People often make decisions based on inferences from sample data without noticing that the process by which the sample is generated could fundamentally challenge those inference. This phenomenon, known as selection neglect, can consequentially lead people astray in their decisions. In this paper, we seek to demonstrate its costly implications in the context of managerial decisions. We conducted four online controlled experiments to illustrate how selection of top performers could easily cause illusory or even reversed correlation in different settings. Specifically, in study 1 and 3, participants engaged in simulated hiring decisions, and were incentivized based on the performance of their selected candidate. The results suggest that people fail to adequately adjust to the transformation of data distribution of selected sample and instead continue to act as if the correlation pattern remain unchanged. Consequently, they end up hiring worse candidates. It also highlights the origination of discrimination, which often stems from such biased belief from selected samples.