It has been widely recognized that investigation data is the foundation of outage cost assessment. However， how to deal with the inevitable fault and concealed information from the original data under investigation remain a challenging problem yet to be solved effectively by applicable measures. Focused on this problem， the Tobit model to truncated process the investigation data is firstly introduced， and then an improved Tobit model suitable for outage loss assessment is proposed. In this model， the data beyond the limits are truncated and the outage cost assessment function is obtained by the regression method based on the maximum likelihood estimation（MLE）. By using this model， the outage cost of all kinds of customers caused by an outage in a specific area can be calculated， and good results can be obtained to identify customers' malicious fault behaviors in an electricity market. At last， a testing scenario of a business customer is done， and the testing results show the correctness of the proposed method.