The model has dummy variables to consider the month of year when the household was interviewed. The regression coefficients were found using an ordinary least-squared method. Although the results do not document how households managed to reduce usage, usage reductions in the order of 10–17% were reported.