To measure how consumption of a good measures with income, I decided to use a time-series approach. The variables I utilized were gasoline/energy consumption as well as disposable income in the United States over the past thirty years. I considered using a cross-sectional approach, but found more data from accurate resources on time-series data. The time-series approach I chose also allowed me to look over a period encompassing two recessions, which was intriguing.
From the Engel Curve portrayed above, we can see there appears to be a relationship between an increase in disposable income and gas consumption over the past thirty years. From the primarily upward sloping curve, we can deduce that gasoline is a normal good.
These regression results do suggest that income does affect gas consumption. For this sample, one percentage point of income was associated with a $17 billion increase in gas consumption. The variable coefficients were both statistically significant (p<.0001). The regression overall fit reasonably well (R2 = .7765; adjusted R2 = .7688) and was highly statistically significant (F = 100.77; p < .0001).
I then performed a Durbin-Watson test and discovered that there was some autocorrelation between my variables. I corrected this by running PROC AUTOREG with /NLAG=1. The regression results after the correction are listed below.