The explosive growth of craft beer is one of the most interesting phenomena in U.S. consumer markets over the past 20 years. It is an almost perfect illustration of the J-curve concept.
We took a look at the characteristics of craft beer demand and found that affluence, importance of the services sector, and size of student population are the key factors driving craft beer penetration. New York is the leading craft beer market.
Craft brew has grown rapidly in the United States and now accounts for 22% of the total beer market in value terms. Today, there are almost as many brewers as in the 1870s; in 1980 there were less than a 100 brewers, today around 4,000.
Geographically, craft brewers today are found all over the country. This is in contrast to the early years when the market mainly existed on the west coast.
What drives the geographic distribution shown in the map above? Which cities have high or low brewer penetration? To understand this we built a statistical model covering 3,875 craft brewers.
We found that affluence is by far the most important demand driver, followed by the importance of the (high value added) services sector, and student population as a share of total population. We also controlled for population density (in a densely populated city like New York, there will be fewer craft brewers since each more easily can serve a larger market), being on the west coast where the craft beer phenomenon started, and being in the south (which has most dry counties / many non-drinking evangelicals).*
The predictions from the model are shown below with number of breweries** plotted against the dominant explanatory factor–affluent population.
Some of the individual cities are shown below. High consumption cities include the famous craft beer cities Seattle, Portland, San Diego and Denver, but also New York, Chicago, and Los Angeles. All these cities have more breweries than expected after controlling for the various factors described earlier.
We also calculated a “love of craft beer” index. New York shows the most love while Provo (arguably for religious reasons) shows the least love. All cities look intuitively correct.
Finally, a few words on the statistical learnings. The model has surprisingly good explanatory power given the modest effort involved in creating it. With more time and resources allocated, and with additional data, it is probably possible to build a highly reliable model.
— — — — —
Note: We have used number of breweries rather than volume sold as the demand variable because volumes are not readily available. We tested an allocation of state-level volumes to the individual breweries and the conclusion was essentially the same, but with more variability.
* The model has an R² of 0.70 and does not suffer from collinearity or heteroskedasticity.
** To make it possible to plot the full model, the number of breweries was adjusted for the independent variables not on the x-axis. This is perfectly sensible, but perhaps hard to understand.