Findings

Even reviewing the map, there are many different ways that the data could be interpreted. To shed some light on this, a multiple regressions model was used. While the information was initially compiled to use percentages a review of the findings leads to the belief that it was wiser to deal with values. The reason for this is that the study intended to test the idea that local demand was driving the location choices of craft breweries. For that purpose the number of people with a certain characteristic made more sense as an estimator of local demand than the percent. For instance, in the case of liberalism as an indicator, an area with only a few thousand people, all of whom voted for Obama, would probably still have less demand for craft beer than an area where several hundred thousand voted for Obama, even if those several hundred thousand are a minority of the population. However, given the huge swings of variation in all of these demographics, the natural logs of these variables are used instead of the numbers themselves. The intuition here is that a brewery, when considering location, would very much prefer a county of 100,000 over a county of 50,000, but be fairly indifferent between a county of 1,050,000 and a county of 1,000,000. Whether or not this assumption is true could be an interesting subject for further research.

The set of regressions run here is done four times. In doing so, it tests the same coefficients when counting number of breweries, with and without weighting for output. It also considers one sample with the entire continental US, assuming that any state not covered by the Brewing News did not have any Craft Breweries at that point, and another that only features the states that the Brewing News covers. From now on, for the sake of simplicity, the set using the unweighted breweries and only the counties from the Brewing News will be considered the main regression model, the set with the same counties but with breweries weighted for output the weighted main regression model, the unweighted set with all counties included the secondary regression model, and the weighted sample of the same to be the weighted secondary regression model.

For each of these sets of data, the basic model being used tests either breweries or weighted breweries against the natural log of the number of 25 to 34 year olds in the county, the natural log of the number of white people living in the county, the natural log of the county’s median income, and the natural log of the number of votes for Obama in 2008. From here, a second regression is done, without controlling for median income. The point of interest here is what demographic trends, particularly with regards to race, can actually be explained by trends in median income. A third test replaces the log of the white population with the log of total population. This is another attempt to gain some insight into just how much the craft beer market is driven by the local white population. Finally, a fourth regression replaces the log of the number of 25 to 34 year olds with the median age, to consider the theory that it is not hip culture-creators, but baby boomers with increasingly fine tastes, that drives the craft beer sector. The results are displayed and described below.