This paper examines whether insider trade clustering is associated with large stock return variations (i.e., crash or jump risk) in Korea. Recent studies argue that not all insider trading is informative and insider trade clustering distinctly yields higher stock price performance than non-clustering. To investigate private information of insider trade clustering, we separate insider trade clustering into sale clusters and purchase clusters and then document whether trading behavior of insider sale (purchase) clusters is related to the likelihood of a crash (jump). We find that insider sale clusters which occurred over the past month of a crash are strongly related to the information flowed over a short period of time. However, we find that insider purchase clusters are less associated with the likelihood of a jump. Our results provide empirical evidence that insiders share private information and insider sale clusters contain robust short-lived negative information. Overall, our findings support that insider trade clustering, in particular insider sale clusters, results from agency problems.