For the Aggregate Engel Curve assignment, I chose to analyze the relationship between overall United States tobacco consumption against median U.S. income per capita from 1995 to 2015. Use of cigarettes has continually shown little change in quantity when prices/taxes increase or decrease, known as having relatively inelastic demand, according to the World Health Organization. However, I was interested in seeing if this inelastic demand was relevant amongst all tobacco products, such as cigars, dip, and hashish products. To determine this, I researched data regarding median income per capita in the United States and compared it to total tobacco consumption (in billions of dollars) over the past 10 years. Although cigarettes are usually considered a normal good, they can also be viewed as an inferior good, if severe addiction is a determining factor. However, with overall tobacco consumption, it is hard to distinguish as a normal or inferior good, due to the range of tobacco products that exist.
For this sample, a one unit increase in disposable income per capita was associated with an additional 0.00308 billion dollar (around 3 million dollars) increase in consumption of tobacco based products a year. Keep in mind, this is in terms of yearly consumption and yearly median income in the U.S. Therefore, an increase in per capita income accounts for approximately 3 million dollars in yearly tobacco sales. On another note, the tobacco data gathered is measured in sales. So, this assumes that, for every unit of tobacco sold, it is consumed.
Looking at the p-value for median income, we can see that it is statistically significant, as it is <.0001. This suggests that there does appear to be a relationship between the consumption of tobacco and disposable income, also meaning that there is evidence to reject the null hypothesis, should this model be incorporated in hypothesis testing. The output also shows a significant R-square static of 0.8884. R-squared is a measure of how close data match the line of best fit. As shown by the scatter plot below, we can conclude that there is a strong correlation between our independent and dependent variables.
*Tobacco Consumption measured in billions ($) per year
*Median Income measured in thousands ($) per year
Correction for heteroscedasticity was the way I wanted to test my data in case of inconsistency, because the data used was cross sectional. Heteroscedasticity essentially means that a model will show small errors in earlier periods, while larger ones are found in the later periods (or vice versa, or no particular order). So basically, non-constant error variance across the independent variable. When looking at the heteroscedasticity estimates (shown on page 1), we can see a low standard error of 0.000190 and a relatively high t-value of 16.13. Therefore, we can conclude that based on the estimates that heteroscedasticity does not appear to be present in the model. To correct for heteroscedasticity, should it be apparent, the weighted least square method can be utilized.
After analyzing the statistically significant data, it seems to be apparent that different tobacco products contain relatively inelastic demand. The overall general increase in disposable income per capita from 1995-2015 didn’t not seem to show any reduction in tobacco sales. The sale of tobacco also saw steady increases from 1995-2015. This is again stating the assumption that each unit of tobacco sold will definitely be consumed. Although, we cannot be 100% sure of this. In fact, it is more than likely that there is sold tobacco that has yet to be consumed, which would affect this particular model. A final prediction, based on this model, is that consumption of tobacco increases as per capita income increases.