In this paper, we study the recent energy use intensity (EUI) in kWh/m²/year of Hong Kong's secondary schools. We found a 22% increase of EUI from, 2003 (86 kWh/m²/year) to 2017 (105 kWh/m²/year) approximately. It may be due to the increase in the number of air-conditioned classrooms. We then use a novel two-stage regression-based approach to analyse the most important set of statistically significant energy performance factors. Then, the energy use intensity (EUI) is normalized with these energy performance factors to find the benchmarking scores used to assess whether the electricity tariff affects schools' energy efficiency. In the first stage of the approach, we first conduct three methods of fitting regression models, backward, forward, and stepwise (OLS), to find a set of normalized factors and construct the benchmarking model. The first stage provides the normalized EUI range, approximately between 50 and 163 kWh/m²/year. In the second stage, if better goodness of fit is desired in the second stage, convex nonparametric least squares (CNLS) regression analyses are employed with the same set of normalized factors. This two-stage regression-based approach is called the OLS-CNLS approach. Using simple EUI (kWh/m²/year) results, we find that high tariff schools' energy efficiency is better than low tariff schools. However, the benchmarking scores (normalized EUIs) of the OLS-CNLS approach show that the null hypothesis cannot be rejected. The energy efficiency of high tariff schools is not significantly better than that of low tariff schools. It implies that the tariff level is not a crucial factor associated with secondary schools' energy efficiency in Hong Kong. As the OLS-CNLS approach allows us to take all the significant energy performance factors into consideration to address the electricity tariff concern, the funding/budget policies probably motivate the school management to look for better EUI.