Data Sources

The data collection for this project varied in its difficulty. All of the county-level data on, age, ethnicity, and population comes from 2010 Census data. In addition, county level median income data was found from the American Community Survey’s 5-year averages, 2006-2010.

For the political data in question, the paper relies on the research conducted by Dr. Mark Newman of the University of Michigan’s Center for the Study of Complex Systems.⁹ It is from his work that the statistics on the number of people who voted for President Barack Obama, Senator John McCain, and “other” are derived. In order to test the possibility that how many people total voted Democrat in a given county matters less than the overall political climate, a variable for the percent of people who voted for President Obama has also been added.¹⁰ The idea of using statistics from the 2010 election, which occurred at precisely the time that the other data in question was being collected was briefly considered again. Ultimately the decision was made to use the 2008 data instead. While the 2010 election may in certain ways be a more accurate predictor of the political affiliations at that moment in time, a problem arose in that the 2010 election did not see any national candidates, and therefore had no single standard with which to judge the political affiliation of a given population. That is to say, New Hampshire and Mississippi may both elect a Republican to the Senate, but that does not mean that they voted with equal degrees of conservatism, since New Hampshire tends to be considerably more moderate. This is evident in New Hampshire’s selection of Obama in 2008, while Mississippi, a far more conservative state, went strongly for McCain. It is also true that there are better indicators of political affiliation than the somewhat binary “did you vote for Obama or not?” test being used in this paper, such as the Cook Partisan Voting Index (PVI). The choice to use the number of Obama voters, rather than the PVI, is based upon the fact that the Census data used for this study is organized on a county basis, while the PVI is organized along congressional district lines. Although the PVI would provide a better indicator of political affiliation, assembling data based around congressional district lines would be an arduous task for the purposes of this paper.

By far the most challenging data collection process was that of assembling a list of craft breweries. One key distinction needs to be made here, because of the unique nature of this paper relative to others on the subject. That distinction is the breweries being counted in this study. The world of beer is full of classifications, and the production side of it is no different. The Brewers Association often acts as the ultimate authority on craft beer classification, and provides a helpful set of classifications (reproduced below):

• Microbrewery: A brewery that produces less than 15,000 barrels (17,600 hectoliters) of beer per year with 75% or more of its beer sold off-site. Microbreweries sell to the public by one or more of the following methods: the traditional three-tier system (brewer to wholesaler to retailer to consumer); the two-tier system (brewer acting as wholesaler to retailer to consumer); and directly to the consumer through carryouts and/or on-site tap-room or restaurant sales.
• Brewpub: A restaurant-brewery that sells 25% or more of its beer on site. The beer is brewed primarily for sale in the restaurant and bar. The beer is often dispensed directly from the brewery’s storage tanks. Where allowed by law, brewpubs often sell beer “to go” and /or distribute to off site accounts. Note: BA re-categorizes a company as a microbrewery if its off-site (distributed) beer sales exceed 75%.
• Contract Brewing Company: A business that hires another brewery to produce its beer. It can also be a brewery that hires another brewery to produce additional beer. The contract brewing company handles marketing, sales, and distribution of its beer, while generally leaving the brewing and packaging to its producer-brewery (which, confusingly, is also sometimes referred to as a contract brewery).
• Regional Brewery: A brewery with an annual beer production of between 15,000 and 6,000,000 barrels.
• Regional Craft Brewery: An independent regional brewery who has either an all malt flagship or has at least 50% of it’s [sic] volume in either all malt beers or in beers which use adjuncts to enhance rather than lighten flavor.
• Large Brewery: A brewery with an annual beer production over 6,000,000 barrels (Brewers Association).

These standards roughly correlate with those that the Brewing News Publications use to classify the breweries on their list, with a couple clarifications. A company that brews beer for a Contract Brewing Company (Pabst being the most prominent of its kind) is referred to as a Contract Brewery. While some of these breweries may, in fact, brew beer for craft breweries, the decision was made to prevent double counting breweries. Furthermore, the narrative of a contract brewery, whose sole purpose is to make beer because somebody else’s recipe has a demand for it, does not seem to provide much insight into the questions being explored here. As such, contract breweries have been omitted from the analysis.

Unfortunately, the Brewing News publications do not distinguish between a Regional Brewery that produces craft beer and one that does not. In the initial round of data entry, a number of Regional Breweries were added to the list which, upon further inspection were discovered to not be craft breweries and were therefore removed. A more complicated problem arose when investigating these breweries, in that a number of them, in addition to producing their own, well respected craft beer, have, in fact, done good deal of contract brewing. This was particularly true for Regional Breweries in the Great Lakes/Rust Belt region. Presumably the cause of at least some of this was old, large brewing companies with huge facilities that over time lost market share to multinational conglomerates and began brewing some beer for said conglomerates in order to cover some of their fixed costs. For the purposes of this paper, however, the decision was made that if the main beer brewed at that facility was, in fact, a craft beer, then they would be included. It seems important to make that fact very clear, because a closer inspection by regional beer experts could find that some breweries within the sample should not be considered Regional Craft Breweries.

Another major omission from the dataset was Brewpubs. This is in some ways unfortunate, because Brewpubs are very much a part of brewing culture. In fact, as of 2011, they make up 1,063 of the total 1,938 breweries in the United States (Brewers Association), and their omission means that the dataset being used for this paper is considerably different than that being used in most other works on this subject. There are three reasons for this. The first reason is that Brewpubs do not just offer beer. They are often interesting, enjoyable restaurants or clubs in their own right. A restaurant may brew their own beer, but that is in fact a somewhat small part of their overall business plan. With a sample size of this magnitude, it is difficult to tell which ones are legitimately breweries in their own right versus restaurants that happen to brew a keg or two of ale for customers who happen to be interested. Second, many brewpubs are franchises, and there was some uneasiness about whether or not counting the same brewery two or three different times in the same area was fair. The most important reason, however, is that the focus of this paper is on whether the behavior exhibited by craft breweries should be described as industry behavior, like in the case of their larger competitors, or as firm behavior, like in the case of a restaurant. Many brewpubs are, in fact, restaurants, and none sell any more than 25% of their beer off premise. It stands to reason that all brewpubs display almost entirely firm-like behavior. The concern is that the inclusion of one thousand or so of them in the sample would mean that even if the behavior of the microbreweries and regional craft breweries in the sample displayed entirely industry-like behavior, the findings of this paper would be more or less in line with the hypothesis.¹¹

Although new editions of the Brewing News magazine are published every two months, this study uses their respective October and November 2010 editions in order to better fit the demographic information given by the 2010 census.

Other than individual breweries, the most serious omissions of data in the sample of this paper comes in the form of geographical regions. Due to the fact that Puerto Rico did not vote for presidential electors in the 2008 election, and given that Dr. Newman’s data on Alaska was somewhat difficult to work with,¹² the decision was made to constrain the sample to the continental United States. In retrospect, this may have been somewhat unfair to Hawaii, which suffered neither of these problems; however, it, as well as several other states ran into trouble when it came to the Brewery Publications. While the Brewing News does cover the majority of the United States, several states are omitted. In some cases, it is difficult to tell whether this is because the authors of the publication are truly unaware of any brewing activity in these states or whether they simply do not have any staff interested in writing about those regions. In addition to Hawaii, there is no information in the publications about brewery activity in Iowa, Kansas, Kentucky, Missouri, Nebraska, North Dakota, or South Dakota. Some of these states may truly have no craft beer presence, but others, such as Missouri, are known to have had multiple craft breweries at the time the maps were published (Kressman). For this reason, any state not featured in the map is simply removed from the Sample, with no assumption made about how many breweries are in those locations. There is another somewhat drastic and noteworthy discrepancy between the data and the figures from the Brewers Association when it comes to the number of breweries active during that period. Provided that no mistakes were made in the manual entry of the data, the Brewing News publications list 457 microbreweries and regional craft breweries in the sample in the period preceding their October-November edition. This is substantially less than the 615 microbreweries and 81 regional craft breweries said to be open during this period by Brewers Association Statistics (Brewers Association), which totals to 696 craft breweries. It is probably worthwhile to consider the nature of these 239 missing breweries.

One of the more obvious solutions is that there are actually a number of breweries located in the omitted states. If those breweries were located in those omitted states, then the sample that excludes those states should still give satisfactory results for that area. It would be concerning however if it is assumed that there were no craft breweries in all the counties within the omitted states. This is particularly true because, excluding Hawaii, the states omitted are all Midwestern states whose many counties share a lot of demographic traits such as being overwhelmingly white, politically conservative, etc. So if, unbeknownst to the people publishing Brewing News, there was a craft beer hot spot in rural Kansas, that could have serious implications for the study.

An alternative theory regarding the missing data concerns the methods of data collection employed by Brewing News, who never explicitly state precisely how they get their information. No data source is explicitly stated, but each map does list contact information for business owners interested in having their businesses listed on the map.¹³ This leads to the possibility that the periodical’s sample of breweries is less than perfect, so much so that 239 breweries across the United States were not included. Whether or not this would be an issue from a methodological point of view is dependent upon whether or not there is a trend among brewers who did not choose to be listed in the publication, or whether those who wanted to be listed were a random sampling. Obviously, a sample of over half of the breweries in the nation, if randomly selected, should be sufficient to test. However, it could be the case that the Brewing News was not widely distributed or well known in certain areas that had a great deal of breweries in them. Denver County, CO, for instance, has a surprisingly small number of breweries given its reputation as somewhat of a hub of craft beer (Frosch), although the low number could also be easily explained by the omitted Brewpubs, as described above.

Finally, it is possible that the Brewers Association’s figure is somehow inflated. It should be noted that, like most of their publications, their statistics on the number of breweries does not include any data source (Brewers Association).¹⁴ Given the small nature of many of these breweries, it may be difficult for a centrally located trade organization such as the Brewers Association to adequately capture certain important facts about breweries, such as whether or not they are actually up and running or whether, while defining themselves as a microbrewery, they are in fact just a brewpub. The same could be said for self-described regional craft breweries that are, in fact, only microbreweries, as the list collected for this paper lists 20 of them, short of the 81 listed by the Brewers Association. These problems merit further research. However, it should be noted that the Brewing News maps were used as a data source only after the Brewers Association did not respond to a request for their data. Were those data made available in the future, it would certainly merit a follow-up study.

⁹ Dr. Newman has produced an excel sheet (as part of his working on a larger, more interesting project about maps of election results) that details the election results in every county in every US state.

¹⁰ Calculated as (Obama/(McCain+Obama+Other)x100.

¹¹ A suggestion made by Professor Gerald Friedman during the defense of this thesis is that brewpubs, while not included as breweries for the sake of the analysis, could be a useful variable in a regression model, as they demonstrate a clear local demand for craft beer. While, as noted above, many brewpubs are much more functionally similar to restaurants than to pubs, or make most of their alcohol sales on beverages other than their own, and thus would not necessarily demonstrate a high demand for craft beer, the point is taken. This would be a very interesting variable to include in future studies on this topic.

¹² The data were organized along metropolitan and micropolitan statistical areas, not by county.

¹³ The magazine does not just cover breweries, but also craft beer stores, bars, and home brew supply shops.

¹⁴ In the time since this paper was written, the Brewers Association’s website has added more data on craft breweries which may have been helpful.