If you think about it, these types of timing decisions are ubiquitous in many different areas of economic decision making. So let me give you a few examples. An important topic in my age group seems to be buying a house for the first time. So suppose you are in that situation and you actually found a house that you like, then the question is whether you want to buy that house immediately at the price that the current owner is asking or whether you want to wait a little bit and see whether there is limited demand and the owner actually goes down with the price somewhat. Another problem would be that of an innovating firm developing a new product. So here a common timing decision is how long to work on the different features of the products before bringing it to the market and the risk of working on these features for too long is clearly that a competing firm may develop a similar or even better product in the meantime. You could also think about political decision making. So for instance when to switch to green energy or when to impose a lockdown in the pandemic in all of these situations you have this timing component and in addition to this you have a situation that is relatively new to the person who must make that decision. So there is substantial uncertainty and this is where our paper comes in. So the question that we are interested in is how do people make decisions in situations where they face this type of uncertainty. So we follow a theoretical approach to address this question which means that we build a game theoretical model and then look for an equilibrium. We consider a very precise setting that resembles a little bit the house buying example. So we have a number of potential buyers who all want to buy the same good and the price for that good is decreasing over time and they must decide when to jump in and buy it. In order to solve this problem we rely on two strengths of the literature. So the first is auction theory because one way to think about this problem is in terms of a descending price auction which is sometimes also called a Dutch auction. So in a Dutch auction the price is decreasing over time. There's a number of bidders present in the auction and the first bidder who stops the auction gets the good at the price that is currently displayed. The second strength of literature that we rely on is decision theory and in particular the literature on ambiguity aversion. This literature has emerged as a consequence of the observation that people seem to be treating known risks such as rolling a dice differently from unknown risks. For example the likelihood that a different buyer has a higher willingness to pay. What has emerged from this literature is that people are more cautious when it comes to unknown risks than for known risks and this is something that we explicitly incorporate into our model. So formally what we do is we consider a descending price auction with bidders who are ambiguity averse and face uncertainty over the willingness to pay of the competing bidders. So a distinctive feature of our setting is that bidders are uncertain about the distribution of competing bids that they will face in the auction. In the context of our housing example it's common that when you start talking about buying a house that you set yourself a limit of how much you want to spend but then when you are actually in the process of buying a house and there is a house that you really like you might get worried that somebody else jumps in and buys the house before you and then it's very easy to go above your initial limit. So more formally what we show is that bidders tend to overweigh the event that they lose the item if they wait for a slightly lower price and this happens at every point in time at the auction. As a result of this they tend to jump in too early and buy the good at a price that is too high from an ex-under perspective. This is what we call in economics dynamic inconsistency. So the one that benefits from this dynamic inconsistency is the seller of the good because he knows that by using this dynamic format he can exploit the dynamic consistency on the side of the bidder and thereby generate a higher price for his good. Our findings highlight certain implications of using dynamic formats in auctions and this is of course important if you think about auction design. For example when you ignore the aspect of ambiguity and use a standard framework a very important property of the Dutch auction is that it is strategically equivalent to a first price sealed bid auction. So a first price sealed bid auction is an auction where all the bidders simultaneously submit a bid and then the highest bid wins. So it's a static auction and this type of auction is actually very popular for example for governments who use these auctions to procure all kinds of services or goods. So one important implication of our model is that choosing a dynamic format rather than a static format can actually be quite beneficial for the auctioneer. The flip side of this of course is that if you care about the consumers the one who are bidding in the auction and you want to avoid that they are consistently over bidding then it's very important to provide sufficient information sufficient data on which they can base their bidding decision. So I view this paper as a first step in addressing the general question that I was outlining in the beginning of how people make timing decisions in unfamiliar situations. One important feature of the model in our current paper is that the stopping payoff so the payoff that our bidders get when they stop the auction is known to them. So in this case they know that they get the good for sure and they know which price they are going to pay. Now if you think of the other examples that I was mentioning in the beginning so for example when to impose a lockdown in a pandemic the stopping payoff is very often also subject to uncertainty. This is something that we want to study in the future and we have already taken some first steps and it turns out that having uncertainty about the stopping payoff actually changes the decision problem quite dramatically. So in particular what you see in this case is that a decision maker rather than wanting to anticipate the action actually wants to delay the action wants to wait. If the decision maker sophisticated he can anticipate this at an earlier stage and try to preempt his later selves from waiting for too long. So it gives rise to pretty rich dynamics and we want to use this model or we hope that we can use this model to address a bunch of interesting questions in the future.