There are many environments where attrition is part of the game. The competitive nature of many aspects of our society often implies that some get ahead while others quit. This applies to education, business, sports, science, art, and nearly every other aspect of the modern system of human interactions. Understanding these processes and incorporating this understanding into the decision-making process is vital for efficient resource allocation. This paper examines the environment in which the decision-maker receives signals about some potential payout and evaluates her prospects. One feature of this environment is failure. I examine the consistences of ignoring attrition. Specifically, the process of information collection is modeled with the information gathering mechanism and its attributes. Implications of informational collection and analyses are of particular importance to the entrepreneur and the firm.

My model shows how seemingly irrational behavior of new entrants can be explained by a lack of complete information. The overestimation of future earnings potential and subsequent over entry is driven by the assumption that income information containing profit levels of firms who have gone under is not accessible. Whenever an observer contemplating entry is not aware of the limited nature of information, she would be bound to over inflate her estimate of earnings potential. This result arises even though the entrant has a correct prior belief. This correct information gets distorted during Bayesian updating, involving an analysis of incomplete information. I also find that increasing the number of signals during sampling is likely to make the resulting bias worse. On the other hand, collecting information about the entire profit population (observing several periods prior to entry) helps to reduce bias. For any agent having more precise information about prior beliefs helps, while having more precise information about current profit distributions is likely to make the bias larger. Dynamic changes in system parameters, such as the variances of prior and/or signal distributions, and the number of incumbents produce results which may explain life cycles of an industry as well as the different hazard rates observed.