Question

. xi: reg lnGas lnP lnInc i.month if date >- 494 & date <- 554 i.month Imonth_1-12 (naturally coded; _Imonth_1 omitted) Source df MS Number of obs 61 F (13, 47) Model Residual 116339448 13 .008949188 Prob > F 47 0 R-squared 1.0000 1.0000 Adj R-squared- Total 116339448 60 .001938991Root MSE lnGas Coef. Std. Err. [95% Conf. Interval] lnP lnInc Imonth_2 Imonth_3 -Imonth_4 Imonth_5 Imonth 6 Imonth7 Imonth 8 -1month9 Imonth 10 Imonth_11 Imonth_12 cons 1 -1.16e-16 6.43e-17 5.63e-18 1.34e-17 -1.58e-18 1.36e-18 -1.46e-17 -1.68e-17 1.97e-17 -9.88e-20 1.83e-17 2.24e-19 4.44e-16 -

The coefficients for the month of observation, _Imonth_2, _Imonth3, etc. are mean effects (dummy variables) that shift the intercept of our demand equation for each month of the sample. In terms of what we know about gasoline demand, why might it be important to model different baseline gasoline consumption by month?

Answer 1

There are a lot of factors that influence gasoline demand and to be able to capture the effect well in the regression analysis we put a dummy variable for months. It could be so that some months observe higher gasoline demand than others. For example, during festive season or during winters, the gasoline demand might go up due to either higher number of trips than rest of the month or lowered ignition of gasoline due to cold. Therefore, when we fit the model and make predictions, the forecasts will be more accurate when such factors have been taken care of.

In general terms, there may be seasonality or spikes in demand of a product through different months and that should be accounted for in the regression analysis using appropriate dummies.