SUMMARY
This study intends to identify the effects of client-related factors and auditor-related factors on the auditor changes. Data is selected using random sampling and purposive sampling. The data collection is performed using mail survey and archaival. The statistic method used to test the hypotheses is regression analysis and RPA (Re-cursive Partitioning Algorithm) model.The study results are as follows: first, The results provide support for the hypothesis that ctient-related factors and auditor-related factors on the auditor changes; second, the normal data test and non response bias using t-ba shows an insignificant result This means that there are non response bias and the normal data; third, this is demonstrated by the multicolinearity number r< 0,8 or VIF mean 1 that shows that the multicolinearity is not dangerous, the Durbin Watson approaches 2 and BG (The Breussh-Godfrey) r = 0 which means that between one variable and the other there is no dependency relationship (independent), and and homoscedacity occur.Keywords: RPA (Recursive Partitioning Algorithm), client-related factors, auditor-related factors and auditor changes