Ideal point estimates based on roll‐call vote results have provided leverage for a variety of theory testing efforts. Recently, scholars have suggested using cosponsorship data as a proxy for roll‐call votes. Conceptually similar to roll‐call votes, cosponsorship data are appealing for a variety of reasons. However, the data‐generating process for cosponsorship is untheorized and little studied. We examine the properties of ideal point estimates from cosponsorship data. We find that the ability to estimate ideal points from cosponsorship data is contingent on the underlying data‐generating process; reliance on such measures requires strong and often unrealistic assumptions.