Another explanation for overbidding and another bias for underbidding in first-price auctions.

First-price auction experiments find often substantial overbidding which is typically related to risk aversion. We introduce a model where some bidders use constrained linear bids. As with risk aversion this leads to overbidding if valuations are high, but in contrast to risk aversion the model predicts underbidding if valuations are low. We test this model with the help of experiments, compare bidding in first-price and second-price auctions and study revenue under different treatments. We conclude that at least part of the commonly observed overbidding is an artefact of experimental setups which rule out underbidding. Constrained linear bids seem to fit observations better