This research is intended to empirically examine the efficiency and productivity change of 26 regional development banks (BPDs) in Indonesia in 2011-2016 using a non-parametric approach of data envelopment analysis (DEA). This research was started by determining input and output variables based on three approaches, i.e., intermediation, operation, and the asset approach. The Multi-stage DEA was adopted to generate the efficiency score, and the inputorientated variable return to scale (VRS) assumption is specified in data analysis. The Malmquist Productivity Index (MPI) was used to measure the total factor productivity change indicating the improvement or deterioration of performance of BPDs over time. The multi-stage DEA result shows a significant difference in the number of efficient BPDs using the three approaches. MPI shows that the highest productivity increase was in the asset approach of 84.0%, supported by the increase in efficiency change and technological change. While in intermediation and operation approach, the increase was only 44.0% and 36.0%, depending on the increase in efficiency change and scale efficiency change. This research suggests that BPDs need to invest more in technology to increase firm efficiency.